Setting Priorities for HIV AIDS Interventions: A Cost-Benefit Approach

Setting Priorities for HIV/AIDS Interventions

To my wife, Elizabeth, my sons, Adam and Matthew, my daughters-in-law, Nancy and Carisa, and my grandchildren, Austin and Victoria

Setting Priorities for HIV/AIDS Interventions A Cost–Benefit Approach

Robert J. Brent Professor of Economics, Fordham University, USA

Edward Elgar Cheltenham, UK • Northampton, MA, USA

© Robert J. Brent 2010 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library Library of Congress Control Number: 2009940730

ISBN 978 1 84720 331 1

02

Printed and bound by MPG Books Group, UK

Contents List of figures List of tables and boxes List of abbreviations Preface PART I

vii viii x xii

WHY COST–BENEFIT ANALYSIS IS NEEDED TO SET HIV/AIDS PRIORITIES

1 Introduction to the book 2 Why not just simply do what is right and try to save lives? 3 Myths and misinformation 4 Counterintuitive results 5 What is wrong with setting any targets? 6 What is wrong with setting the particular MDG targets? 7 Cost–benefit analysis 101 8 Cost–benefit analysis 201 PART II

9 10 11 12 13 14 15 16 17 18 19 20

3 7 11 17 21 24 27 30

HIV/AIDS AS A HUNGER AND ECONOMIC DEVELOPMENT ISSUE

Introduction to Part II HIV and hunger Nutrition and HIV at the individual level Nutrition and HIV at the country level Income as a factor raising HIV rates Education as a factor raising HIV rates Islam as a factor lowering HIV rates Impact of HIV on agricultural households Agricultural policy and HIV interventions Sex and HIV I: the role of transmission Sex and HIV II: the role of concurrency Sex and HIV III: the role of networks

v

35 41 44 49 53 58 64 68 75 78 84 88

vi

Setting priorities for HIV/AIDS interventions

PART III

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

Introduction to Part III Threshold analysis theory Threshold analysis practice: the effectiveness of HIV education Threshold analysis practice: the benefits of avoiding HIV Threshold analysis practice: the costs of a possible HIV/AIDS vaccine Willingness to pay theory Willingness to pay practice: the benefits of condoms Cost minimization theory Cost minimization practice: the costs of treating TB Cost-effectiveness theory Cost-effectiveness practice: the benefits of ARVs Human capital theory Human capital practice: the benefits of female primary education Value of a statistical life theory Value of a statistical life practice: the benefits of VCT

PART IV 36 37 38 39 40 41 42 43

COST–BENEFIT METHODS AND APPLICATIONS

113 117 122 128 132 137 141 146 150 153 158

SOCIAL CONSIDERATIONS IN CBA

Introduction to Part IV Commodification: everything is seen as a commodity to be bought and sold What is so “social” about CBA? Fundamentals of CBA Social and private perspectives in CBA CBA and equity I: allowing for ability to pay CBA and equity II: allocating by time and other non-price methods Conclusions I: how not to set priorities for HIV Conclusions II: using CBA to set priorities for HIV

References Index

95 101 105 109

167 172 176 180 185 191 195 201 207 215

Figures 7.1 The marginal benefits MB and marginal costs MC of bread 28 8.1 The marginal benefits MB and marginal costs MC of malaria reduction 31 10.1 Vicious circle of malnutrition and HIV 42 20.1 Sexual network with a single bridge 89 20.2 Sexual network with three bridges 89 26.1 The market for condoms in Tanzania 118 28.1 The costs and benefits of alternative TB treatments 129

vii

Tables and boxes TABLES 3.1 5.1 5.2 9.1 9.2 11.1 12.1

18.1 21.1 25.1 27.1 27.2 29.1 30.1 31.1 32.1

Percentage knowing that a healthy-looking person can have HIV in five Demographic and Health Surveys Cost–benefit outcomes for alternative levels of prevention: Cost Case 1 Cost–benefit outcomes for alternative levels of prevention: Cost Case 2 2007 HIV infections, rates and deaths by region (2001 figures in brackets) HIV prevalence rates (per 100 000 population) in the United States by race/ethnicity and gender, 2006 Prevalence of micro-nutrient deficiencies in different HIV populations Prevalence of nutritional deficiencies and estimates of deaths and DALYs lost of children aged birth through four by region, 2004 Sexual behavior and actual versus predicted HIV prevalence rates in the United States and SSA Impact of preventative interventions on behaviors, percentage changes WTP for an HIV/AIDS vaccine in Mexico according to income Private benefits and costs of a condom social marketing program in Tanzania (in TZSH) Social benefits and costs of a condom social marketing program in Tanzania (in TZSH) The cost per patient from diagnosis to completion of DOTS treatment in 2001 (in US$) The cost-effectiveness of selected HIV interventions around the world (in US$) Cost-effectiveness ratios for antiretroviral therapy (US$ cost per DALY) The benefits of a life in Tanzania as the present value of 26 years of earnings (in TZSH) viii

13 23 23 37 38 45

51 82 96 115 125 126 134 138 144 147

Tables and boxes

33.1 34.1 34.2 35.1 35.2 39.1 41.1

The cost per student of seven years of primary education in Tanzania 1994/95 to 2000/01 (in TZSH) US occupational fatality rates by industry, 1992–95 national averages Estimates of the value of a statistical life in various countries Cost–benefit outcomes for VCT testing using the VSL method for estimating benefits (m = millions of TZSH) Cost–benefit outcomes for VCT testing using the HC method for estimating benefits (m = millions of TZSH) Private (employer) calculation of benefits and costs of HIV testing of employees and hiring an HIV negative individual Estimated value of waiting list time (£ per month)

ix

152 154 155 161 162 181 193

BOXES 23.1

Threshold ratio for balancing costs and benefits of an education intervention

107

Abbreviations ABC AIDS ANCs ARVs B C CBA CDC CEA CSM CSWs CV DALYs DAW DOTS E GCE GFATM GNP HC HH HIV IDUs LDCs MB MC MC MDGs MSMs MTCT NGOs NIDA PCP PEPFAR PLWA

Abstinence, Be Faithful and Condoms Acquired immunodeficiency syndrome Antenatal clinics Antiretroviral drugs Benefits (total) Costs (total) Cost–benefit analysis Centers for Disease Control and Prevention (US) Cost-effectiveness analysis Condom social marketing Commercial sex workers Contingency valuation Disability adjusted life years Division for the Advancement of Women Directly observed treatment short course strategy Effects (total) Global Campaign for Education Global Fund to Fight AIDS, Tuberculosis and Malaria Gross national product Human capital Head of household Human immunodeficiency virus Intravenous/injecting drug users Least developed countries Marginal benefit(s) Marginal cost(s) Male circumcision Millennium Development Goals Males having sex with males Mother-to-child transmission Non-governmental organizations National Institute on Drug Abuse (US) Pneumocystis pneumonia (US) President’s Emergency Plan for AIDS Relief People living with AIDS x

Abbreviations

PNP Public–private NGO partnership (model) PPP Public–private partnerships PSI Population Services International PWP Public–private workplace participation (model) ROAH Research on Older Adults with HIV SD Standard deviation SSA Sub-Saharan Africa STDs Sexually transmitted diseases STIs Sexually transmitted infections TB Tuberculosis THIS Tanzania HIV/AIDS Indicator Survey TZSH Tanzanian shillings UNAIDS Joint United Nations Programme on HIV/AIDS UNICEF United Nations Children’s Fund UNIFEM UN Development Fund for Women UNFPA United Nations Population Fund VCT Voluntary counseling and testing VSL Value of a statistical life WFP World Food Programme WHO World Health Organization WTP Willingness to pay/willing to pay

xi

Preface HIV/AIDS has claimed at least 25 million lives so far and many more deaths are to come. Some recent progress has been achieved, but it is time to take stock of what we have learned about the epidemic and examine what needs to be done. The starting point for the book is the realization that HIV/AIDS is much too complex a phenomenon to be understood only by reference to common sense and ethical codes. For years the infection was treated as if it were basically a medical problem, with an emphasis on biology, laboratories and medication. All these factors are important, but the realities of HIV/AIDS extend much wider and involve an understanding of social and economic determinants, especially as HIV transmission is largely determined by human behavior and the choices made by individuals. Since individual choices are central to the dynamics of the disease, one needs to have available a framework for choice that can unify the field. That framework is cost–benefit analysis (CBA). This book, therefore, will present the CBA framework in as simple a fashion as possible, given that HIV/AIDS is anything but simple, and that policies to combat it have to ensure that all the important considerations are recognized and incorporated into the analysis. CBA has its technical aspects that take time and effort to master. Nonetheless, the aim in this book is to strip away the barriers that would exist if the reader were to consult the many textbooks and handbooks that are devoted to CBA. At heart, CBA is a way of thinking about public policy issues. The thought process will be exposed and examined in such a way that the contribution of CBA can be readily appreciated. CBA is inherently an applied economics field; theory and practice are so entwined that one part cannot exist without the other. This feature of CBA is very important because, as we have just pointed out, logical reasoning can take us only so far in understanding what to do about HIV/AIDS. Data must be used to reinforce the reasoning, because without empirical support for our assumptions, interventions will not be successful. There are many in the HIV/AIDS field who recognize that HIV/AIDS policies need to be “evidence based”. But what form must the evidence take? CBA identifies not only what sort of evidence (data) needs to be collected; it also explains how that data should be assembled to help determine the HIV/AIDS policy decisions. xii

Preface

xiii

The book has four parts. Together the four parts are intended to be a re-examination of most of the important controversies that have cropped up both in the HIV/AIDS and in the CBA fields. Often the resolution of the controversies will follow as a natural consequence of applying the CBA framework to the issues. The first part introduces the main themes. It explains why understanding the disease and its transmission is not so simple and why CBA is needed. Part II summarizes what we know about HIV/AIDS, basically that in Sub-Saharan Africa it is a hunger issue and in the United States it is a sexual issue, especially when it involves minorities. Part III presents the main CBA evaluation methods and shows how they have been applied. Part IV goes into detail about what is “social” about CBA as constructed by economists. Much of this final part is devoted to non-economists, but mainstream economics is not always comfortable with CBA practice and so some attention is given to bringing economists on board the CBA train. Overall then, the audience for the book is anyone who is concerned about trying to improve health care interventions in the HIV/AIDS field, whether they be activists, policy-makers, economists, non-economists or concerned citizens. This book has its origins from the time I visited the University of Dar Es Salaam for a semester as part of the Fulbright research award on the cost–benefit analysis of HIV/AIDS interventions in Tanzania in 2003. Much of my understanding of HIV in Tanzania stems from my stay at the University. I wish to thank Longinus Rutasitara, the chair of the Economics department at the time, for all his help and encouragement. I would also like to thank Fordham University (which gave me the Spring semester off from teaching in 2007 to help write the book) and the Health Economics Unit at the University of Cape Town in South Africa (where I spent part of the semester in 2007 and was able to interact with a number of researchers and learn from their extensive HIV/AID experience, especially Edina Sinanovic, Susan Cleary and Michael Thiede).

PART I

Why cost–benefit analysis is needed to set HIV/AIDS priorities

1.

Introduction to the book

The human immunodeficiency virus (HIV) that causes the acquired immunodeficiency syndrome (AIDS) has continued to thrive in the world over the last two or more decades despite all our efforts to restrain it. Our interventions are either not working, or not working fast enough. So millions have died, are dying, and will continue to die for years to come. The time has come to step back and reflect on what we know and don’t know about the HIV/AIDS pandemic so that priorities for HIV interventions can be set. The central argument of this book is that HIV/AIDS is much too difficult a problem to try to tackle it with best guesses, common sense, medical and ethical principles alone. Decisions need to be made that rely on data that relate to individual behavior and preferences as contained in a formal economic evaluation. HIV is not like other public health disasters. Most public health hazards attack and kill off the most vulnerable people in society, especially the young and the old. HIV on the other hand targets largely prime age working adults. If you are a poor person living in a crowded home where others have TB or cholera, you are likely to catch these diseases. But with HIV there is more of an individual choice involved in its transmission. Economics can be defined as the science of choice. How are choices made in economics? The answer is: on the basis of benefits and costs. This is true for individuals deciding how many hamburgers to buy, as well as for governments who have to decide how much to spend on health care. If the benefits of a good or service are greater than the costs, then one chooses more of it. If the benefits are less than the costs, then one chooses less of it. And if the benefits are exactly equal to the costs, then one has the right amount of the good or service and so one continues consuming or utilizing the good or service at the same levels as before. So this book will be all about how the benefits and costs of various interventions need to be identified, measured and compared in order that HIV/AIDS policy priorities can be set.

TYPES OF HIV EPIDEMIC Countries can be classified into two categories in terms of HIV. The first category, which comprises the United States and most Western European 3

4

Setting priorities for HIV/AIDS interventions

countries, can be called “localized”, where HIV just affects high-risk groups, such as intravenous/injecting drug users (IDUs), males having sex with males (MSMs) and commercial sex workers (CSWs). The other category relates mainly to Sub-Saharan African countries where HIV is “generalized” in the national population. The main reason for drawing this distinction is that what constitutes an HIV/AIDS intervention differs in the two sets of countries. In countries with localized HIV populations, specific groups are to be targeted. Providing condoms and clean needles would seem to be the interventions that need to be evaluated first. However, in countries with generalized HIV populations, almost any kind of public policy change can be viewed as an HIV intervention. Agricultural, transport, trade and educational reforms must all be looked at with an “HIV lens” (Gillespie and Kadiyala, 2005). In these countries cost–benefit analysis (CBA) needs to be applied routinely to many different kinds of intervention, whether it is the time that a commodity market opens, or whether women’s education should be subsidized. This book will therefore cover CBAs related to both sets of population category. For generalized HIV populations we will discuss HIV policy as one primarily dealing with a hunger issue and not one only of sexual behavior modification.

WHY SETTING HIV PRIORITIES IS NOT SIMPLE Two fundamental reasons why the setting of HIV policies is not a straightforward exercise is due to the fact that HIV/AIDS epidemics are not unicausal and that the problems to be solved do not stay the same over time. Transmission can be due to heterosexual contact, MSMs, IDUs, blood transfusions and mother-to-child transmission (MTCT). HIV epidemics are also long-wave phenomena. The five waves are HIV infection, opportunistic infections, AIDS, death and impact. Some countries appear to be over the first wave, including the United States, Uganda, Thailand and Brazil. But no country is over the death wave and the impact wave is only just beginning (Gillespie and Kadiyala, 2005). As a consequence there are a host of different possible interventions to pursue and what one chooses to do at one point in time is not necessarily optimal at all periods of time. One size does not fit all; nor does one time fit all times. Again this points to using CBA to evaluate many different interventions and reusing CBA on many different occasions. This book aims to act as a complement (two goods with quantities moving in the same direction) to that by Jeffrey Sachs’s (2005) book entitled The End of Poverty. He argues that, if developed countries just

Introduction to the book

5

devoted 0.7 percent of their national income to foreign aid, then poverty could be eliminated. But, poverty will not be eliminated unless major successes related to HIV/AIDS are achieved. And ensuring sufficient funds is only half of the problem. How they are to be spent is just as important and CBA needs to be enlisted in the task of deciding what to invest in. Sachs’s book has a chapter outlining the details of a specific project that would reduce poverty. But it is not shown that this project actually would be better than other development projects. Part of our book will be evaluating HIV intervention projects in countries where poverty is widespread (such as Tanzania). Obviously, reducing poverty can be expected to be worthwhile. However, the object is still to make countries better off and only the use of CBA will establish this.

OUTLINE OF THE BOOK We will devote Part I of this book to spelling out in greater detail what is needed and why only CBA can be relied upon. We will first explain why the identification of HIV priorities is not straightforward and list some of the things that we think we know, and what we think makes sense, which are, in fact, things that are not true. Myths and counterintuitive results abound. Current HIV strategies assume that we already know what to do in any given country and that the task now is simply to ensure that this is “scaled up” to all the population. Hence the emphasis given to the Millennium Development Goals (MDGs), which simply set dates by which specified targets are to be achieved. Scaling up is not as obviously desirable as it is assumed, and we analyze its role in Chapter 7. Our second task in Part I is to explain why it is that setting any goals in the abstract devoid of CBA is not helpful in general, and also not helpful in the case of the particular goals set out in the MDGs. We close Part I with a summary of some of the key principles of CBA and how they relate to the setting of HIV policy. Part II switches from things we do not know to an account of some of the things that we do know about the HIV epidemic. We focus on SubSaharan Africa (SSA) where most people living with the HIV disease are located. The main theme is that in SSA, HIV/AIDS is primarily a hunger issue. Many people think that the reason HIV is so widespread in Africa is because, in a context where heterosexual activity is the main transmission mechanism, sexual activity must be higher than elsewhere. It turns out that when people are healthy, sexual transmission of HIV is inefficient. But, when people are not healthy, and suffer from malnutrition and parasitical

6

Setting priorities for HIV/AIDS interventions

diseases, immune systems are greatly compromised and HIV transmission is greatly facilitated. Focus is then given to how multivitamins can be used as a micro-nutritional supplement. From there we turn to macronutritional issues and how agriculture policies can be viewed using the HIV lens. The emphasis here is largely on national and regional evidence. The main message will be that in countries with widespread epidemics, almost any change in institutional arrangements surrounding agriculture can be viewed as a potential intervention for HIV. Policies involve trade-offs and these need to be identified and quantified. Since the whole raison d’être for CBA is to deal with trade-off situations where advantages need to be lined up and compared alongside disadvantages to see where the net position lies, the scene is now set to cover specific CBAs of various HIV interventions. Part III is therefore devoted to the presentation of a number of CBAs of HIV/AIDS interventions. We deal with some related to countries with localized epidemics and others pertaining to countries with generalized epidemics. As we shall explain, cost-effectiveness analysis (CEA), the technique of choice in the health care evaluation field, cannot be used to set HIV intervention priorities, so we shall show how it can be reconstructed as a CBA criterion. Since there is no real alternative to using CBA for the setting of HIV priorities, we will present a number of different methodologies for measuring the benefits. Readers should then be able to choose and embrace at least one CBA methodology that they are comfortable with. Having seen how CBA operates in practice, Part IV will draw together some of the conclusions regarding the strengths and weaknesses of CBA. We examine the welfare economic base to CBA and discuss equity considerations as they relate to setting HIV priorities. CBA may be a terrible way to carry out health care evaluations. But, it will be seen to be the best way there is.

2.

Why not just simply do what is right and try to save lives?

When I visited Tanzania, and was beginning my research on CBA and HIV/AIDS, I met a director of a non-governmental organization (NGO) that had run a number of interventions in one region of the country. When I asked what CBAs of their programs had been undertaken I was told that no CBAs had been undertaken because they were not necessary. When I asked why there were unnecessary, the reply was that the NGO’s activities were worthwhile because they saved lives. In response to my next question as to which of the interventions were necessary to save lives, I was informed that all of them were necessary. To an economist this judgment was thought to be highly suspect. Was it really the case that if one person in the NGO did something else in the region that everything would grind to a halt? And without any empirical evidence, how could one be sure that, relative to doing something else, lives were being saved? Economics is all about considering increments in activities. Economists ask, what would happen if there were a little more of this and a little less of that? If everything that was being undertaken was necessary then a little less would eliminate everything of value. Even if this were true, surely the NGO should try to find out how many extra lives would be saved if activities in the organization were expanded? So some evaluation would be necessary even if saving lives was the only yardstick by which success were to be judged.

THE NEED TO MAKE CHOICES The key point here is that setting HIV policy is not “simply” a matter of trying to save lives, but a question of how many lives one can save given the resources available. Economics is fundamentally about trying to obtain more rather than less. If more lives can be saved by expanding or rearranging activities then this should take place notwithstanding that currently some lives are being saved. Economists express this idea more formally by pointing out that if one uses resources in one particular way, then one is foregoing the opportunity of using resources in some other way. 7

8

Setting priorities for HIV/AIDS interventions

This leads to the notion of cost in CBA as one of “opportunity costs”. So the decision to devote resources to a particular HIV intervention needs to recognize that other ways of saving lives are being turned down. It is not just that if one provides condoms to prevent HIV, then one will be giving up resources that could be use to provide antiretroviral drugs (ARVs) for those with AIDS. There are many diseases that kill people other than HIV/AIDS. More children in Africa die due to malaria than AIDS. Furthermore, spending on health is not the only way to save lives. Food keeps people alive too. Thousands die every day due to malnutrition throughout the world. One needs to have a comprehensive evaluation framework that can recognize the many different ways that lives, and the quality of lives, can be affected. CBA is the most comprehensive framework because it uses monetary terms to express outcomes and in this way bring every possible alternative use of funds into alignment. Expenditures on housing, education, transport, the environment, criminal justice, national security, among others, can all be evaluated in monetary terms such that one spends the most on those uses of funds that give the highest monetary net benefits.

ARGUMENTS FOR TRYING TO AVOID MAKING CHOICES It is at this stage when there is talk of monetary valuation that the noneconomist starts to get annoyed. How can one possibly put a monetary value on a life saved? And who says that one cannot have all the basic needs such as food, housing and health at the same time? We deal with these two questions in turn. First, notwithstanding that in some approaches to CBA monetary values are directly applied to lives saved, best practice in CBA is not to go around and ask people how much they would be willing to pay to save their lives. Obviously, the answer for most people to this question would be an infinite amount, or almost equivalently, all the money in the world. Instead, best practice in CBA asks the question, for a specified risk of losing your life, how much would be required as compensation? These kinds of question individuals routinely face in their daily lives. The choice of occupation often depends on such a compensation mechanism. Why are people in Southern Africa willing to work underground in mines and possibly die? It is because they expect to get a higher wage than if they did not work in the mines. Some people live near environmental hazards because house prices are cheaper. In many different ways, such as choosing which mode of transport to use to make one’s journeys to work, people trade off risk for financial

Why not try to save lives?

9

gain. So it is not an actual life that one is trying to value but what is called a “statistical life” (after Thomas Schelling, 1968). In the course, say, of heading a large multinational company, those involved may be more susceptible to a heart attack because there is great pressure to perform. If out of 10 000 CEOs, ten did die during a year due to stress, then there would be an annual 1 in 1000 chance of dying, on average, that someone thinking about being a CEO would expect to face. Should the salary be $10 000 p.a. more than someone not under so much pressure, then a 100 percent chance of dying would be 1000 times greater than a 1 in 1000 chance of dying, and so 1000 × $10 000, that is, $10 million, would be the value of a statistical life. Note that the identities of which CEOs will die are not known. If things turn out well, perhaps no CEO would actually die. But given that a person did accept the extra $1000 for doing the CEO job with the specified expected risk of dying, then, implicitly, that person would be valuing his/her life at $10 million. As we shall see in Part IV, if that is how much people value their lives, then by what authority is an economist to disagree with that valuation? The second question, as to why one cannot have condoms and ARVs, or health and education, has a simple answer. It is a fact of life that resources are limited and so one cannot do everything that one would like to do. One has to make choices and, as stated earlier, economics is the science of choice. The need to make choices does not go away just because people think it is wrong to have to make choices. The old Soviet Union considered that economics was bourgeois and therefore expendable. But they eventually realized that gold medals in Olympics and sending people into outer space had an opportunity cost in terms of foregone agricultural output and this “price” was too high to pay. One author who is clearly exasperated by the need to have to make choices when determining HIV policies is Stephen Lewis. In his book on HIV entitled Race Against Time, Lewis (2005, pp. 157–8) records an interchange he had with a very senior member of the World Bank. The official told him: “The people with AIDS are going to die. The money would probably be better used for prevention. It’s all a matter of tradeoffs”. Lewis then tells us his response: “I couldn’t believe what I was hearing. ‘Trade-offs,’ I sputtered. ‘You speak to me of trade-offs? You have drugs to keep people alive, and you’re going to let them die because of a trade-off? Why don’t you find more money and do both treatment and prevention, and screw the trade-off?’” We can all share the frustration of Stephen Lewis, but resource constraints are a fact of life and not just a figment of the imagination of economists. In fact, resource constraints are a part of Stephen Lewis’s reality. Just six pages later in his book, after expressing his support for

10

Setting priorities for HIV/AIDS interventions

the World Food Programme’s (WFP) advocacy of school feeding programs he writes: “It’s hard to find anyone who’s associated with school feeding who doesn’t feel that it should be universal. At the moment, alas, such thoughts are the stuff of fantasy. The WFP simply doesn’t have the money”. So Lewis does actually accept, even though he does not like the fact, that if additional money is not made available then choices will have to be made. To conclude: even though it may seem that economists are “trying to play God” by using CBAs to help to determine who lives and who dies, this is not the case, seeing that resources are scarce in this world and this reality means that we cannot save everyone we would like to. It is because individuals need to be compensated for the risk of losing their lives, and the lives of others, that we can be confident that matters of life and death are not ignored by CBA and that these considerations will be captured by the monetary outcomes.

3.

Myths and misinformation

CBA attempts to quantify the benefits and costs of any intervention. It does not just assume that these benefits and costs are known. In the case of HIV/AIDS this is very important as there is a lot of information circulated that is either misunderstood or flat out false. In part, this is due to the stigma associated with the disease, which makes denial very prevalent. If people say that they are not HIV positive, or do not know (or do not want to know) whether they are HIV positive, how can one know what is best in terms of treatment or prevention? In part, the lack of knowledge is an inherent ingredient of HIV/AIDS as it is especially complex, both biologically and socially. One sign of the biological complexity is that AIDS is not a clearly specified condition in itself. If your immune system is compromised due to HIV you become susceptible to a number of opportunistic infections – for generalized epidemics it is likely to be TB, while in localized outbreaks it may be some rare forms of cancer of the blood (such as Kaposi’s sarcoma) or rare kinds of pneumonia (such as pneumocystis pneumonia, PCP). So when someone “dies of AIDS” it may not always be recorded as AIDS and could be classed by the proximate cause of death, for example, as due to TB. An example of the social complexity of the disease is that, prior to ARVs, the disease AIDS would normally follow HIV only after a period of five to ten years. During that period, one may have been able to function normally even though HIV was present. During that period, the capacity to infect others sexually would not be constant. An HIV positive person was highly infectious for the first few months and then again for the last few months when AIDS was present. In the meanwhile the probability of infecting others was very low. In the current AIDS era having unprotected sex is always risky. But, the risks are not the same over time, so what is rational behavior would change over time. Society is often very uncomfortable with the idea that rules governing behavior need to be nuanced. Knowledge of the means of transmission of HIV is important in enabling people to protect themselves and others from the disease. Surveying how knowledgeable people are about the transmission mechanism is often used as an indicator of the effectiveness of HIV educational programs.

11

12

Setting priorities for HIV/AIDS interventions

For example, in the Tanzania HIV/AIDS Indicator Survey 2003–04 (THIS, 2005) knowledge of three HIV transmission facts were tested. Did the respondent know that: (1) people can reduce their chances of getting the AIDS virus by having sex with only one uninfected, faithful partner and by using condoms – the so-called “ABC” (Abstinence, Be Faithful and Condoms) of HIV; (2) a healthy-looking person can have the HIV virus; or (3) HIV cannot be transmitted by mosquito bites or by sharing food with a person who has AIDS? It is instructive (and sobering) to know that the bar for knowledge of HIV transmission is set so low that as late as 2004 in a country such as Tanzania where 2.3 million were living with HIV and 150 000 had died of AIDS in just the previous year, that the THIS (2005, p. 57) stated: “It is encouraging [my italics] that 44% of young women and 49% of young men know all of these facts about HIV/ AIDS”. Let us look a little deeper into why there is this knowledge deficiency by examining the determinants of one of the questions related to HIV transmission.

WHO KNOWS THAT A HEALTHY-LOOKING PERSON CAN HAVE HIV? De Walque (2006) analyzed recent Demographic and Health Surveys in five African countries. He focused on the one question regarding whether someone who looks healthy can have HIV. This question was viewed as a good indicator of the general state of knowledge in a country concerning the HIV/AIDS epidemic and a piece of knowledge with important implications for prevention. Given that the disease does take time to transform into AIDS, the answer to the question should be: “Yes, someone who looks healthy now could still have HIV”. Table 3.1 reports the percentage of people who answered the question correctly in the five countries, giving the response by gender. Although there is a lot of variation across countries, uniformly it was the case that females were less informed. The one variable that explained the variation across all five countries and for both sexes was the number of years of education that a person had. Since females on average have fewer years of education, this is one explanation for the gender–knowledge disparity. Other than years of education, there was no determinant that was statistically significant for all ten groups (both sexes in each of the five countries). In a few cases the results had different signs for the same determinant. For example, coming from an urban and not a rural area had a positive effect on HIV knowledge for males and females in Burkina Faso, but it was negative for males in Kenya (while

Myths and misinformation

Table 3.1

Percentage knowing that a healthy-looking person can have HIV in five Demographic and Health Surveys

Country Burkina Faso 2003 Cameroon 2004 Ghana 2003 Kenya 2003 Tanzania 2004 Source:

13

Percentage of Males Who Know

Percentage of Females Who Know

71.28 79.84 77.03 90.33 84.44

57.35 68.68 63.32 86.16 78.76

Based on De Walque (2006) Table 14 (for the World Bank).

still positive for urban females in Kenya). On the whole then we see that what is known in one country or group may not be known in another country or group.

MYTHS In addition to the existence of information that is true that people don’t know about, there is also information that people think is true, but is actually false, that is, myths. Irwin et al. (2003) have written a book listing a number of myths concerning HIV/AIDS. Here we cover just four of the myths that they examine. We mention them because, as explained below, if believed, they all seemingly present obstacles to an awareness of the need to carry out CBAs. However, note that two of the ten alleged myths mentioned by Irwin et al. are not myths. That is, they claim as myths (p. 134) that: “Financial resources for global health are extremely limited, so public health officials in poor countries should prioritize programs” and (p. 73) “Very few poor countries have adequate facilities and services” and “trying to deliver ARVs without sufficient infrastructure” has the danger that “drug resistant strains could develop” because of these limitations. As we explained in the last chapter, there is a shortage of funds, so not everything we would like to do can be funded. Also, poverty does reduce the effectiveness of ARV treatment in developing countries relative to richer countries. There are side-effects with ARVs and there are shortages in trained health care staff to monitor these side-effects in poor countries. This does not, of course, mean that ARVs cannot be worthwhile. Again it just points to the need to carry out a CBA to see whether they are worthwhile without adequate monitoring.

14

(i)

Setting priorities for HIV/AIDS interventions

AIDS is Primarily an African Problem

The claim is that other countries are unlikely to be affected to the same extent as African countries so their experience with HIV/AIDS is not of much relevance to other countries. The lie to this claim is readily apparent when one is aware of the fact that the country that (as of 2007) has the third largest absolute number of HIV cases is not in Africa, but is in Asia, that is, India with 2.4 million people infected (just 0.2 million below Nigeria, which has the second highest number; South Africa has by far the highest total with 5.7 million cases). Also note that it is the poor and marginalized populations in all countries, including the United States, who are most affected by HIV. The CDC HIV/AIDS Surveillance Report for 2008 shows that HIV/ AIDS rates for African American females were 18 times the rates for white females and four times that for Hispanic females – see Chapter 9. (ii)

Prevention is Better than Cure

Many think that available resources should be given to prevention and that costly treatment via ARVs should wait until prevention programs have been fully funded. As Irwin et al. (2003, p. 64) point out, there are limits to what prevention efforts can achieve, especially as the spread of HIV has not been halted. “Understanding how social and economic factors determine individuals’ vulnerability to infection does not necessarily mean that public health officials will be able to alter these patterns”. Moreover, prevention and treatment are not mutually exclusive. As we shall see in greater depth in Part II, prevention and treatment (and mitigation) all support each other. (Some analysis of how to compare treatment and prevention takes place in Chapter 8.) (iii)

There is Nothing to be Gained from Helping Other Countries

Since it is perceived that the constituents of wealthy countries are not affected, politicians in these countries think that AIDS is a low-priority issue. This perception by politicians is false seeing that we all now live in a very globally interconnected world. All countries are affected by foreign trade, tourism, labor migration and cross-border refugee flows. Nowhere is the need to devote resources to prevention efforts abroad as well as at home more necessary than in the United Kingdom, where persons infected in Sub-Saharan Africa were the group most affected. More than threequarters (77 percent) of newly diagnosed HIV infections in 2004 were contracted in high-prevalence countries (UNAIDS, 2006a, p. 47).

Myths and misinformation

(iv)

15

There is Nothing We Can Do

The AIDS crisis is considered by some to be just too big. According to the UNAIDS (2008), there were 2.7 million persons newly infected with HIV and 2 million AIDS deaths worldwide during 2007. The history of HIV is not just a list of failures. There are also successes. UNAIDS (2008) reports that, in 14 of 17 African countries with adequate survey data, the percentage of young pregnant women with HIV has declined since 2000–01. Also, thanks to ARVs, HIV is now no longer automatically a death sentence. There were around 1.5 million people who were on ARVs by the end of 2006 according to Lewis (2005). Nattrass (2006) reports that in South Africa the numbers on ARVs rose a hundredfold between October 2003 to the end of 2005 (from less than 2000 to almost 200 000). Nor is it the case that every HIV/AIDS target is not reached. Nattrass cites the Western Cape as having reached 130 percent of the ARV target set for it by the Department of Health’s Operational Plan. The Global Fund to Fight AIDS, TB and Malaria (GFATM) and the United States’s President’s Emergency Plan for AIDS Relief (PEPFAR) have brought billions of additional dollars to assist in the fight against HIV/AIDS. The challenge is to ensure that these extra funds are directed by CBA to the best uses and not wasted.

THE RELEVANCE OF MYTHS AND MISINFORMATION FOR CBA The problem with believing the four myths just identified is that one would think wrongly that one can easily formulate HIV priorities without reference to CBA. If AIDS is purely an African problem then other countries do not need to have an HIV plan. If prevention is better than cure, then why bother evaluating ARVs? If people in the United States have nothing to gain by fighting AIDS in the developing countries, then benefits in the United States can never outweigh costs. And if there is nothing that the rich countries can do to help the struggle against AIDS in poor countries then again why carry out a CBA if it cannot possibly uncover any intervention that is worthwhile? The position we take on the other hand is that we need to carry out CBAs of all kinds of HIV intervention and let the results reveal what is worthwhile or not. Why prejudge the outcome when, as we have seen, general information abut HIV/AIDS is so unreliable? Even if general information were reliable, it would still be necessary to carry out CBAs in order to allow for divergences in known trends. Let us consider one implication of the De Walque findings reported in Table 3.1.

16

Setting priorities for HIV/AIDS interventions

Say you are considering persuading females in a high school in an African country to use free condoms that your agency is providing. If the female at the school thinks that condoms are not necessary, because their sex partner does not have HIV because he looks so healthy, then the condom promotion program would not work as the condoms would not be used. On the other hand, if the females know that a healthy-looking person can have HIV, then the condoms might be used. One should therefore expect that in a country like Burkina Faso where 43 percent of the females don’t know the HIV facts, that the outcome of a CBA of condom promotion in schools would be much worse than in a country like Kenya where only 14 percent of the females don’t know the facts. To conclude: allow the data to tell you what is worthwhile rather than assume that a condom program that worked in one country in one time period (with a given level of HIV general knowledge) can bring the same results as a condom program in another country at some other time period (with a different level of HIV general knowledge). What really confirms the fact that we cannot assume that we already know what needs to be done to combat HIV is the fact that in some countries not only is knowledge of the epidemic incomplete and wrong, the knowledge base can also actually deteriorate over time simultaneously with the worldwide epidemic growing even larger. The UNAIDS (2006b, pp. 56–7) update tells us that in a 2005 survey in the United Kingdom, 79 percent of respondents nationally knew that HIV can be transmitted through unprotected sex when earlier (in 2000) the percentage had been as high as 91 percent. Condom promotion that was done as recently as five years ago may not now be worthwhile or may have to be repeated.

4.

Counterintuitive results

We have just seen in the previous chapter that there is a lot of misinformation about HIV/AIDS. Even when we obtain the correct information, we may not make the best use of the information if we go straight to conclusions without going through a systematic evaluation exercise using the data. Let us look at some strategies that intuitively would seem to make a lot of sense as ways of preventing the spread of HIV, but do not end up as furthering the cause. In each case we explain the logic and why the facts may not fit the logic.

ABSTINENCE Consider a country in which there are two groups: a high-risk group that has ten or more partners and a low-risk group with one sex partner. Given that those in the high-risk group have more partners than those in the low-risk group, they are more likely to be HIV positive than those in the low-risk group who are more likely to be HIV negative. Persons in the high-risk group have sex with others in the high-risk group, but they also have sex with persons in the low-risk group. Those in the lowrisk group partner only people in the high-risk group. As a result of all the pairings that have one partner HIV positive and one partner HIV negative, there is a national HIV prevalence rate of, say, X percent. Now introduce an informational program that encourages abstinence. If the program has an impact only on persons in the low-risk group, then this group will no longer interact with the high-risk group. The new HIV prevalence rate could now actually rise to greater than X percent. How could this possibly happen? The only people having sex with those in the high-risk group are others in the high-risk group. Their chances of getting the infection have risen and this would lift the prevalence rate. It is true that the new abstainers can no longer get HIV and this would bring the prevalence rate down. But, the net result would be a higher rate if the numbers now getting infected exceeded the numbers now no longer getting infected. Kremer (1994) was the first to model the possibility that withdrawing

17

18

Setting priorities for HIV/AIDS interventions

the low-risk group from the pool of partners could raise the prevalence rate. He followed up his theory with an application related to data in the United Kingdom, which showed that, for this country, raising the prevalence was not just a possibility; it actually took place. For HIV to continue over time in a population, on average over a person’s lifetime, every person infected must also infect at least one other person. If less than one other person is infected, the disease will die out. Kremer supplies some figures to illustrate how abstinence can be counterproductive by raising the chances that more than one other person will be infected per infected person. Assume that for the disease to continue, each person must have seven partners a year to end up infecting one other person. Say that all sexual partners meet at a bar. This bar is frequented by two groups each consisting of the same number of members, a high-activity group that finds a partner eight times a year, and a low-activity group that finds a partner two times a year. On any given night, the bar will be populated more by the high-activity group, in the ratio of 80 percent to 20 percent. A person going to the bar to meet someone for sex, who chooses the partner at random, would have an 80 percent chance of meeting someone with eight partners and a 20 percent chance of meeting someone with two partners. The weighted average number of partners would be 6.8 partners a year (that is, 8 × 0.8, plus 2 × 0.2). HIV would die out at this partner rate. Now let the low-activity group cease to go to the bar at all. The only people at the bar would be the high-activity group. There would be a 100 percent chance that they would be meeting someone with eight partners. So on average there would be eight partners a year. Even though the low-activity group does not in any way add to the population average number of partners of eight, this number (as it is greater than seven) is sufficient to ensure that the disease would now rise over time instead of falling. What the numbers in this example make clear is that the average number of partners in the population is largely determined by the high-activity group and that if this group never meets partners with lower partner rates, the overall rate will go up. So in terms of containing the disease, abstinence by the lowactivity group is not helpful.

MARRIAGE For many people, abstinence cannot be regarded to be the single long-term solution to the HIV/AIDS problem. Apart from the desire to experience sexual pleasure, a fundamental reason to have sex is to have children. From this perspective neither abstinence nor the use of condoms is useful.

Counterintuitive results

19

One reason why some people get married is to have children. Hence abstinence is often linked with marriage as a part of a package. In this context, someone would practice abstinence for the period prior to marriage, then get married and have children. If one’s partner followed the same path, then HIV/AIDS could be largely avoided. Why would postponing sex until one is married not be helpful? The trouble is that a person has full information of, and one can try to control, only one’s own sexual behavior. One’s marriage partner’s sexual history may not be known. Typically, males are older than females when they get married. In Sub-Saharan Africa the gender–age difference at marriage can be as much as ten, 20, or even 30 years. It is when the husband is so much more sexually experienced than the wife that marriage can become a problem for younger females. This possibility had become so much a reality that UNAIDS (2004) even considered that for many women their main HIV risk factor is being married to a husband with previous or current other sex partners. It point outs that among sexually active girls aged 15–19 years in the cities of Kisumu (Kenya) and Ndola (Zambia), HIV levels were 10 percent higher for married than for sexually active unmarried girls, and that in rural Uganda, among HIV-infected women aged 15–19 years, 88 percent of the girls were married. It concludes (p. 10) that, “persuading girls to abstain from sex until marriage is of little help”.

TESTING UNAIDS (2006a) reports that in the United States around 25 percent of those with HIV do not know they have the disease. In which case, the chances of them infecting others were very high. One might think, therefore, that if more people were tested, and the governments were to subsidize the tests to encourage this to take place, then HIV transmission would be slowed. However, in the US context at least, this subsidy argument can be questioned. There are two main reasons. First, more risky sex and not less might take place as a result of testing. As explained in Philipson and Posner (1995), one of the selfish motives for getting tested would be in order to obtain unprotected (and hence risky) sex when without the test one was previously only getting protected sex. If a person were known to have many other partners, then a prospective new partner may insist that sex will take place with that person only provided that a condom is used. Taking a test would be beneficial here because, if it were found to be negative, then this result could get the new partner to change his/her mind about requiring that a condom be used. In line with this type of reasoning,

20

Setting priorities for HIV/AIDS interventions

we would expect to find (as we do actually see in many countries) that those being tested were high-risk groups (a higher percentage were found to test positive than in the population as a whole). Second, the tests may increase the number of sex partners. According to Boozer and Philipson (2000), one tests if one expects to gain from the knowledge revealed by the test. How much knowledge one gains depends on what information a person has prior to the test. Say a person thinks they are HIV positive going in to the test, because risky sex is practiced with a number of partners. In this case, an HIV test would just confirm what they already know. Sexual behavior would already have changed if it was going to. So why have the test? Similarly, if someone practiced abstinence, one would expect to be HIV negative. A negative HIV test would lead to the exact same behavior as before the test. Again, if nothing is gained by testing, the test is not worth doing. The Boozer and Philipson policy conclusion is therefore that the government should not subsidize the test for everyone. Only those who might learn something new from the test and would change their behavior in a way that lowered transmission should be encouraged to test. In their empirical work for the United States they found that, indeed, those who did not learn anything did not change the number of partners. For those who thought they were negative, but tested positive, their behavior did not change. However, those who thought they were positive increased their number of partners by 20 percent when they learned that they did not have HIV. Overall, testing could be expected to raise the transmission rate. The point about highlighting these three counterintuitive results is not to argue that abstinence, marriage and testing cannot be a part of the solution to HIV/AIDS transmission. Rather, the point is that only by considering them in the context of an actual behavior intervention exercise, which will be evaluated on the basis of results of any behavioral change, will progress be made. In particular, counseling services will probably have to be provided in each case to ensure that behavior change in the “right” direction takes place, whether the person is getting HIV tested, or getting married, or belongs to a high-sexual-activity group. Behavior change cannot be assumed to take place just because we think this is what should take place.

5.

What is wrong with setting any targets?

We will examine the desirability of setting targets in the context of the MDGs, though the considerations discussed apply to any set of targets, such as the recent “3 by 5 Initiative” whereby the WHO (World Health Organization) set out the target to put 3 million people into HIV/ AIDS treatment by 2005. The MDGs were derived from the Millennium Declaration, unanimously adopted by world leaders at the 2000 Millennium Summit. They have become an organizing framework for UN development work for both donors and developing nations. We will focus just on Goal 6, which is aimed at combating HIV/AIDS, malaria and other diseases. There are two targets assigned to Goal 6 (UN Millennium Project, 2005, p. xvi): Target 7: Have halted by 2015 and begun to reverse the spread of HIV/AIDS Target 8: Have halted by 2015 and begun to reverse the incidence of malaria and other major diseases

There are two main problems with setting any targets. The first involves their feasibility and second their desirability.

FEASIBILITY What happens if the set of targets one specifies cannot be achieved given the resources, technology and time at hand? Say one can achieve Target 7 or Target 8, but not both. Which one does one try to achieve completely? Is achieving one target completely better than tackling both targets and achieving some, but not full success for either? Targets only give guidance for policy-makers when they can be met fully.

DESIRABILITY Even if targets are feasible, are they desirable? If achieving full success is not feasible, is some success better than none? Only CBA can answer these 21

22

Setting priorities for HIV/AIDS interventions

two questions. So targets only make sense when they are backed up by numbers that come out from undertaking a CBA. To illustrate how CBA can inform the process of setting targets, let us assume that we know the total costs and benefits of various levels of preventing HIV. To keep the arithmetic simple and easy to interpret, assume that there are 100 persons per year who newly get HIV and these numbers can be reduced by the government’s provision of free condoms. Benefits per person are $10 throughout, so total benefits depend simply only on the numbers benefiting. As there are 100 persons infected, the total benefits of completely preventing the disease would be $1000. (All dollar amounts can be expressed in millions if numerical simplicity is a barrier to understanding the logic of the illustration.) The costs have first a fixed and then a variable component. Forty of the potentially infected persons live in a town and 60 live in a rural area. In town there are no transport costs. The only cost is $200 for setting up a clinic that distributes the condoms. So this $200 is fixed and has to be incurred irrespective of whether all 40 living in the town next to the clinic, or fewer than 40 people, are cared for. The variable costs just apply to the 60 persons living in the rural area who have to travel to the clinic. We consider two cases for the transport cost per person. In Cost Case 1 the travel costs are $5 per person (so the variable cost for all 60 rural persons would be $300, making a total cost of $500 including the clinic cost); and in Cost Case 2 the travel costs are $15 per person (so the variable cost for all 60 rural persons would be $900, making a total cost of $1100). The costs and benefits for all levels of prevention for the two cases are shown in Tables 5.1 and 5.2. With Cost Case 1 in Table 5.1, “halting” the spread of the HIV, that is, preventing 100 cases, is not only desirable, it is also the most desirable intervention level. The net benefits of $500 are greatest with 100 cases prevented. However, targeting runs into problems with the higher transport costs shown in Table 5.2. Building the clinic for 40 townspeople is the most desirable intervention level. Preventing 80 breaks even, but it is not the most desirable level. Halting the spread of HIV is not now desirable at all. The point then is that with knowledge of costs and benefits, policymakers can not just set targets, they can set the most desirable targets. Without knowledge of the costs and benefits some target-setting may not make any economic sense. Does it ever make sense to set a target for prevention less than 40? In the absence of the information in Tables 5.1 and 5.2, caring for 40 may appear to be too “ambitious” compared with ten, and ten may seem to be a good “conservative” level to start with. But, if one knows the costs and benefits to be either those in Table 5.1 or those in Table 5.2, any level less than 40 is not the best target to set.

What is wrong with setting any targets?

Table 5.1

Number Prevented 10 20 30 40 50 60 70 80 90 100 Source:

Number Prevented

Source:

Cost–benefit outcomes for alternative levels of prevention: Cost Case 1 Total Benefits ($B)

Total Costs ($C)

Net Benefits ($B – $C)

100 200 300 400 500 600 700 800 900 1000

200 200 200 200 250 300 350 400 450 500

–100 0 +100 +200 +250 +300 +350 +400 +450 +500

These numbers were created by the author.

Table 5.2

10 20 30 40 50 60 70 80 90 100

23

Cost–benefit outcomes for alternative levels of prevention: Cost Case 2 Total Benefits ($B)

Total Costs ($C)

Net Benefits ($B – $C)

100 200 300 400 500 600 700 800 900 1000

200 200 200 200 350 500 650 800 950 1100

–100 0 +100 +200 +150 +100 +50 0 –50 –100

These numbers were created by the author.

6.

What is wrong with setting the particular MDG targets?

We have just examined the weaknesses of targets in general terms. Let us now look at problems in detail with Target 7 of the MDGs. Recall that this wanted to “Have halted by 2015 and begun to reverse the spread of HIV/ AIDS” (see Chapter 5). Even the task force set up by the UN to recommend strategies for implementing the MDGs asked the question, “What does ‘halting the spread of HIV/AIDS’ mean and how will we know when it has been achieved?” (UN Millennium Project, 2005, pp. 24–6). It points out that if one takes a literal interpretation, that is, bringing the number of new infections to zero, the target is unsatisfactory for two reasons. It is not feasible to achieve this by 2015 and it focuses solely on prevention and sets no target for treating those already with HIV. The criticism that Target 7 ignores treatment is obviously valid and warrants no further discussion. So let us examine the feasibility issue further.

FEASIBILITY AS AN ECONOMIC ISSUE The only sure way to ensure that there are no new cases of HIV is to have a vaccine that is 100 percent effective and that this vaccine is given to everyone. As of 2009, such a vaccine does not exist. If it did exist, there is no way that the world’s population will all have been vaccinated by 2015, given that the BCG vaccination for TB already exists and there is not 100 percent compliance (perhaps because it is not 100 percent effective). But, let us assume that it is feasible. It would probably require all the world’s resources to achieve it. Would this be worthwhile? Did anyone at the UN carry out a CBA to demonstrate this? Interestingly, the UN task force thinks that achieving Target 7 would be a failure even if it were achieved! It states, “Stabilizing incidence at anywhere near current levels in the hardest-hit countries cannot be considered success” (ibid., p. 24). Instead it proposed replacing Target 7 with, “Reduce prevalence among young people to 5 percent in the most affected countries and by 50 percent elsewhere by 2015” (ibid., p. 25, Box 1.3). Since there is a fundamental misconception here that lies at the heart 24

What is wrong with setting MDG targets?

25

of the problem of setting HIV prevalence or incidence targets let us clear this up now. Say a country has a 25 percent prevalence rate among its young people. How could anyone possibly baulk at setting a target for this number to be reduced to 5 percent in 2015? The central problem with the 5 percent target in the context of a current 25 percent reality is that we do not know what would happen to the prevalence rate in the absence of any policy interventions for HIV/AIDS. If the epidemic has not peaked in this country then it could be that the rate would go up from 25 percent to 40 percent by 2015 without any interventions. In fact, the rate may well rise to 30 percent even with policy interventions. But, note that in this latter case HIV/AIDS policy would have been spectacularly successful for a 40 percent rate to be cut to 30 percent in just a few years. Yet the task force claims that, in line with the original target specification, keeping the rate at 25 percent when there would have otherwise have been a 40 percent rate “cannot be considered success” (ibid., p. 24). Worse still, with its new specification that the rate has to be cut to 5 percent, reduction from 40 percent to 10 percent by 2015 would also be judged a failure. In countries where the epidemic has not peaked, and this is most of Sub-Saharan Africa, Asia and Latin America, setting targets at rates below current levels is almost setting up the HIV/AIDS policy system for failure. Sensible targeting then can only be made in the context of what would happen in the absence of policy intervention. So, if one has to have targets, all countries should have one simple target, that is, to lower rates below what they might otherwise have been. In this formulation target-setting and CBA would be mutually reinforcing. For it needs to be understood that when a CBA is undertaken and no specific alternative is considered (as in Tables 5.1 and 5.2 in the previous chapter) the benefits and costs are calculated relative to what they would have been in the absence of intervention. So positive net benefits for an intervention means that the intervention is worthwhile relative to doing nothing. As long as all interventions are subject to a CBA test, then a target of reducing HIV/AIDS relative to doing nothing (called the “with and without” approach in CBA) would be successful only if the interventions were worthwhile. Conversely the target would not be met only if interventions did not take place, and interventions would not take place if a CBA test says they are not worthwhile.

FEASIBILITY AS A POLITICAL ISSUE While the difference between feasible and non-feasible targets is all important from an economic point of view, from a political point of view, targets

26

Setting priorities for HIV/AIDS interventions

have a momentum of their own even if they are not feasible. Stephen Lewis (2005, who we mentioned in Chapter 2) called the WHO’s “3 by 5 Initiative” to put 3 million in treatment by 2005 “brilliant” even though it was clear to him by the middle of 2005 that only 1.5 million would be treated by the end of 2005 (Lewis, 2005, p. 154). He pointed out that it did not matter that the target will not be reached as the international community: “was stagnating in its response to the pandemic until the WHO initiative vaulted upon the scene. ‘Three by five’ has had a remarkable impact: countries everywhere falling over themselves to introduce treatment, the numbers are increasing on a daily basis, and hope has re-emerged where only fear and despair held sway” (ibid., p. 155). As long as the setting of targets does not get in the way of undertaking CBAs, and it is seen as a political movement to galvanize international support for HIV/AIDS interventions, there is no harm in the exercise. However, the economic perspective is what is more important, as Stephen Lewis ultimately acknowledges. He pointed out that even though 1.5 million being treated was not 3 million, it was still 1.5 million more people than would have been treated otherwise. It was the “with and without comparison” saving 1.5 million additional lives that really justified Stephen Lewis calling the 3 by 5 Initiative “brilliant”.

7.

Cost–benefit analysis 101

There are lots of frills with CBA (such as discounting future benefits and costs to convert them to current values). But fortunately the essentials are few and can easily be summarized. In this chapter we cover some of the basic principles as included in a first-level course (101) and in the next chapter deal with some of the extensions as included in a second-level course (201). We leave to Parts III and IV the addition of some of the refinements of CBA. When considering a change in the level of an activity, one should compare the marginal (additional) benefits with the marginal (additional) costs. If marginal benefits (MB) are greater than marginal costs (MC), one should do more of the activity, for one is gaining more than one is losing, leading to positive net benefits. If marginal benefits are lower than marginal costs one should do less of the activity as one is losing more than one is gaining. Net benefits are negative in this case. And if it happens that marginal benefits just equal marginal costs, one should do neither more nor less. One should stay where one was as no other level of activity would bring higher net benefits. Obviously the principles are very general and can apply to any activity. The principles apply to the private sector as well as to governments, though how one would measure the marginal benefits and costs would differ according to who is undertaking the CBA of the activity. Let us see what we can learn from applying these principles to a basic health intervention. Poverty is widespread in all the countries where HIV/AIDS severely reduces life expectancies. So let us consider a basic nutrition supplement program, giving out loaves of bread, which increases the daily calories a person consumes. Say the bread program distributes the bread in reverse order of the body weight of the person being assisted. So the marginal (additional) benefits are highest for those first receiving help. From there the marginal benefits start to decline as less needy, but still needy, people get extra food. The decline in marginal benefits continues until those receiving the bread have a body weight that is judged to be normal, that is, the recommended daily intake of calories is reached. At this point the bread supplement program provides zero marginal benefits. The cost of a loaf of bread is the compensation that has to be paid to the bakers to supply the bread. Assume that this is a fixed amount per loaf.

27

28

Setting priorities for HIV/AIDS interventions

MB

Monetary values

0

100

1000

MB – MC > 0

MC

2500

3500

5000

Loaves of bread

MB – MC < 0

Source: Author’s own work.

Figure 7.1

The marginal benefits MB and marginal costs MC of bread

To this must be added the transport costs of delivering the bread to those in the program. If the needy live in more and more remote places, then the cost per loaf delivered will be rising as the program expands. Figure 7.1 shows the falling marginal benefits and the rising marginal costs. The differences between MB and MC are represented by the dotted arrows. Focus on the situation where 100 loaves are provided where the marginal benefits are high and the marginal costs are low, corresponding to the case where the neediest are receiving the food, and they are nearby, making it relatively cheap to service them. The difference between marginal benefits and marginal costs is both positive and very large. A CBA would record the large, positive difference and, because of this calculation, judge this scale of operations a huge success. As the food program expands to the level corresponding to 1000 loaves, less needy people are being helped and it is more costly to provide them with food. However, a CBA would again judge this scale worthwhile as the marginal benefits still exceed the marginal costs. The food program would then continue expanding until 2500 loaves are provided. At this scale of operations MB is just equal to MC. A CBA finding neither a positive nor a negative difference would decide that this is where the program should stop expanding and that the best scale has

Cost–benefit analysis 101

29

been obtained. Although nothing is gained on the last loaf provided, all the previous 2499 loaves have generated net gains and so this is the most gain that can be achieved. If the program expands to 3500 loaves, MB is less than MC. A CBA would not sanction this program level and signal that the program should be cut back. If the program were expanded even further to 5000 loaves, where everyone who is underweight gets assistance and the marginal benefits are zero, MC far exceed MB, and a CBA would certainly not approve of this scale of assistance. There are two main lessons to be learned from considering this evaluation exercise. First, we understand that expanding the food program is not always worthwhile. It depends on the size of marginal benefits and costs. We cannot just state that providing extra bread must be worthwhile because food keeps people alive for it is a necessity. At the levels of 100 and 1000 loaves, MB exceeds MC so these levels are worth providing. But, at the levels of 3500 and 5000 loaves, MB is less than MC and these levels are not worth pursuing. Second, just because we undertake a program evaluation and find positive net benefits (say we are at 1000 loaves), it does not necessarily mean that we move from there to provide bread to everyone who needs it. “Scaling up” is only desirable until we reach the optimum of 2500 leaves. After that level, we should be “scaling down” not scaling up. In any case it definitely does not mean that we provide bread on the maximum scale. At some point (5000 loaves) there are no additional benefits to be achieved and this cannot possibly be worthwhile if marginal costs are positive. Diminishing marginal benefits must be expected and factored in whenever scaling up is to be contemplated.

8.

Cost–benefit analysis 201

Looking at whether a little more, or a little less, of an activity is worthwhile is mainly what CBA is about. So a comparison of marginal benefits and marginal costs is the main social decision-making test. However, sometimes the evaluation must switch from looking at the margin to analyzing the effects in terms of totals. Totals are just the sum of the marginal effects. When an activity proceeds only in large jumps (for example, it is not useful to build only half a bridge) an all or nothing evaluation must be carried out. The criterion now for determining whether the activity is worthwhile is whether total benefits exceed total costs. If so, then the project for all the levels is considered worthwhile. The complications that arise for CBA by switching from a marginal to a total emphasis can be appreciated by considering two policy interventions for the one health improvement. Note that now we will be considering situations where MC is falling and not rising as in Figure 7.1 in the previous chapter. Countries in Sub-Saharan Africa where the HIV/AIDS epidemic is at its height are also countries where malaria is rampant. Reducing malaria would therefore help to keep alive those infected with HIV/AIDS. Consider the case where in a community there are ten people (cases) with malaria as depicted in Figure 8.1. There are two main interventions for malaria: prevention and cure (treatment). One way of preventing malaria is to spray areas that are breeding grounds for mosquitoes. There is a large initial outlay for setting up the program to spray mosquitoes with insecticides. Thereafter, as more of the swamps are sprayed, the marginal cost would fall as the number of beneficiaries expands and the number of cases of malaria declines. We therefore characterize the marginal costs of prevention as falling from right to left in Figure 8.1 from $5 to 20 cents, as the number of cases decline from ten to zero. Treatment on the other hand is individualized. A person comes into a clinic and receives medication (say quinine). If we ignore transportation cost, we can regard treatment cost as being constant, say at $2 per person. So in Figure 8.1, treating two people (giving two people quinine) is simply twice the total cost of treating one person (giving one person quinine). The MB curve in Figure 8.1 is, like that in Figure 7.1, depicted as falling as the scale of a malaria intervention is increased. The MC curve

30

Cost–benefit analysis 201

$5

a

$3

b c

$2

d

31

MC of treatment e

$1

f MB g MC of prevention

$0.2 10 0 Source:

6

5

Cases of malaria

4

Malaria reduction

10

Author’s own work.

Figure 8.1

The marginal benefits MB and marginal costs MC of malaria reduction

is assumed to be the same irrespective of by how much malaria is being reduced. We again think of the worst cases (or areas) being targeted first and less serious cases being targeted later. In terms of the treatment alternative, the evaluation can follow marginal lines as people can be treated one at a time. Equating marginal benefits and the marginal costs of treatment would lead us to point c in Figure 8.1. A malaria treatment reduction of four, so that six cases remain, is worthwhile as marginal benefits exceed the marginal costs of treatment up to this level. However, for the prevention alternative, marginal analysis gives the wrong answer. Equating MB and MC would place the intervention program at point e with a six-person reduction and four cases remaining. Note that for each of the six-person reduction, the MC of prevention exceeds the MB. So not only would the six-person reduction be rejected, one would not have started the intervention program in the first place, as MB (point b) is below MC (point a) when there are ten cases of malaria. Interestingly, although one would not have started intervening with prevention, scaling up prevention actually could be worthwhile. For if greater than six persons are prevented, then marginal benefits exceed costs for all levels until malaria is eradicated and no cases remain. One

32

Setting priorities for HIV/AIDS interventions

should therefore never stop at preventing six people, but consider whether preventing all ten persons with malaria is desirable. Is total eradication worthwhile? That depends on whether the size of the negative net benefits from cases where prevention is not worthwhile at the margin (area abe between 10 and 4) can be overcome by the size of the positive net benefits where prevention is worthwhile at the margin (area egf between 4 and 0). This is precisely what a total benefits and costs evaluation assesses. That is, for all the ten persons prevented from 10 to 0, do the total benefits (as given by the area under the MB curve) exceed the total costs (as given by the area under the MC of prevention curve)? The traditional proverb is that “prevention is better than cure”, or in the situation we are considering, “prevention is better than treatment”. However, only a CBA can tell us whether that is true or not for a particular intervention. The MC of treatment curve intersects the MC of prevention curve at point d. To the left of this point, at level 4, treatment is less costly, while to the right of d, prevention is cheaper. Thus, labeling an intervention as “prevention” or “treatment” does not help to decide which is better. What is important is the size of the benefits relative to the costs of each alternative. In fact, if the total costs exceed the total benefits for both interventions, then prevention may be better than cure, but prevention actually would not be worthwhile. We are now in a position to complete the picture about the desirability of scaling up first discussed in the context of Figure 7.1. When MC is rising, and MB is falling, as in Figure 7.1, then scaling up should be considered whenever an evaluation of an initial level of an intervention produces an MB that exceeds MC. Scaling up is a non-starter if the MC of the initial intervention exceeds the MB. But, with MC falling, as with prevention in Figure 8.1, the evaluation of the initial intervention does not predetermine whether scaling up should subsequently take place. Even though MB may be declining, if at some level of intervention the MB is above MC, then increasing the scale may lead to total benefits that exceed total costs.

PART II

HIV/AIDS as a hunger and economic development issue

9.

Introduction to Part II

For a CBA to be undertaken there has to be an intervention effect to evaluate. For the purposes of this book, an intervention is anything that reverses what has contributed to HIV/AIDS and its effects on people’s lives. Thus, given that unprotected sex causes people to get HIV, then providing condoms to reduce the incidence of unprotected sex would be an intervention. When the sharing of needles contaminated by HIV transmits the disease, supplying clean needles would be another intervention. The first step then in examining CBAs of HIV interventions is to identify what has contributed to the progression of the disease. As we explained in Chapter 1, there are many causes of HIV. To avoid an encyclopedic list of everything that has played a role, we will organize our discussion in Part II of this book around the fundamental question of why there is so much more HIV/AIDS in countries in SSA that have experienced a generalized epidemic than in a country such as the United States where HIV is largely just in localized populations. Once we have established this, we will then look at a second HIV distributional question, that is, why is HIV prevalence in the African American community so much higher than the rest of the United States? By trying to answer these two distribution questions we can see exactly what is similar and what is different about the epidemics in Africa and the United States. In this chapter we first summarize the important facts regarding HIV/ AIDS throughout the world and in the United States, which will provide the backcloth for all of the discussion in Part II. Then we provide a guide to the rest of Part II. Before we begin this chapter we need to make some general remarks about HIV/AIDS numbers as there are many different data sources and many different ways of reporting the numbers. The main points are these: ●

The Joint United Nations Programme on HIV/AIDS (UNAIDS) is the major source of country HIV data and it gives periodic updates. The important updates occur in December in every year. The main drawback with the UNAIDS data is that they are not collected on a uniform basis for every country. In addition, each individual country has its own measurement initiatives and the numbers that 35

36

Setting priorities for HIV/AIDS interventions

●

●

●

●

come out from these sources may not correspond with those that UNAIDS reports. UNAIDS uses two sources of data, depending on whether a country’s epidemic is generalized or localized (“concentrated”). For generalized epidemics UNAIDS uses estimates primarily from surveillance among pregnant women attending antenatal clinics (ANCs). Obviously, these numbers exclude women who are not pregnant and/or not sexually active and omit all men. For localized epidemics UNAIDS relies on studies among key populations that are at highest risk of HIV exposure, such as IDUs, MSMs and CSWs (see Chapter 1). Instead of relying on ANC data, a number of SSA countries have initiated population-based HIV prevalence surveys. Most of these population-based surveys have generated HIV estimates that are lower than those for ANCs. For example, in a sample of 14 SSA countries, 13 of them had lower HIV prevalence rates measured by their population-based surveys than previously estimated using ANC data (UNAIDS, 2006a, Figure 2.1). UNAIDS has recently changed its reporting system. Apart from incorporating new data from population-based surveys, UNAIDS has also changed the coverage of its surveillance system. Surveillance has expanded into rural areas where prevalence is known to be lower. So even though HIV prevalence rates may look like they have decreased between 2005 and 2006, most of this difference is due to a changed methodology (revised data). This means that one must always be careful about making comparisons of HIV prevalence rates in a country over time. Lastly, as a rule of thumb, if you see an estimate of HIV prevalence for a country that is low, and you thought you had seen a number published that was much higher, this would be because the specification of the group involved is different. High rates relate to adults aged 15 to 49 years, while low rates relate to the population as a whole. For example, the adult rate in mid-2006 for South Africa was 18.3 percent when the prevalence rate in the total population was 11.2 percent.

PREVALENCE RATES OF HIV (i)

HIV/AIDS Prevalence Rates Worldwide

In Table 9.1 we present the numbers as of December 2007 as reported in UNAIDS (2008). The majority of people living with HIV/AIDS, nearly

Introduction to Part II

Table 9.1

37

2007 HIV infections, rates and deaths by region (2001 figures in brackets)

Region

Adults and Children Living with HIV/AIDS

Adults 15–49 Prevalence Rate (percent)

Deaths in Adults and Children

Sub-Saharan Africa

22 000 000 (20 400 000) 380 000 (300 000) 4 200 000 (4 200 000) 740 000 (490 000) 74 000 (25 000) 1 700 000 (1 400 000) 230 000 (210 000) 1 500 000 (650 000) 730 000 (610 000) 1 200 000 (1 100 000) 33 000 000 (29 500 000)

5.0 (5.7) 0.3 (0.3) 0.3 (0.4) 0.1 (0.1) 0.4 (0.2) 0.5 (0.5) 1.1 (1.1) 0.8 (0.4) 0.3 (0.2) 0.6 (0.6) 0.8 (0.8)

1 500 000 (1 300 000) 27 000 (22 000) 340 000 (250 000) 40 000 (15 000) 1 000 (.. . .) 63 000 (47 000) 14 000 (15 000) 58 000 (6 700) 8 000 (9 600) 23 000 (18 000) 2 000 000 (1 700 000)

N Africa and Middle East South and SE Asia East Asia Oceania Latin America Caribbean E Europe and Central Asia W and Central Europe North America Total (global)

Source:

Constructed by the author from UNAIDS (2008) Annex A.

two-thirds of the total, are in SSA. This is true even though SSA has only 10 percent of the world’s population. With 2.7 million new infections and 2 million dying each year, the total of 33 million is going to rise over time, at least for the next few years. The total in 2007 was up from 29.5 million in 2001. Within SSA there is enormous variation around the 5.0 percent average, with rates as high as 26.1 percent in Swaziland and as low as 0.1 percent in Mauritania. Southern Africa is the epicenter of the global epidemic with about one-third of the world’s cases living in this area. East Africa is next in line after Southern Africa, and then comes West Africa where Côte d’Ivoire has the highest rate at 3.9 percent. The lowest HIV rates in Africa are in North Africa. Nearly 60 percent of people infected in

38

Setting priorities for HIV/AIDS interventions

Table 9.2

HIV prevalence rates (per 100 000 population) in the United States by race/ethnicity and gender, 2006

Race/Ethnicity

Males

Females

White Black Hispanic Asian/Pacific Islander American Indian/Alaska Native

395 2388 863 220 340

63 1122 263 46 127

Source: Created by the author from data released in CDC (2008).

SSA are women, which seems to indicate that heterosexual activity is the main transmission mechanism. In North America and Western Europe, around a quarter of the people with HIV are females. (ii)

HIV/AIDS Prevalence Rates in the United States

Table 9.2 gives the HIV prevalence rates for various groups in the United States expressed per 100 000, as opposed to per 100, which is the usual way that UNAIDS and most other agencies report the numbers. Since rates are so much lower in the United States and elsewhere it is often more convenient to express them per 100 000 to avoid a large number of decimal places. The rates are as of the end of 2006 (beginning of 2007). HIV is more prevalent amongst blacks than for other races and groups in the United States. Although blacks are only 12 percent of the population, they make up 46 percent of the 1 106 400 total number of persons that were living with HIV; whites were 35 percent and Hispanics were 18 percent (leaving 1 percent Asian/Pacific Islander and less than 1 percent American Indian). Females in the United States constitute only 25 percent of the HIV total, much lower than the 50 percent worldwide and the 60 percent in SSA, though the share is rising over time. The concentration of HIV is so large for black females that their rates outnumber white females by a ratio of nearly 18 to 1 (and 4.3 to 1 relative to Hispanic females). In total in 2006, there were 1 106 400 individuals living with HIV, up by 112 400 from the 2003 total of 994 000. Some of the increase was expected due to the fact that people now live longer with HIV, and are not dying at the same rate as before, due to treatment with ARVs. Also, if people are living and not dying they can, of course, still transmit the disease. The elderly in the United States (people over 50) make up 24 percent of the total number living with HIV/AIDS (up from 17 percent in 2001) and constitute 15 percent of new HIV/AIDS diagnoses.

Introduction to Part II

39

The disease is virtually absent now in children (under the age of 13 years). So MTCT is not a major transmission mechanism, unlike for SSA where it is how 5 percent of the infections are generated. The breakdown of the total number of HIV cases in the United States by means of transmission was as follows: ● ● ● ● ● ● ●

48 percent were male via MSMs; 18 percent were female high-risk heterosexual; 12 percent were male IDUs; 9 percent were male high-risk heterosexual; 7 percent were female IDUs; 5 percent were male via joint MSMs and IDUs; 1 percent were other.

For males, the main ways that HIV was transmitted was via MSMs and IDUs. Relatively few males were infected by having unprotected sex with females. Females in the United States are like their counterparts in SSA by having heterosexual sex as the main transmission mechanism. So in Part II, when we discuss sexual activity and its role in explaining similarities and differences in regional and racial HIV prevalence rates, we will focus exclusively on heterosexual transmission and largely ignore MSMs and IDUs. Pisani (2008) examines these latter two transmission modes in detail, which are especially important for explaining HIV prevalence in many Asian countries.

OUTLINE OF PART II We see in Table 9.1 that the HIV prevalence rate in SSA is 17 times that of North Africa and of Western Europe, ten times that of Latin America and eight times that of North America. We explain why the rates are so high in SSA by arguing that HIV in these countries is primarily a hunger issue and not simply a sexual issue. To make this case we have to explain why it is that nutrition is more important and sexual behavior less important in African epidemics. Table 9.2 reveals that blacks in the United States are almost nine times more likely to be infected by HIV than whites (and twice as likely as Hispanics). Although black males are twice as likely to be infected as black females, it must be remembered that in the United States the racial difference is starker for black females, as black females are 18 times more likely to be infected than white females, while it is “only” six times higher for black males than for white males. We shall see that the role of sexual

40

Setting priorities for HIV/AIDS interventions

relations is more important for accounting for the second, racial, HIV distributional question than the first distribution question related to region. Apart from accounting for regional and racial differences in HIV prevalence rates, the other main task for this second part of the book is to make the reader aware of the wide variety of interventions that are possible to try to stem the rates of HIV infection in countries where there are generalized epidemics. Some of these interventions will be evaluated in cost–benefit terms in Part III. Many of the non-standard HIV interventions involve seeing how to reduce the problems associated with a given HIV prevalence rate (that is, targeting mitigation) and this is why we shall be focusing our efforts on detailing the implications of the fact that HIV in SSA is mainly a hunger and economic development issue. We proceed as follows. Chapters 10–12 make the case for the importance of hunger in SSA. Chapters 13–15 provide the social, cultural and economic background. Chapters 16–17 identify the possible interventions that play a role in agricultural economies. We close with Chapters 18–20, which are devoted to the contribution of sexual behavior to HIV transmission in individual and social settings, comparing and contrasting SSA with the United States.

10.

HIV and hunger

Table 9.1 in Chapter 9 informed us that HIV/AIDS is concentrated in SSA. We will see evidence in the next two chapters that hunger, in the form of malnutrition, also has a major presence in SSA. In this chapter we are going to argue that this positive association between HIV and malnutrition in SSA is not an accidental relationship and that understanding this association is crucial for explaining why HIV is highest in SSA. Malnutrition occurs when the nutrients available to a person are insufficient to meet the body’s needs. This nutrient deficiency between what the body needs and receives can be because a person does not receive sufficient nutrients or because, even though the level may be sufficient, the nutrients are not properly absorbed. What the body needs is not constant over time and is itself a function of the extent to which disease is present. One of the many reasons why HIV and malnutrition go hand in hand is that people living with HIV have higher than normal nutrition requirements. Gillespie and Haddad (2002, p. 10) report that people with HIV need up to 50 percent more protein and up to 15 percent more calories than those who are uninfected.

THE DIRECT ROLE OF MALNUTRITION IN HIV TRANSMISSION There are two main types of malnutrition: (1) macro-malnutrition exists when there is insufficient protein, energy (carbohydrates) and fat and (2) micro-malnutrition involves a shortage of vitamins (such as A, D and E) and minerals (for example, calcium, sodium and potassium). These two types of malnutrition interact and create what Semba and Tang (1999) call a “vicious circle” whereby malnutrition and HIV work together to deplete the immune system. One mechanism that Semba and Tang use to explain the vicious circle involves highlighting the existence of free radicals in the body that damage healthy cells. Antioxidants are chemical compounds or substances (such as vitamins E, C and beta carotene) that prevent the free radicals from operating. When the production of free radicals in the human body exceeds the body’s ability to neutralize them via the

41

42

Setting priorities for HIV/AIDS interventions Insufficient dietary intake Malabsorption and diarrhea Impaired storage and altered metabolism Increased HIVreplication Progression of disease Increased morbidity

Micro-nutrient deficiencies

Increased oxidative stress Immunosuppression Source: Semba and Tang (1999).

Figure 10.1

Vicious circle of malnutrition and HIV

antioxidants this is defined as “oxidative stress”. So micro-nutrient deficiencies increase oxidative stress and this is one half of the vicious circle shown in Figure 10.1. With cells that are damaged, their ability to fight the HIV virus is reduced. There is increased HIV replication and illness (morbidity) leading to insufficient dietary intake and malabsorption of what is consumed (macro-malnutrition) and this is the second half of the circle. What is particularly significant about this vicious circle is that malnutrition is both a cause and a consequence of HIV transmission.

THE INDIRECT ROLE OF MALNUTRITION IN HIV TRANSMISSION We have just seen that malnutrition leads not only to greater HIV replication once the virus has been introduced, but it is also associated with increased sickness more generally. This is important because being infected with other diseases, especially parasitical ones, can lead to higher HIV transmission. Mosquitoes do not spread the HIV disease but they break the skin, and because HIV is transmitted from one person’s bloodstream to another’s, anything that breaks the skin makes it easier for HIVinfected blood to enter into another person. Stillwaggon (2006) was one of the first to emphasize the importance of parasitic diseases and malnutrition in explaining why the transmission rate

HIV and hunger

43

in SSA was higher than elsewhere. Her main point was that it was wrong to just focus on the HIV virus and ignore the characteristics of the host who was being attacked by the virus. HIV is like any of the other public health threats such as TB, smallpox and cholera from the point of view that it is the poor and malnourished who suffer most. Similarly, the parasitic diseases are more prevalent among the malnourished. Hookworms, roundworms and amebas are widespread among the poor in SSA. In fact, 80 percent of the 200 million people infected with schistosomiasis are in SSA. Water in dams and lakes are home to the snails that host the schistosome worms. Using these resources for drinking, washing and fishing turns out to be “high-risk” activities from an HIV perspective. Breaking skin is one thing; breaking skin around the genitals is clearly more important. But this is exactly what these worms do as they cause genital lesions and inflammatory effects in genital areas. The fact that breaking the skin around the genital areas is important for HIV transmission is not a new finding just related to parasitical diseases. Sexually transmitted diseases (STDs) such as chancroid, syphilis, gonorrhea and chlamydia also open up the skin. STDs, in common with schistosomiasis, have an inflammatory effect on genital tissue, which attracts T cells to the site and makes them vulnerable to being infected by HIV (Stillwaggon, 2006, p. 63). The point that we wish to emphasize about STDs is that they are also more likely to be present in populations where malnutrition is widespread.

THE SIGNIFICANCE OF MALNUTRITION AS A CAUSE OF HIV TRANSMISSION With hunger as a main determinant of HIV transmission in SSA, a whole new set of interventions for HIV are opened up. It is not just sexual behavior that one seeks to influence but also factors that influence, and are influenced by, malnutrition. Vitamin supplementation and food programs take center stage. SSA economies are mainly agriculturally based. Anything that enhances or restricts the flow of food can be a target for an HIV intervention. This theme that HIV policy in SSA must not be limited to trying to change sexual behavior and must be looked at broadly in terms of going outside the health sector is one that will be explored throughout Part II and is covered in great detail in Chapters 13–15. But first we look specifically at the evidence for micro- and macro-nutrition deficiencies in SSA and see how they have been remedied.

11.

Nutrition and HIV at the individual level

Many countries have recognized the importance of micro-nutrients in supporting the immune system in its fight against all kinds of diseases. For example, South Africa has a national program that provides vitamin A supplementation in schools. Vitamin A is particularly important for SSA as there is more likely to be a deficiency due to the type of foods eaten and the fact that the tropical sun increases the demand for vitamin A (Stillwaggon, 2006, pp. 34–5). In the HIV context we would expect vitamin A to be important as it is required for the production of T cells. In this chapter we look at some evidence of micro-nutrient deficiencies in HIV populations and report the results of an attempt to assess the effectiveness of supplementing vitamins in SSA.

THE EXTENT OF MICRO-NUTRIENT DEFICIENCIES IN THOSE WITH HIV Semba and Tang (1999) report the extent of micro-nutrient deficiencies in various HIV populations and the percentages are shown in Table 11.1. In the United States, homosexual men and heterosexual adults have the lowest levels of micro-nutrient deficiencies, and IDUs from large inner cities have the highest. There are not many studies for developing countries, but the ones that do exist show that pregnant women are most at risk. Note that vitamin A in particular is deficient in HIV populations in developing countries and in some groups in the United States.

THE EFFECTIVENESS OF MULTIVITAMIN SUPPLEMENTATION Fawzi et al. (1998, 2004) undertook a serious of studies of multivitamin supplementation of pregnant women in Tanzania. They used a controlled clinical trial whereby persons (pregnant women) were randomly assigned to two groups: an experimental group (who were given specified vitamins) 44

Nutrition and HIV at the individual level

Table 11.1

45

Prevalence of micro-nutrient deficiencies in different HIV populations

Location

Risk Group

Criteria

New York City

Heterosexual adults

Miami

Homosexual men

Miami Baltimore/ Washington Baltimore

Adult men Homosexual men

Vitamin A Vitamin E Vitamin B6 Zinc Vitamin A Vitamin E Vitamin B6 Zinc Selenium Vitamin A Vitamin E Vitamin A

0 4 2 4 11 19 30 26 11 3 11 15

Vitamin A Vitamin A

65 24

Malawi Kenya Source:

Injecting drug users Pregnant women Pregnant women

Deficient (percent)

Semba and Tang (1999).

and a control group (who were given a placebo) to see if there were any differences in outcomes between the two groups. Any differences, by the random design, would be due only to the effectiveness of the treatment (the multivitamins). One thousand and seventy-five HIV-1-infected pregnant women at between 12 and 27 weeks’ gestation received the multivitamins A, B1, B2, B6, niacin, B12, C, E in combination and with vitamin A on its own. The main results from the first study were: ● ●

● ●

Thirty fetal deaths occurred among women assigned multivitamins compared with 49 among those not on multivitamins. Multivitamin supplementation decreased the risk of: low birth weight (<2500 g) by 44 percent; severe preterm birth (<34 weeks) by 39 percent; and small size at birth by 43 percent. Multivitamins led to a significant rise in T-cell counts (CD4, CD8 and CD3). Vitamin A had no significant effect on any endpoint in the study.

It was good that birth outcomes improved. But from the HIV perspective what was encouraging was that the T-cell count increased. Interestingly, vitamin A on its own was not helpful.

46

Setting priorities for HIV/AIDS interventions

In the follow up study, that is, Fawzi et al. (2004), where the HIV-1infected pregnant women were now studied for multivitamin effects six years later, the multivitamins used were B, C and E, and vitamin A was again tested on its own and in combination with the other vitamins. The main results were: ●

● ● ●

Survival was longer and progression of the disease was slowed by multivitamins. Of 271 with multivitamins, 67 progressed to WHO stage 4 or died compared with 83 of 267 with the placebo. Multivitamins led to a significant rise in T-cell counts and lower viral load. Vitamin A had no significant effect. Adding it to multivitamins lowered some endpoints. On its own it increased risk of death. Multivitamins were a low-cost way of delaying the initiation of ARV therapy.

The multivitamins not only helped the HIV-infected mothers slow down the progression of the disease, they would also help prevent others from being infected because the viral load decreased, and the viral load determines the extent to which HIV can be transmitted to others. Surprisingly, vitamin A, which is the one that we highlighted in the introduction to this chapter, was not only unhelpful as in the earlier study – this time it was actually detrimental. We return to the main theme of this book that just because an intervention seems sensible we cannot just proceed and implement it. One needs to look at data to see whether at the levels it will be applied, and in the context of the circumstances of a particular project in a specific country, the intervention will actually be effective and hence potentially worthwhile.

THE IMPLICATIONS OF THE FAWZI ET AL. WORK FOR SOUTH AFRICA There are two aspects of the Fawzi et al. research that relate specifically to South Africa HIV policy and we examine them in turn. (i)

Multivitamins as a Cure for AIDS

In the early years of his presidency of South Africa (1999 to 2000), Mr. Thabo Mbeki championed a small group of denialists who claimed that malnutrition and not HIV is the cause of AIDS. Since 2002, the Health Minister Dr. Tshabalala-Msimang has continued Mbeki’s agenda by

Nutrition and HIV at the individual level

47

resisting the introduction of ARVs and instead supporting unproven alternative therapies (such as Virodene, a freezing solution) – see Nattrass (2006). One organization that followed the denialists’ doctrines was the Rath Health Foundation. It appeared to point to Fawzi’s research as supporting evidence for its strategy to market multivitamins as an alternative to ARVs. However, the Fawzi et al. conclusion was that multivitamin supplements, because of their cost-effectiveness, could be used before ARVs are introduced. They never claimed that multivitamins should be used instead of ARVs. To leave no doubt in the reader’s mind of the effectiveness of ARVs, we will refer to a study of the effects of ARVs on the health of children with HIV in South Africa by Eley et al. (2006). The study looked at the extent of malnutrition in children under 15 years of age and monitored the changes recorded one year after having received ARVs. They used three measures of malnutrition: ● ● ●

underweight: this measures children’s weight for their age; wasting: this measures children’s weight for their height; stunting: this measures children’s height for their age.

By all three malnutrition measures ARVs gave dramatic improvements. In the sample, after just one year: underweight went down from 56 percent to 18.2 percent; wasting declined from 20.5 percent to 2.4 percent; and stunting fell from 67.4 percent to 46.6 percent. ARVs will save many children’s lives, on a scale that could not possibly be matched by multivitamins. (ii)

The Role of Vitamin A in HIV Prevention

As pointed out earlier, South Africa has a vitamin A supplementation program in its schools. Do the Fawzi et al. studies imply that this program should now cease, as vitamin A did not slow down the disease for pregnant women and new mothers? Of course, the program could still be beneficial because vitamin A is vital for a number of health concerns even if HIV is not one of them. Vitamin A is essential to vision, fetal development and generating an immune response for other diseases. But what do the Fawzi et al. findings for vitamin A signify in the context of HIV? It would seem that a reduction in vitamin A is a marker for reaching a stage in the development of the HIV disease. When the disease has reached a certain stage, vitamin A goes down. As it is just a marker for the stage of the disease, it does not therefore follow that replacing the vitamin A will reverse the infection. One can use the analogy of hair loss and hair

48

Setting priorities for HIV/AIDS interventions

transplants. When humans reach a particular stage in life, their hair might drop out. The hair loss is thus a marker for getting old. If they received a hair transplant, the hair loss would be remedied. But the hair replacement would not make the humans become younger. In the same way, vitamin A supplementation would remedy the vitamin A deficiency, but this would not mean that the attained HIV stage would be reversed.

12.

Nutrition and HIV at the country level

HIV can reside in an individual and in a group of individuals, a country population. So interventions can be carried out targeting individuals or the whole nation. Similarly, malnutrition can be treated individually and at country levels. In this chapter we follow up the previous chapter with an analysis of the malnutrition/HIV link at the national and regional levels.

UNDERSTANDING THE MALNUTRITION MEASURES The three measures of malnutrition we introduced in the last chapter for measuring malnutrition in children under 15 in South Africa (underweight, wasting and stunting) can also be used to establish the extent of malnutrition in a national population and region. Children’s weight or height by age and sex in a particular country can be compared with the weight or height by age and sex in a reference population, usually the United States. A child in the United States would be judged malnourished if the child was well below the weight for height that was typical in the United States. If the height and weight distributions were normally distributed (a symmetrical bell-shaped distribution), approximately 95 percent of the children would be within plus or minus 2 standard deviations (SDs) from the mean or median in the United States. This means that 5 percent of the children would not be typical. Since being way above what is typical is not a hunger problem, only those in the bottom tail of 2.5 percent would be considered malnourished. So –2 SDs from the United States mean or median is the standard benchmark for being malnourished. We will now apply this benchmark for each of the three nutrition measures – see Caulfield et al. (2006). Wasting takes place if a child’s weight for height is less than –2 SDs of that typical in the United States. Weight can be adjusted over time with interventions and so wasting is regarded as a short-term measure of malnutrition, that is, it records inadequate nutrition over a short period of time. This is in contrast to stunting, which is height for age that is less than 49

50

Setting priorities for HIV/AIDS interventions

–2 SDs of that typical in the United States. Height may not adjust even in the long run. So stunting measures chronic malnutrition. Underweight, weight for age that is less than –2 SDs of that typical in the United States, is a mixture of the two as it combines wasting and stunting. Although the stunting and wasting measures are more useful as guides to malnutrition, most of the data globally on malnutrition refer to the underweight measure, so we shall be using these to quantify the relative extent of malnutrition in Africa.

THE EXTENT OF NUTRITION DEFICIENCIES AT THE COUNTRY LEVEL We have seen that HIV/AIDS is greater in SSA than elsewhere. Now we document the extent to which malnutrition is more prevalent in SSA. To make a relevant comparison among regions we need to distinguish mild to moderate malnutrition from the standard measure of malnutrition. Mild to moderate malnutrition will be if a country is 5 percent below the average in the United States, while as we saw earlier, standard malnutrition exists if the chances are less than 2.5 percent that one’s nutrition is normal (like that typical in the United States). The impact of malnutrition in a country can be felt in two main ways. Children die every year from malnutrition, so measuring the number of deaths is one way of recording the importance of malnutrition. The other way is to recognize that both the quality of life as well as the quantity of life (the number of deaths) is important. A measure that combines both quantity and quality of life is the number of disability adjusted life years (DALYs). Quality is ultimately converted into quantity terms by recognizing that a year of life with a disability is worth less than a year of life in full health. Say, two years in a wheelchair is worth one year with full health. Then a disease that causes someone who otherwise would be healthy to die, and lose 30 years of life expectancy, would cause the loss of 30 DALYs; while if the person who dies from a disease loses 30 years of life expectancy that would have otherwise been spent in a wheelchair, the loss from the disease would be recorded as 15 DALYs. Table 12.1 shows the prevalence of malnutrition by region in terms of mild and severe underweight and the impact of both types of underweight on the number of deaths and DALYs. In terms of severe underweight, South Asia is the region with the highest prevalence of malnutrition. SSA however, has the most mild and severe underweight problems. This is important because the consequence of mild underweight, when combined with serious underweight, is that there is an aggregate impact on the

Nutrition and HIV at the country level

51

Table 12.1 Prevalence of nutritional deficiencies and estimates of deaths and DALYs lost of children aged birth through four by region, 2004 DALYs Deaths due Prevalence Prevalence lost due to to Mild of Mild of Severe & Severe Mild & Severe Underweight Underweight Underweight Underweight (percent) (percent) (thousands) (thousands)

Region

Sub-Saharan Africa N Africa and Middle East South Asia East Asia and Pacific Latin America and Caribbean Eastern Europe and Central Asia High-income countries Source:

32 21

38 35

1 334 305

45 131 10 308

46 18 6

44 29 23

870 125 22

27 879 5 777 725

6

21

14

489

2

14

0

0

Caulfield et al. (2006).

quantity and quality of life that is greater in SSA than in any other region in the world. Both the number of deaths and the number of DALYs lost due to malnutrition are much higher in SSA than in Asia.

THE EFFECTIVENESS OF REDUCING MACROMALNUTRITION BY INCREASING CALORIES Given that SSA has more HIV/AIDS cases and more malnutrition consequences than any other region, if HIV/AIDS is to be primarily a hunger issue in SSA as claimed, then it is important that there be evidence that decreasing malnutrition has lowered HIV/AIDS rates. Eileen Stillwaggon (2002) undertook a statistical study of 44 Asian, African and Latin American countries. One dependent variable was the rate of HIV prevalence in urban low-risk populations. She found these statistically significant effects (in order of importance): 1. 2.

A rise in calories between 1970 and 1995 led to a decline in HIV. The more unequal the distribution of income (measured by the Gini coefficient) the higher the rate of HIV.

52

3. 4.

Setting priorities for HIV/AIDS interventions

A rise in the urban population resulted in an increase in HIV. The level of real GDP in 1995 was not significant, but the change in per capita GDP between 1960 and 1995 had a significantly positive effect on HIV.

When she disaggregated her sample for 20 countries in Latin America and the Caribbean (in Stillwaggon, 2006, Table 4.2), she found basically the same set of results as in her earlier study except that the size of per capita income was now significant and positively related to HIV prevalence (and international migration raised infection rates). It is the rise in calories (between 1970 and 1995) that we wish to emphasize here as we will be discussing the other factors that Stillwaggon found significant in subsequent chapters. Obviously there are many different ways that countries can increase the quantities of calories consumed, by raising average incomes and increasing food consumption in this way, or by carrying out food programs for the poor and focusing on those who originally consumed least. Irrespective of the exact mechanism, Stillwaggon found that increasing calories had the largest impact on HIV/ AIDS of all the factors she tested (given the typical unit changes that occurred for each factor). The direction of the impact was as expected seeing that the infection rate declined as calorie consumption increased. HIV/AIDS was indeed found to be a hunger issue in her study. Very few of a country’s development goals could be achieved without an increase in per capita income. So the role of national income levels and changes, and its effects on HIV/AIDS prevalence rates, needs to be examined further and we cover this topic in the next chapter.

13.

Income as a factor raising HIV rates

Even if one accepts that HIV is primarily a hunger issue in SSA, this is not the end of the story because other factors do play independent, even if lesser, roles, and these other factors could have secondary roles in so far as they first impact hunger and then go on to influence HIV. As we shall see, there are three other contributing causes of HIV that need separate examination: income, education and religion. We devote a chapter to each of the three factors, starting with income in this chapter, once we have described their joint impact.

INCOME, EDUCATION AND RELIGION AS DETERMINANTS OF HIV IN SSA Brent (2006) undertook a statistical study of HIV rates in 31 SSA countries. His dependent variable was the national female adult prevalence rate. Among the findings were these three statistically significant effects: ● ● ●

the higher the income of the country, the higher the HIV rate in that country; the more females that were enrolled in schools, whether primary or secondary, the higher the HIV rate; the greater the share of the population that was Muslim, the lower the country’s HIV rate.

We will subsequently attempt to analyze each fact separately. But, at this stage we just want to point out that these three factors were highly interconnected in the results. To see this, consider two of the countries. At one extreme was South Africa, which had one of the highest HIV rates (over 20 percent). It was one of the richest in the subcontinent. Perhaps because of the high income, the country had high levels of female school enrollment. However, less than 5 percent of its population was Muslim. At the other extreme was Niger, which had about 95 percent of its population Muslim, while having one of the lowest HIV rates (less than 1 percent). Perhaps 53

54

Setting priorities for HIV/AIDS interventions

because of its Muslim culture, national income was low and female school enrollment was less than in most other countries. Once more we come across the main theme of this book. The more we study the HIV/AIDS pandemic, the more we find results that are counterintuitive. The finding that income was positively related to the infection rates is especially puzzling because we have just presented in the two previous chapters a strongly plausible case that HIV in Sub-Saharan Africa is mainly a hunger issue. How can hunger be paramount if it is the high-income countries that have the highest prevalence rates? Surely rich countries can afford to feed themselves? Reconciling this paradox is the main task for this chapter. We will start at the beginning by looking at the first study that undertook a multi-country comparison of HIV prevalence rates and proceed to examine more recent work.

THE INITIAL EXPLORATORY STUDY OF HIV RATES IN DEVELOPING COUNTRIES A background study was carried out by Over (1998) for the World Bank’s (1999) review of the literature on how to reduce the impact of HIV/AIDS in developing countries. The sample consisted of 72 countries, covering a mix of African and Latin American nations. He found that national income (gross national product per person) had a negative effect on HIV prevalence rate. This is in line with the idea that HIV is a poverty and hunger issue. Poor people cannot afford to have nutritious diets and poor women may have to engage in risky sex for cash in order to survive. The facts seemed to match the intuition of what was determining HIV. But, the facts were very much a product of the sample used. Recall, from Table 9.1 in Chapter 9 that Latin American countries, which are richer countries than those in SSA, have lower HIV rates. So lower prevalence rates are bound to be associated with higher income levels when the experiences of countries in these two regions are combined together. The information contained in Table 9.1 can make the sampling problem even clearer. If one were to look at HIV rates across Western Europe as well as across Latin America and SSA, a negative relation would appear very strong. As the region we are considering increases in income (SSA has mainly low-income countries, Latin America consists of middleincome countries and Western Europe contains high-income countries) the HIV rates fall (SSA has high rates, Latin America has medium rates and Western Europe has low rates). What the numbers reveal therefore is that the more developed the country, the lower the prevalence rate. This need not necessarily mean that a country, such as Tanzania in SSA, that

Income as a factor

55

experiences an increase in national income, will have a declining infection rate as, even with the rise in income, it will remain a developing country.

LATER STUDIES OF HIV RATES IN DEVELOPING COUNTRIES Recall from the previous chapter that when Stillwaggon (2002) combined countries in three different regions (Asia, Africa and Latin America) she did not find a negative relation between income and HIV. The relationship was not statistically significant. However, when she disaggregated her sample just for 20 countries in Latin America and the Caribbean (2006) she found that the size of per capita income was now significant and positively related to HIV prevalence. The positive relation held when just medium-income countries were considered as a group. Brent’s (2006) study of 31 SSA countries can therefore be viewed as complementing the later Stillwaggon work by disaggregating the total sample for just lowincome countries and in this single-region sample he also found a statistically significant positive relationship between income and HIV. All the studies referred to so far involve data that related to a number of countries at a single point in time. Studies that use such data are called cross-section analyses. An alternative approach is to use data for a single country over a large number of years. These are called time series studies. When there are a number of countries or regions, and a number of years, the data are called panel data. Brent (2009d) used panel data for 20 regions over eight years in Tanzania to try again to uncover the relationship between income and HIV. Using panel data was important because, as we saw earlier, the three-way links between income, education and religion were so strong in SSA that it was difficult to be able to identify independent effects. In a particular region of Tanzania, religion and race and other cultural variables are unlikely to change much from year to year. So one could see, holding religion relatively constant, the effect of income on the HIV rate in a region over time. The percentage of blood donors with HIV was the dependent variable. The results confirmed that income and HIV were again significantly positive.

REASONS FOR A POSITIVE INCOME/HIV LINK It turns out that it is reasonably easy to explain the positive relation between income and HIV. How does an SSA country become rich in the first place? There are many reasons. Here we just focus on two that are

56

Setting priorities for HIV/AIDS interventions

particularly important in the context of HIV. The rich African countries, particularly the ones in the south of the continent, are the ones where mining contributes a large share of national income. Miners live apart from their wives and families and often live in single sex dormitories, especially when, as in South Africa, it was illegal for a miner to bring his family with him. In these circumstances being faithful to one’s wife is extremely difficult. Also, part of the process of a country transforming from a poor agricultural economy and becoming a richer industrial economy involves people moving from rural to urban areas. Rural–urban migration typically involves just the males and not the whole family, who again live apart from their wives and families in the towns for a period, even if it may be temporary. Large urban populations attract commercial sex workers. And who can afford to pay for the services of CSWs, or set up a girlfriend in addition to supporting a wife and family? Obviously, it is the richer members of a community who have the resources to do these activities.

RESOLVING THE HIGH-INCOME/HIGH-HIV PARADOX The paradox is this. There is good reason, and a lot of evidence to support it, to argue that HIV is a hunger issue and so one would expect income and HIV to be inversely related. At the same time, there is good reason, and also a lot of evidence to support it, to argue that HIV is a disease determined by the rich and so one would expect income and HIV to be positively related. The problem seems to be, how can both of these arguments be correct? The answer is that they are both correct and they both hold at the same time! High income leads to high HIV prevalence rates, and low income also leads to high HIV infection rates. One can begin to understand this when one recognizes that income inequality is one of the determinants of HIV rates that is consistently found in statistical explanations of HIV rates. We saw this in the Stillwaggon (2002) analysis. The pioneering statistical study by Over (1998) also found this result. The books by Epstein (2007) and Barnett and Whiteside (2006) stress the importance of inequality. The levels of income inequality in South Africa and Botswana are some of the highest in the world. And what does an economy with high income inequality have that one with low inequality does not have? The answer is that the economy with inequality has larger numbers of rich and poor who both contribute to there being more persons with HIV/AIDS. An analogy with another disease, obesity, may be helpful. Rispel and Setswe (2007), Table 15, report that in South Africa, 23.3 percent of female children in 2003 were obese, while 6 percent were underweight. How can

Income as a factor

57

South Africa have both large numbers of female children with malnutrition and large numbers who are obese? Malnutrition is associated with low income and obesity with high income. As their per capita income is much lower than the United States, it should have problems with malnutrition and not obesity. But, it is possible for a country with one of the highest rates of inequality in the world to have an obesity level that is comparable to the United States, which is one of the richest nations in the world, yet also have a high rate of malnutrition, as there will be large numbers of rich and poor coexisting when there is so much inequality. To summarize: does low income lead to high HIV prevalence? The answer is yes. Barnett and Whiteside (2006, p. 253) write: “Africa is the only continent in which overall per capita food supply has fallen over the past 30 years”, the period when HIV has exploded in Africa. So HIV is primarily a hunger issue in SSA. Does high income lead to high HIV prevalence? The answer is also yes. So both high income and low income contribute to HIV. What guarantees that both high income and low income are important for explaining the highest rates of HIV occurring in SSA is the existence of widespread inequality. The conclusion that both high income and low income lead to high HIV rates is consistent with the finding from the second Brent study (2009d) related to panel data for Tanzania referred to in this chapter. Just like the cross-section of the 31 SSA countries, he found that the regions of Tanzania that had the higher levels of income (Dar es Salaam, Arusha and Kilimanjaro) also had the higher HIV rates. But as income increased, the rate of HIV decreased. So although the level of income is positively associated with the level of HIV, it is changes in the level of income that are negatively associated with changes in the level of HIV. The policy significance of all this is therefore that because HIV is primarily a hunger issue, raising income can help to make it less of a problem.

14.

Education as a factor raising HIV rates

Here we examine the second of the paradoxes (income was the first) that was found in the Brent (2006) study of 31 SSA countries. This paradox involves the positive relation between female education and HIV rates. We will focus on just female education given that the majority of HIV cases in SSA involve women and gender-specific interventions would seem to be more likely to be important for the African HIV epidemic. As with the income paradox, there are many good reasons for thinking that the relationship between education and HIV should be negative. As the initial attempt to reconcile the paradox turns out to be incomplete, we will fill in the blanks to provide a second explanation that has more immediate policy significance. The story has a “happy ending” because of the joint interaction between income and education that was highlighted in the last chapter.

REASONS FOR A POSITIVE EDUCATION/HIV LINK In the exploratory statistical study by Over (1998) of what determines HIV rates in 72 countries, he not only found the income effect discussed in the last chapter, he also found a particular education effect – that the greater the gap between female and male enrollment rates, the higher the country’s HIV rate. So if female enrollments could be made to rise relative to those of males, HIV rates can be reduced. A subsequent World Bank group study (World Bank, 2002, p. xvii), took the result for the male– female literacy gap seriously and gave it greater emphasis: “A general basic education – and not merely instruction on prevention – is among the strongest weapons against the HIV/AIDS epidemic”. It adds: “the education of children and youth merits the highest priority in a world afflicted by HIV/AIDS. This is because a good basic education ranks among the most effective – and cost-effective – means of HIV prevention” (p. xv). It gives six main reasons why increasing female general education would lead to prevention. All of these reasons increase women’s empowerment in such a way that one would expect that HIV rates would be reduced. Better educated women are more likely, in comparison with their peers: 58

Education as a factor ● ● ● ● ● ●

59

to delay marriage and childbearing; to have fewer children and healthier babies; to enjoy better earning potential; to have stronger decision-making and negotiation skills; to have higher self-esteem; and to avoid commercial sex.

All of these reasons make a lot of sense. The only problem is that the facts do not support the argument that HIV and female education are typically negatively related.

THE POSITIVE RELATION BETWEEN HIV AND FEMALE EDUCATION IN SSA Brent (2006) was so convinced by the World Bank group’s arguments that he set out to explore the strength of the negative relationship between income and HIV for 31 African countries. He did not expect to find that the relationship would be positive. However, he found the “wrong” sign for the education variable in all his equations, no matter the controls he used, no matter the statistical technique employed to carry out the estimation and no matter the education specification. The wrong sign was found for the female literacy rate; the gross enrollment rate (including people at school of all ages) for both primary and secondary schooling; the net enrollment rates (including those only of school age) for both primary and secondary school. For all these variables Brent also tested the differences (gender gaps) between the numbers for males and females to try to reproduce the Over result that sparked the World Bank group’s interest in female education as a way of preventing HIV. All of the gap specifications also had the wrong sign. Even a new specification, the difference between gross and net enrollment rates, to try to capture the idea of a non-standard student (mainly those overage students who might be more likely to engage in sex), did not change the unexpected sign for the education variable.

THE PARADOX AND ONE ATTEMPT TO RECONCILE THE PARADOX In fact, the knowledge of a perverse sign to the education variable was known long before the Brent (2006) study. The World Bank’s (1999) review of the HIV/AIDS literature in developing countries recognized that there were two conflicting sets of results. Its statement of the paradox

60

Setting priorities for HIV/AIDS interventions

was not just in terms of education and HIV, but also included the role of income, which also had a perverse sign in terms of its impact on HIV infections. The two sets of conflicting results were these: (1) low income reduces the ability to afford treatment for STDs or to buy condoms. The less educated have less knowledge of high-risk behavior. Hence least developed countries (LDCs) with higher income per capita will have less HIV. (2) At the individual level, the probability of infection is greater for those with higher income and education. The women attending family planning clinics in Dar es Salaam who had partners with more than 12 years of schooling were five times more likely to be infected than those with partners with no schooling. This was thought due to the fact that educated partners can attract and support commercial and casual sexual partners and have traveled more. The World Bank’s review tried to reconcile the two positions by pointing out that when many became infected in the early and mid-1980s, awareness and knowledge of HIV prevention were low. So the greater knowledge that comes with income did not apply. What did apply was the fact that higher-income people had a higher number of partners, and this explains the positive association. It was conjectured that over time this correlation will be reversed. There was some evidence for this. In Brazil, through 1985, three-quarters of those newly diagnosed with AIDS had a university degree. By 1994, only a third had a degree.

FROM RECONCILIATION TO VACCINE This reconciliation process (that educated people were not fully aware of the HIV transmission mechanism early on in the epidemic, and that when this information became known, the more educated would be the ones most likely to change their behavior subsequently) is entirely plausible and, in the fullness of time, undoubtedly will turn out to be a correct prediction. But, at this point in history, as an explanation of the data, it is premature. In a review of the education and HIV literature by Hargreaves and Glynn (2002) they found that, of 27 studies, only one reported a significant negative relation between HIV and education. The Brent (2006) study of 31 SSA countries used HIV data for 2000 (that is, end of 1999). It did not relate to the mid-1980s and early 1990s. Similarly, Brent’s (2009d) study of 20 regions in Tanzania found a positive relation between HIV and female primary enrollments for the period 1994–2001. Again this was late 1990s and early 2000s and not for a decade earlier. For the reconciliation process to take place over time as suggested by the World Bank, the relation between HIV and education must first be

Education as a factor

61

transformed from positive to neutral and then from neutral to negative. Since the Hargreaves and Glynn survey there has indeed been increasing evidence that the education and HIV relation is becoming more neutral. Glynn et al. (2004, p. 4) find for four African cities, namely, Cotonou (Benin), Yaoundé (Cameroon), Kisumu (Kenya) and Ndola (Zambia) that: “there was no evidence of an increased risk of HIV infection associated with education as seen in earlier studies”, and De Walque (2006) reports that in cross-section studies in five African countries (Burkina Faso, Cameroon, Ghana, Kenya and Tanzania) there was no “robust” association between education and HIV infection. So the tide has indeed started to turn for some countries. But, this does not mean that the relationship has become consistently negative. Yet there are some who cannot wait and have declared that the education/HIV relation is now so clearly negative that education is like a vaccine and education for all would save millions of lives – see Global Campaign for Education, GCE (2004). The claim that millions of lives would be saved by universal primary education is based on the following calculation. De Walque (2007) found when tracking a cohort of the general population in Uganda for 12 years that young people with little or no education were 2.2 times more likely to contract HIV as those who have completed primary school education. This 2.2 differential was then applied to primary school enrollment rates across the globe to obtain an estimate of 700 000 HIV cases avoided per year. In a decade this becomes 7 million cases prevented. Apart from questioning the wisdom of extrapolating from one year to a decade, we have to point out that the 2.2 differential is a best case scenario that is not yet applicable to all countries. In fact, De Walque (2007, p. 688), acknowledged explicitly that his study was a special case. He writes: “This paper seems to be one of the first to report robust evidence that for young cohorts in Africa there is now, at least for females, a negative gradient between HIV infection and education”. If this study is one of the first, and it relates to Uganda, which is a special case (in that it is one of the few success stories in Africa where rising HIV rates have been reversed nationally), then one cannot assume that the 2.2 differential is typical and can be applied to all countries. The inability to apply the Ugandan experience to all countries is confirmed in the latest survey of studies of female education and HIV in SSA by Hargreaves et al. (2008). The vast majority of studies, including preand post-1996, found a positive or no relation between female education and HIV rates. Only three of 35 studies had a negative relation in line with the existence of a vaccine. All three did occur post-1996, but still in this more recent period, 17 of the 20 studies did not support a negative relationship.

62

Setting priorities for HIV/AIDS interventions

HOW THE EDUCATION VACCINE IS THOUGHT TO OPERATE If the education vaccine were to exist, what would make it effective? According to GCE, the education vaccine works via a person’s increased ability to evaluate, understand and apply facts. So an educated person is better able to evaluate HIV messages that now take place than a less educated person and is able to adjust her/his behavior accordingly. In particular, young females would have better negotiating skills and are more likely to require that their partners use condoms. The GCE is right to point out that study after study do show that educated females are more likely to use condoms with casual partners than non-educated females. But, what is not so clear is whether condoms are always used by educated females within marriage. Brent (2009d) found that in a sample of 585 females in Tanzania in 1999, those with primary education were less likely to use condoms with their spouses than for those without primary education. Thus, one reason why the education vaccine is not yet effective could be because of this reduced condom use in marriage. Acting on this information could be one way of strengthening the “vaccine” so that it exists sooner rather than later.

AN ALTERNATIVE ATTEMPT TO RESOLVE THE HIGH-EDUCATION/HIGH-HIV PARADOX The World Bank referred to the HIV/education relation as a “correlation”. As any statistician will tell you “correlation does not imply causation”. However, in the case of the 31 SSA countries, a two-way test revealed that the relationship was causal because female education determined HIV rates, but HIV rates did not determine education levels. When one moves away from two variable correlations to consider more variables, a number of possible interactions are possible. Brent’s (2009f) panel data study for Tanzania worked with the three variables: HIV rates, education and income (in a two-equation framework). In this set up, one could estimate the direct effect of education on HIV rates and also an indirect effect of education working through income. The direct effect for the 20 regions of Tanzania was the same as for the 31 SSA countries, and the same as for most of the studies in the Hargreaves and Glynn survey. That is, there was a positive relation between education and HIV rates. On the other hand, one of the major reasons why people pay for education is that schooling increases one’s income. This education/income positive relation is almost universal. It applied also to Tanzania. So regions that had rising primary

Education as a factor

63

school enrollments were ones with rising income levels. Now recall from the previous chapter that it was found that increasing a region’s income level (not having a high income level) lowered its HIV rate. So the indirect effect of education on HIV was negative (education goes up, bringing income up, leading to HIV going down). It turned out that the positive direct effect of education on HIV was overwhelmed by the negative effect of education working through income. So the net effect on HIV rates of education was ultimately negative. One need not be overly concerned with a positive direct effect if the net effect is negative. The paradox can be circumvented. To conclude: it is likely that the effect of education on HIV infection rates, holding income constant, is positive for many developing countries and will continue to be positive for a few more years at least. Over time one can expect that this will be reversed. But, in the meanwhile, an increase in female education, allowing income also to rise, could still be effective in lowering HIV rates, if Tanzania’s experience is a good example. Of course, effectiveness is just one element in an evaluation of an intervention even when the direct effect itself becomes negative. It must also be shown that the benefits flowing from that effectiveness outweigh the costs. A full CBA of education needs to take place and this will be covered in the next part of the book.

15.

Islam as a factor lowering HIV rates

Although Gray (2004) reports that in six out of seven studies that he surveyed there was a significantly negative relation between HIV prevalence and being Muslim, and Brent (2006) also found this inverse relationship in 31 SSA countries, one does not really need to undertake a detailed statistical analysis to be aware of the relation between Islam and HIV in Africa. One can see the relation visually on any geographical HIV prevalence map of Africa by comparing it with a geographical map of the percentage rate for the population that are Muslim in that region. HIV prevalence rates are least (between 0 and 2 percent) in North Africa where most predominately Muslim countries exist. HIV prevalence rates are between 2 percent and 5 percent in West Africa, where there are more predominately Muslim than non-predominately Muslim nations; and between 5 percent and 15 percent in Central and Eastern Africa, where there are more non-predominately Muslim than predominately Muslim nations. HIV prevalence rates are greatest (between 15 percent and 37 percent) in Southern Africa where most African nations with Muslims as a minority can be found. The question that needs to be answered is: why is there this negative relationship between the percentage of a country’s population that is Muslim and the HIV prevalence rate? We shall see that some reasons (norms) are inherent to the faith of Islam, while others are more due to the religious environment that Islam sets up rather than the faith itself.

MUSLIM NORMS THAT LIMIT THE SPREAD OF HIV There are some tenets of the Islamic faith that would seem to be risk factors for HIV, such as permitting up to four wives and allowing divorce relatively easily, which together would increase the number of lifetime sexual partners. But, the fact that the overall effect was a negative HIV impact means that the Islam risk-reducing factors must have outweighed those increasing the risk. So we will just concentrate on the norms reducing the spread of HIV, namely prohibitions on alcohol consumption and 64

Islam as a factor

65

the requirement that male Muslims be circumcised (female circumcision is not a requirement). (i)

The Prohibition of Alcohol

Islam prohibits the consumption of alcohol. Obviously, alcohol does not cause HIV directly. But, what it does do is impair sexual decision-making abilities leading to the promotion of risky behavior. Under the influence of alcohol, the ability to use a condom goes down. Commercial sex workers would appear to be a more attractive option if one is drunk at the time. (ii)

Male Circumcision

Male circumcision (MC) is also required by the Muslim faith – it is a precept of the Koran. Its prevalence is therefore greater the larger the percentage in a country that is Muslim. Drain et al. (2006) report that among 49 countries with “high” male circumcision prevalence (having a circumcision rate greater than 80 percent of the population), the mean percentage of the Muslim population was 69 percent and the mean percentage of the Christian population was 16 percent. The question is: to what extent does being circumcised lower HIV rates? A distinction needs to be drawn between MC efficacy and MC effectiveness – see Boily et al. (2008). Efficacy refers to the effect of MC under ideal conditions (for example, with 100 percent compliance and adherence) and efficiency refers to the effect in the “real world” under incomplete compliance. There is no doubting MC’s efficacy (if one recognizes that MC may impact HIV on its own or via its impact on other sexually transmitted diseases). In three randomized controlled clinical trials in Kenya, Uganda and South Africa, the reduction in HIV prevalence rates were between 50 and 60 percent. As for the effectiveness of MC, of the 81 variables tried one by one by Drain et al. (2004) to explain HIV rates in 122 developing countries (including development indicators, sexual behavior, reproductive health, economic factors, population and religion) circumcision had the strongest association with HIV prevalence and its effect was to lower the HIV rate. However, the link between being circumcised and being Muslim was so strong in the data set that one could not separate the two effects as determinants of HIV prevalence in a country. In the context of just the African continent, it is highly likely that MC does have a separate effect as HIV rates are lower in West Africa than in East and South Africa and this would be because MC is more widespread in West Africa and not just limited to Muslim groups. Another important distinction to draw in the connection of MC is that

66

Setting priorities for HIV/AIDS interventions

between its effect on physiology and its impact on behavior. De Walque (2006) points out that the clinical trials have confirmed the physiological impact of MC. Biologically, the underside of the foreskin is rich in HIV target cells (CD4 T cells), so the removal of the foreskin would reduce the risk of HIV. But, there is nothing in the circumcision procedure that necessarily impacts behavior. This is important because De Walque found that, in five SSA countries there was no association between MC and HIV prevalence (with or without controls for religion). So the positive physiological effect of MC was offset by a negative behavioral effect whereby those with MC were less likely to be faithful, less likely to be abstinent and (if single) more likely to have had earlier sexual initiation.

CHARACTERISTICS OF MUSLIM NATIONS NOT RELATED TO THEIR FAITH Apart from tenets of religion, some of the reduction in HIV prevalence rates in Muslim countries could be due to capital punishment that exists in some Muslim countries that would cause a reduction in HIV risk behavior. Mackay (2001) points out that eight countries (all in the Middle East) have the death penalty for homosexuality and four countries (Iran, Pakistan, Saudi Arabia and Yemen) have the death penalty for adultery. Even though capital punishment is not always enforced in these Muslim countries, one can assume that extra-marital affairs are socially frowned upon and will not be tolerated.

SUMMARY AND CONCLUSIONS One main reason why Muslim countries (and Muslim regions within a country) have lower HIV prevalence rates is due to their requirement of male circumcision, and MC has been shown to have a preventative effect for HIV transmission. So although no one has suggested that one become Muslim to reduce the risk of HIV, there is a strong movement towards introducing MC as a way of preventing the spread of HIV. Bailey et al. (2008) report that one study has estimated that the potential impact of MC would be to avert 5.7 million new infections in SSA, and another study shows that (with 50 percent MC uptake over ten years) HIV prevalence in men would decline in Nyanza province in Kenya from 18 percent to 8 percent. There are two main issues that arise from all this that have relevance to the undertaking of HIV intervention evaluations: Although MC is effective in general terms, there are a number of steps

Islam as a factor

67

that have to be demonstrated before it can be shown to be worthwhile to scale up MC in any given context. Apart from measuring the costs of circumcision as it is carried out now (and Bailey et al. do report studies that show MC is as cost-effective as other HIV prevention tools) one also needs to look at the side-effects of circumcision in both traditional and modern settings in order to ensure that resources are in place to limit the side-effects of increasing the scale of MC efforts. The concern therefore is whether MC can be provided safely when provided to a large number of males in developing countries. Bailey et al. found in their study of over 1000 MCs in Bungoma, Kenya, that 35 percent of the traditional circumcisions had adverse side-effects (such as infections, swelling, profuse bleeding and lacerations), many of which were serious and permanent. Even in modern settings, there were adverse side-effects (22.5 percent in private facilities and 11 percent in public facilities). Many of the adverse side-effects were due to a lack of sharp instruments and a lack of adequate sterilization equipment. Bailey et al. suggest that much additional resources and staff training needs to take place before scaling up of MC should take place. MC is effective, all other things held constant (as they are in a clinical trial). However, subsequent behavior can offset the preventative effects of MC. If a circumcised person has more sex partners just because he is circumcised, then the MC intervention will have nothing to show for itself. In this way MC can mirror ARVs. The advent of ARVs meant that HIV was not a death sentence. But, in the United States at least, ARVs have been one cause of an increase in the number of sex partners for people who are infected as MSMs. In the same way, MC can increase new HIV infections if people try to exploit the effectiveness of circumcision by reverting back to past risky sex behavior.

16.

Impact of HIV on agricultural households

Even if one may doubt that hunger is the main cause of HIV transmission in Africa, it is clear to all that a major consequence of HIV/AIDS in SSA is hunger for many of those remaining alive. HIV affects prime age adults. So when these adults die (and 2 million adults died of AIDS in 2007 alone) this adversely affects the food consumption of others. We look at a typical downward spiral initiated by an HIV infection in an agricultural household and use this to help identify the main groups affected. From there we take these groups and show how they are adversely affected so that one can get a handle on ways to mitigate the harmful effects of the deaths from AIDS.

A TYPICAL DOWNWARD SPIRAL FOR AN AGRICULTURAL HOUSEHOLD INFECTED WITH HIV Gillespie et al. (2001) have compiled a list of the typical steps in the progression of the HIV disease on an agricultural household, which is the main social grouping in SSA. These steps are: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

Adult becomes sick and he or she reduces work. Replacement labor is “imported”. Adults work longer hours on the farm. Health care expenditures rise and household food consumption is reduced. Switch to labor-extensive crops and farming systems. Nutritional status deteriorates. Adult stops work and increased care given to sick adult. Divisible assets disposed of, for example, livestock. Debts increase and children drop out of school. Adult dies and funeral expenses incurred. Household fragments as other adults migrate for work. Cultivation of land is reduced, as more land is left fallow. 68

Impact of HIV on agricultural households

13. 14. 15. 16. 17. 18. 19.

69

Inappropriate natural resource management may lead to pests and disease. Effects of knowledge loss intensify. Increased mining of common property resources. Household property rights affected (re: surviving widow). Solidarity networks strained. Partner becomes sick. Downward spiral follows previously identified steps for first sick adult, but accelerates.

From this list of steps, we can see that there are three main sets of people affected by the death of an adult in a household: other prime age adults, parents and children. Other prime age adults have to adapt their labor and work practices to the loss of labor caused by the deceased. They may have to migrate to join or form new households. Wives of the deceased may also have to migrate and join other households if their property rights are not protected. The parents of the deceased may have had to pay for the medical services and give up their time in order to care for their adult offspring prior to their death, and pay for the funeral expenses after their death. The adult’s parents may have to look after the children of their offspring, for example, their grandchildren. The children may also have to change households, whether going to their grandparents’ home or living with some other adults. We now outline some of the important studies that have focused on the three sets of people affected by an adult HIV death.

IMPACT OF WORKING-AGE ADULT DEATH ON SMALL-SCALE FARM HOUSEHOLDS IN KENYA Yamano and Jayne (2004) used a two-year panel of 1422 Kenyan households before and after an adult death in 22 districts between 1997 and 2000. They studied household size and composition, crop production, asset levels and off-farm income. They estimated the proportion of deaths due to AIDS by looking at the death rates of HIV negative adults with similar characteristics to calculate the number of deaths to be expected in the absence of HIV/AIDS. Thus, the HIV negative sample was the control group. They found: ●

Half of the deceased prime age men were in the highest per capita income quartile in 1997 and likely to be household heads (HHs). This income result was examined in detail in previous chapters

70

Setting priorities for HIV/AIDS interventions

●

●

●

so should not be a surprise for us now. But, because of this fact there were severe reductions in off-farm income (of the order of 79 percent) when adult males died. Some of this reduction was due to the fact that when a male HH died, the number of adult women living in the household also declined. So the loss of labor of other family members contributed to the loss of income and not just that related to the deceased male. There were no statistically significant income shocks from female HH deaths. The decline in household size was greater than one (the person who died). Older daughters in households afflicted by male HH deaths were very likely to leave the household (and possibly get married). The number of younger boys and girls declined when there were female HH deaths. Grain crops were adversely affected when female HHs died, while it was “cash crops” (coffee, tea and sugar) that were affected when male HHs died. Farm equipment was sold off when male adults died, while small animals were sold when females died.

Overall then, the results showed that the effects of an adult death very much depended on the gender and position of the deceased person in the household (whether they were an HH or not) and the household’s initial asset levels. Poor households did not recover quickly, at least not over the three-year period. Yamano and Jayne concluded that any assistance to HIV-affected households should target those with the recent death of a male HH.

IMPACT OF WORKING-AGE ADULT DEATH ON LABOR SUPPLY IN THE SHORT RUN IN TANZANIA Kagera was a region of Tanzania affected early and greatly by HIV/ AIDS. It is therefore a good region to study to discover the impact of AIDS deaths on agricultural households. Note that in Africa a prime age adult death is quite rare in the absence of AIDS. Beegle (2005) looked at labor supply decisions in terms of the number of hours worked and the types of employment undertaken. She analyzed data related to 800 households interviewed four times over the period 1991 to 1994. What was particularly notable about her work was that she looked at labor supply decisions in the year prior to death as well as the year after death. This is important because some of the impact of an adult death could be experienced prior to death and not just after death, which was

Impact of HIV on agricultural households

71

the Yamano and Jayne focus. This is a particularly important distinction for examining AIDS deaths relative to other adult deaths because: (1) an AIDS death can be anticipated, which is not the case for deaths occurring by accident, and (2) prior to an AIDS death there is illness, with all the effects of this illness occurring, which again is different from accidental deaths. Overall, the number of hours worked did not change either a year before or year after an adult death. Children’s hours worked did not rise as one might have expected. This could have been because when an adult death occurs, a new adult joined the household. Rather, the impact of adult deaths was felt in types of employment. These impacts were largely experienced prior to death and not afterwards. Future deaths of adult males or females in a household led to significant decreases in wage employment by males, but neither wage employment nor non-farming employment (for example, doing chores) fell in response to past adult deaths. As for types of crops grown, coffee and banana farming was lowered in the six months following a male death, but these crops were unaffected for periods greater than six months after a death. This suggests that some farm activities are only temporarily reduced as a result of male deaths.

IMPACT OF WORKING-AGE ADULT DEATH ON CONSUMPTION IN THE LONGER RUN IN TANZANIA If adult deaths cause significant adverse effects in the short run, what happens if we consider a much longer period? Beegle et al. (2008) looked at a cohort of households in the Kagera region of Tanzania at baseline (1991–94) and again in 2004, over ten years later. Because it was a cohort, consisting of the same individuals in both periods, it meant that they had to track those in the original 832 households to see where they ended up. The 832 became over 2700 households in 2004. The individuals who were in the original baseline and who moved to new households where an adult death did not take place were the control group for the individuals in the original households who resided in households where an adult death did take place. Thus, the two groups had the same history and initial characteristics except that one group had the adult deaths and the other did not. Tracking was important because migration and the dissolution of households are responses to escape poverty following the death of a prime age adult. Between 1991 and 2004, 22 percent of the households experienced an adult death (a person dying aged 20–55). Households with such a death

72

Setting priorities for HIV/AIDS interventions

had a 7 percent drop in consumption in the first five years after the death. After the first five years the negative effect became smaller and statistically insignificant. This is evidence of a strong recovery. The impact of an adult death did not vary according to the income level at baseline; nor did it matter whether the person who died was a blood relative or not. The ability of the household to cope in the longer run was probably due to households being reorganized, whether by migration or household reformation. In a third study of the Kagera region, it was found by Beegle and Krutikova (2006) that female orphans from male deaths had a 25 percent greater probability of marrying young, which is an adverse effect seeing that they are more likely to have children soon after marriage and younger mothers are more likely to suffer micro-nutrient deficiencies and be unaware of the health risks associated with pregnancy at an early age (that is, increased risk of maternal and infant mortality). The greater marriage probability did not hold if the females orphaned were at school, living in households in non-farming sectors, or if they were in rich households.

IMPACT OF WORKING-AGE ADULT DEATH ON WOMEN’S LAND RIGHTS IN KENYA According to Aliber and Walker (2006), Kenya has the most extensive rural registration system for land in SSA. The country was therefore a good case study to see what happens to widows’ land ownership when their husbands die of AIDS. They looked at three sites scattered throughout Kenya. Over three-quarters of those acquiring land did so via inheritance. A crucial distinction for registration was between inheritance with subdivision and inheritance without subdivision. For registrations of inheritance without subdivision, individuals do not receive formal transfer of ownership. Much of the land was held “non-formally” such that it was registered in the name of someone (usually the parents) other than the household that resides on the land or farms it. In the three areas studied, married women traditionally accessed their land through their husbands. Given that their husbands do not have land registered in their name, the married women would appear to be especially vulnerable to land ownership threats if their husbands died of AIDS. Although women who did not have title deeds were only weakly protected by law in Kenya, there was an informal system set up of local leaders (such as elders and sub-chiefs) who felt an administrative responsibility to ensure that women’s land rights were not discriminated against. There were many threats to women’s land ownership on the deaths of their husbands, including land disputes, tenure threats and tenure loss.

Impact of HIV on agricultural households

73

However, there was not a statistically significant difference in the proportions of land ownership threats for households affected by HIV/AIDS than non-affected households. More important was the fact that the tenure security threats to the HIV/AIDS-affected households that did take place did not usually result in actual, sustained loss of land rights. To summarize: although most women did not have property rights in Kenya, their husbands often did not either. So they did not have rights to lose in the first place. The women ended up continuing on the land that they were brought up on, or the land they moved to when they got married.

IMPACT OF WORKING-AGE ADULT DEATH ON PARENTS’ FINANCES IN CAMBODIA AND THAILAND Knodel (2008) reports that in both Cambodia in 2004 and Thailand in 1995, losing a child to AIDS results in a larger percentage of parents reporting that they have experienced a worsening of their financial situation in the last three years, holding constant education, age, marriage and so on. In particular: 52.8 percent had a worsened financial situation in Thailand after an AIDS death relative to 47.1 percent with no death; and 63.3 percent had a worsened financial situation in Cambodia after an AIDS death relative to 56.9 percent with a non-AIDS death and 40.9 percent with no death. Thailand’s experience makes clear that an AIDS death, like any other death, will lower a household’s financial situation. Cambodia’s experience adds the fact that an AIDS death is not like any other death as it worsens a household’s situation even further. This could be because there will be sickness prior to death, or because an AIDS death has a stigma that other deaths do not have. The Knodel study reminds us exactly why HIV/ AIDS has a more impoverishing effect on parents in poorer countries. In Thailand, a largely Buddhist country that is classed as middle income by the World Bank, the government provides an extensive public health system and there is a welfare system that supplies a small monthly income for elderly people who are indigent; while in Cambodia, a largely Buddhist country that is classed as low income by the World Bank, few of the poor have health insurance and government assistance. In other words, in a poor country like Cambodia, parents of adults with HIV/AIDS are more likely to be adversely affected because there are fewer alternatives to parental assistance available.

74

Setting priorities for HIV/AIDS interventions

RELEVANCE OF THE IMPACT OF HIV ON AGRICULTURAL HOUSEHOLDS FOR CBA From the perspective of carrying out a CBA, the material in this chapter is relevant in terms of its role in expanding considerably the list of what one can call an HIV intervention to be evaluated. An HIV intervention need not involve a condom or antiretroviral drugs. It can be anything that interferes with, or reverses, the downward spiral that exists when a person becomes HIV positive in a household. Thus, for example, one needs to evaluate giving financial assistance to parents of infected adults; providing technical advice as to which labor-saving crops to introduce; or altering property rights for women.

17.

Agricultural policy and HIV interventions

In this chapter we continue the theme began in the last chapter that nonmedical, and even non-sexual interventions, which involve influencing agricultural household’s behavior, can be effective and worthwhile HIV policies. Pisani (2008) makes the claim that HIV/AIDS cannot be mainly a developmental issue because it is the more developed countries in SSA that have high HIV prevalence rates and not the less developed countries. We have already addressed this issue in Chapter 13 where we argued that both low and high income can lead to high infection rates. Here we will explain some agricultural interventions that have had an impact, though not all favorable to the aim of reducing HIV rates. The fact that some agricultural policies lead to trade-offs is another important reason why CBA is necessary for setting HIV priorities as one needs to measure costs as well as benefits. We close by drawing together material from this and the previous chapters that reinforce the argument that HIV in SSA is mainly a hunger and economic development issue.

AGRICULTURAL POLICIES CAN LOWER HIV/AIDS Loevinsohn and Gillespie (2003) cover in detail a large body of work examining the link between HIV and agriculture. Here are some of their main findings: ●

●

●

In Malawi, opening times of rural markets were said to favor sexual relations more than commercial ones. Tanzania changed opening and closing times of rural markets to make the trade in sex more difficult, that is, having the markets early and in daylight so that people would have to hang around for a long while before they can use the cover of dark to conceal their activities. Some medicines need to be taken “on a full stomach”. Antiretroviral drugs are toxic, and may be particularly toxic to the malnourished. So a full stomach makes ARVs more effective. The provision of family accommodation rather than single-sex 75

76

Setting priorities for HIV/AIDS interventions

●

dormitories on plantations and other rural industries can reduce sexual activity. Marketing arrangements for commodities like tobacco, sugarcane and fish often draw men and women away from their families for extended periods leading to situations of risk.

Loevinsohn and Gillespie conclude that one needs to develop an “HIV/ AIDS lens” when considering agricultural policies. HIV must be “mainstreamed” into development policy.

AGRICULTURAL PRACTICES CAN INVOLVE A TRADE-OFF Some ways that agricultural practice adjusts to the adverse economic impact of HIV end up increasing the risk of HIV: ● ●

● ●

Breastfeeding leads to some nutritional benefits for the child, but has a risk of increasing MTCT. Some changes in crop selection (such as cassava) make sense because they do not require a lot of labor, and can bolster income, but these alternative crops have a lower protein concentration at a time when HIV is raising protein requirements. Better transport infrastructure facilitates the marketing of food surpluses, but may increase travel-related susceptibility. Liberalization of food markets may increase income, but it may increase risk if there is more travel and overnight stays at spatially scattered markets.

The policy conclusion that follows from an awareness of the HIV tradeoffs that agricultural changes can generate is that we need to employ CBA to evaluate impacts. Since some changes may increase the risk of HIV, and do so in different time periods, we have to weigh up the costs and benefits of short- versus long-term responses by agriculture to HIV.

AGRICULTURAL POLICIES AND HIV INTERVENTIONS To respond to the HIV epidemic one can do three things: ●

prevent HIV with behavior modification (for example, condoms);

Agricultural policy and HIV interventions ● ●

77

treat HIV with medications (for example, ARVs); mitigate harmful effects on others (for example, assist orphans).

The main focus for our discussion in the last and previous chapters was how agricultural policies can impact mitigation. However, as Loevinsohn and Gillespie point out, the three are very much interconnected and agriculture/nutrition has a key role in all of these: ●

●

●

Treatment can be preventative: nutritional supplements that reduce the viral load reduce the transmission in an unprotected sexual encounter, even without a change in their behavior. Mitigation can be preventative: social innovations that secure widows’ entitlements, allowing them to exchange a fair share of the production for the young adult’s labor would benefit both and keep both on the farm. Treatment can be mitigating: ARVs that lead to a longer period between HIV and AIDS ensure that the person can continue to be productive on the farm and support the rest of the family.

We can now reformulate our argument that HIV/AIDS is mainly a hunger issue in SSA from the point of view of intervening to combat the disease. There are some, like Pisani, who state that as HIV is largely transmitted sexually in SSA then sexual behavior must be the main cause. Let us assume that this is correct. It is clear that altering sexual behavior relates to the category of “prevention” and not to treatment, or mitigation. What is even clearer is that intervening via treatment and mitigation must be related to low income and poverty. ARVs are still expensive (equivalent to a person’s per capita income in many SSA countries) and they are an expense throughout a person’s lifetime (ARVs need to be obtained each and every year). Many individuals in SSA cannot afford to take the HIV tests to access the ARVs even if they get ARVs free of a user charge. When it comes to mitigation, no one argues that a poor person/country is better able to cope with the disease than a rich person/country. So hunger and poverty is an issue with two out of three of the intervention categories. Given that the three categories are interdependent, the role of hunger and poverty must be central. Agricultural policies in particular, and development policies in general, become key intervention areas even in a world where HIV is sexually transmitted.

18.

Sex and HIV I: the role of transmission

Although HIV is largely a hunger issue in SSA, this does not mean that it is unnecessary to discuss the role of sexual activity. As we know, the main type of transmission in SSA is through unprotected heterosexual sexual intercourse. So an examination of sexual behavior in SSA cannot be kept out of the equation. In the next three chapters we will analyze what is known about the role of sexual activity in the heterosexual transmission of HIV. In this chapter we focus on individual behavior and in the next two chapters we extend the analysis to cover sexual behavior in a social context. Because we will be concentrating on heterosexual transmission, we will proceed from a discussion of why HIV prevalence is so high in SSA relative to other regions, to an examination of why HIV is so prevalent among African Americans relative to other races in the United States. To help understand why HIV in SSA is more of a hunger issue than one of extreme sexual behavior, we will cover in detail the study by Oster (2005) and tie this in with the work of Eileen Stillwaggon (2002, 2006) referred to in earlier chapters.

WHY SEXUAL BEHAVIOR CANNOT BE THE CHIEF DETERMINANT OF THE HIGH HIV PREVALENCE RATES IN SSA Stillwaggon (2002) makes clear that sexual activity, in an otherwise healthy person, is an inefficient way of transmitting HIV. She cites these facts related to sexual transmission in the United States and Europe: (1) transmission from an HIV positive female to a healthy male occurs in one in 1000 contacts; (2) transmission from an HIV positive male to a healthy female occurs in one in 300 contacts. So let us do the arithmetic. On average, an uninfected woman would have to have sex with an infected man every day for one year to get HIV; and on average, an uninfected man would have to have sex with an infected woman every day for three years to get HIV via unprotected heterosexual activity. However, this greatly overstates the chances of transmission as, typically, couples do not have 78

The role of transmission

79

sex every day. For example, MEASURE (2001) estimates for Tanzania that an average man (albeit including those who are not sexually active at all) has 30 sexual acts per year. At this rate, the average man would have to have sex for over 30 years before he was likely to contract HIV from an infected female partner. It would then take another ten years for him to infect a new female partner (assuming that he had another one). A heterosexual HIV epidemic would not be sustained with these transmission rates. The key phrase then in the transmission numbers quoted above is the requirement that the person be “otherwise healthy”. Since this is often not the case in Africa, one should expect the transmission rate in SSA to be much higher than these figures suggest. We have already covered in depth the presence of malnutrition and how this compromises the immune system and makes people in SSA more vulnerable to HIV. Stillwaggon (2006) emphasizes the role of malnutrition and adds the fact that parasitic diseases are also much more prevalent in SSA. Remember that HIV is transferred from one person’s blood to another. Anything that breaks the skin and brings people’s blood in more ready contact with another’s must make them more susceptible to getting HIV from having sex with an infected person. We made these points earlier in Chapter 10. Since we intend to build on this understanding let us review them again now. Nearly everyone knows that one cannot get HIV directly from being bitten by mosquitoes, as they extract your blood and do not inject others’ blood into you. But, indirectly, mosquitoes can play a role in HIV transmission as they open up one’s skin and bring blood to the surface. Malaria is higher in SSA than elsewhere and so is schistosomiasis (a disease caused by flatworms), which is another way that skin can be opened up. Many women in SSA are immersed in water in lakes for long periods during various activities (for example, fetching water to drink and cook with) and then get worm eggs that penetrate their skin and generate genital lesions. Parasitic diseases compound the effect of malnutrition as they lower the immune system as well as opening up one’s skin. Stillwaggon’s central thesis (2006) is therefore that HIV in SSA is influenced by the same factors that promote the transmission of other infectious diseases. Nutritional deficiencies, parasitic diseases, poor general health and little access to health services all occur because people are poor. So HIV transmission is mainly due to these factors and not because people in SSA are more sexually active.

80

Setting priorities for HIV/AIDS interventions

SEXUALLY TRANSMITTED DISEASES AS THE TRANSMISSION MECHANISM IN SSA The presence of sexually transmitted diseases, STDs (also called sexually transmitted infections STIs) is another reason why the skin gets broken. So STDs that manifest themselves as open genital sores greatly increase the probability of blood appearing during sex. Stillwaggon acknowledges that STDs are a co-factor in HIV transmission and play a more significant role in SSA because of the kinds that are prevalent, especially genital ulcers, which are common in areas where water is scarce as is the case in most of Africa. The main reason why STDs are a co-factor in SSA is because, due to poverty and the inadequate access to health services, they are more likely to be untreated. Oster (2005) argues that it is because STDs are largely untreated that one should expect to find a higher transmission rate in SSA than in the United States. Using data on hemophiliacs and transplant recipients, Oster places the per partnership transmission rate in the United States as 10 percent male-to-female and 5 percent femaleto-male; and, based on figures for the general population in three East African countries, she sets the per partnership transmission rates in SSA to be 27 percent male-to-female and 12 percent female-to-male.

THE OSTER SIMULATION MODEL The unit of analysis in the Oster model was a partnership. Some partnerships were more likely to be involved with risky sex (unprotected sex among casual partners) than others. If one knows how many partnerships engage in risky sex in a population, and if one knows the transmission rate for each partnership, then one can estimate what the overall HIV prevalence rate would be. Oster reported that the actual United States HIV prevalence rate was around 0.15 percent and it was 11.9 percent in a sample of 12 SSA countries. Oster wanted to find out, given her assigned difference in transmission rates between SSA and the United States, how much of the disparity in HIV prevalence rates between the two regions can the per partner transmission rate difference explain. If it is not the transmission rate, then it must be the type and extent of sexual behavior that would account for the difference in prevalence rates. Oster’s analysis came in three parts: 1.

First she estimated (simulated) what would be the HIV prevalence rates in the United States and SSA if they had the sexual behavior that they did have and had the per partner transmission rates that she had

The role of transmission

2.

3.

81

assigned. She wanted to see if her model could predict current prevalence rates. Second she estimated what would be the HIV prevalence rates in the United States and SSA if they both had the same assigned transmission rate, but had the actual differences in sexual behavior. Third she estimated what would be the HIV prevalence rates in the United States and SSA if they both had different transmission rates, but had the same pattern of sexual behavior.

The types of sexual behavior that would affect the spread of HIV were assumed to depend on four characteristics: gender, age, marital status and types of sexual partnership. Gender was important not just because it led to differences in transmission rates; the number of partners within and outside marriage, the type of partners (casual or regular) and the age of marriage were all functions of gender. The type of partner was important because condoms would be more likely to be used with casual partners. The top row of Table 18.1 records the assumptions about transmission rates used in the model. The middle rows show the actual sexual behavioral differences in the United States relative to the SSA countries in terms of the share and number of casual partners by gender and marital status. We see that sexual behavioral differences are not large. A greater share of United States individuals have casual partners before marriage, but they have a lower share of casual partners within marriage. The bottom rows of Table 18.1 present the current and predicted HIV rates. Given the actual sexual behavior in the middle rows, and the transmission rates in the top row, we see that the model predicts that HIV prevalence would be 0.2 percent in the United States and 12.7 percent in SSA, very close to the current prevalence rates of 0.15 percent and 11.9 percent. So the model gives a good fit to the data. The final two rows of Table 18.1 show what the model predicts given different sexual behavior and different transmission rates. If the United States had the SSA transmission rates given in the top row, but retained the sexual behavior differences shown in the middle of the table, then the United States at 12.2 percent would virtually have the same prevalence rate that SSA currently has, that is 12.7 percent. While if the United States had adopted SSA sexual behavior patterns and retained its low transmission rates, its prevalence rate at 0.2 percent would be indistinguishable from the 0.15 percent prevalence rate that the United States currently has. In other words, almost all of the difference in prevalence rates between the United States and SSA is driven by differences in the transmission rates and close to none of the disparity is due to differences in sexual behavior. Although condom use in the United

82

Setting priorities for HIV/AIDS interventions

Table 18.1

Sexual behavior and actual versus predicted HIV prevalence rates in the United States and SSA

Transmission rates

United States

Sub-Saharan Africa

Male-to-female 10% Male-to-female 27% Female-to-male 5% Female-to-male 12% Share with casual partners Single women Married women Single men Married men Number of casual partners Single women Married women Single men Married men Actual HIV prevalence rate Simulated HIV transmission rate Both having SSA transmission rate Both having SSA sexual behavior

Source:

72.7% 2.4% 81.4% 6.2%

27.6% 8.9% 53.6% 21.1%

1.49 1.52 2.33 2.13

1.18 1.24 1.80 1.57

0.15% 0.2% 12.2% 0.2%

11.9% 12.7% 12.7% 12.7%

Table II of Oster (2005) with items added and deleted.

States at 63 percent was much higher than the 44.5 percent condom use in the SSA sample, differences in condom use do not explain differences in the HIV epidemic. Oster could not apply the same three types of analysis she developed to account for United States and SSA prevalence differences to determine why the rates are higher in East than in West Africa because data on West Africa transmission rates were not available. Difference in prevalence rates across the African region were mainly explained by the start date of the epidemic (the year when the HIV rate first reached 1 percent). This finding is consistent with the work of the World Bank (1999) and Brent (2006). What is particularly novel about Oster’s analysis is that she could explain why the starting dates were what they were. She cites evidence that the Democratic Republic of Congo was where the epidemic originated. The year the epidemic started in a country was therefore regressed on the distance of that country from the former Zaire. Oster found that around 60 percent of the variation in country HIV starting

The role of transmission

83

dates was associated with the distance of a country from the presumed origin of the epidemic. A country’s starting date was largely outside of its control; a country cannot choose how close it is to the Democratic Republic of Congo.

19.

Sex and HIV II: the role of concurrency

The last chapter dealt mainly with individual behavior, such as the number of partners and whether the partner was their spouse or a casual relationship. The transmission rate analyzed was per partnership. The more partners that individuals have, especially the more casual partnerships, the higher the HIV prevalence rate. Now we look at sexual behavior more in a social setting and recognize that the sexual behavior and history of a partner may be just as (or even more) important than the number of partners. Moreover, partnerships can interact and so some partnerships are more pivotal than others in spreading the disease throughout the population. In this chapter we deal with the timing of partnerships and in the next examine how some partnerships are more strategic than others. The information in Table 18.1 that Oster (2005) used in her analysis showed that the number of casual partners for men and women, whether they were married or not, were higher in the United States than in the 14 countries in SSA. This result is consistent with a number of other studies that find that Africans do not have more sex partners than elsewhere. For example, Caraël (1995) found that men in Thailand and Rio de Janeiro were more likely to report five or more casual sex partners in the previous years than men in Tanzania, Kenya, Lesotho or Lusaka in Zambia, and very few women in these African countries had five or more partners. So it is not the number of sexual partners that people have in SSA that can account for the higher HIV prevalence rate in Africa. There are three ways that people can select partners. If someone is “exclusively monogamous” then the partnership is isolated and does not affect other partnerships. Then there is “serial monogamy”, where people go into relationships one at a time. Other partnerships are affected, but sequentially. Finally, there is “concurrent partnerships” where partnerships overlap as individuals have more than one partner at the same period of time. With the first type, there can be no transmission from partnership to partnership. With serial monogamy a period of time must elapse before subsequent partnerships can be affected by earlier ones. Transmission is largest with concurrent partnerships for two main reasons. First, no time is used up in waiting for one relationship to end and a subsequent one to 84

The role of concurrency

85

begin. So transmission can be simultaneous. Second, as we mentioned in Chapter 3, the viral load is greatest for the first three months of HIV (and at the later stages when AIDS takes place). Concurrency means that the middle stage, when transmission is least likely to occur, could be skipped altogether if partnerships are simultaneous. Thus, two early-stage, highinfection relationships can coexist for an individual. Morris and Kretzschmar (1997) have compared the spread of HIV in two hypothetical populations, each with the same number of sexual relations. In one, serial monogamy was practiced, while in the other concurrency was widespread. The epidemic was ten times greater after five years in the population with half of the partnerships concurrent. These striking results suggest that concurrency has an important role to play in the spread of HIV. We will now highlight the existence of concurrency to account not only for the higher prevalence rate in SSA, but also to explain why the HIV rate is greater among African Americans than whites in the United States.

CONCURRENCY IN SSA It is the presence of greater concurrency that is the most convincing sexual explanation as to why the prevalence rates are higher in SSA. Halperin and Epstein (2004) give the percentages of males with concurrent partners as 55 percent in Lesotho, 36 percent in Côte d’Ivoire, 22 percent in Lusaka, 18 percent in Tanzania and 13 percent in Kenya, while it is only 2 percent in Singapore and Sri Lanka, 3 percent in Manila and Thailand and 7 percent in Rio de Janeiro. Halperin and Epstein then point out an important implication of this greater concurrency for HIV prevention efforts as they relate to condoms. We will examine the case for condom subsidies in great detail in Part III. Here it is only necessary to point out that condom promotion has been a mainstay of donor-financed HIV prevention in Africa for a long time, yet the greater utilization of condoms has been painfully slow. For example, in Tanzania, in 1989/90, 5 percent of sexually active women and 10 percent of men had ever used condoms, and these numbers rose only to 13 percent and 32 percent, respectively, one decade later (MEASURE, 2001, Table C.2). Almost everywhere, condoms are less likely to be used with a regular partner than one who is casual. Given that Africans via concurrency are having more regular, long-lasting partnerships, then it is going to be harder to get Africans to use condoms than elsewhere. It is interesting that Shelton et al. (2005, p. 1058) seemingly try to connect the dots between the importance of concurrency in explaining the

86

Setting priorities for HIV/AIDS interventions

epidemic in Africa and the positive relationship between income and HIV in SSA that we examined in Chapter 13 (except that they refer to the relationship as being between wealth and HIV and not just income and HIV). Wealth, or income, is the key because it is “associated with the mobility, time and resources to maintain concurrent partnerships”.

CONCURRENCY AMONG AFRICAN AMERICANS Adimora et al. (2006) report that national data regarding the five-year prevalence of concurrency among African Americans in the United States are published only for women. Data in the National Survey of Family Growth show that concurrency among black women aged between 15 and 44 was 21 percent as compared with 11 percent among whites (and 8 percent among Hispanics). Most of this disparity could be explained by the low marriage rates among blacks, which in turn can be explained by the fact that African Americans are more often in poverty and they are the only racial group in the United States where the women outnumber the men. The low male–female ratio is due to many factors, such as migration and the higher mortality rate of African American males. But the gender ratio determinant that we will be focusing on is male incarceration, which Phill Wilson of the Black AIDS Institute in Los Angeles claims is the single biggest driver of the heterosexual spread of HIV among African American women (cited by Jon Cohen, 2004). As we now explain, incarceration and concurrency are very much linked together. Phill Wilson sees parallels between incarceration in the United States and migration in South Africa in terms of causing concurrency. In the “prison industry” and the mining industry large groups of men are taken from their families for an extended period of time. Both the men who leave and the women who stay behind often have new sexual partners. The partners in prison may be coerced, but they are new partners nonetheless, and they are also likely to be high-risk partners (MSMs and IDUs). To begin to quantify the extent to which incarceration causes concurrency, Adimora et al. (2004) studied the African American population in the predominately rural counties of North Carolina where both incarceration and sexually acquired HIV were high. Concurrency was measured in terms of having overlapping partnerships related to the last three partners and incarceration was being in prison for more than 24 hours. The concurrency rate for women with no history of incarceration was 31 percent and it rose to 43 percent for women with a history of incarceration. This rise was small relative to the increase for males, where it almost doubled (the

The role of concurrency

87

concurrency rate rose from 43 percent without a history of incarceration to 80 percent with a history of incarceration). Unlike concurrency in SSA, higher income caused concurrency to go down and not go up. Men coming from a household with less than $16 000 had a concurrency rate of 69 percent and this would fall to 45 percent when household income was above $16 000, and the respective numbers for women was a drop from 41 percent to 27 percent. A certain amount of income is necessary to sustain a marriage and avoid concurrency. Married people were much less likely to have concurrent partners than those unmarried and this applied to both men (37 percent concurrency rate if married versus 71 percent unmarried) and women (23 percent concurrency rate if married versus 36 percent unmarried). Note that, just as in the African context, a person’s level of education did not prevent the spread of HIV. There was no statistically significant difference in concurrency rates for African American males and females between those with, and those with less than, high school education. Lagarde et al. (2001) point out that concurrency is just one of the factors that fuel HIV epidemics and that it may not always be present. In their study of five SSA cities, those with the higher HIV prevalence rates (Ndola with 28.4 percent and Kisumu with 25.9 percent) had the same average rates of concurrency as those cities with the lower HIV prevalence rates (Dakar with 1 percent, Cotonou with 3.4 percent and Yaoundé with 5.9 percent). Epstein (2007) criticizes this study because it did not control for differences in circumcision rates among the five cities. Actually, the Lagarde et al. study had no controls whatsoever. So, for example, differences in income may have generated the concurrency rates they observed. But, more generally, their study highlights the importance of being aware of how one measures the time period over which overlapping partnerships take place. In the Lagarde et al. article, the time period was specified as one year prior to being interviewed. As they themselves point out, HIV could have been acquired many years earlier and the “propensity” to have concurrent partners adjusted to having already been infected. The danger then of using a short period, 12-month measure of concurrency rather than the five-year period employed in the National Survey of Family Growth (referred to earlier in this chapter) is that reported concurrency rates could be the consequence of HIV rather than its cause.

20.

Sex and HIV III: the role of networks

We have seen that greater concurrency is one reason why HIV prevalence can vary among countries and among groups. Concurrency is an important component in a network analysis of the spread of HIV. Here we outline other key features of a network analysis that we will use to give a fuller explanation of why it is that African Americans have higher rates of HIV than whites (and Hispanics). We shall see that patterns of population exposure, and not simply individual-level behavior, can determine the spread of HIV.

KEY CONCEPTS OF NETWORK ANALYSIS The University of California San Francisco (2003) have produced a publication (prepared by Dan Wohlfeiler and John Potterat) detailing how sexual networks and partner selection can help to explain why African Americans have higher HIV prevalence rates than other races in the United States. We can explain the central ideas by referring to Figures 20.1 and 20.2 based on their work. The network in Figure 20.1 consists of eight individuals (circles), basically in two groups of four. Six of the individuals have two sex partners. Two persons (individuals 4 and 5) have three partners and they are called “core” members. These two core members are also the “bridge” linking the two sets of four. Consider individual 1 to be the index person who has HIV. In just two steps, individual 1 working through individuals 2 and 3 can cause half of the network to be infected. An effective prevention program targeted at the bridge between individuals 4 and 5 would ensure that half of the network would be infection free. Next consider Figure 20.2, which again has a network with eight individuals, six having two partners and two persons having three partners. The network in Figure 20.2 is configured differently from Figure 20.1 as the core members are now the index person 1 and individual 8 and there are three bridges: between individuals 2 and 5, between individuals 3 and 6, and between individuals 4 and 7. 88

The role of networks

89

2

6

1

4

5

8

3 Source:

Based on University of California San Francisco (2003).

Figure 20.1

Sexual network with a single bridge

1

Source:

7

2

5

3

6

4

7

8

Based on University of California San Francisco (2003).

Figure 20.2

Sexual network with three bridges

Individually then, sexual activity in the two networks is exactly the same. However, in just two steps from the index person 1 in Figure 20.2 there would now be seven persons infected – everyone except individual 8. An effective prevention program in this network would be much harder to achieve as all three bridges would have to be targeted in order to isolate the two groups.

USING NETWORK THEORY TO EXPLAIN HIGHER HIV RATES AMONG AFRICAN AMERICANS We have just seen that it is the existence of core members and those who are bridges that can largely determine the extent to which HIV can spread in populations. We will examine these two network ingredients in turn as they relate to African Americans relative to whites in the United States. Laumann and Youm (1999) define a core member as one who has had at least four sexual partners in the same 12-month period. People with only one partner in the past year are called “peripherals”. They found in a nationally representative survey (the National Health and Social Life Survey) that African American peripherals were five times more likely to have sexual relations with an African American core member than for peripheral whites with white core members. So infections are not restricted to just those in the core and peripherals are more likely to be affected. This

90

Setting priorities for HIV/AIDS interventions

differential matching of core members with non-core members among African Americans is all the more significant given that African American core members on average had more sexual partners than their white core counterparts. Interracial sex was the bridging mechanism that was examined in the Laumann and Youm study. Although being a bridge causes prevalence rates to rise in a national population, not being a bridge to other racial groups by one racial group causes the infection to stay within that group and not leak out to other racial groups. White core members were more likely to partner with Hispanic core members than were African Americans. So white members were more effective at spreading the infection to other races than were African Americans. Within the African American group, there are not only more bridges between the core group and non-core group members, there are also more bridges with high-risk groups. We have already referred to the North Carolina study that highlighted the role that incarceration plays in spreading HIV among the African American community (Adimora et al., 2004 – see Chapter 19). People in prison are clearly a high-risk group as their HIV prevalence is eight to ten times that of the general US population (see Thomas and Sampson, 2005). Adimora et al. (2004, p. S43) also report that African Americans in North Carolina were more likely to partner IDUs and MSMs (almost half of the respondents reported that at least one of their last three partners had smoked crack and 12 percent of women had a partner who they thought had sex with other men). On the basis of all the content in Part II we can summarize the differences and similarities between HIV in SSA and HIV in the United States. The HIV/AIDS epidemics in Africa and the United States are different in the following ways: ● ● ● ● ●

●

Hunger, malnutrition and parasitical diseases are the prime causes of HIV in SSA. Those infected with HIV, and affected by HIV, are much younger in Africa. There are 12 million AIDS orphans in Africa. The majority of those infected in Africa is female, while in the United States the majority infected is male. Males affected in the United States are in special high-risk groups, while those in Africa are typical of the general population. The epidemic is still growing in many African countries and in the United States the epidemic has stabilized (at approximately 40 000 cases per year). The United States has legal safeguards reducing job discrimination and female property rights.

The role of networks

91

The HIV/AIDS epidemics in Africa and the United States are similar in the following ways: ●

●

●

Africans and African Americans have approximately the same number of sex partners. Greater concurrency with sex partners is a problem for both African Americans and those living in SSA. The more vulnerable groups in society are the ones who are affected most. Nearly one in four United States blacks and the majority in SSA are living in poverty. Those persons that are HIV positive are stigmatized. This explains why people don’t get tested even though a quarter of those infected in the United States and the majority in SSA don’t know they are infected.

PART III

Cost–benefit methods and applications

21.

Introduction to Part III

From the analysis covered in Part II we have the requisite background to appreciate the range of considerations that determine the transmission of the HIV/AIDS disease and an understanding of some of the factors that influence the effectiveness of interventions to impact the transmission process. Now is the time to go into detail about how CBA has been used to evaluate the interventions. In this chapter we look at some of the evidence of the effectiveness of possible interventions and explain why effectiveness needs to be put into a broader evaluation framework in order to be useful for setting priorities. Then we give a guide to the rest of Part III. In subsequent chapters we go through the basic principles of CBA and show how these principles can be, and have been, applied. In outlining the main evaluation theories and practice of CBA as they relate to HIV we will just focus on a few main points. All the details can be found by consulting the actual studies themselves.

ESTIMATES OF THE EFFECTIVENESS OF VARIOUS HIV/AIDS INTERVENTIONS In Table 21.1 we present a summary of 151 studies given by Bollinger (2008) related to the average effectiveness of the kinds of HIV preventative interventions planned by the World Bank and UNAIDS. There are four behavioral outcomes monitored for 12 different interventions. The behavioral outcomes are the use of condoms, treatment for STIs, the number of sexual partners and the age of first sex. All outcomes except age of first sex are defined such that a decrease (a negative sign) is to be viewed as an indicator of the extent of desired effectiveness and all outcomes are measured as percentage changes. Each outcome has three possible target groups: high risk (involving MSMs and IDUs), medium risk (people with more than one partner in the previous year) and low risk (people with only one partner in the previous year). When numbers do not appear in the table it is either because only one (or two) of the risk groups is targeted by a particular intervention or because there are no studies giving reliable results. In terms of the outcome of increasing condom use (that is, decreasing

95

96

–17.1

–18.5

–20.6

–16.6

–15.7

–11.6 –23.4 –10.0

Medium risk

–39.0

–42.5

–44.2

High risk

–7.5

–1.0

–17.0 –16.1 –2.5

Low risk

Reduction in Non-use of Condoms High risk

–17.7

Medium risk

Low risk

Reduction in STI Non-treatment

–93.0

–11.2

High risk

0.0

–35.5

–18.3

–4.2 –13.3

Medium risk 0.0

Low risk

Reduction in Number of Partners

Impact of preventative interventions on behaviors, percentage changes

Mass media VCT Community mobilization Sex worker outreach School-based programs Programs for out-of-school youth Workplace programs Condom social marketing

Intervention

Table 21.1

High risk

0.11

0.08

–0.3

0.0

Medium risk

Low risk

Increase in Age of First Sex

97

Source:

–17.0

–22.0 –36.3

–85.1

–1.9 –37.5

–55.1

30.2

–5.3

Based on Table 3 of Bollinger (2008).

Public sector condom distribution IDU outreach MSM outreach STI treatment Other peer evaluation –63.0

–30.6

–11.0

–1.2 –10.1 –52.0

–13.0

–10.0

98

Setting priorities for HIV/AIDS interventions

condom non-use) public sector condom distribution is, on average, the most effective program for high- and medium-risk individuals. For lowrisk individuals, workplace programs are the least effective. Workplace programs are, however, the most effective for high-risk individuals if decreasing the number of sex partners is the desired outcome. Not surprisingly, the most effective way of ensuring that people get treated for STIs is to actually provide an STI treatment program directly, rather than have recommendations or information for STI treatment as part of other programs (such as in schools). When the aim is to increase the age of first sex, effectiveness definitely varies by intervention as some are effective, some non-effective and some even counterproductive.

THE ROLE OF ESTABLISHING EFFECTIVENESS IN THE EVALUATION OF AN INTERVENTION There are four main ingredients that make up an evaluation in the form of a CBA. They are the inputs, the outputs, the values for the inputs and the values for the outputs. The inputs and their associated values make up the costs, and the outputs and their associated values determine the benefits. The benefits and costs are linked through the relationship between the inputs and the outputs (called the “production function” by economists). If the inputs do not generate positive output levels, that is, the intervention is not “effective”, then it does not matter what the values of the inputs and outputs are from the point of view of CBA. Without positive outputs there cannot be positive benefits (no matter how high the values for the outputs may be); and without positive benefits, no costs are worthwhile (no matter how low the values for the inputs may be). So establishing effectiveness is the starting point for undertaking a CBA. Unfortunately many economists also stop their analysis of interventions with an estimate of effectiveness. Because they are ignoring the two ingredients entailed in valuing the inputs and outputs, what many economists do by establishing effectiveness through an estimate of the production function is supply a necessary condition for determining the worthwhile intervention; their work is not a sufficient condition for a CBA. Two ingredients of a CBA are useful, but one needs all four ingredients to carry out an evaluation determining whether an intervention is worthwhile. The summary of 151 studies of effectiveness shown in Table 21.1 provides empirical support for three points that we wish to highlight about CBA and its role in making intervention evaluations. First, even as late as 2008, there are interventions that have never had a CBA. We know this simply because there has never been an effectiveness study for them

Introduction to Part III

99

and, as we have just said, an effectiveness study is a necessary condition for a CBA. So we just do not know whether certain interventions are worthwhile. For example, we know that VCT (voluntary counseling and testing) increases condom use (reduces condom non-use) for all risk groups, but we do not know whether it is effective in increasing the age of first sex. Second, following on from the first point, what happens when, as with mass media, an intervention is effective for one or more of the four effectiveness categories, but not all of them? An effectiveness study has to choose which one of the categories is more important, while a CBA would have a benefit measure that depends, either directly or indirectly, on all four of the outcome categories. Third, whether an intervention is desirable, and also the extent to which it is desirable, depends on the details of the intervention, that is, the type of intervention and the risk group it is targeted at. Thus, VCT is not effective, and hence not worthwhile, from the point of view of studies aimed at reducing the number of sex partners for low-risk groups; while this same intervention could possibly be worthwhile for medium-risk groups as it is effective for them.

OUTLINE OF PART III For all of the interventions that we will be covering in Part III, we make sure that the issue of effectiveness is examined and, for at least one of the chapters, showing how to estimate effectiveness will be the central purpose of the chapter. But, as has been pointed out earlier, effectiveness deals only with two of the four CBA ingredients. A different approach to evaluation is called “cost-effectiveness analysis” (CEA) and this relies on three of the CBA ingredients. The price of inputs is applied to the inputs to form the costs and this is then compared with the objective of effectiveness, which can generally be thought of as the output (or physical effect). CEA recommends choosing the intervention that produces the output/effect at lowest cost. Many of the chapters in Part III include a discussion of what to include as costs and how to measure them. When we focus just on costs, this will lead to an evaluation method called “cost minimization”, and here the outcome is the output of the intervention. When the outcome measure is regarded as an effect, and we also include a comparison of costs, this is the realm of CEA proper. Four chapters will be devoted to explaining cost minimization and CEA. The rest of Part III will tackle the fourth and last ingredient of CBA, that is, how to value the output to form the benefits. CBA involves comparing the benefits with the costs. Everyone involved in studying HIV/ AIDS is inevitably aware of the costs, even if they just pertain directly to

100

Setting priorities for HIV/AIDS interventions

the costs of running a program. Supplies, drugs, salaries, equipment and so on, have to be budgeted for, and paid for, if any kind of program is to take place. The question is: how much value does one get from the costs that one incurs? So Part III concentrates more on the benefits side, that is, how one can put a monetary value on the output that generates the costs. To this end we will present a different benefit methodology for every application we will be analyzing, except for the first benefit method, which is the threshold approach. This is the most general of estimation methods and so we give three examples that cover effectiveness, as well as valuing outputs and inputs. The other benefit estimation methods include the willingness to pay (WTP) approach, risk compensating wage differentials, lifetime earnings and the revealed preferences of public decision-makers. The applications cover specific and general education, condoms, VCT, ARVs, treating HIV patients with drugs for TB and valuing an actual and a “statistical” life.

22.

Threshold analysis theory

We begin our review of the various methods for estimating benefits and costs by looking at the most general approach. This is appropriate when we have the least idea as to how to proceed with estimation. The clearest case when this method is most useful is when the intervention has just taken place. The time when the effects of the intervention will occur and can be observed is going to be in the future and not now when one may want to make the evaluation. In this situation, the best one can do is place limits on what the unknown value can or cannot be and then try to assess whether those limits are likely to be too high or too low given some other information that we are not unsure about. These limits are called the threshold values. They are also called “switching values” as they determine whether the intervention will switch from being judged successful to being declared unsuccessful. The general method is outlined in this chapter and there are applications in the next three chapters.

THE BASICS OF THE THRESHOLD METHOD At the threshold, benefits B equal costs C. This is because if benefits were slightly higher, or costs slightly lower, net benefits would be positive and we would know that the intervention would be worthwhile. Similarly, if benefits were slightly lower, or costs slightly higher, net benefits would be negative and we would know that the intervention would not be worthwhile. So, at the threshold one is unsure whether the intervention should be accepted or rejected. Any small change in the estimates of benefits or costs would tip the balance, that is, make the intervention pass over or fall under the threshold. To see how the method works, let us divide and multiply the total benefits B by the number of units affected E (effects) to obtain: B = [B/E] × [E]. In this case, equating benefits with costs, that is, setting B = C produces the fundamental threshold relation that has three ingredients: [B/E] × [E] = C

101

102

Setting priorities for HIV/AIDS interventions

If the effect is the number of HIV persons (cases) averted, then the fundamental relation tells us that the benefits per person multiplied by the number of persons benefiting should equal the costs. Say the benefits per person are $200, there are five beneficiaries, and the costs are $1000, then we would have: [$200] × [5] = $1000 What this means is that if we had full information for a particular intervention, such that we could estimate all three ingredients of a CBA accurately, and we happened to find that the benefits per person are $200, there are five beneficiaries and the costs are $1000, then we would have the result that B = $1000 and C = $1000 and so net benefit would be equal to zero. We would then be indifferent about accepting or rejecting the project. So far all of this is pretty obvious. What makes this interesting is when we are missing something and have information related to only two of the ingredients in the fundamental relation. In this case we can use the two pieces of information that we do know to place limits on what the possible value for the third ingredient can be. There are going to be three possibilities. We will deal with each one in turn. First assume that we cannot estimate the costs, but we can estimate the benefits per person as $200 and that there will be five beneficiaries. Then we can calculate the left-hand side of the fundamental relation B/E × E. That is, total benefits B are $200 × 5 = $1000. The fundamental relation tells us that costs need to be $1000 in order to have zero net benefits. Second, take the case where we cannot estimate the benefits per person in an evaluation. Again we have costs of $1000 and five beneficiaries. Using the fundamental relation we have B/E × 5 = $1000. So we can deduce that the threshold value for B/E is $200. Finally, let us be uncertain about effectiveness E so that we do not know how many beneficiaries there will be from an intervention. With costs of $1000 and benefits per person equal to $200, the fundamental relation becomes: $200 × E = $1000 and we deduce that the threshold value for E is 5 for this evaluation. To summarize the threshold method: by a process of arithmetic (multiplication or division) from the fundamental relation one takes what can be estimated and deduces what we are unable to estimate. This unknown then is the switching value for the evaluation to either just pass or just fail the cost–benefit test.

Threshold analysis theory

103

THE STRENGTHS AND WEAKNESSES OF THE THRESHOLD METHOD Some benefits and costs are unknowable as they relate to the future. One advantage of the threshold method is that one can use it when data is not available or even too costly to collect. Sexual behavior is a very private activity and people are uncomfortable telling others about it. Just as important an advantage of the threshold method is that one can apply it when the project evaluator has difficulty in fixing a value parameter, such as how much to value a dollar today against a dollar to someone in the future. This is a judgment that has to be made when one uses a discount rate in a CBA. If a dollar today is worth the same as a dollar next year, then the discount rate would be zero. A positive discount rate means that a dollar today is worth more than a dollar next year, that is, next year’s value is discounted to make it equivalent to today’s dollars. Standard CBA practice in the health care field is to discount a dollar in the future at the rate of 3 percent. Previously, 5 percent was the recommended discount rate. If the evaluator is unsure about whether to use 3 percent or 5 percent then he or she might want to treat the discount rate as an unknown and see what rate will transform a positive into a negative outcome for a particular intervention. The key issue then for threshold analysis is what to do with the threshold value once one has estimated it. The hope is that the threshold value that one uncovers is obviously too high or too low. Say one is valuing a life saved by an HIV intervention and comes up with a switching value of $100. There is hardly anyone on the planet who would not think a life would be worth saving if it could be achieved for $100. On the other hand, say the threshold value for a particular intervention is $1 trillion. Then, even though you might be tempted to say that a life is priceless and hence must be worth $1 trillion, on reflection this valuation would mean that only about 14 persons could be saved from HIV if the total annual income of the United States was devoted to this particular intervention. Thus, $1 trillion must be too high a value for a life, as a program saving 14 lives would mean that 270 million or so Americans would have no resources to live on for the whole of 2009 in order to finance the HIV intervention. The problem then arises if the threshold value obtained is not obviously reasonable or unreasonable then what does one do about the unknown value? In general, one would then have to devote more time and resources to finding how reasonable the threshold value is. The process will be different in each case. In the following three chapters we will give three examples of how the process operates. Here we will just illustrate the process for the discount rate issue. The literature suggests 3 percent or 5 percent, which

104

Setting priorities for HIV/AIDS interventions

really means that only a threshold value in the range 3 to 5 percent would complicate matters. So a threshold value of zero would mean that the intervention should be rejected, as any positive discount rate (including 3 to 5 percent) would lead to negative net benefits; while a threshold rate of 6 percent would mean that the intervention should be approved as any lower value for the discount rate (such as 3 to 5 percent) would generate positive net benefits. For a threshold value that is in the range 3 to 5 percent, the evaluator has to actually decide explicitly whether the old or the new literature on the discount rate is the more convincing. The second main problem with threshold analysis is that it can only be used once per evaluation. If there is more than one ingredient that one wants to treat as an unknown, then the approach cannot be used. In terms of the fundamental relation it would mean that one of the three (benefits per person, the number of beneficiaries or the size of the costs) could be estimated. Moreover, even if the method were used for just one category, the approach is unlikely to be useful if applied to more than one intervention. This is because one might be lucky that the threshold value obtained will be obviously wrong or right for one particular intervention. But, the more it is used, the more likely it will be that a threshold estimate will be derived for which it is not clear how reasonable it is. For example, if the switching value for the discount rate is employed for a whole series of projects, there is bound to come a time when a value in the range 3 to 5 percent will appear.

23.

Threshold analysis practice: the effectiveness of HIV education

As we stressed previously, to do a CBA there must be an effect from the intervention that we are trying to evaluate. If there is no effect, one need not do an evaluation because we can avoid incurring any cost at all simply by not undertaking the intervention. However, as we also saw in Parts I and II, effectiveness cannot be taken for granted as we kept coming up with counterintuitive results. So, the first step in carrying out a CBA is to quantify effectiveness. In this chapter we use threshold analysis to illustrate one way of measuring effectiveness. This will be applied to an HIV education intervention in the United States as described by Norton et al. (1998). In principle, an HIV education program is a multi-option intervention. Once one is told all the ways that HIV can be transmitted (for example, mother-to-child transmission, unprotected sex, blood transfusions and sharing needles when taking drugs), the recipient of the information can act on it and stop all the ways that HIV is transmitted. In practice, only one or two types of behavior can be expected to be altered, leading to a reduction in one type of transmission (for example, using condoms when having sex). It all depends on the audience to whom the education program is being addressed. The education program we will be examining was geared to injecting drug users and crack cocaine users. There are two main transmission mechanisms involved with these two groups – the sharing of injecting equipment and through unprotected sex.

THE INTERVENTION The North Carolina Cooperative Agreement Program (NC CoOP) was one of 23 cooperative sites funded by the National Institute on Drug Abuse (NIDA). Its aim was to decrease the spread of HIV infection by implementing community-based outreach and intervention. It targeted hidden populations of IDUs and crack users not in treatment in order to recruit them for extensive risk reduction. Once recruited, and there were 347 in the sample, the clients were then interviewed and data were 105

106

Setting priorities for HIV/AIDS interventions

collected on their demographics and HIV risk behavior. Client behavior was recorded at intake and then at follow up. In between the two periods the intervention took place, which involved two sessions testing the clients’ knowledge about HIV transmission mechanisms and prevention practices and then showing them cue cards that had basic HIV information on them. Information given emphasized behavior change. The clients were given the opportunity to practice cleaning needles and using a condom on a dildo. All persons contacted, whether they participated in the intervention or not, were given prevention packets containing male condoms and risk-reduction kits for injectors, together with information as how to use both items in the packets.

THE THRESHOLD ESTIMATE OF EFFECTS The rationale given by Norton et al. for using threshold analysis in their study was that they were evaluating the intervention very early on, well in advance of knowing accurately the full information about effects and benefits. Costs were known as they mostly occurred during the intervention period. Around 85 percent of the cost involved labor, consisting of two outreach workers and an individual trained to draw blood (phlebotomist). Blood work was involved because all clients were offered a free HIV test and a discussion of test results (if requested) was a part of the follow up sessions that focused on behavior change. The non-labor costs involved utilities, supplies, rent and laboratory equipment. Research costs were excluded. In total, the costs came out to be $94 791 per year. The benefits per person had two components. The first was the benefits from avoiding a case of HIV, which was $443 214, being an updated figure from the estimate taken from Holtgrave and Qualls (1995). As we shall see in the next chapter, the benefit of avoiding a case of HIV consists of the saved individual’s satisfaction from the added life expectancy that is generated. The second component of benefits was the direct medical costs that were saved if the person did not get infected by HIV and this amounted to $59 584 per year. Total benefits were therefore $502 798 per year. In terms of the fundamental relation presented in Chapter 22, that is, benefits per case averted (B/E) × the number of cases averted per year (the effect, E) equals the cost per year of the program (C), or: [B/E] × [E] = C Norton et al. knew B/E and C, so E was the unknown. This makes sense because the intervention depended on behavior change and it was always

The effectiveness of HIV education

BOX 23.1

107

THRESHOLD RATIO FOR BALANCING COSTS AND BENEFITS OF AN EDUCATION INTERVENTION

Costs Cost per year Benefits Total per beneficiary per year

$94 791 $502 798

Cost/total benefits per beneficiary $94 791/$502 798 = 0.189 HIV infections avoided per year to break even Source:

Based on Norton et al. (1998).

hard in HIV programs to get people to alter their behavior. Effectiveness could not be guaranteed. The unknown E could be obtained from the two known elements in the fundamental relation by dividing both sides by B/E to obtain: [E] = [C] ÷ [B/E] This expression for E was used to derive the threshold value presented in Box 23.1. On average there would need to be nearly one-fifth of an HIV case avoided for the program effectiveness to justify the costs.

THE USE OF THE THRESHOLD VALUE IN THE EDUCATION EVALUATION If more than one-fifth of an HIV case was avoided then the education program would be worthwhile; while if less than one-fifth of an HIV case was avoided the education program should be rejected. So what effectiveness outcome was the more likely? Norton et al. did not know precisely the extent to which the future transmission rate would drop. But for there to be future HIV case reductions, the clients would have to be reducing the number of partners and increasing their use of condoms when having sex, and decreasing their use of drugs. Norton et al. found strong evidence for all these types of behavior change. In their follow up survey of the 347 clients, the vast majority of whom returned, it was found that:

108 ● ● ● ●

Setting priorities for HIV/AIDS interventions

43 percent decreased the number of encounters with another sexual partner; 31 percent increased condom use; nearly two-thirds of IDUs decreased their use of injecting drugs; and nearly three-quarters of non-injectors decreased their drug use.

Given this risk-reduction behavior change, reaching the effectiveness threshold level was thought achievable. As Norton et al. (1998, p. 205) said: The behavior change data “imply that reaching the threshold is a plausible goal for the NC CoOp program”.

24.

Threshold analysis practice: the benefits of avoiding HIV

As we know, there are many ways to prevent the spread of HIV. It would be useful if we could obtain a broadly applicable, general estimate of the benefits of avoiding an HIV infection, which then can be compared with the costs of a particular intervention. Holtgrave and Qualls (1995) came up with such an estimate, which they argued would help overcome the tendency in the United States in the mid-1990s to use CBA primarily for treatment rather than prevention programs. According to them, the reluctance to use CBA for prevention stemmed from the extra challenges posed by aiming to change sexual behavior that are not present with treatment evaluations. Thus, prevention CBAs involve the thorny issue of trying to measure program-related changes in sexual behavior when people do not like to reveal information about their behavior in this area. As we shall see, the Holtgrave and Qualls study has strong parallels with the method to determine effectiveness presented in the last chapter. We will, however, emphasize more the benefit estimation part of the work.

THE THRESHOLD ESTIMATE OF BENEFITS Recall that the fundamental relation for the threshold method involves finding values for elements where benefits (B) equal costs (C): [B/E] × [E] = C So it is clear that the left-hand side of the relation determines the benefits. In which case it follows that a threshold estimate of the benefits can be determined as: B = [B/E] × [E] In this relation, E is the effect. This can be defined in many different ways. In the last chapter we defined E as the number of persons who were not going to contact HIV due to an intervention. Now we are going to 109

110

Setting priorities for HIV/AIDS interventions

define it in terms of a health intervention outcome, albeit a result that is a per-person outcome. In Chapter 12, when dealing with the effects of malnutrition, we first introduced the idea of a health outcome that merged the quantity (mortality) and quality (morbidity) of life called a disability adjusted life year (DALY). The example we referred to there was one where, if two years in a wheelchair is worth one year with full health, then in the case of a disease that causes someone in a wheelchair to die who otherwise would be healthy, and loses 30 years of life expectancy, the loss would be recorded as 15 DALYs. Now we look in detail how the DALY measure can be fixed for someone getting HIV/AIDS as it is a DALY that will be used to estimate E in the threshold relation for the subsequent application. Note that the utility of being healthy is on a scale of 1 for healthy and 0 for being dead. So the utility of living with a disease is somewhere in between 0 and 1; while the scale for the disability (disutility) effects of a disease are 1 minus the utility of being healthy. This means that the disutility (disability weight) of each year being dead is 1 and the disutility of each year being healthy is 0.

ESTIMATING E AS A DALY FOR SOMEONE INFECTED WITH HIV/AIDS Holtgrave and Qualls assume that a person who has HIV/AIDS would otherwise have a life expectancy of 65 years in 1993. This is lower than the life expectancy for people in the general population because, for example, injecting drug users usually do not live as long as most people. The first 25 years are infection free. At the end of the 25th year, the person gets the HIV virus. For six years, from ages 26 to 31, the person carries on as normal and he or she is supposed to be unaware of the presence of HIV. A normal person’s life year is valued at 1, so the disutility of being a normal (healthy) person is 0. The person gets tested positive at the end of the 31st year and now becomes aware of the HIV infection. Each of the three years in this state, from ages 32 to 34, are valued at 0.9 years of a healthy life year, presumably not 1 because of the anxiety and stigma of knowing that one is HIV positive. The disutility that the HIV causes is 0.1 life years. The person is in this state for three years. Then the HIV virus turns into fullblown AIDS at the end of the 34th year and this lasts three years. During this period, from ages 35 to 37, a whole host of opportunistic infectious diseases appear. There are 26 possible conditions in all, such as skin infections and TB. In the first year of the AIDS period, at age 35, the infectious diseases are not too severe and the year is weighted equal to 0.65 of a healthy year, which makes the disutility 0.35 of a healthy year. For the

The benefits of avoiding HIV

111

last two years with AIDS, from ages 36 to 37, infectious diseases are severe and each year in this state is only valued at 0.4 of a healthy year. So for these years the disability weight is 0.6. At age 38 the person dies. Each year of death is, quite understandably, given a utility weight of zero. So all 27 years from years 38 to 64, that is, after the person’s death and up to his/ her life expectancy of 65 years when the person was otherwise expected to die, generate no healthy years and all 27 years are counted as 27 disutility years. Let us ask the question: if the person avoided getting infected by HIV, how many years of equivalent healthy years would be saved and not lost? Lost years are measured in DALYs and it is in these units that one saves years. The answer to the question is 28.85 DALYs. This total is made up of four parts: 1.

2. 3.

4.

In each year between ages 32 to 34, when the person is aware of HIV, the person gets 0.9 of a healthy year. So 0.1 of a year is lost each year due to the disability of having HIV, or 0.3 DALYs are lost over the three-year period and these are now saved. At age 35, in the first year with AIDS, the person gets 0.65 of a healthy year. So 0.35 DALYs are lost due to the disability of having HIV. In both of the latter years with AIDS, between ages 36 and 37, a person gets 0.4 of a healthy year. So 0.6 of a year is lost each year, or 1.2 DALYs for the two years. For the 27 years between ages 38 and 64 when the person is dead from AIDS, no healthy years are experienced. Not getting HIV means saving these 27 DALYs.

Thus: 28.85 DALYs = 0.3 DALYs + 0.35 DALYs + 1.2 DALYs + 27 DALYs. The 28.85 DALY total ignores the process of discounting. In CBA, a dollar’s worth of benefits next year are worth less than a dollar’s worth of benefits this year, if for no other reason than because one could put the dollar that one has today in the bank and earn interest. So if the interest rate is 5 percent, then a dollar today would be worth $1.05 next year. Alternatively, one could say that $1 next year was equal to $1/(1.05) or 95 cents in today’s value terms. It is the division of next year’s dollar by 1.05 that is called discounting and it converts future values into today value (present value) units. The DALYs that were calculated above are going to be expressed in dollar terms to form the benefits. So these need to be discounted too. To keep matters simple we have assumed that the bank interest rate was 5 percent because this was exactly the same rate that Holtgrave and Qualls used as their discount rate.

112

Setting priorities for HIV/AIDS interventions

As we have seen, a DALY next year is worth 0.95 DALYs today. A DALY in two years’ time would have to be divided by 1.05 times 1.05 to convert it into a present value DALY. In which case the DALY two years into the future would be worth 0.91 DALYs today. For each additional future year, discounting involves dividing by 1.05. In the 28th year, one divides 1.05 exactly 27 times to get a present value of 0.268 DALYs. The end result is that the 28.85 undiscounted DALYs when converted into present value terms at a 5 percent discount rate amounted to 9.26 DALYs. It is that number that will be used to correspond to E in the threshold relation.

CALCULATING THE BENEFITS AND APPLYING IT TO CDC’S PAST BUDGET ALLOCATIONS Holtgrave and Qualls use an estimate of $45 000 for B/E taken from the literature, that is, Owens et al. (1993). Multiplying B/E by the E estimate of 9.26 forms the threshold value for benefits equal to $416 700 in 1993 US dollars. With this sum as the benchmark, Holtgrave and Qualls turned to the US government’s Centers for Disease Control and Prevention (CDC) budget for 1993 of $356.6 million to see how many HIV infections it should have averted. The budget supported both a diverse set of prevention programs (including community-level involvement, information dissemination, prevention case management, school-based HIV-infection education, street outreach and counseling, testing and partner notification) and surveillance activities. Using the logic of the threshold equation, the authors divided $356.6 million by $416 700 to come up with the figure of approximately 860 HIV infections that should have been averted. Given the threshold nature of the benefits (they are such that B exactly equals C, and does not exceed it), one can say that at least 860 HIV cases should have been averted for the sum of $356.6 million that it spent. Although Holtgrave and Qualls did not attempt to estimate the actual number of cases averted by the 1993 prevention program, they did point to an earlier analysis of the CDC’s budget related to VCT services, which cost $102 million and averted 7000 HIV infections in 1990 alone, which was many times the 860 figure for less than a third of the 1993 prevention budget. So there is evidence that CDC-funded HIV-infection prevention programs did generate effectiveness results well in excess of the benchmark levels required.

25.

Threshold analysis practice: the costs of a possible HIV/AIDS vaccine

The surest way of ending HIV/AIDS is for a vaccine to exist and be widely distributed. However, as Whiteside (2008) points out, scientific advances have been slow in this area as global spending on an AIDS vaccine is less than 1 percent of research and development spending on health products. One major reason for this lack of resources for research and development for an AIDS vaccine is that there are doubts that uninfected persons in developing countries would be able to afford to pay for it when it does get created. Even if governments want to get involved and subsidize the vaccines that are invented and produced, they also need to know how much private individuals would be willing to pay for them, as the government can then ascertain how much of a subsidy would be required to fill the gap between what people can afford and what the vaccine costs. The WTP for a vaccine can be used in the context of threshold analysis because evaluating a vaccine is a prime example of undertaking a CBA where one is missing a vital piece of information that is, in the current state of knowledge, unknowable. One cannot begin to estimate the cost of something that does not yet exist. However, one can ask what the threshold level for costs must be to ensure that the benefits that can be estimated are not exceeded. This is the question that we will attempt to answer in this chapter for one country at least – Mexico. Mexico was chosen by Whittington et al. (2002) because it is a middle-income developing country where HIV is threatening to become a public health hazard, but as yet there are relatively few who are actually infected with the virus. Therefore, there could potentially be a large number of unaffected individuals who would be willing to pay to avoid getting the disease.

113

114

Setting priorities for HIV/AIDS interventions

THE CONTINGENCY VALUATION METHOD FOR MEASURING BENEFITS Contingency valuation (CV) is a technique for benefit estimation that involves asking people in a questionnaire for their valuations for something, usually a policy change. What makes the elicited valuation contingent is the conditional nature of the context in which the valuations are to relate. For example, one could ask, if this type of change were to occur, and if these were the circumstances in which the change were to occur, how much would you be willing to pay for the change to occur (or not occur, if the change were not beneficial)? CV is an obvious method to use to estimate the benefits of a vaccine because one would be asking what one is willing to pay for a hypothetical vaccine. The specified circumstance for the valuation that we will be analyzing is one where the vaccine is assumed to be 100 percent effective.

THE THRESHOLD ESTIMATE OF COSTS Let us use the threshold relation one last time, but this time we shall reverse it and instead of setting B = C we will use C = B: C = [B/E] × [E] If we divide both sides of the relation by E we obtain: [C/E] = [B/E] In this relation, E is the number of persons getting the vaccine. Since there need be only one vaccine per person, the number of persons E is also the number of vaccines produced. So C/E in the above relation is the cost per unit (or average cost). The relation tells us that the threshold value for the cost per unit is equal to the benefit per person. Whittington et al. estimate the benefit per person from a survey they undertook for this purpose.

ESTIMATING THE WILLINGNESS TO PAY FOR AN HIV VACCINE IN MEXICO Whittington et al. asked 234 adults (aged 18–60) in Guadalajara, Mexico two questions regarding individual monetary evaluations of a hypothetical HIV vaccine that would be available as an injection or as oral drops.

Costs of a possible HIV/AIDS vaccine

Table 25.1

WTP for an HIV/AIDS vaccine in Mexico according to income Minimum WTP Estimate (pesos)

Full sample Household income (pesos per month) <4500 4500–6400 6500–12 400 >12 400 Source:

115

Mid-point WTP Estimate (pesos)

6358

9858

2533 2916 6681 11 969

5257 7194 11 817 14 163

Based on Whittington et al. (2002) Table 2.

The vaccine would not have any side-effects, would last a lifetime and with it there would be no chance of contracting HIV. But, the vaccine would be in limited supply so everyone who wanted a vaccine would have to pay a common, fixed price. The first question was a valuation that asked for an individual’s minimum willingness to pay (WTP) for the vaccine. The second question was a valuation that asked for an individual’s maximum WTP for the vaccine. The results reported in Table 25.1 present the results for the minimum WTP estimates and the mid-points of the difference between the maximum and minimum estimates (to reflect what is typical within the range of uncertainty over the two sets of estimates). The results are sample means given for the sample as a whole and by four household income groups. The table shows that there was a large range of uncertainty over the WTP estimates, as on average there was typically a 9858 pesos (US$1038) difference between the sums reported as the minimum and maximum amounts. So Whittington et al. used the minimum amount of 6358 pesos (US$669) as a conservative summary estimate of what people were WTP for the vaccine. Because this amount was about 8 percent of mean annual income (roughly equal to one month’s income), the conclusion was that there was likely to be a large private demand for a vaccine in Mexico. The extent of the market would depend on the cost of the vaccine. If the cost of the vaccine was 100 pesos (US$10.87), 90 percent of the sample would be willing to pay for the vaccine; while if the cost of the vaccine was 6000 pesos (US$652), less than 25 percent of the sample would be willing to pay for the vaccine. Using threshold analysis we can therefore say that

116

Setting priorities for HIV/AIDS interventions

if a vaccine could be invented that cost 100 pesos, most Mexicans would be willing to pay for it. At a cost of 6000 pesos, most Mexicans would purchase it only if there were a large subsidy. Table 25.1 makes clear that the WTP for a vaccine is very much a function of a household’s income. The poor (with an income less than 4500 pesos per month) would require a 60 percent subsidy if the mean cost of 6358 pesos were charged. The rich (with an income of higher than 12 400 pesos per month) would demand the vaccine even if came onto the market at a cost of 10 428 pesos. This positive relationship between WTP and income is a general one, applicable to all countries, and its implications will be highlighted in the next two chapters.

26.

Willingness to pay theory

The threshold approach that we have just covered is an appropriate method when we have little information available on which to base our evaluation of an intervention. The WTP approach is particularly useful when we have access to the most information. It is the method that is considered to be best practice in CBA because it is the approach that fits in best with standard microeconomic principles. Values in the presence of competitive markets are determined by demand and supply, and demand is based on willingness to pay. The mechanics of using marginal benefit and marginal cost curves to make cost–benefit evaluations were illustrated in Chapters 7 and 8. Here we explain in greater detail what lies behind the construction of the MB curve. For simplicity we will stick with the simple case where costs per unit are constant.

COMPETITIVE MARKETS AND CBA Competitive markets allocate resources according to demand and supply. That is, at the point where demand and supply intersect the equilibrium price and quantity are determined. Demand reflects MB and supply MC. In this way the market functions just like a cost–benefit mechanism. To see this, look at the market for condoms that is shown in Figure 26.1 (based on Brent, 2009c). The market relates to a condom marketing program in Tanzania, so the prices are expressed in Tanzanian shillings (TZSH). The quantity (number of packs of condoms) that the market would authorize would be 12 at a price of 290 TZSH. This outcome corresponds to point c where demand meets supply. Note that this is also the point where MB = MC and hence this would be the outcome that would obtain if a CBA test were applied to the condom market. At a quantity less than 12, MB is greater than MC, and this implies that more should be produced; and at a quantity greater than 12, MC is greater than MB and so less should be produced. Only where quantity is 12 is quantity right according to a CBA test, as there would be no net gain from differing from this quantity.

117

118

Setting priorities for HIV/AIDS interventions

Price (TZSH) 2000

a

b

290

c

e

100 0

Source:

d

2

12 Packs of condoms

Marginal costs = supply Marginal benefits = demand

39

Based on Brent (2009c).

Figure 26.1

The market for condoms in Tanzania

THE LINK BETWEEN DEMAND, MB AND WTP Why does the demand curve reflect MB and how does the demand curve reflect willingness to pay? The answers lie in the basic microeconomics principles related to consumer utility maximization. Individuals get satisfaction, or utility U, from consuming goods that they buy, of which a condom is an example. Marginal utility MU is the extra utility that is obtained by consuming one more, or one less unit of the good. To express units of utility into units of money, which is by definition required to obtain marginal benefits MB, that is, benefits are outcomes expressed in monetary terms, one must divide the marginal utility of the pack of condoms, MU(C), by the marginal utility of income, MU(Y). Thus, we have the first equality: MB = MU(C)/MU(Y) From the individual consumer’s perspective the cost of the good is the price P. This is already in monetary terms. The extra satisfaction in monetary terms is, as we have just seen, MU(C)/MU(Y). To maximize total

Willingness to pay theory

119

satisfaction for a given income Y, the individual has to equate the extra cost and the extra gain, and hence we have a second equality: P = MU(C)/MU(Y) From the two equalities we can deduce that: MB = P So we have the result that the MB curve is given by the price curve. But, the price curve, that is, the relation between quantity demanded and price, is by definition the “demand curve”. From this we obtain the result that the MB curve and the price curve are one and the same thing. For each unit of quantity, we interpret MB to be the price that an individual is willing to pay for it. This then establishes the identification of benefits with WTP. The final piece of economic theory that we will refer to is a principle that for historical reasons is called a law and this is the “law of diminishing marginal utility”. This states that the more that one consumes of any good, the less is the additional satisfaction. This means that, typically, MU will decline as quantity expands. If we now assume that for a particular individual the marginal utility of income is constant, then MU(C)/MU(Y) will decline as MU declines. In other words, the demand curve (and hence the MB curve) slopes downward as output increases from left to right on the quantity axis. This is a very important result for CBA as it explains and justifies the case for repeatedly undertaking economic evaluations that is made throughout this book. It is not enough just to show that, for a specified quantity, the benefits are greater than the costs in order to justify scaling up a particular HIV/AIDS intervention. Given that the MB will decline with expansion, the net benefits cannot be relied upon to be the same. As we saw in Chapter 8, the MC could also fall from left to right. But, the point still holds. Net benefits are likely to vary with the output level.

THE STRENGTHS AND WEAKNESSES OF THE WTP METHOD The major advantage of the WTP approach is that it fits in well with the idea that people’s valuations are reflected by their market behavior. How can we be sure that someone really values something? The answer is, if they are willing to pay for it. So if the price of a hamburger is $4 and the person actually buys a hamburger at this price, then we can be sure that

120

Setting priorities for HIV/AIDS interventions

the individual values the hamburger at least $4. The valuation could not be only $3 or the person, if rational, would not have paid $4 for the hamburger. The valuation could have been $6, or even more than this. But, it cannot be less than $4. This reasoning also applies for condoms. If a Tanzanian pays the market price of 290 TZSH per pack (of three), then this is the behavior-revealed MB amount. A less obvious advantage of the WTP approach, one that has enormous practical use (and will be examined further in the next chapter), is that a purchased price is a total valuation. Instead of asking how much an individual values avoiding getting pregnant, and then asking the same person how much he or she values avoiding getting HIV, one has the answer to both questions in a single source, that is, in the price that was paid for a condom. The price paid reflects the use value whatever the use, and for all uses, even if some of the uses involve using condoms for non-sexual activity purposes (for example, water storage). In the health care field, in which HIV/AIDS interventions can be located, there is often an absence of formal markets for health care products. This is especially the case outside of the United States where health goods and services are usually provided by governments free of user charges. Here, “no price” does not translate to “no value”. In the absence of markets, one can use two alternate routes to extract WTP estimates. One can look to valuations in parallel markets. For example, one can use the market price for skin creams to value the benefits of avoiding skin infections from AIDS. Or, as we saw in the last chapter when we examined the demand for a vaccine that does not yet exist, one can undertake a questionnaire that asks people hypothetical questions for their contingent valuations of their WTP as if a market did actually exist for the product or intervention that one is trying to evaluate. The absence of markets is one reason why WTP estimates are rarely found in the health care evaluation field. But, a more fundamental reason is that willingness to pay is questioned on equity grounds. Some health care economists, and nearly all non-economists, think that one should not use WTP even when markets do exist because willingness to pay depends on ability to pay, and many people do not have the ability to pay for goods and services even if they are really needed. We take this criticism very seriously and much of Part IV is devoted to examining the role of equity in CBA. It turns out (as we shall see in Chapter 40) that one can combine willingness and ability to pay when calculating the social benefits of interventions. But, even though this weakness can be overcome, here we just want to record that it is, nonetheless, a weakness and it does need to be overcome.

Willingness to pay theory

121

The reason why the equity weakness must exist can be seen by reexamining the second equality presented above as: P = MU(C)/MU(Y) Previously, when discussing this ratio we assumed that, for an individual, the marginal utility of income MU(Y) could be treated as a constant. But, across individuals, MU(Y) cannot be a constant because of the law of diminishing marginal utility. Income is like any good – the more of it you have, the lower is the marginal utility. Since the poor have little income, their marginal utility of income will be high. So if MU(Y) is high, the ratio will be lower for a poor person than for a rich person with the same MU(C). Hence, the price P that the rich will be willing to pay for any good can be expected to be higher than for the poor. As this is a general result, we would expect to see the positive relationship between WTP and income in almost every application. In fact, this is exactly what we found in the last chapter when we looked at the WTP for an HIV vaccine in Mexico. In Table 25.1 we saw the positive relationship for both measures. For example, the overall average minimum WTP estimate was 6358 pesos. For the four groups the relationship was as follows: ● ● ● ●

For the poorest group, the WTP was 40 percent of the average. For the second lowest group, the WTP was 46 percent of the average. For the second highest group, the WTP was 105 percent of the average. For the richest group, the WTP was 188 percent of the average.

So, this positive relationship held throughout the four income groups.

27.

Willingness to pay practice: the benefits of condoms

When most of the HIV transmissions involve heterosexual couples, as in SSA, promoting condoms has been the major intervention instrument. However, according to Potts et al. (2008), while condoms have been successful in bringing about some declines in generalized epidemics, they have not yet played a primary role. Thus, Epstein (2007) argues for Uganda, which was one of the first countries to reverse a widespread HIV/AIDS epidemic, few people used condoms at the time the HIV infection rate began to decline. When HIV is present just in a few high-risk groups, such as CSWs, condom promotion can have a large impact. For example, the 100 percent condom use requirement in brothels in Thailand was largely responsible for “nipping in the bud” a widespread epidemic in this country. For most countries, condoms have not been used by a sufficiently large share of the population for all sexual partners and for consistently long periods. However, it is not obvious that, if condom use actually did increase on a larger scale, then large reductions in HIV prevalence would not result. So a scheme to promote condom use could be important. The intervention that will be evaluated is the condom social marketing (CSM) program in Tanzania administered by Population Services International (PSI). We will refer to numbers and calculations that appear in Brent (2009c).

PSI’S CONDOM SOCIAL MARKETING PROGRAM IN TANZANIA The CSM program has existed in Tanzania since 1993. In recent years, PSI has aimed to sell condoms to promote behavior change among groups at risk of contracting HIV. The program has four main components: ● ●

educating high-risk groups about the need to use condoms to avoid HIV; promoting its own brand, Salama, which was designated as a highquality condom; 122

Willingness to pay practice ●

●

123

increasing access to the Salama condom by selling them in nonstandard locations (for example, bars and kiosks) after regular shopping hours; and subsidizing the price of the Salama condom; the subsidized price was around a third of the cost of the condoms.

The subsidy component of the CSM program is especially important given the criticism of relying on market-based WTP mechanisms mentioned in the last chapter, that is, people may not be able to afford to buy the condoms. However, whether people will actually buy the condoms at the subsidized price is an empirical matter. If customers are not sensitive to price changes, then CSM programs would not work. So to evaluate the program one first needs to estimate the demand curve for condoms – the relationship between the prices of the condoms and the respective quantities demanded at these prices. In order to provide data on the relationship between prices and quantities of Salama condoms, PSI undertook a survey in 1999 of (male) condom purchasing behavior for 1272 persons aged between 15 and 49 years living in five major townships in Tanzania. Each person who usually bought condoms was assumed to buy one pack of condoms at a time. So the number of persons who were willing to pay a particular price (and there were 18 different prices that people usually paid) corresponded to the number of packs purchased. A curve was fitted to the 18 different price and quantity combinations using regression analysis and the resulting demand curve took the form depicted in Figure 26.1 presented in Chapter 26. Three combinations are shown in the figure: ● ● ●

point a where two were bought at the price of 2000 TZSH; point c where 12 were bought at the price of 290 TZSH; and point e where 39 were bought at the price of 100 TZSH.

The official exchange rate at this time was 805 TZSH (to 1US$). The three prices were therefore equivalent to $4.97, $0.36 and $0.12 (respectively) for a pack of three condoms.

MEASURING THE PRIVATE BENEFITS AND COSTS OF THE CONDOM SOCIAL MARKETING PROGRAM As we saw in the previous chapter, the WTP approach identifies the MB curve with the demand curve. It was explained in Chapters 7 and 8 (when we covered the basic principles of CBA) that when considering a range of

124

Setting priorities for HIV/AIDS interventions

outputs, the total benefits B are the sum of the MBs. This means that total benefits over a range of outputs are given by the area under the demand curve between the beginning and ending quantity levels. In Figure 26.1 we take point a as the starting price and quantity, since some people were willing to pay a high price for their condoms. So without a pricesubsidizing CSM program, two packs would be purchased at a price of 2000 TZSH. With a CSM program, the price people pay is subsidized. At a price of 100 TZSH for a pack of Salama, point e is the price and quantity that results with the program in place. With the program, the quantity that corresponds with the subsidized price is 39 packs. The CSM program therefore involves moving the purchases of condoms along the demand curve from point a to point e so that the number of condoms bought increases from 2 to 39. The total benefits of this price reduction are represented in Figure 26.1 by the area bounded by 2ae39. This area was calculated to be 11 415 TZSH. The cost per pack was estimated by a World Bank group to be around 290 TZSH, which meant that for the 37 extra packs (39 – 2) the total costs were approximately 10 794 TZSH. The net benefits were positive at 621 TZSH, but they were rather small, being only 6 percent above the costs – see the first column of numbers in Table 27.1. The net benefits would have been even greater if the subsidized price charged would not have been so low. If the market price of 290 TZSH were charged, the program would have involved moving from point a to point c (instead of point e) in Figure 26.1. The quantity change would not have been so large. The number of packs would have increased by ten (from 2 to 12). The total benefits would have been the area under the demand curve given by 2ac39. This area was calculated to be 7187 TZSH. With costs of 3222 TZSH, the net benefits would have been 3965 TZSH. These alternative results are reported in the second column of numbers in Table 27.1. The net benefits would have increased enormously to be 123 percent above the costs. In the last chapter we explained that a market operates like a CBA test. Now we have just seen that the market is not just any CBA test. It is the one that generates the highest net benefits. It is instructive to understand why this is the case. “Consumer surplus” is defined as the difference between what the consumer is willing to pay and what he or she has to pay. Consider again Figure 26.1. With the price charged equal to the market price, the consumer has to pay 290 TZSH. What the consumer is willing to pay is given by the demand curve. For the second pack, someone is WTP 2000 TZSH and has to pay 290 TZSH. So for this second pack, consumer surplus is 1710 TZSH. For units up to 12, consumer surplus (and also MB – MC) is positive. At the market equilibrium quantity of 12,

Willingness to pay practice

Table 27.1

Private benefits and costs of a condom social marketing program in Tanzania (in TZSH)

Benefit and Cost Categories

Private benefits Costs Net benefits Source:

125

CSM Program: Price Lowered From TZSH 2000 to TZSH 100

From TZSH 2000 to TZSH 290

11 415 10 794 621

7 187 3 222 3 965

Based on Brent (2009c) Tables 6 and 7.

there is no longer any consumer surplus. The total consumer surplus for the quantity change from 2 to 12 is represented by the area bac. This area amounts to 3965 TZSH, which is also equal to the net benefits of increasing condoms up to this level. As we can see in Figure 26.1, bac is the largest area of consumer surplus that can be obtained by increasing output from the initial quantity of 2. In other words, the competitive market equilibrium maximizes consumer surplus. It was because the expansion of the actual CSM program to the quantity of 39 was producing negative consumer surplus (for example, the difference between point d and e was minus 280 TZSH) that the net benefits were lower for the existing price subsidy in Table 27.1.

MEASURING THE SOCIAL BENEFITS AND COSTS OF THE CONDOM SOCIAL MARKETING PROGRAM WTP is the correct concept to use to measure benefits. However, the market demand curve measures only private WTP. It does not necessarily measure the WTP of everyone as there are some people who do not buy and sell on a particular market, yet they are affected and may be WTP something. Economists call this non-market, indirect effect “externalities”. In the case of condoms, there may be others, that is, the partners of those purchasing the condoms, who benefit from the condoms. For the WTP of those who do the purchasing to also reflect the WTP of others, they must be both informed, and care about, the effects on others. In the case of condoms, it is the sex partner(s) of the purchasers that is of interest. Brent concentrated only on the immediate partners (and not the partners of the partners, who all could get HIV if the original couple

126

Setting priorities for HIV/AIDS interventions

Table 27.2

Social benefits and costs of a condom social marketing program in Tanzania (in TZSH)

Benefit and Cost Categories

Social benefits Costs Net benefits Source:

CSM Program: Price Lowered From TZSH 2000 to TZSH 100

From TZSH 2000 to TZSH 290

17 168 10 794 6 374

10 809 3 222 7 587

Based on Brent (2009c) Tables 6 and 7.

transmitted the disease through unprotected sex). In the context of sexual relations it seems plausible to argue that, if the sex partner is a regular partner or a spouse, then persons purchasing the condoms would reflect their partners’ preferences when they reveal their preferences by their WTP. However, for occasional partners, or partners one has just met, it is reasonable to argue that their WTP preferences are being ignored by the person purchasing the condoms. In the PSI sample, 50.4 percent of the sample were not regular partners or spouses. If these partners have the same WTP as the purchasers, then the private benefits can be scaled up by 1.504. Table 27.2 takes the private benefits from Table 27.1 and multiplies them by 1.504 to obtain the social benefits. We see that both the actual CSM program, and the one that would result from charging a price equal to the market price, have very high net benefits. Again, the program that lowers prices only so far as they cover costs would have higher net benefits than the actual CSM program that charged a price one-third of the costs.

SUMMARY AND CONCLUSIONS The three main points that come out of our examination of the CSM program in Tanzania based on the WTP approach that have general relevance for CBA theory and practice are these: ●

Competitive markets operate like CBA tests. They lead to the highest net benefits as they maximize consumer surplus. Note that maximizing revenues (price × quantity) is not consistent with CBA as it ignores consumer surplus. The price that consumers are willing to pay is greater than the price that they are actually charged.

Willingness to pay practice ●

●

127

However, the demand curves that are used to generate the consumer surplus estimates only reflect private benefits and do not necessarily represent the WTP of society as a whole, that is, those in the market and those external to the market. External WTP must be added in if the market affects others positively and must be subtracted out if others are affected negatively. CBA is concerned with maximizing net benefits. It does not just try to maximize benefits. On the basis of estimates of social benefits, the actual CSM program in Tanzania produced very large net benefits, around 59 percent higher than the costs. But, these net benefits resulted from the actual CSM program aiming to sell the largest number of condoms, effectively trying to generate the most benefits. A policy of trying to maximize net benefits would have led to even higher net benefits than the actual CSM program, with net benefits 135 percent over costs.

28.

Cost minimization theory

If one can ignore the existence of consumer surplus, then a benefit is simply a negative cost. Since cost data are nearly always available, many health care studies evaluate interventions by just using costs. That is, if the current method costs $X and the alternative way of doing the same thing costs $Y, then the new method would be better then the old as long as $X > $Y. The difference $X – $Y would be the cost savings of the new method and this difference would also equal the net benefits of changing methods. This is the logic of cost minimization. With the application in the next chapter in mind, we will explain issues in terms of TB treatment.

COST MINIMIZATION AS A CBA The theory underlying cost minimization is basically the same as for evaluating a quantity change brought about by a price reduction that we explained in Chapters 26 and 27, except that we are now dealing with a quantity change that is not necessarily market clearing (one where demand equals supply). Figure 28.1 summarizes how the cost minimizing evaluation is to take place. We start with a situation where we are treating 20 patients with TB using the standard regime, Method 1, at a cost of $200 per patient. Then a new regime is introduced, Method 2, that lowers costs to $100 per person. The question is: how beneficial is the new method? The total costs of Method 1 are represented by area 0bc20. This is equal to $4000, being the product of $200 × 20. The total cost of Method 2 for the same quantity 20 is shown by the area 0de20, which amounts to $2000 ($100 × 20). The difference is the area dbce, a sum of $2000 ($4000 – $2000). This is the size of the cost saving. As it is positive, Method 2 should be used to replace the current treatment regime. But, how can we be sure that the cost minimization method is giving us the right answer? We can check this by going back to our usual CBA framework and verifying the outcome. A CBA evaluation of Method 1 would show that the 20 patients produce total benefits of area 0ac20 and total costs of area 0bc20. The net benefits (the difference between the two areas) are given as area bac. Similarly, the total benefits of Method 2 are area 0ac20 and total costs are

128

Cost minimization theory

129

a

Price

$200 $100

b

c

d

e

Marginal costs of Method 1 f

Marginal costs of Method 2 Marginal benefits = demand

0

20

25

Patients treated for TB Source:Author’s own illustration.

Figure 28.1

The costs and benefits of alternative TB treatments

area 0de20 with net benefits equal to dace. As Method 2’s net benefits dace are dbce greater than Method 1’s net benefits of dace, the area dbce reflects the change in the net benefits. This area is equal to the cost savings area calculated earlier. This “happy” result comes about because the consumer surplus area abc is common to the benefits of Method 1 and Method 2. When the difference is taken, this area simply cancels out. So all one is left with is the area under the demand curve that consists of costs. The difference in costs then approximates the difference in net benefits. Note that this happy result would not come about if the quantity level associated with the two methods were different. If 20 was the quantity associated with Method 1 and 25 was the quantity related to Method 2, area abc would be the old consumer surplus and adf would be the new consumer surplus and the two areas would not now cancel out.

WHAT TO INCLUDE IN THE MEASUREMENT OF THE COSTS Although everyone involved with undertaking a health care intervention is inevitably aware of costs, this does not mean that everything that participants include as costs is meaningful. Nor does it mean that some things

130

Setting priorities for HIV/AIDS interventions

that they do not include can be ignored. Cost in economics is “opportunity cost”. If one uses a resource for one purpose, then one misses the opportunity of using it for some other purpose. It is the value of the resource in its next best alternative use that one should use to measure costs. There are two main principles involving the measurement of costs that need to be understood: First, and this point will be emphasized in the application in the next chapter, what is a cost to a particular agency may not be an opportunity cost. For example, when purchasing inputs for a health care intervention, taxes may be incurred. The taxes do not involve any deprivation of the use of resources by others. The tax payments are simply transfers from the users to the government. The agency loses the exact same amount that the government gains and the two sums cancel out. This means that the taxes can be ignored when making a CBA. On the other hand, many health care activities gain from services being freely donated by volunteers. These services could be used to contribute to some alternative activity so this value should be included as a cost even though no one is asking any money for the services. Second, cost in CBA is “social” cost. That is, costs are the sum of costs of all persons affected by the activity. These costs include those affected by the agency actually carrying out the intervention (the “private costs”), but also include those outside the agency (the “external costs”). Just as we saw in the case of the benefits of condoms, where there were partners who benefit from condoms being used by a purchaser, there can be costs on others affected by the purchase and use of a resource. An important category of external cost that frequently appears in health care evaluations is the time given up by patients and their families when treatment is being provided by a physician or nurse. In order to get treated, people need to give up their time travelling and attending a health care facility that could have been used, say, to earn wages. These foregone wages are costs even though they do not appear on the budgets of the health care facilities.

THE STRENGTHS AND WEAKNESSES OF THE COST MINIMIZATION METHOD The cost minimization approach appeals very much to those who think that the whole idea of measuring benefits makes no sense. These include physicians, nurses, psychologists, psychiatrists and a whole host of health professionals, workers and researchers. They think that all health care expenditures are necessary. The only recognition of economic principles for these groups is that they realize that costs must be involved. If no one

Cost minimization theory

131

incurs the costs, and funds the agencies, then the services cannot be supplied even though they are “obviously” worthwhile. Hence, if there is any way that costs can be reduced, then there is an awareness that adopting this cost saving activity must clearly be worthwhile. There is nothing basically wrong with this line of reasoning, except that it is too limited a way to be looking at health care evaluations. Most times, quantities change and then the method breaks down. After all, HIV/AIDS is a major pandemic. One wants resources for treatment, care and mitigation to increase, and to increase drastically. To see what difference a quantity change makes, refer back to Figure 28.1. With quantity fixed at 20, the cost saving area is the net benefits dbce. If quantity were to change to 25, the net benefits would be equal to dbce plus the area cef. The additional net benefits cef is the amount by which the WTP, given by the demand curve, exceeds the costs. Thus, when quantities change, measuring the demand curve to quantify the benefits is essential and cannot be avoided. The practical question then reduces to one of knowing whether quantities can ever be assumed to be constant. We give two reasons for doubting whether quantities can remain unchanged. (1) It is rare that when comparing two interventions that they will give exactly identical outputs. Even two drugs targeting the same disease or complaint will not achieve this to the same degree or with the same side-effects. (2) Even when quantities could be exactly the same, the cost savings themselves will most likely induce a quantity change. Consider the cost reduction of treatments from $200 to $100 shown in Figure 28.1. Consumers would not be content to demand the same number of treatments, that is, 20. Instead they would demand 25. To keep quantity at 20, some form of rationing would be required. There is one implication of the cost minimizing analysis shown in Figure 28.1 that involved producing a fixed quantity of 20 at least cost, that is worth generalizing. We saw that the consumer surplus area abc, because it was common to the two methods, could be ignored. Anytime there is a common ingredient in two interventions, that ingredient does not need to be valued. So if Methods 1 and 2 both have a side-effect of losing two lives, these two lives do not need to be estimated, at least not for this particular CBA. The loss would be the same no matter which intervention was chosen.

29.

Cost minimization practice: the costs of treating TB

It has been estimated that as many as one-third of the world’s population has a latent form of TB, although most do not have any symptoms of the disease – see UN Millennium Project (2005, p. 100). As we learned in Part II, people with HIV have a compromised immune system, which means that they are more susceptible to any disease including TB. So we should not be surprised that people with HIV are more likely to get TB. What we should be surprised with is the extent of this susceptibility. According to Ngamvithayapong (2007), people who are HIV positive are 100 times more likely to contract TB than those who are HIV negative. So TB is one of the main opportunistic infections that people with HIV can expect to get. In fact, in SSA, TB is the number one cause of death for those with HIV. Conversely, if one has TB, one is also very likely to be infected with HIV. So the WHO and UNAIDS recommend that everyone with TB living in a country with a more than a 1 percent HIV prevalence rate should be given counseling and testing for HIV. As TB and HIV are so intertwined in SSA, that is, there is a dual TB/HIV epidemic, one can class TB treatments as HIV interventions. The TB intervention evaluation that we will be examining is by Sinanovic and Kumaranayake (2006) that relates to South Africa.

TB TREATMENT IN SOUTH AFRICA In line with much of SSA, TB and HIV are going hand in hand in South Africa. Sinanovic and Kumaranayake report that in 2002 there were 6.5 million people infected with HIV and that 60 percent of TB adult patients also had HIV. The cure rate for TB was only 55 percent, when the WHO target was 85 percent. One of the reasons for the low cure rate was that patients did not always have access to treatment and testing, whether from public health facilities or from private resources (such as private doctors working in employer health services). The idea was to test whether a partnership between the public and private sectors could provide treatment at 132

Cost minimization practice

133

comparable costs as the public sector on its own, which used highly paid nurses to supervise the treatments. There were two types of private sector institutions involved in the public–private partnership (PPP). The first type of private sector participant was the for-profit mining companies who supplied worksite TB services. This PPP arrangement was called the PWP model, for public–private workplace participation. Involving the mining sector was important because the TB incidence rate was nearly ten times larger here than in the rest of the population and HIV prevalence was also high among miners. The second type of private sector participant was the non-profit, non-governmental organizations (NGOs) who supplied treatment out in the community (in the townships). This PPP version was called PNP, to stand for the public–private NGO partnership model. In this model, and in the PWP, supervision was done by health care workers following national guidelines. The standard treatment for newly diagnosed pulmonary TB patients is the directly observed treatment, short course strategy (DOTS). As its name implies, drugs are supplied and someone has to observe them being taken. Non-compliance is a major problem with TB treatments and the aim is to monitor the extent of compliance. The standard drug regimen involves supplying the correct dosage five days a week for six months.

ESTIMATING THE COSTS For each of the three models, that is, purely public, PWP and PNP, two separate sites were chosen in order to allow for sampling variation. Costing over a 12-month period was undertaken for all three models under two categories. Financial costs were the actual expenses incurred in providing the services. As we pointed out in the last chapter, these expenses may not reflect opportunity costs. So a second cost category was estimated called social costs. These are the opportunity costs of all parties that cover both the institutional providers and the patients. The main cost items were hospital stays, clinic visits by patients and monitors, and carrying out the testing (the sputum smears and cultures). Table 29.1 shows the cost estimates for the two categories for the six sites. On the basis of the social cost figures in Table 29.1, we see that the lowest cost per patient, irrespective of site, were for the community-based PNP model – see the costs for sites C and D. For this model, the largest cost component was for the supervision of the program. If one does not take the societal perspective, then the least cost model would vary enormously by participant. From the public provider and patient perspectives,

134

Setting priorities for HIV/AIDS interventions

Table 29.1

The cost per patient from diagnosis to completion of DOTS treatment in 2001 (in US$) Public–Private Workplace

Public–Private NGO

Site A

Site B

Site C

Site D

Site E

Site F

Social costs Public provider Private provider Provincial TB program Patient Total societal costs

0 609 46 0 654

0 708 36 0 744

59 121 71 39 290

63 114 76 37 290

445 0 62 102 609

506 0 62 122 609

Financial costs Patient Non-patient costs Total financial costs

0 544 544

0 669 669

2 203 205

2 209 211

26 457 483

20 530 550

Source:

Purely Public Model

Based on Sinanovic and Kumaranayake (2006, Tables 2, 3 and 4).

PWP does not involve any costs at all, though the provincial government does incur some costs. Obviously, from the private provider perspective, the purely public model is cheapest as the private sector does not participate. The financial costs are also lowest for the PNP model. But, it is instructive to see why the financial costs differ from the social costs across the models. There are two main cost items that are worth examining in greater detail: patient costs and drug costs. (i)

Patient Costs

Patients do not travel at all for the PWP program as they get the services at work. Patients are served out in the community in the PNP model, so again travel costs are small. But, for the purely public model, patients have to travel to the clinics. This involves monetary travel costs, which are basically the same in economic and financial terms, and time costs, which are zero in financial terms, but positive in economic terms as time is valued at the foregone wage rate. The total patient costs at sites E and F amounted to around a third of the entire PNP model costs.

Cost minimization practice

(ii)

135

Drug Costs

The TB drug costs in financial and economic terms were estimated to be $49 at most of the sites, being the state-tendered price that the provincial TB control program paid for the drugs. However, the market price was over seven times as large as the state-tendered price. At site B, the private provider paid the market price, but the state only reimbursed the provider $36 for the drugs. This meant that drug costs at this site were $383 in social terms and $36 lower than this in financial terms. As a result, site B had the highest treatment costs of all the sites. If site B had received the drugs free of charge from the provincial program like site A, instead of just getting the $36 reimbursement amount, site B’s total per patient costs would have been approximately halved.

SUMMARY AND CONCLUSIONS Apart from its use as a guide to decision-making in South Africa, there are three points that come out of the PPP application that inform us about the cost minimization methodology: ●

●

●

Cost minimization means different things according to the perspective (patient, provider and so on) of the party involved with the provision and supply of the good or service being evaluated. So strictly, one could have many different answers to the question, which form of PPP provides the services at lowest costs? Only if the social perspective is adopted that is central to the CBA method will the answer be unambiguous. Financial costing is clearly inadequate as a means of estimating social costs. Take the case of the TB drugs. The social cost is set at the foregone values of the resource. This cost is the same at all sites. But, the financial cost varies enormously even if all sites have the same level of efficiency. The financial cost would be $0 if the province provided the drugs free of charge. The cost would be $10, if the private provider bought the drugs at the state tendered price of $46 and was given a subsidy of $36 by the province. If the province was not involved at all and the private provider had to pay the market price, then the financial cost of the drugs would be $350. Cost minimization is not independent of WTP considerations even though this evaluation methodology does not attempt to measure benefits. Why is the market price for drugs $350 even though the state would pay only $46 for them? The answer is, because people

136

Setting priorities for HIV/AIDS interventions

are willing to pay $350 for the drugs. If the price charged is $46 and consumers are WTP $350, then they receive a consumer surplus of $304 if the province buys the drugs for them. This is not just true for drugs. It is true for any resource that is costed by using the market price. In the South African PPP application, supplies (for example, tests and X-rays) were costed using market prices and thus dependent on WTP. Note that the costing that took place in Table 29.1 for the PPP application was in terms of all the inputs that are used in treatment from the beginning of the six-month period until the end. There was no mention of outputs. A possible output would be the cost per successfully treated patient. In this application it has to be assumed that the success rate was the same across all the six sites for the cost minimization method to be valid (as explained in the last chapter). What happens when outputs and not just inputs can vary? We are in the realm of cost-effectiveness analysis, which is the evaluation method that will be examined in the next two chapters.

30.

Cost-effectiveness theory

When there is more than one alternative way of undertaking an intervention and not only the costs differ by alternative, but the quantities produced by the alternatives also differ, then the evaluation method switches from being cost minimization to becoming cost-effectiveness analysis (CEA). In principle, the quantity produced could be any outcome that results from an intervention that one is evaluating, such as the number of HIV tests undertaken or ARVs administered. However, we will concentrate on the most general of the health care outcomes, that is, the disability adjusted life year (DALY), which was the outcome measure we used to measure the effects of malnutrition in Chapter 12, and the effect we used to apply the threshold method to estimating the benefits of avoiding HIV in Chapter 24. Strictly, a CEA that uses a DALY as the only outcome measure can be called a “cost-utility analysis”. But, this distinction is not always adhered to in the health care evaluation literature, so we shall simply stick to the term CEA to refer to this method where effects and costs vary.

CARRYING OUT A COST-EFFECTIVENESS ANALYSIS There are a number of steps that have to be undertaken to carry out a CEA. The main steps are: 1.

Establish the sum of money that is going to be spent on the possible interventions. Say the sum available is $500. 2. Identify a series of interventions, measure the costs and DALYs they generate, calculate the cost per DALY (the cost-effectiveness ratios) and list them in order from lowest to highest. 3. Start at the least costly intervention and see whether it can be financed by the funds available. If the cost is less than the budget, one can approve the project. Then one works up the list and approves all those that are feasible with the funds budgeted. One stops at the last project that just can be financed with the $500. This is the set of approved projects. All those alternatives that cannot be covered by the $500 are rejected. 137

138

Table 30.1

Setting priorities for HIV/AIDS interventions

The cost-effectiveness of selected HIV interventions around the world (in US$)

Intervention

Cost per DALY Gained

1. Hospital-based blood screening 2. Female condoms targeted to prostitutes 3. Home-based care done through the community 4. Therapy for TB involving isoniazid for six months 5. Prevention of mother-to-child transmission (PMTC) via formula provision 6. Antiretroviral therapy for adults

1 12 99 169 218 1800

Source: Based on information in Creese et al. (2002) Table 5.

The process can be illustrated by looking at a small selection of the costeffectiveness ratios that Creese et al. (2002) calculated on a common basis for a number of developing countries for 2000. Refer to Table 30.1. Project 1, blood screening, is the most cost-effective intervention at $1 per DALY and this gets approved first. Next to be approved is Project 2, female condoms, and this is followed by home-based care (Project 3), TB therapy (Project 4) and formula provision to avoid transmission from mothers to their children (Project 5). If we sum the cost of all five projects this amounts to a total cost of $499. There is only $1 left of the $500 in the budget, so effectively the funds have been exhausted by the five projects. As there are no more funds available to finance any additional interventions, Project 6, involving ARVs, would get turned down. This process is called “cost-effective” because it guarantees that no other allocation of resources could produce more DALYs for the same total cost.

CEA AS CBA The CEA process can be expressed in CBA terms by first defining a benefit as an effect measured in monetary terms. Since a DALY is the effect, and if we regard the conversion of the effect into monetary terms as “pricing” the DALY, then we can obtain the identity: B = [Price of a DALY] × [Number of DALYs]

Cost-effectiveness theory

139

Our CBA decision rule is always that one approves projects provided that the benefits B are greater than the costs C. So using the benefit identity in the CBA decision rule leads to: [Price of a DALY] × [Number of DALYs] > C Since dividing total costs C by the number of DALYs is just the cost per DALY, then dividing both sides of the equality by the number of DALYs produces the CBA criterion: Price of a DALY > Cost per DALY But, where does the price of a DALY come from in the CEA exercise? The answer is that the price of a DALY is a threshold value just like the ones that appeared in Chapters 22 to 25. It is the cost per DALY of the last project that was just marginally approved in the fixed budget. So with $500 as the budget, the cost-effectiveness ratio of Project 5 is the threshold cost per DALY. This is given as $218 in Table 30.1. Thus, if the price of a DALY is $218, then for a different intervention to replace one in the approved set, it must produce a DALY at a cost of less than $218.

THE STRENGTHS AND WEAKNESSES OF CEA The main advantage of CEA is that it is relatively easy for an evaluator to undertake. The medical team produces the health outcomes, the comptrollers list the costs and one just divides the costs by the outcomes. It is true that there is more work to do if the outcome is in DALYs and not just persons treated or tested and so on. But certainly no direct estimation of benefits seems to be involved. Also, from the point of view of one department in a large agency, CEA appears to be relevant to the world of administrators who are presented with a fixed annual budget. The argument goes like this: if the funds are limited to those budgeted, what is the point of evaluating projects that cannot be financed out of this budget? The biggest problem with CEA is that the idea of a fixed budget constraint is not valid in theory or practice. From a social point of view, it is not the budget of an individual department that fixes the threshold CEA ratio. It needs to be the total funds of all agencies dealing with HIV/AIDS interventions. This is no easy figure to calculate because, as we have seen in Part II, when one is dealing with generalized HIV epidemics, an HIV/ AIDS intervention could involve transport, nutrition and education as well as health expenditures. Moreover, does it ever make sense to think of

140

Setting priorities for HIV/AIDS interventions

funds being fixed independently of considering the benefits of what is to be financed? $5 for a tooth-pick is not much of a deal, but, $5 for a diamond would be a great bargain. One cannot decide whether to cap spending at $5 unless one knows what could be obtained for $5, or even more than $5. In any case, the world of the early 2000s is very different from the world of the early 1980s. With the advent of the Global Fund and the US President’s PEPFAR program, there are more funds available to be spent on HIV/AIDS interventions than ever before. Even if (or especially if) these funds were to dry up in the future, the point is the same. The budget cannot be regarded to be fixed as theory requires. In practice, evaluators nearly always undertake their CEAs without mentioning an explicit budget constraint. How can one know whether an intervention is cost-effective or not if one is not told the amount of funds available? The answer is that the evaluators often use thresholds that others have thought decisive. In the World Bank’s World Development Report for 1993 it was suggested that any intervention producing a DALY at a cost of $50 or less was cost-effective in the context of the poorest developing countries. However, as we shall see in the next chapter this benchmark cannot be generally used to determine what is socially desirable even if one accepts it as being a good guide to what is cost-effective. Another possible line of criticism of CEA, especially by Arnesen and Nord (2006), has been its reliance on DALYs as the sole outcome measure. There are, indeed, measurement problems involved with estimating DALYs. But, the main alleged conceptual criticism is Arnesen and Nord’s claim that the DALY assumes that the lives of disabled persons are worth less than the lives of healthy people. To understand their claim, say that a person is at the final stages of HIV, that is, AIDS is taking over. If a DALY for those with AIDS is valued at 0.2, then saving their lives would contribute 0.2 DALYs; while saving the life of a healthy person would generate 1.0 DALYs. The healthy person’s life would always be saved first. But, this argument ignores the flip side. If a person is mildly ill and has a DALY recorded as 0.95, then spending on curing his/her illness would contribute only 0.05 DALYS, much less than the 0.8 DALYS that would be produced by spending on a cure for AIDS. Thus, the correct way of formulating the Arnesen and Nord concern is to say that a DALY biases interventions in favor of cures over preventions when DALYs are low due to disabilities.

31.

Cost-effectiveness practice: the benefits of ARVs

There is a common saying: “Put your money where your mouth is”. More prosaically, the expression is: “Where your treasure is, there is your heart”. What these sayings imply is that if you say you care for something, then you should be willing to back up the words by giving up your money to support it. This is the basic idea behind the revealed preference methodology that we are going to use in this chapter to carry out an evaluation of antiretroviral drugs (ARVs). To use the CEA method, one needs to have an estimate of the price of a DALY. We present a study that estimated the value of a DALY derived from the revealed preferences of the Global Fund to Fight AIDS, Tuberculosis and Malaria (GFATM).

THE REVEALED PREFERENCE APPROACH APPLIED TO INDIVIDUAL AND SOCIAL DECISIONS The revealed preference approach is not something that should be entirely new to the reader as we have already used it in Chapters 26 and 27 to justify the WTP method when dealing with individual preferences as embodied in market behavior. Recall that we said that if someone was willing to pay $4 for a hamburger, then that amount (at least) is what the hamburger must have been worth to the person paying that amount for it. It was this reasoning that we invoked when we argued that the demand curve for condoms is also the marginal benefit curve for condoms as it records the WTP that was revealed by people when they make their purchases. We are now going to extend the argument to cover two goods and not just one and then deal with social, rather than market, revealed preferences. Say we already knew that a person bought one hamburger for $4 and market research now tells us that this same person also bought a book for $8. Then we can deduce using the revealed preference methodology that the person valued the book as equivalent to two hamburgers. We can deduce this because the person spent twice as much on the book as he or she spent on one hamburger. That is, the person valued the book twice as 141

142

Setting priorities for HIV/AIDS interventions

much as a hamburger, given the two prices established by markets and the income that the individual had available. As was explained in Chapters 26 and 27, market revealed preferences, especially when they are augmented by valuations of third parties who are also affected, is best practice for estimating benefits in CBA. However, when it comes to DALYs, not only do markets for these outcomes not exist, one can also argue that, even if explicit markets were used, they would not be reliable as individuals have no experience of purchasing DALYs. Thus, it may be true that purchasing cough syrup for $10 could increase people’s quality of life, such that they avoid the loss of 0.01 DALYs. But how could anyone know this? Certainly the number of DALYs saved is not specified on the label of the cough medicine. Nor would a person know how many DALYs could be saved if instead of buying the cough mixture, some alternative medicine (say aspirin) was purchased. In these circumstances, it may be useful to consider an alternative approach that tries to extract an expression of revealed preferences by social decision-makers. The starting point for a revealed preference study is a statement of objectives (preferences) by the person or group making the social expenditure decisions. For a government, the objective could be part of legislation or, for an international body, the objectives would be specified in the constitution or rules of association that set up the organization. Then there needs to be some information on spending and the extent to which the objectives were changed by the spending. Finally, there needs to take place a statistical analysis (regression equation) that examines the relation between the changes in the measured objectives and the changes in measured spending. So if, for example, a unit change in a first objective increases spending by $1, and a unit change in a second objective causes spending to go down by $2, then spending is unchanged if two units of the first objective are exchanged (or traded) for one unit of the second objective. The rate of exchange between one unit of one objective for one unit of the other is the trade-off of the second objective in terms of the first objective. When one of the two objectives is expressed in monetary terms, and the other is not, then the trade-off between the two objectives will effectively be a “price”, that is, converting the non-monetary objective into monetary terms. We will now explain how these ingredients for the revealed preference approach were made operational in the context of GFATM expenditure decisions as analyzed in Brent (2009b).

Cost-effectiveness practice

143

THE GLOBAL FUND TO FIGHT AIDS, TUBERCULOSIS AND MALARIA (GFATM) The Global Fund was set up in January 2002 to finance projects for the three diseases to be carried out by nationally organized entities represented by governmental ministries, groups of ministries, international organizations, civil society and NGOs. By the end of 2005, the amount approved in grants was US$9.6 billion. Over the period 2002–05, there were 350 grants given to 128 countries. The grants going to 48 African countries will be analyzed, amounting to around 60 percent of the total commitments. In the mission statement for the Global Fund it was stated that funds would be given to: “Countries and regions in greatest need, based on the highest burden of disease and the least ability to bring financial resources to address these problems”. The burden of disease could be measured by the number of DALYs lost to each of the three diseases and the ability of a country to finance these diseases could be reflected by a country’s per capita GNP. So there were two objectives that were assumed to be important for the GFATM, that is, DALYS and per capita GNP. The higher the number of DALYs lost, the greater the need; and the lower the per capita GNP of a country, the greater the need. It is the trade-off between these two objectives that leads to an estimate of the price of a DALY in terms of dollars of national income. The Brent study had separate statistical equations for four DALYs lost-to-disease specifications, that is, DALYs lost from any disease at all; DALYs lost from the three diseases AIDS, TB and Malaria; DALYs lost from the two diseases AIDS and TB and DALYs lost just due to AIDS. We will focus just on the case where DALYs lost to AIDS was involved. The statistical relationship between the amounts of grants committed and the two objectives revealed that: (1) one more DALY lost per capita to AIDS caused GFATM spending to rise by US$72.67, and (2) $11 871 of additional per capita GNP caused GFATM spending to fall by US$72.67. We can put these two pieces of information together as follows. One more DALY lost per capita would cause GFATM to spend $72.67 more (as they would be more in need). This could be offset by per capita income for a country going up by $11 871 seeing that this would cause GFATM to spend $72.67 less (as they would be less in need). The two changes being considered as equivalent in spending terms would mean that one more DALY would be traded for an additional $11 871. In other words, a DALY was priced at $11 871 by the Global Fund.

144

Table 31.1

Setting priorities for HIV/AIDS interventions

Cost-effectiveness ratios for antiretroviral therapy (US$ cost per DALY)

Type of ARV Intervention Single interventions First line drugs only First line drugs with intensive monitoring First and second line drugs only First and second line drugs with intensive monitoring Combined interventions First line drugs only First line drugs with intensive monitoring First and second line drugs only First and second line drugs with intensive monitoring

CE Ratio (US$) 350–2029 596 2010 1977 547–563 1144 2011 5175

Source: Based on Brent (2009b) Table 3.

CEA OF ANTIRETROVIRAL DRUGS Table 31.1 reports estimates of the cost-effectiveness ratios for ARVs that appear in the health evaluation literature. The single interventions are for ARVs assuming that no other interventions are taking place. The combined interventions are when ARVs are added to a package that includes prevention interventions first. Without a specified budget constraint we cannot tell whether any of the ARV options in Table 31.1 are worthwhile. If we use the World Bank’s (1993) cut-off value mentioned in the last chapter, that producing a DALY at a cost of $50 or less was cost-effective in the poorest of countries, then none of the ARV options would be approved. However, as we also saw in the last chapter, that if we use the CBA version of the CEA criterion, we would pass an intervention if: Price of a DALY > Cost per DALY If we use the price of a DALY that was implied by past Global Fund decisions we would have to approve any intervention that produced a DALY at a cost of less than $11 871. We would now have to decide that instead of none of the ARV options in Table 31.1 being acceptable, any one of them could be approved using a CBA.

Cost-effectiveness practice

145

SUMMARY AND CONCLUSIONS Spending reveals preferences. When one is spending on two different items, the relative spending reveals the relative preference of the two items, that is, the trade-off of one item for the other. If one item is a physical good while the other is in monetary terms, then the trade-off expresses the value or price of the physical item in monetary units. In this way a DALY was priced using the spending behavior of the Global Fund. We then placed this DALY price into a CBA framework that involved comparing the price with the cost that was incurred in generating a DALY produced by an intervention. We applied this framework to a consideration of whether ARVs were worth financing. We found that the price of a DALY was much higher than any of the cost per DALY estimates that currently exist in the literature, which involved many different ways of carrying out the ARV intervention. The main conclusion we reached was that without an explicit statement of the funds that are to be made available, a CEA based on commonly applied cost-effectiveness guidelines did not give reliable decision outcomes. We saw that although ARVs are routinely found to be the least cost-effective of all HIV interventions (see Table 30.1 in Chapter 30 as well as Table 31.1) they were in fact highly socially desirable in CBA terms. Logically then, the converse can also be true. An intervention that is the most cost-effective for some other category of HIV intervention may not be actually worthwhile using a CBA test (if, for example, it provides the DALYs at a cost greater than $11 871). One last point. Why should we treat seriously an estimate of the price of a DALY based on the revealed preferences of the Global Fund? This estimate has social significance for two main reasons. (1) The mission statement of the Global Fund said that its allocation of grants would be based on social considerations and these considerations explicitly included the burden of disease, which is typically measured by reference to DALYs lost. (2) The Global Fund is simply a financing agency. It did not have any control over how the money was to be spent, which was to be decided by a Technical Review Panel, consisting of scientists. There is no reason to think that the Global Fund members were making decisions that would serve their self-interest or the interests of any particular political group. The valuation could therefore be interpreted to be social and not private or political.

32.

Human capital theory

In Part I we came across the claim that saving lives is what HIV/AIDS interventions is, or should be, all about. There we pointed out that not all HIV interventions saved lives; and even if they all did, there may be other uses of resources (such as for nutrition or treating other diseases) where even more lives could be saved, or the same number of lives could be saved at lower cost. So one needs to evaluate all interventions, no matter how obviously beneficial they may seem to be in terms of saving lives. For the rest of Part III we will accept that saving lives is the main objective of HIV/ AIDS projects and see how CBA can be used to evaluate the projects from this perspective. In this chapter and the next we deal with the most common way of dealing with the estimation of lives saved and lost in CBA, and in the following two chapters we will examine best practice for valuing lives.

THE HUMAN CAPITAL APPROACH The basic idea behind this method for estimating benefits is that the value of a person’s life that you save can be measured by the present value of the lifetime earnings of the individual. The wages of an individual are determined by the value of the product that the worker produces. The greater the quantity of goods generated, and the amount that people are willing to pay for the goods and services produced, the more valuable the worker. When the person dies, society loses the value of the output that would have been produced. There are four main steps that have to be undertaken to calculate the present value of lifetime earnings: 1. 2. 3. 4.

The current age and the expected remaining working life of the person whose life is being saved have to be identified. The earnings for each and every working year for the individual then have to be estimated. Each year’s earnings must be discounted to its present value equivalent by using a specified social discount rate. The present value of each year’s earnings must be summed to obtain the lifetime earnings. 146

Human capital theory

Table 32.1

Year

147

The benefits of a life in Tanzania as the present value of 26 years of earnings (in TZSH) Growth Adjusted Earnings

Discount Factor

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Totals

335 613 346 018 356 744 367 803 379 205 390 960 403 080 415 576 428 458 441 741 455 435 469 553 484 109 499 117 514 589 530 542 546 988 563 945 581 427 599 451 618 034 637 194 656 947 677 312 698 309 719 956 13 118 105

0.605 0.587 0.570 0.554 0.538 0.522 0.507 0.492 0.478 0.464 0.450 0.437 0.424 0.412 0.400 0.388 0.377 0.366 0.355 0.345 0.335 0.325 0.316 0.307 0.298 0.289

Source:

Created by the author based on Brent (2009d).

Present Value 203 052 203 249 203 446 203 644 203 841 204 039 204 237 204 436 204 634 204 833 205 032 205 231 205 430 205 629 205 829 206 029 206 229 206 429 206 630 206 830 207 031 207 232 207 433 207 635 207 836 208 038 5 343 913

We will illustrate these four steps in the context of valuing the benefits of saving the lives of females by providing them with primary education in Tanzania, which is the evaluation we will be examining in detail in the next chapter. Table 32.1 shows all the base figures and the resulting calculations. Illustration from Tanzania 1.

Tanzania provides primary education free of charge for seven years for all females (and males) aged seven to 13. Many over-aged females

148

Setting priorities for HIV/AIDS interventions

also attend primary school. For example, in 2001, most pupils (85 percent) were over the age of 13 in the top grade (grade 7). We will assume that schooling typically spans the years eight to 14 to ensure that these over-age females are being represented. HIV infection will be assumed to take place in the 15th year if the female had not attended primary school. For ten years, for years 15 to 24, HIV infection will exist, but not affect whether the female works or not. Without ARVs, the uneducated female is assumed to die at the age of 25 and it is from this year that foregone earnings will start to accrue. Life expectancy in Tanzania is around 50 years. So it is the earnings in years 25 to 50 that are lost for the uneducated female and would be saved if she had attended primary school. 2. Average yearly earnings for females were estimated to be 239 880 Tanzanian shillings (or around US$400). This figure can be expected to increase in line with the recent growth rate in the economy as a whole (which was 3.1 percent for the years 1992 to 2000). So the earnings figures rise over time in column 2. The 26 years of earnings from years 25 to 50 amount to just over 13 million TZSH (or US$21 863). 3. Because a dollar in the future is worth less than a dollar today (as, for example, a dollar today can be put in the bank and earn interest) any future dollars are worth less than a dollar today. How much less depends on the interest rate. We will be considering a 3 percent interest rate because this is the rate that is the recommended rate for use with health care evaluations. A dollar today at an interest rate of 3 percent would be worth $1.03 next year. So $1.03 when discounted at a rate of 3 percent (which means that it is divided by 1.03) is worth $1 today. Alternatively we can say that the future value is multiplied by 1 over 1.03, that is, 0.97, to find the worth today, in which case 0.97 is called the “discount factor” for next year’s values. The conversion of future values to today value terms (when the evaluation of the intervention is taking place) is called finding the “present value”. Because the earnings in year 25 are 17 years into the future from the year that primary school education began, there would be 17 years of discounting at 1.03 to find the present value and so the discount factor for $1 that is 17 years into the future would be 0.605. The discount factors are recorded in column 3 of Table 32.1. Each year’s discount factor is 0.97 of the previous year’s discount factor. This discounting continues to year 50 where the discount factor is 0.289. 4. Multiplying each year’s earnings by each year’s discount factor produces the present values of earnings. Total discounted earnings are equal to 5.3 million TZSH (or US$8907). To conclude: the value of a

Human capital theory

149

life saved for a female provided with primary education in Tanzania is $8907 using the human capital approach.

THE STRENGTHS AND WEAKNESSES OF THE HUMAN CAPITAL APPROACH The main (and perhaps only) advantage of the human capital approach is that data on earnings is usually readily available and so it is very easy to use and interpret. So much so that legal systems throughout the world use the approach to fix compensation for injuries and deaths. If a person is attacked and is off work for six months, then six months’ earnings are awarded in compensation (in addition to the sums for pain and suffering). Similarly, if the government’s negligence has led to the death of a spouse whose present value of lifetime earning would have been $3 million, then this is the amount of damages that the remaining spouse can expect if he or she successfully sues the government for the negligence. Most non-economists and economists are against using the human capital approach to measure the benefits of life-saving interventions. The main complaint by non-economists is that the approach violates social justice or equity. A millionaire’s life would be valued in the millions while the life of a homeless person without earnings would be worth nothing. Similarly, women (who earn less than men) and the elderly (who have fewer years left to work or are retired) would be undervalued if all one looked at is their lifetime earnings. People in developing countries have much less physical capital to accompany their labors. So their productivity and hence their earnings are bound to be low. Many of these issues can be seen in the measurement of the value of saving a female’s life in Tanzania via an investment in primary education that we just examined. A female’s life was valued at only $8907. Many US citizens earn more than this in a day, never mind a lifetime. While there has been an attempt to measure human capital in terms of time rather than earnings, see Brent (1991), and this would eliminate a number of the equity issues, this new way of valuing a life has not been widely adopted by evaluators. The biggest problem with the human capital approach from an economist’s perspective is that it has nothing to do with the fundamental willingness to pay foundations of CBA. The preference of the person whose life is being saved and evaluated is not being considered in any way. At best the human capital approach is a measure of an external benefit, that is, the effect on others if a person’s life is lost or saved.

33. Human capital practice: the benefits of female primary education There are two main types of education intervention. The first type basically involves providing information about HIV/AIDS and its transmission, which people can use to their advantage. This was the nature of the application covered in Chapter 23 in connection with threshold analysis. The information program could be disseminated out in the community or in a school setting. In a school environment the education program could be run by a teacher or by a peer of the student body. The second type of education intervention takes the form of a general basic education and need not specifically have any course or instruction that relates to HIV/ AIDS. This chapter is concerned with explaining how one goes about evaluating this second type of education intervention. It is this type of program that the World Bank (2002, p. xvii) thinks is one of the “strongest weapons against the HIV/AIDS epidemic”. The reasons why the World Bank thought this way were summarized in Chapter 14. Here we give the details of the evaluation of the provision of female primary education in Tanzania by Brent (2009d), which we introduced in the last chapter.

FEMALE EDUCATION AND ITS EFFECTIVENESS IN REDUCING HIV/AIDS IN TANZANIA As we have stressed a number of times, there is no point in spending time evaluating an intervention that is ineffective, as it would never pass a cost– benefit test. There would be costs and no benefits to show for the costs. As we saw in Chapter 14, there is at this time no general presumption that HIV rates will go down when female education goes up. Effectiveness first has to be established. As was also mentioned in Chapter 14, in the case of Tanzania, effectiveness was found using statistical techniques related to 20 regions over seven years. We will just report the results as they appear in the best estimates. The direct effect of educating a cohort of 20 507 females was, consistent with most of the literature, to raise the number of HIV cases, in this case by 463. But, there was also, simultaneously, an indirect effect that meant that, as the cohort of females became educated, 150

Human capital practice

151

the incomes to them and others would rise, and this income effect lowered the number of HIV cases by 1879. The net effect of female primary education was the difference between the direct and indirect effects, that is, a reduction of 1408 HIV cases.

THE CBA OF FEMALE PRIMARY EDUCATION IN TANZANIA From the effectiveness part of the study we have just seen that, if 20 507 females were provided with primary schooling in Tanzania, then around 1408 HIV cases would be saved. So the CBA in its simplest terms was one of seeing whether the benefits of saving 1408 lives were worth the expense of educating 20 507 females. In the previous chapter, we sketched out the timing of the expenses and the timing of the benefits using the human capital approach that entailed measuring benefits by the present value of lifetime earnings. To summarize the time profile of benefits and costs: ● ● ●

A female goes to school for years 8 to 14. If infected, she is symptom free for ten years and earns income until 25 when she dies. If uninfected (because she is educated), she earns income for 26 years (from years 25 to 50) and then she dies.

The present discounted value of the 26 years of earnings was already calculated in Table 32.1 in Chapter 32 to be 5.3 million TZSH (or US$8907). With each of the 1408 lives saved being valued at 5.3 million TZSH, the total benefits were 7522 million TZSH (or US$47 million). What we now have to explain is how the costs of schooling were calculated. Table 33.1 gives the details. The primary schooling years took place in years 8 to 14. The Tanzanian government paid for most of it. The government cost rose over time. Expenses incurred by the females (or their families) were mainly school books and uniforms and these were assumed to be relatively constant. The total cost per year is shown in column 4 of Table 33.1. When we multiply these figures by the discount factors, which like the benefits were based on a discount rate of 3 percent, we obtain the present value amounts appearing in the final column of the table. The sum of these present value costs were 127 743 TZSH. Since 20 507 females was the cohort size that was the basis for the generation of the benefits, the total costs for the cohort was 20 507 times 127 743 TZSH, that is, 2620 million TZSH (or US$16.38 million). With total benefits of 7522 million TZSH, and total costs of 2620 million

152

Table 33.1

Year 8 9 10 11 12 13 14 Totals

Setting priorities for HIV/AIDS interventions

The cost per student of seven years of primary education in Tanzania 1994/95 to 2000/01 (in TZSH) Government Cost

Private Cost

Total Cost

Discount Factor

Present Value

12 830 13 215 15 902 17 039 18 983 21 710 31 290 130 969

1 512 1 512 1 512 1 512 1 512 1 512 1 512 10 584

14 342 14 727 17 414 18 551 20 495 23 222 32 802 141 553

1 0.971 0.943 0.915 0.888 0.863 0.837

14 342 14 298 16 414 16 977 18 210 20 032 27 471 127 743

Source: Created by the author based on Brent (2009d).

TZSH, the net benefits of providing primary education to the cohort of females were 4902 million TZSH (or US$30.6 million). The Tanzanian primary education project was clearly worthwhile with benefits almost three times as large as the costs.

SUMMARY AND CONCLUSIONS From the CBA methods point of view, what is noteworthy about the application of the human capital approach to Tanzanian primary education was the result that the project was found to be very worthwhile even using the very conservative, and hence very controversial, estimate of a female life to be valued only at $8907. This result is not hard to explain because although earnings (and hence the value of life using the human capital approach) are very low in a developing country, the costs of schooling are also expressed in local value terms and therefore very low. The cost of seven years of primary education in Tanzania was only about $213. So it is the difference between the scaled down benefits and the scaled down costs that determines evaluation net benefits for developing countries, not the size of benefits or costs on their own. This means that the desirability of a health care intervention is not necessarily predetermined by the benefit methodology one adopts for making the evaluation. The human capital approach will not automatically rule out every project that affects the poorly paid in society. Notwithstanding this caveat, the human capital approach is not best practice for valuing a life in CBA and it to this end that we next turn our attention.

34.

Value of a statistical life theory

The human capital approach is the most often used method to value a life in health care evaluations. But as we have just explained in the last chapter, it is not best CBA practice. Individual preferences are not being considered. If anything, it is the foregone output to the rest of society that is being recorded, not the value to the individual him or herself. In the value of a statistical life (VSL) approach, it is the preferences of individuals regarding the amount of money that is required to compensate them for putting up with reduced safety that is the relevant valuation. This fits in well with the WTP base of CBA as the individual is choosing to pay for something (usually a job) that has a lower risk of a fatality.

THE VALUE OF A STATISTICAL LIFE APPROACH As first argued by Schelling (1968), the wrong question to ask in a CBA is: how much are you willing to accept to compensate you for losing your life? The most common response would be that there would be no finite amount that would adequately compensate. After all, if you are dead, what can one do with the money (other than give it to a friend or family member)? Rather, the question to ask is: how much are you willing to accept for a specified risk of losing your life? If the risk is 0.001 and you would accept $1000, then you are implicitly valuing your life at $1 million. That is, you would need to receive 1000 times more than $1000 to accept a risk that is 1000 times larger than 0.001. Obviously, a risk 1000 times larger than 0.001 accumulates to a risk of 1 that a person would die. The risk of 1 is a statistical death not an actual death. By accepting $1000 for the 0.001 risk of dying you would, of course, hope that you would not actually die. The $1 million dollar valuation is the statistical implication of the fact that if 1000 people accepted a risk of 0.001 of dying, then on average there would be one death that would result from the collective risk-taking. It is in this sense that the valuation can be called the “value of a statistical life”. Government decision-making nearly always involves considering small risks that affect a large number of persons (for example, operations for hip replacements) rather than a certain death for a known

153

154

Setting priorities for HIV/AIDS interventions

Table 34.1

US occupational fatality rates by industry, 1992–95 national averages

Industry

Fatality Rate per 100 000 Workers

Agriculture, Forestry & Fisheries Mining Construction Manufacturing Transportation & Utilities Wholesale Trade Retail Trade Finance, Insurance & Real Estate Services Source:

17.0 24.5 12.8 3.6 10.4 3.5 2.8 1.1 1.5

Based on Viscusi and Aldy (2003) Table 1.

person (which would happen if there were, say, one donor heart available and two potential heart recipients and the government had to choose the one person to give the heart to). There are few studies that focus on the VSL in the context of HIV/ AIDS risks and even fewer related to SSA. Most research has involved risks regarding the choice of occupation in the United States and we will examine a typical study in this area. Occupational choice is assumed to involve comparing industries with different risks of fatalities against the extra wages that one can obtain in the riskier industries. Let us look at data issued by the National Traumatic Occupational Fatality Project on US fatality rates by industry for 1992–95 as shown in Table 34.1: The two points to note in Table 34.1 is that there are sizeable differences in fatalities by industry (the rate in Mining is over 22 times that in Finance, Insurance & Real Estate) and the fatality rates are small in absolute size even in Mining (of the order of 0.000245). Given the fatality rate differences by industry, by how much did the wages paid to workers to be employed in risky industries vary? According to a study by Moore and Viscusi (1988), the typical worker required an extra annual compensation of $43.4 to accept a job in an industry with an additional 1 in 100 000 chance of dying during the year. If 1/100 000 equals $43.4, then multiplying both sides by 100 000 means that a chance of dying of unity (that is, it would be certain) requires compensation of $43.4 times 100 000, equivalent to $4.34 million. This would be an estimate of the VSL in the United States involved with the choice of industrial occupation.

Value of a statistical life theory

Table 34.2

Estimates of the value of a statistical life in various countries

Countries

South Korea (1985) Taiwan (1985) New Zealand Canada Australia United Kingdom France Austria Sweden United States Denmark Switzerland Japan Source:

155

1997 Gross Domestic Product (US dollars per capita)

Mean Value of a Statistical Life (VSL) (millions of US dollars)

2 630 5 901 15 100 19 225 20 316 20 831 22 795 24 481 24 670 28 206 30 834 34 397 36 399

0.620 0.956 1.625 3.518 2.126 2.281 3.435 3.253 3.106 3.472 3.764 7.525 8.280

Based on Miller (2000) Table 2 and an abridged version of Table 5.

The Moore and Viscusi figure of $4.34 million was just one of the 39 VSL estimates that Miller (2000) found for the United States and thought suitable by him for further examination. The average of the 39 studies for the United States was around $3.47 million. There were 12 other countries that had undertaken VSL studies. All of the non-US countries had much fewer studies in which to extract a typical value. For many of the 12 countries there was just one study per country. In total there were 69 studies with VSL estimates in the 13 countries. The average VSL estimate for each of the 13 countries is presented in Table 34.2. The table also shows the per capita income for each country. Because only 13 countries have VSL estimates, how can the approach be used in countries not represented in Table 34.2? Note that the countries in Table 34.2 are sorted into ascending order by income per capita. It is clear that, on the whole, the higher a country’s income per capita, the higher is the VSL estimate. The general relationship between income and VSL was estimated by Miller based on 68 of the estimates and this took the form: VSL = 136.7 × GDP per capita. Thus, as long as one knows a country’s GDP per capita, one can generate an estimate of the VSL for any country by multiplying by 136.7.

156

Setting priorities for HIV/AIDS interventions

THE STRENGTHS AND WEAKNESSES OF THE VALUE OF A STATISTICAL LIFE APPROACH Individuals make risky decisions all the time. But, they do so for a reason. They expect to gain something in return for taking the risk. This applies no matter the particular risk one is considering, whether it be the choice of job, the choice of mode of travel to take to work (bus, train or car), or the choice of whom to have sex with and whether to use a condom or not. How large the gain has to be to compensate for the risk is a matter of individual preferences. So if one is carrying out a CBA based on individual preferences, risk preferences need to be incorporated just like any other values related to project inputs and outputs. Thus, generating a value of a life that is derived from these risk-return preferences is best practice for CBA. CBA would break down if one were to ask how much the life of an identifiable individual (like you or a family member) is worth. But, if the unit one is trying to value is an unknown person of the general public who is on average likely to die in the course of carrying out advantageous activities that inevitably incur some risk of fatality, that is, one is evaluating a “statistical life”, then meaningful values can be obtained that are less controversial. Having said that, there are a number of problems in applying the VSL approach to evaluate a particular intervention – see Viscusi and Aldy (2003) for a full list. Some of the difficulties are: ●

●

●

●

We know that if a person has to consider a situation where they are certain to lose their life, that is, the probability of dying is 1, then a finite dollar evaluation will not be forthcoming. So only “low” risk situations are feasible using the VSL approach. In Table 34.1, risk rates were expressed in units of 1 in 100 000. Different risk units have been shown to give different VSL estimates. To obtain a VSL of $4.64 million for an individual that values a 1 in a 100 000 chance of dying at $4.64, one simply multiplies the $4.64 amount by 100 000. This assumes linearity. People often value risk non-linearly (so a rise of risk by 1 in 100 000 from a starting base of 0.5 would lead to a lower compensating wage adjustment than when the starting probability is 0.6). What happens if the individual is not well informed and he or she thinks that the risk level in mining is not 24.5, but instead thinks it is much lower? Do we use their actual probabilities or their subjective probabilities to estimate the VSL? Can individuals understand fully the meaning of very low probabilities? If they think that a probability of 0.00001 is really no different

Value of a statistical life theory

157

from 0 (that is, 0.00000), then the VSL estimate would be an unacceptable amount, that is, zero dollars. Other problems mainly stem from the fact that (1) all of the estimates existing in the literature that Miller considered relate to developed countries and none to developing countries and (2) many of the estimates comes from wage-risk choices and none of them relate to risks related to HIV/ AIDS. So if one tries to take VSL values found in developed countries and apply them to HIV/AIDS interventions in developing countries, one will be ignoring the fact that perceptions of risk depend on context and culture. Mining may have a risk of 24.5 per 100 000 workers in the United States, but the figure could be radically higher in Africa. Moreover, what relevance would a low mining risk fatality have for a high-risk activity like sharing needles among IDUs? As Grüne-Yanoff (ND, p. 7) points out, individuals “evaluate fatality risk according to the kind of death this risk may bring about: whether it occurs through illness or accident, whether it is slow or quick, consciously experienced or not”. This point is very relevant to understanding why attempts to apply preferences for work-related deaths, which usually involve sudden accidents, would not be appropriate for preferences for AIDS deaths that are sometimes not strictly accidents and are long drawn-out events. We have come across the law of diminishing marginal utility a number of times. Applied to income, it means that as one’s income increases, the less is the additional satisfaction from having more of it. The relevance of this principle to estimating the VSL is that it is obtained from preferences to trade off higher risk for higher income. If the rich value additional income lower, they will require greater amounts of income to compensate them for the higher risk. So one should expect that the estimated VSL would have to increase with the income level of the person whose life one is valuing. This is clearly seen in Miller’s equation: VSL = 136.7 × GDP per capita. Thus, because people in developing countries have low per capita incomes, their lives will be valued low using the VSL approach. The conclusion therefore is that, while it is good that the VSL approach fits in well with the WTP base behind CBA, the approach will inevitably share the equity drawback of WTP – that it is going to be greatly affected by the income level of the preferences for whom it is used to estimate the benefits.

35.

Value of a statistical life practice: the benefits of VCT

In this chapter we will explain how the VSL approach has been used to value VCT services in Tanzania as covered in Brent (2009a). In that country, at the time of the evaluation, VCT services were available to a very small percentage of the population. One issue then, after carrying out a CBA of an existing program, is how to evaluate VCT programs if they were to be scaled up to the population as a whole. As we have seen in Part I, scaling up HIV/AIDS interventions is a main objective of UNAIDS and the World Bank. So we will be examining results for both existing and scaled up VCT programs in Tanzania. Because evaluating VCT completes our study of alternative methods for evaluating CBA interventions for HIV/AIDS, and as a kind of conclusion to Part III of this book, we will compare and contrast the VSL results for VCT with those using a competing benefit methodology, that is, the human capital (HC) approach. As usual, we begin the outline of a CBA application with a discussion of the effectiveness of the intervention.

ESTIMATING VCT EFFECTIVENESS We will start with a framework that provides estimates of the number of lives saved if we assume that VCT is effective. Then we refer to evidence supporting this assumption for Tanzania. VCT services involve testing, and then counseling on the basis of the test results, to change behavior. The effectiveness of the services depends crucially on the existence of discordant couples – couples where one partner is HIV positive and the other is HIV negative. If both partners are HIV negative then behavior does not need to change. Similarly, if both are HIV positive then it is not so essential that behavior changes (though this is not always the case as people can get reinfected with different strains). When one partner is HIV positive and the other is HIV negative, then risky behavior has to change to prevent the positive person infecting the negative one. The existence of a discordant couple is a joint event: one must be HIV 158

Value of a statistical life practice

159

positive and the other HIV negative. The probability of there being one individual in a discordant couple is therefore the probability of this joint event, which in the simplest case is the product of two probabilities, that is, the probability of one being infected P and the probability of the other person not being infected 1– P. That is, the joint probability is: P (1 – P). If, as in the Tanzanian VCT program between 1997 and 2001, the percentage tested HIV was 70.8 percent, this would fix P = 0.708 and 1 – P as 0.292, so the joint probability was 0.2067. In a program that tested a group of 3586 individuals, there would be 0.2067 × 3586 people – equal to 741.4 – who could be infected without behavior change if, say, condoms were not used during sexual relations. Therefore, with behavior change following the counseling part of VCT, there would be 741.4 lives saved by the intervention. This number assumes that the VCT program involves an individual being tested and counseled on his or her own. If there were dual testing, then both the individual and the partner would be changing their behavior in the event that one of them was HIV positive and the other was HIV negative. So there would be twice as many lives saved under a dual testing VCT program (to accompany the twice as much testing costs). In this case 1482.8 lives would be saved. The existing VCT program in Tanzania at the time served very few individuals. The ones who were served were more likely to be HIV positive, as the main reason why they got tested in the first place was that they were involved in risky sexual behavior and they wanted to confirm their HIV status. This was why the HIV positive rate was so high with P = 0.708. A scaled up program would be serving the average Tanzanian and not just the high-risk groups. Since the HIV prevalence rate in Tanzania at the time was 7 percent, the appropriate probability for a typical Tanzanian in a VCT program extended to the general population would be P = 0.07 and not P = 0.708. Hence the joint probability of a couple being discordant with this lower probability would be 0.065. This means that 0.065 times the 3586 people being tested would result in 233.4 lives saved in a scaled up, separate VCT program, much less than the 741.4 lives saved under the existing program. A scaled up dual program would again save twice as many lives, that is, 466.9. To summarize: the existing VCT program saved 741 lives if it was a separate testing program and 1483 if testing was dual. A scaled up separate testing program would generate 233 lives saved and this would rise to 467 under a dual testing program. These numbers were estimates that were derived from the P (1 – P) formula. These numbers were not unrealistically high because both the Voluntary HIV-1 Counseling and Testing Efficacy Study Group (2002) and Brent (2009a) found empirically that, for Tanzania, VCT was highly effective.

160

Setting priorities for HIV/AIDS interventions

THE CBA OF VCT SERVICES IN TANZANIA The evaluation of VCT services in Tanzania involves taking the estimates of the number of lives saved from the previous section and multiplying them by the value of a life saved to form the benefits. Then one has to estimate the costs to produce the net benefits. We explain these steps in turn. All the cost–benefit ingredients and results are shown in Tables 35.1 and 35.2. Table 35.1 relates to the case where the VSL method was used to measure the benefits and Table 35.2 is the counterpart with the HC approach used to measure the benefits. (i)

The Benefits

The VSL method explained in the last chapter came up with an estimation equation for the value of a life that was dependent on the national income of the country (GDP per capita). The equation was: VSL = 136.7 × GDP per capita. The 1998 figure for GDP was 170 844 TZSH, leading to a VSL of 23.354 million TZSH. Multiplying this amount for the VSL by the number of lives saved for the various VCT programs produces the total benefit figures given in Table 35.1. Following work by Haacker (2006), an estimate of the value of a life using the human capital (HC) approach can also be obtained from an equation dependent on a country’s national income. This relation for Tanzania is (which assumes that the life one is saving has 24 years remaining): HC = 24 × GDP per capita. With the same 1998 GDP figure, the HC estimate of the value of a life is 4.1 million TZSH. Multiplying this figure by the number of lives saved for the various VCT programs gives the total benefit figures in Table 35.2. (ii)

The Costs

There are two costs associated with VCT services. The first one relates to the costs of testing and counseling. Obviously if one expects to obtain behavior change after testing, one must expect to pay for counseling to accompany the testing. The cost per client in the program (whether they accepted to be tested or not) was estimated to be 17 358 TZSH. With 5535 clients in the program, the testing and counseling costs were 96.1 million TZSH. These costs did not vary by VCT program except that they would be double the amount for dual testing. The second cost was the foregone benefit of unprotected sex. If one is now to use a condom with one’s partner, the loss of satisfaction from having sex without a condom must be factored in. An estimate of the

161

Source:

96 m 192 m 96 m 192 m

16 796 m 9 809 m 1 661 m 3 089 m

Foregone Benefits

Based on Brent (2009a) Tables 2 and 3.

Existing, separate Existing, dual Scaled up, separate Scaled up, dual

VCT Costs 16 893 m 10 001 m 1 757 m 3 281 m

Total Costs 741 1 483 233 467

Number of Lives Saved 23.4 m 23.4 m 23.4 m 23.4 m

Benefits per Person

17 314 m 34 628 m 5 452 m 10 904 m

Total Benefits

1.03 3.46 3.10 3.32

Benefit– Cost Ratio

Cost–benefit outcomes for VCT testing using the VSL method for estimating benefits (m = millions of TZSH)

VCT Program

Table 35.1

162

Source:

96 m 192 m 96 m 192 m

Based on Brent (2009a) Tables 2 and 3.

Existing, separate Existing, dual Scaled up, separate Scaled up, dual

VCT Costs 16 796 m 9 809 m 1 661 m 3 089 m

Foregone Benefits 16 893 m 10 001 m 1 757 m 3 281 m

Total Costs 741 1 483 233 467

Number of Lives Saved

4.1 m 4.1 m 4.1 m 4.1 m

Benefits per Person

3 040 m 6 080 m 957 m 1 914 m

Total Benefits

0.18 0.61 0.54 0.58

Benefit–Cost Ratio

Cost–benefit outcomes for VCT testing using the HC method for estimating benefits (m = millions of TZSH)

VCT Program

Table 35.2

Value of a statistical life practice

163

value of unprotected could be deduced from the differential price charged by commercial sex workers for protected and unprotected sex, which was estimated to be 3627 TZSH. For a 24-year remaining lifetime (and an average of 28 sexual acts per year) this cost becomes 3.3 million TZSH. This sum is then multiplied by the number of people who are expected to be changing their behavior, which varies by the particular VCT program, and this leads to the foregone benefits figures in Tables 35.1 and 35.2. The total costs in these tables are then the sum of the two costs (the VCT costs and the foregone benefits). (iii)

The Benefit–Cost Results

On the basis of the figures in Table 35.1, which uses the VSL method to measure benefits, we can see that existing separate VCT services in Tanzania were only marginally worthwhile. Benefits were just 3 percent larger than the costs (that is, the cost–benefit ratio was 1.03). However, for dual programs, benefits were over three times larger than the costs. Note that foregone benefits are by far the larger element in the costs of VCT. This category of cost was estimated in this study for the first time. Its importance is not just in its absolute size, but how it varies with the VCT program considered. It is because foregone benefits were so much lower in the existing dual program that this version was so much more worthwhile than separate testing. The reason why foregone benefits were so much lower can be easily understood. When a person gets tested separately and is found to be HIV positive, that person will feel obliged to use condoms with his or her partner since the HIV status of the partner is unknown without that person being tested. With dual testing, partners need to give up unprotected sex only if they are part of a discordant couple, that is, the partner has a different test result. Scaled up programs were always highly worthwhile with benefits over three times the costs. The difference between separate and dual programs disappears, in contrast to the evaluation of existing programs. Again, the main explanation lies with the foregone benefits category. When, probabilities related to the average person are used (instead of just the high-risk individual) there are fewer discordant couples and thus less need to give up unprotected sex. When we replace the VSL estimate of benefits with the HC valuation method, which is the only difference between Tables 35.1 and 35.2, we see that no VCT program is worthwhile. Benefit–cost ratios are always less than 1, that is, benefits are less than costs and so net benefits are negative. These results are simply a product of the fact that the formula used to estimate the value of a life as a function of a country’s per capita GDP had a multiplier that was five times larger when applied to the VSL method than

164

Setting priorities for HIV/AIDS interventions

it did when used with the HC approach. Hence if benefits are five times larger using the VSL method, then the benefit–cost ratios also will be five times larger.

SUMMARY AND CONCLUSIONS We now summarize results and issues with all of Part III in mind. The main points are: ●

●

●

●

●

●

Only CBA can determine priorities. There are no real alternatives. Some proposed alternatives, such as cost minimization and costeffectiveness analysis, are, at best, partial and incomplete CBA methods. Since there is only CBA, it is up to the reader to rely on the method for valuing the benefits that best reflects his/her values, preferences and understanding of how the world operates. There are a number of different methods. CBA is not restricted to using just one benefit method. Having said that, some methods are better than others. This is important because different methods may give opposing outcomes. We saw that in the evaluation of VCT services in Tanzania. The services were worthwhile using the VSL method and not worthwhile using the HC method. What makes one method of estimating benefits better than another is whether it relies on individual preferences or not. WTP and the VSL approaches are best practice because they are based on these preferences. Thus, for the VCT evaluation for Tanzania one can conclude that these services are, in fact, worthwhile because the VSL results showed this to be the case. When considering evaluations, it is important to ensure that alternative versions of the program are being considered. It is not the case that, for example, every VCT program is going to be highly worthwhile. Altering how a program is to take place (for example, using dual rather separate testing) will affect the results. Similarly, when considering evaluations, one should also ensure that different scales of operations are being considered. Again, it is not the case that if one scale is worthwhile then another scale will be equally worthwhile. With VCT services in Tanzania we saw that existing programs were marginally worthwhile, but scaled up programs were highly worthwhile.

PART IV

Social considerations in CBA

36.

Introduction to Part IV

By now it should be fully clear why the setting of MDGs was not very helpful. It is true that seeking to halve the spread of HIV/AIDS by 2015 does at least try to ensure that that there is some accountability for all the monies that have been devoted to this disease. In this way the effectiveness of interventions would be an issue. There cannot be no improvement. But, why a 50 percent improvement should be feasible when no mention is made of how the improvement is to be achieved makes no sense. Moreover, we need to know why a 50 percent improvement is better than any other rate of improvement. CBA on the other hand, would enable us to answer all the relevant questions, not only about whether particular interventions are effective or not, but also about whether any or all of them are worthwhile. We add up all the worthwhile interventions so that we can find what the resulting level of improvement can be and then determine what the total necessary expenditures should be. Part III covered some of the necessary ingredients of CBA. Many of the technical details were omitted, especially a discussion of the statistical methods that need to be employed to analyze the data so that they can estimate the inputs, outputs, effects, benefits and costs. The aim was not to prepare the reader to carry out a CBA for him/herself. Rather the object was to show the many ways that a CBA evaluates an HIV/AIDS intervention, so that the reader can appreciate what CBA has to offer for the setting of priorities to deal with this disease. In this introduction to the last part of the book we will summarize the main points from Part III and present an outline to Part IV, which is devoted to explaining what is so “social” about an economist’s cost–benefit calculations.

WHAT HAS BEEN LEARNED FROM PART III There are three main themes that were covered in Part III that need to be highlighted so that one has the necessary base to understand the arguments that will be made in Part IV:

167

168

(i)

Setting priorities for HIV/AIDS interventions

There is More than One Way to Carry Out a CBA

Since measuring outcomes in monetary terms to form the benefits is the distinctive part of CBA, as it separates the discipline from cost minimization and cost-effectiveness analysis, Part III concentrated on presenting many different methods for measuring the benefits of HIV/AIDS interventions. Many think that CBA relies solely on valuing lives and other effects of health care interventions by a person’s earnings (that is, the human capital approach). In the health care evaluation field this was the main valuation method and this approach has its uses. But, outside the health care field the human capital approach is not best CBA practice and this needs to be understood. The willingness to pay approach is best practice in CBA generally, whether valuing inputs to form the costs or valuing the outputs to determine the benefits. The main purpose of Part IV is to explain why it is that using WTP is considered to be best practice. Even though WTP is best practice, nonetheless the reader was presented in Part III with a selection of alternative methods for valuing benefits and these various methods had applications, so that the reader could fully appreciate the implications of using each method. CBA is the only way to determine whether a project is worthwhile. But, the reader should be aware that there exists a whole range of possible ways that have been used to carry out CBAs of HIV/AIDS interventions (and many others in the CBA field that have not yet been applied to interventions for this particular disease). Exposed to a range of options and their implications, the reader should be able to adopt a CBA method whose assumptions he or she feels comfortable with. (ii)

The Details of Interventions are Important

Apart from giving the reader an appreciation of the implications of what it means to adopt a particular benefit methodology, the purpose of pairing the various approaches with actual applications was to make the point that details matter to CBA outcomes. Some of the important details are how an intervention is implemented (for example, on what scale it is to be carried out), while other details relate to local circumstances that impact projects (such as the wage rates that exist in a particular area). This point that the details are important in determining outcomes must be realized whenever it is suggested that, because an intervention was worthwhile on a particular scale in a specific country, that therefore the project must be worthwhile when implemented on a different scale in some other country. The outcomes depend on the particular numbers; if the numbers are different, the outcomes could be different. It is as simple as that. As obvious

Introduction to Part IV

169

as this point is, it has not been understood by the international agencies making HIV/AIDS decisions who are determined to scale up effective programs whether they are worthwhile or not. To see how details mattered let us look back at the evaluation of the condom social marketing program in Tanzania in Chapter 27. The program involved charging a price for a pack of condoms that was one-third of the costs. At this price, social benefits were about 60 percent higher than the costs (the cost–benefit ratio was 1.59). If the price charged had not been so low, and raised equal to the costs, then output would have been lower and hence benefits would have been smaller. But, with output lower, costs would also have been much reduced. So the social benefits would have been over three times the size of the costs (the cost–benefit ratio would have been 3.35) with the higher-priced program, making the smaller-scale project much more socially worthwhile than the actual program that was in place. The existing condom social market program was itself just one variant of a whole host of possible condom promotion schemes. One alternative, that existed in Tanzania and was run by the Ministry of Health, supplied condoms free of charge. Zero pricing is every non-economist’s delight. Why not everyone’s delight? This is because the details of a program are important. The free condoms were supplied at government clinics that operated during the daylight hours. People going to the daytime program would be highly visible to all and thus open to stigma. People could ask: “Why was so-and-so going to get free condoms when he or she is supposed to be happily married?” Instead of being subject to such stigma, people could go to the condom social marketing program. This did not provide condoms free, but did supply the condoms in bars and clubs where customers could go late at night under the cover of dark. Half of the total condoms used in Tanzania were obtained in this anonymous fashion that mitigated the effect of stigma – see Brent (2009e). Clearly the price charged for condoms, and where and when they can be bought, are important details that would affect the outcomes of any condom promotion program. (iii)

CEA is Not Useful when Budgets are Variable

Most health care evaluations, and this includes those involving HIV, are carried out as cost-effectiveness analyses. As we explained in Part III, CEA cannot always be relied on to determine whether projects are worthwhile. CEA requires a fixed budget constraint. All interventions are then to be ranked from lowest to highest according to their cost-effectiveness ratios. The decision-maker is to first approve the intervention with the lowest cost per unit of effect and go up the list until all the funds in the budget have just been exhausted. If the budget is not fixed, then all the decisions that were

170

Setting priorities for HIV/AIDS interventions

predicated on the fixed budget may have to be altered if funds are variable. With more funds, additional projects with higher cost-effectiveness ratios that were originally rejected can now be approved. With fewer funds, some of the previously approved projects would now be ruled out. The issue then is, how realistic is it to assume that there is a fixed budget constraint? Superficially, a fixed budget is what many agency departments actually do seem to experience. A department may get an annual allocation of funds and be told do the best job (that is, carry out the most costeffective projects) with these funds. But, this is often an illusion because not all the allocation of funds will be spent, and if more funds are required, the agency as a whole may be able to do some fundraising to secure added financial resources. In general, the reality is that additional funds will be found if a project is thought worthwhile and not otherwise. So it matters what the extra money is going to be spent on. A project that saves lives will often be able to attract additional funds; while a project that seeks to build a statue to honor a political leader will not be as easy to finance. It frustrates many people that large sums of additional monies have been raised for HIV/ AIDS in the last few years, especially by PEPFAR and the Global Fund, and comparable sums have not been available for other diseases (such as malaria and malnutrition) where millions of lives are also at stake. The reality though is that more funds are available for HIV/AIDS only because it is HIV/AIDS that is being targeted by the funds. Donors do not seem to be willing to make the same commitment to other public health issues. If donors think that HIV/AIDS is exceptional, then HIV/AIDS is exceptional, whether project evaluators like it or not. To conclude: CBA compares the benefits and costs of any available level of health care expenditures. It can deal with situations where funds are variable, unlike CEA where it must be assumed that funds are fixed. We suggested that budgets are usually flexible, depending on how the funds are to be spent. Note that for the last few years, funds for HIV/AIDS have been increasing so they have not in fact been fixed. But, reductions in funds can also be accommodated by CBA and not just increases. At the time of writing (summer 2009) the world is in the midst of a financial crisis. One should therefore expect that charitable giving will decrease and there will be fewer funds available for HIV/AIDS interventions.

OUTLINE OF PART IV Part IV will present some of the arguments in favor of going the cost– benefit route to make HIV/AIDS expenditure decisions. Isn’t using

Introduction to Part IV

171

economic reasoning like treating people as if they are in a meat market – commodities to be bought and sold? The next chapter will try to answer this question. It will be argued that there is no simple classification scheme that identifies “higher” from “lower” goods and services. After all, meat can keep you alive, and there can be markets for goods that save lives and not just those that can satisfy the taste buds. The next two chapters deal with the issue of what economics has to offer for an understanding of social decisions. What makes CBA “social”? The role of WTP will be highlighted and the difference between making a decision from a private rather than a social perspective will be detailed. Next comes two chapters on equity. These two chapters deal with the heart of the non-economist’s complaint of CBA that relying on WTP is inherently unfair. We will show how WTP can and has been adjusted for an individual’s ability to pay in CBA using distribution weights. Relying on this weighed WTP will be seen to be a lot fairer than trying to avoid efficiency considerations altogether by using time and other non-price methods for rationing health care resources. The final two chapters present the main conclusions of the book. First we explain how not to make health care expenditure decisions for HIV/ AIDS and then we say how one should go about setting priorities in this area. How not to make expenditure decisions is to do what actually takes place now, basically, simply rely on common sense. Part I was devoted to explaining how things are so complicated in the HIV/AIDS field that guesswork cannot be relied on. We will now build on this understanding by showing why HIV/AIDS interventions do not go about with labels on them saying: “Look at my anti-HIV label! I’m a project that was specially designed to combat the disease, so I am bound to be always worth supporting”. One should label the outcomes of the interventions “worthwhile”/“not worthwhile” and not rely just on the description of the interventions to make this judgment. How to make HIV/AIDS decisions is to rely on CBA. This uses data that reflects actual human behavior and assembles these data in a form that shows whether something is worthwhile or not.

37.

Commodification: everything is seen as a commodity to be bought and sold

Non-economists (and many health economists) take the view that they are content to use cost minimization and cost-effectiveness analysis to try to evaluate HIV/AIDS interventions, but they balk at the use of CBA for the same purposes. Using monetary values for inputs to determine costs is fine, but using values for outputs to determine benefits is not fine (in fact, it is unacceptable). In Part III, when we looked at these alternative evaluation methods, we already pointed out that putting monetary values on inputs and outputs involves the same set of ethical and social considerations. So if one accepts one set of valuations (on the costs side) one should, logically, accept the other set of valuations (on the benefits side). But, now let us examine in detail exactly what the values are that underlie the estimation of benefits in CBA. Best practice in CBA is to use willingness to pay to obtain the monetary values for outputs and inputs. When we went through the theory behind WTP we drew the analogy with hamburgers. If people actually pay $4 for a hamburger then they must value the hamburgers at least $4 or else they would not have been willing to pay that amount for them. It is at this point that you can imagine the critics of CBA saying: “That is exactly the problem with CBA. It is treating health care and HIV/AIDS interventions as if they were commodities like hamburgers! How can you use the idea of WTP to value something like a human life, or helping people who are needy?” Our answer will be that one cannot easily distinguish, and hence separate out, “higher”-level and “lower”-level goods and services. If one tries to make such a separation, this often ensures that “higher”-level goods and services get ignored from consideration in an evaluation.

172

Commodification

173

WHETHER TO TREAT HIV/AIDS INTERVENTIONS AS IF THEY WERE “COMMODITIES” The case against treating everything as if it were a commodity has been stated best by Os Guinness (2001, p. 67): Not everything can or should be given a market price. The sign of a good society is the level and number of things that are acknowledged to be beyond market values – and thus appreciated for their own sake and not for extrinsic, especially financial rewards. The line between “For Sale” and “Not for Sale” is a key indication of a nation’s values.

One can agree with the general sentiments expressed by Guinness’s words. Unfortunately they are not as helpful as they seem from the point of view of carrying out a health care evaluation. There are three main difficulties that need to be recognized. First, the idea that there are two types of goods: one type that is really special and above the need to value them in monetary terms, and another type that is in some way ordinary and therefore alright to allocate a price, is not descriptively true of how human beings run their lives. Individuals are willing to pay, and actually do pay, for almost everything. Art and culture are not immune from monetary evaluation. Most museums, opera houses, art galleries, zoos and botanical gardens experience no philosophical difficulties in charging an admission fee for entry. In the process individuals have to decide whether seeing nature, a work of art, a virtuoso performance, or an endangered species, are all worth paying for relative to (yes!) buying a hamburger. And we have seen in Part III, that individuals are even willing to trade off a chance of losing their own lives as long as the monetary gain is large enough. In fact, trying to claim that certain parts of an intervention are above monetary valuations is a position that turns out to be counterproductive. If one treats providing health care, peace and quiet, preserving culture and the environment as activities or states of the world that should not be priced, this actually ensures that they do not play a part in the formal evaluation. One quantifies and values all the tangible items and excludes, and hence ignores, the intangible items as they are “difficult” to value. If one regards something as “priceless”, evaluators take this much too seriously and literally give it “a zero price”! It is better for an evaluator to give his/ her best shot at valuing something, no matter how imperfectly, rather than to exclude it entirely from consideration. One needs to act as if everything has a price in order to ensure that everything gets included. Second, giving a value to something is indeed a sign of appreciating it for its “own sake”. Take the case of housework. In many cases throughout

174

Setting priorities for HIV/AIDS interventions

the world, housework is undertaken by women and no explicit remuneration is given. This leads to the contribution of women to society and the economy being undervalued. For example, home-making services are not usually recorded as a part of national income (GDP). One reason why macroeconomists have underestimated the impact of HIV/AIDS on countries in Africa is that the epidemic has killed many women and their work is outside the monetary sphere. On the other hand, it is standard in the design of a health care CBA (if not always in the practice in Africa) to measure how much time is given up, say, in visiting a doctor for testing and treatment, and to put a monetary amount on that time given up. We have seen this consideration play a role in the evaluation in Chapter 29 (see Table 29.1) where valuing time was a part of the patient costs of visiting TB clinics. These patient costs were different for different forms of public–private partnership giving treatment because their locations were different, causing travel times to vary. The point is that CBA sets out to value all inputs and these valuations are not excluded because they relate to women. In principle, a CBA would value each and every one of the contributions of a woman working at home, whether she is a cleaner, a cook, a child minder, a planner of household finances, or a counselor. Putting a price on all women’s home-making services is recognizing their contribution and ensures that the evaluator is not taking them for granted. Third, CBA is not culturally neutral; it does recognize a “nation’s values”. A cost–benefit evaluation does depend on the legal system that defines what should and should not count. Net benefits of, say, $1 million dollars can be generated by a magazine containing child pornography, or from a certain form of cancer research. From a private perspective, $1 million dollars is $1 million dollars, no matter how it is obtained. But, a social CBA would not treat the two sums equivalently. Increases in cancer research would be positively valued as this would lead to reductions in cancer cases. Increases in child pornography would not have a positive net benefit because the effect of pornography on the children would be a cost, and the benefits (WTP) of the magazine readers would be given zero social value because it is illegal. Nonetheless, because something is illegal does not mean that it does not take place. CBA needs to be involved in setting priorities for all types of intervention, whether they be increasing “goods” or reducing “bads”. For example, drug treatment programs need to be evaluated even though the taking of some drugs is illegal. A CBA would justify switching a person from the addiction to heroin to the addiction to methadone if the harm to others is lower. The relative legality of the two forms of addiction would not be decisive. The relative effectiveness and net benefits of treatments would be the crucial factor for a CBA.

Commodification

175

SUMMARY AND CONCLUSIONS These are the main points of the chapter: ●

●

●

HIV/AIDS is a behavioral disease. Certain behaviors will give a person the disease, help to transmit the disease to others and reduce the adverse effect once the disease has been detected. All these behaviors determine monetary prices and are affected by them. Interventions that try to alter behaviors work through prices. The prices of condoms, commercial sex work, clean needles, ARVs, getting tested and being counseled are elements that need to be considered in an evaluation of an intervention. CBA ensures that these considerations are included as part of the evaluation, whether in the form of the benefits or as a part of the costs. When we price something we give it value. We are trying to estimate its inherent worth. Valuing something does not “devalue” it somehow. Placing a zero price on a good or service is the surest way to devalue it. Many risky behaviors that contribute to the transmission of HIV/ AIDS are illegal to some extent, for example, injecting drugs, commercial sex work and even (in some developing countries like Tanzania) males having sex with males. When interventions (like condoms and clean needle exchanges) are promoted to change behaviors, in the process of reducing transmission, they reduce the riskiness of the behaviors and inevitably lead to an increase in the illegal activities. Although, CBA operates within an existing legal system, the extent to which illegal activities continue is not the outcome that is to be used to judge the value of an intervention. If transmissions decline and net benefits are positive, then “harm reduction” interventions are socially worthwhile even if (unintentionally) the number of lawbreakers increase. After all, if saving lives is valuable, then all lives are worth saving – those who are involved in legal activities (like the wives who subsequently have sex with their husbands who first visited CSWs and will not now be infected because condoms were used) and even those doing things to themselves that society does not approve (injecting themselves with newly provided clean needles).

38.

What is so “social” about CBA? Fundamentals of CBA

What makes CBA “social” is that it tries to specify when any intervention makes society better off. There are two main objectives that are to be used to make the determination as to whether something is a social improvement. The first is economic efficiency and the current chapter is devoted to this objective. The second is equity, and the next two chapters will examine this objective. In the process of outlining the principles behind economic efficiency, we will explain what the crucial value judgments are that need to be made to justify the methodology of CBA, that is, we will be identifying the fundamentals of CBA.

ECONOMIC EFFICIENCY AND WILLINGNESS TO PAY Mainstream economics assumes that society is individualistic. For an intervention to make society better off, one first has to make individuals better off. How does one know whether individuals are better off? One looks at their net benefits. If the willingness to pay for all the outputs that individuals receive exceeds their willingness to accept compensation for all the inputs they supply, then net benefits are positive and they are better off. To find out whether society is better off one just sums the net benefits (both positive and negative) for all the individuals to obtain aggregate net benefits. If this is positive then the intervention is worthwhile. After all interventions that have positive net benefits are undertaken, then net benefits are highest and “economic efficiency” has been realized. In other words, the idea of economic efficiency is bound up with the objective of maximizing the satisfaction (net benefits) of all individuals. The shorthand way of referring to economic efficiency is in terms of maximizing the size of the “economic pie”. The value judgment here is that “more is better than less”. In the absence of concerns about equity, economic efficiency seems straightforward as it would seem strange to advocate that one should waste resources and not put them to their best use. Why not make someone better off if one can do so and not make anyone else worse off? 176

Fundamentals of CBA

177

ECONOMIC EFFICIENCY AND CONSUMER SOVEREIGNTY People get satisfaction from receiving the output from an intervention if they are willing to pay for it. With resources (income) limited, purchasing health care goods and services precludes the purchase of other goods and services that also give satisfaction. There would be no point in people purchasing the health care if it did not make them better off. Because it is the individuals themselves who are to decide how to spend their income, and in the process make themselves better off, the most important value judgment in CBA is the assumption of “consumer sovereignty”, that is, the individual is assumed to be the best judge of his/her own welfare. In economics, outside the health care field, the assumption of consumer sovereignty is generally accepted. The main violations of this principle would be related to age. We do not ask fiveyear-old kids how much they are WTP to stay up after midnight. The parents decide for the kids that they have to be in bed fast asleep well before midnight. Nor do we always accept the preferences of the elderly when their minds seem to go and they are not able to look after themselves. Family members or paid professions are occasionally the deciders of what is best for the elderly. In between the very young and the very old, individuals are left to decide what is best for them, whether it is to join the army, who to vote for, whether or whom to marry and what career path to follow. Loosely speaking, consumer sovereignty is the democratic way. In the health care field the assumption of consumer sovereignty is strongly questioned. Does the individual really know what is best for them? And even if this is the case, does the individual want to make health decisions for him/self? We address the second question first. It is true that sometimes we would like to leave certain decisions to the doctors, for example, when to stop life support for a brain-dead relative. But, generally, who knows better than the patient how much pain he or she is in, and whether the risk of dying from an operation is worth undertaking? Most of the time the elderly are physically but not mentally impaired. The reality is that less than 5 percent of health care expenditures involve emergency situations where patients are unconscious or otherwise incapacitated to such an extent that someone else (like a surgeon in the ER room) has to decide what treatment to apply. As for the issue that individuals are not well informed about their health conditions and their alternatives, one can point out that there is nothing wrong with an individual consulting a physician and asking that person to act as an advisor as to the best course of treatment. After all, individuals

178

Setting priorities for HIV/AIDS interventions

usually do not know much about the law, yet they do not leave all their legal decisions to lawyers. One could argue that for HIV/AIDS, the lack of information to make informed decisions can no longer be an excuse for saying that individuals do not know what is best for themselves. If there is one success story of all the millions that has been spent on this disease it is education and information programs. The whole point about establishing the vast number of surveillance systems through the world is that people can know what the risks are concerning behavior that can transmit HIV/AIDS. If 60 percent of IDUs in your country are infected, the probability of getting infected from sharing a needle is not unknown. When youths are told that there are certain myths involved with the disease, for example, that a healthylooking person can still be infected, there is a purpose in mind. This is to let the teenager know that a “pretty face” can still transmit the disease to them through sexual relations. The complaint concerning HIV/AIDS interventions is not that people still do not know what is involved with certain risky behaviors. Rather, the criticism is that behavior has not changed enough in face of this information to prevent the pandemic. It would seem that this points to a glaring weakness of assuming consumer sovereignty. But, the reality is exactly the opposite. Western donors have spent over two decades deciding what is best for Africans concerning HIV/AIDS transmission. Africans must be abstinent, delay first sexual intercourse, have a single partner, use condoms every time they have sex (no matter the partner, except the spouse), never share needles, and so on. In the year 2007, there were still 1.9 million new infections in SSA. Telling people what to do has not worked. Now is the time to find out the preferences and WTP of those on whose behavior the future of the epidemic depends. Even as of 2009 we do not know the best way to control HIV/AIDS because we have not bothered to use extensively the evaluation methodology that is built on asking what people want and how much they are willing to pay for it. In a sense, the reason why mitigation is a third intervention alternative to prevention and treatment is the recognition that consumer sovereignty cannot be ignored, no matter how hard one tries to do so. Given that we seem to be unable to stop people choosing not to abstain from sex and be faithful, and individuals are going to go to CSWs, and injecting drugs into their veins, this has set up the need for harm reduction strategies. Looking after orphans, preventing mother-to-child transmission, providing clean needles for IDUs and circumcising males are all interventions that have accepted that risky behavior will take place and that the objective now is to reduce the cost to others of this risky behavior.

Fundamentals of CBA

179

SUMMARY AND CONCLUSIONS It is a fact of life that resources are limited. It clearly does not make sense to anyone that resources be wasted. Economic efficiency is a centrepiece of any economic evaluation methodology and is not controversial. The other fundamental ingredient that goes hand in hand with economic efficiency and its reliance on WTP is consumer sovereignty. If one does accept this assumption, there is no good reason not to accept willingness to pay as a measure of benefits, and therefore no good reason not to use CBA. We have spent more than 20 years largely ignoring consumer sovereignty in devising HIV/AIDS interventions. It is now time to give consumer sovereignty a try and use the economic evaluation methodology that is based on this assumption.

39.

Social and private perspectives in CBA

The perspective in CBA is very broad and very long. It embraces the effects on everyone in society, now and in future generations. A social evaluation does not consider just the parties that are directly involved with an intervention, that is, the firms (the hospitals and the physicians) and the consumers (the patients as clients). It also covers those indirectly affected, including the family members of the patients and even the general taxpayer. These third-party effects are called “external effects”. Strictly then, CBA should be called “social” cost–benefit analysis to recognize the all-inclusive nature of the evaluation. However, this usage is not widespread either within or outside the health care field (it has the connotation of “socialism”). So we have, throughout this book, just referred to the analysis as CBA, leaving the social connotation implicit. In the health care field it is considered good practice to make the perspective explicit at the outset of a study. It is thought necessary to specify whose perspective the study is from: is it that of the hospital, the client, the government or society. Although, as we show below, an economic evaluation has usefulness from any perspective, it is only the social perspective that is important for social decision-making. A study that ignores the costs for the families of care-giving to those sick with opportunistic infections due to AIDS is not very useful for deciding on the best place to treat AIDS patients. In this chapter we will go through a particular application, HIV testing, to compare and contrast the private and social perspectives in carrying out an economic evaluation. The evaluation is based on the study by Bloom and Glied (1991).

A PRIVATE EVALUATION OF HIV TESTING A firm is interested in HIV testing mainly from the point of view that, before employing someone, it can screen them out if they are likely be very costly to the firm in terms of claiming health benefits (and thus raising future premiums for the firm) and in lost production because of absences 180

Social and private perspectives

Table 39.1

Private (employer) calculation of benefits and costs of HIV testing of employees and hiring an HIV negative individual

Source of Avoided Costs

Health insurance Life insurance Disability insurance Pension Discounted total cost if develops AIDS Probability that HIV+ person develops AIDS during tenure Expected value of costs avoided (benefits) Testing costs range

Source:

181

Large Firm

Large Firm

Small Firm

Small Firm

High-cost City

Low-cost City

$80 000 $21 800 $13 400 –$2400 $90 900

$40 000 $21 800 $13 400 –$2400 $58 900

$32 000 $900 $600 –$360 $29 300

$16 000 $900 $600 –$360 $17 500

0.35

0.35

0.15

0.15

$31 000

$20 600

$4400

$2 300

$22 400– $310 300

$ 22 400– $310 300

$3700– $10 800

$3700– $10 800

High-cost Low-cost City City

Based on Table 1 and numbers given in the text in Bloom and Glied (1991).

due to illness. The costs averted become the benefits of screening to the firm. These benefits are reduced if the firm gets any savings due to an HIVinfected person dying and not claiming his/her pension. Of course, these costs averted are benefits only if the applicant is HIV positive. So the probability of being infected has to be applied to the costs averted to form the expected benefits. The costs of screening are the financing of the tests that have to be given to everyone who applies for employment. The magnitude of the firm benefits from HIV testing depend on the type of city where the firm is situated, since some cities (like New York City) provide generous health care packages, while other cities (like San Francisco) provide only minimal health care benefits. The amount of the firm costs mainly depend on the size of the firm. Large firms have to test a large number of potential employees. The resulting costs and benefits for different sized firms living in low- and high-cost cities are displayed in Table 39.1. We see that for most small firms, HIV testing is not beneficial (that is, profitable). This also would be true for many large firms except those who can experience low testing costs, because they: (1) purchase tests on a large scale and receive discounts for doing so, and (2) routinely give employees blood tests, so the HIV test would not require additional blood work. But,

182

Setting priorities for HIV/AIDS interventions

the point remains that HIV testing could be profitable for some firms who recruit workers in high prevalence cities (where the probability of a person being infected is high).

A SOCIAL EVALUATION OF HIV TESTING Because testing costs are likely to be the same from the social perspective as for the private perspective, the big difference between private and social outcomes of firm HIV testing come from the benefits side. The main private benefit (that the firm does not have to pay the health costs if an infected person is not employed) is basically a cost-shifting exercise. The health care costs still have to be incurred by someone. The difference is that if the person is employed the firm pays the costs, while if the person is not employed the government and others have to incur the costs. So from the CBA perspective the private benefits are not social benefits (they will not be costs avoided). There are, though, other categories of effects that are social benefits. The main ones are prolonging the life of an infected person (by giving drugs if people get a positive test result); removing a person’s uncertainty by having the test and preventing the transmission to others if people’s test results change their behavior. To estimate the social benefits of HIV testing, in terms of preventing transmission and saving lifetime costs of AIDS, Bloom and Glied made the following assumptions: (1) if the present value of direct and indirect costs of a newly infected person is $600 000; (2) if each infected person infects, on average, 0.8, individuals per year and (3) if one seropositive person who is tested now infects 10 percent fewer (that is, 0.72 persons) then the present value of savings over an 11-year testing period would be $25 million. This is much greater than the costs of testing. However, this positive social result is very sensitive to the assumptions made: (1) if the infection rate is 0.08 (and not 0.8), and the reduction is 3 percent (and not 10 percent), then the savings would be only $18 300. This is less than the cost of testing in low prevalence areas. (2) In some studies in the United States, individuals informed of their HIV status actually increased their high-risk behavior. So social benefits could be negative.

THE PRIVATE AND SOCIAL PERSPECTIVES OF FIRM HIV TESTING COMPARED Bloom and Glied summarize their overall findings as showing that the private net benefits could be positive, while the social net benefits are

Social and private perspectives

183

most likely negative. In this special case the policy prescription is clear. If social outcomes are negative, firm HIV testing should be banned as it is in a number of US states (on the grounds that it would discriminate against disabled people getting jobs). But, if firms are prevented from doing what is in their own best interests they will adjust their behavior and this could greatly impact outcomes from a social perspective. For example, if firms have to employ people who might be infected with HIV, they have an incentive to reduce the package of health and other benefits that they provide with employment; or the firm could simply not employ anyone who they think could potentially be affected (for example, young African Americans). So Bloom and Glied point out that there would be real social costs if the government bans all firms for HIV testing on equity/ discrimination grounds. Given the possible difference between private and social interests, subsidizing the health care package of firms (or providing national health insurance) could be necessary to reduce the social costs of banning firm testing.

SUMMARY AND CONCLUSIONS One should always adopt the social perspective in an economic evaluation if one wishes to try to influence social decisions. Even though the social perspective is primary, other perspectives are not irrelevant. In a mixed economy, where the government makes decisions recognizing its interaction with the private sector, it is important to know outcomes from a narrower perspective. For example, if it is not socially worthwhile for people to be tested for HIV, but it is worthwhile from an individual firm’s point of view, then there could be an “incentive compatibility problem”, which is to say that the socially desirable outcome will not in this case be chosen by the individual. Therefore some government incentive must be given to induce individuals to adopt the socially optimal outcome if they do not think it is in their best interests. Note that the social dimension of CBA does not mean that it endorses “socialism”. The American Heritage Dictionary defines socialism as: “Any of various theories or systems of social organization in which the means of producing and distributing goods is owned collectively or by a centralized government that often plans and controls the economy”. CBA provides a basis for deciding whether public or private expenditures are worthwhile or not. It does imply that the government has to own and run the activity. Whether the firm does the employment testing or the government does the testing is not predetermined. The choice is decided by the relative sizes of the benefits and costs. If testing is socially worthwhile, then the firm should

184

Setting priorities for HIV/AIDS interventions

do the testing if the net benefits are greater; and the government should do the testing if it can organize the intervention with greater net benefits. We already saw this in Chapter 29 where three ways of treating TB patients were being evaluated. The public–private NGO model was preferable to the purely public model (and the public–private workplace model) only because it carried out the treatment at lower costs. Bloom and Glied (1991, p. 1798) conclude that because their analysis indicates that firm HIV testing is not likely to be socially worthwhile this means that; “Existing state and federal legislation related to HIV testing has been motivated primarily by concerns over social equity”. It is now time to deal explicitly with the equity issue in CBA, which is the subject matter of the next two chapters.

40.

CBA and equity I: allowing for ability to pay

We now address the heart of the complaint from non-economists about CBA – that it is not fair to value goods and services by a person’s willingness to pay. If someone has low WTP it may be because the person is poor and not because he or she does not want to have health care. Actually, many health care economists also think that WTP is unfair. What is more surprising is that mainstream economics assumes that, even if CBA can allow for equity, it should not do so, as it is more efficient to help the poor in other ways than by biasing heath care evaluation outcomes. In this chapter we will explain and examine the validity of all these viewpoints. Although much of the discussion appears controversial, the points that will be made to resolve the controversies are really obvious if one views CBA from a social perspective. The main argument will be that to throw out WTP from health care evaluations is equivalent to the proverbial “throwing out the baby with the bathwater”. WTP is to be an essential ingredient even if it will be shown to be not the only essential ingredient.

NON-ECONOMISTS’ CONCERN OVER EQUITY The concern that WTP does not allow for equity does not mean that WTP should not be a part of a health care evaluation. WTP reflects people’s preferences in general and the strength of these preferences in particular. If someone is WTP $30 for a skin cream, to deal with one of the opportunistic skin infections that accompany AIDS, one can plausibly assume that the skin infection really bothers that person, and bothers that person so much more than can be alleviated by just giving that person $10 to live with the inconvenience of the skin infection. WTP is a very good measure of how much people benefit from health care interventions. So the problem with WTP is not what it includes, but what it excludes. The critics are right to the extent that WTP does not automatically consider equity; this needs to be introduced into CBA as an additional ingredient. But how? 185

186

Setting priorities for HIV/AIDS interventions

HOW TO INCORPORATE EQUITY INTO CBA: INTRODUCING DISTRIBUTIONAL WEIGHTS The problem with willingness to pay is that it does not allow for ability to pay. Everyone agrees that a dollar to a homeless person is worth much more than a dollar to a millionaire. So if the homeless person is willing to give up a dollar to pay for health care, then that dollar should not be treated the same as a dollar given up by a millionaire, or even the average person who is not homeless. The dollar to the homeless person should be worth more in the evaluation of the intervention than a dollar to an average person. The solution then in order to allow for equity in CBA is to weight the willingness to pay for the inputs and outputs of an intervention by the ability to pay of those receiving the benefits or incurring the costs. Let weightB be the weight on the benefits and weightC be the weight on the costs. Then the cost–benefit criterion becomes that the weighted benefits should exceed the weighted costs: (weightB) Benefits > (weightC) Costs The way to think about the weights is as numbers that can be equal to, less than, or more than unity (that is, 1). To simplify matters, assume that the costs are financed by the average person and so have a weight of 1. Since multiplying anything by 1 does not make a difference to its absolute or relative value, the cost–benefit criterion becomes: (weightB) Benefits > Costs The issue now is how to determine the distribution weight for the beneficiaries.

HOW TO DETERMINE THE DISTRIBUTIONAL WEIGHTS There is a whole literature on how best to determine distribution weights for CBA – see Brent (2006, Chapter 10) and Brent (1998, Chapter 3). Distribution weights have also played a role in CEA in the health care evaluation field – see Brent (2003, Chapter 10). Here we just want to outline a simple method that has been widely applied in CBA especially for developing countries so that we can illustrate how one can determine the weights in practice. The method was first devised by Squire and van der Tak (1975), two

Allowing for ability to pay

187

researchers working for the World Bank. The idea was that there should be diminishing social marginal utility of income, that is, the higher a person’s (or group’s income), the lower should be the social value of an increment in income. In other words, the distribution weights should decline with the person’s income. An easy way of ensuring this property is simply to have the weight inversely related to income. To ensure that the weights average to 1, the inverse relation was expressed relative to average income. So the distribution weight formula became: weightB = average income/income of beneficiary Obviously if the beneficiary is an average income earner, then the numerator of the formula would be the same as for the denominator and so the weight would equal 1 as expected. But, in line with how distribution weights have been applied in practice, we will interpret this formula as applying to people’s income in groups that are defined in multiples of average income. Say a person is poor if they are in an income group that has less than, or equal to, one-third of average income, and define a rich person as someone in an income group that has greater than, or equal to, three times the average income. A poor beneficiary would have a weight of 3 (the inverse of 1/3). This means that the poor would have each dollar that they received as benefits from the intervention valued in the social evaluation as $3. Conversely, the rich would have each dollar that they received as benefits from the intervention valued in the social evaluation as 33 cents (the inverse of 3/1). In this weighting scheme, the poor homeless person’s weight would be nine times that of the rich millionaire’s weight; quite a lot of preferential treatment. Note that whether an intervention that favors a low-income person will actually be approved does not depend solely on the values of the distribution weights. The size of the benefits and costs are also important. It the cost of the cream to treat a skin infection is $30, then the WTP of the poor would have to be at least $10 for the cream treatment to be judged socially worthwhile.

APPLICATION TO CBA OF CONDOMS IN TANZANIA As there are no published studies of CBAs of HIV/AIDS interventions that have used distribution weights, we can for illustrative purposes reconsider the condoms social marketing project for Tanzania covered in Chapter 27 and apply weights to this evaluation. Recall (from Table 27.2) that the original project involved a price subsidy that lowered

188

Setting priorities for HIV/AIDS interventions

condom prices from 2000 TZSH to 100 TZSH. The social benefits were 17 168 TZSH and the costs were 10 794 TZSH, producing a worthwhile project judged on economic efficiency grounds (with positive net benefits of 6374 TZSH). Now let us weight the benefits and see what happens to the outcome. A distribution and efficiency CBA would weight a person’s willingness to pay by his/her ability to pay. As we have just seen, a group with an income higher than average would attract a weight less than 1; while lowincome groups would be assigned a weight more than 1. The PSI survey did not include an income question. But, Badru (2000) did state that the people surveyed were from urban groups that were “socio-economically better off” compared with the national 1996 demographic survey. So the weight attached to the benefits for the CSM intervention should be less than 1. The National Bureau of Statistics Tanzania (2002, p. 101) reported that households in urban areas (other than Dar Es Salaam) had average incomes of 30 426 TZSH when those in mainland Tanzania were 17 928 TZSH shillings. If we assume the Squire and van der Tak distribution weighting scheme that is inversely related to average income in Tanzania mainland, then the distribution weight to be applied to beneficiaries is 0.59 (that is, 17 928/30 426). The weighted social benefits would have been 10 116 TZSH. The decision outcome would have been altered by the distribution weight as the social benefits were not now sufficient to cover the costs of 10 794 TZSH. For the CSM program to be worthwhile with distribution weights, the alternative subsidy scheme mentioned in Chapter 27 would have had to be implemented. This involved increasing the subsidized price to 290 TZSH instead of implementing the original subsidized price of 100 TZSH. Weighted benefits at this reduced level of condom provision would have been 6369 TZSH, which would have far exceeded the costs of 3222 TZSH.

ECONOMISTS’ CRITICISMS OF DISTRIBUTION WEIGHTS Economists have two main concerns with the use of distribution weights. We will explain and respond to each criticism in turn. First, mainstream economists prefer that the tax system be used to redistribute income to the poor, rather than applying distribution weights in projects for the same purpose. If the concern is that distribution weights are too “subjective”, then using the tax system to redistribute incomes does not avoid this

Allowing for ability to pay

189

problem, as one still needs to ascertain everyone’s social marginal utility of income to be able to know how much to tax a rich taxpayer relative to a poor transfer recipient. Second, mainstream economists think that using WTP is fundamental to economics as it recognizes that individuals have preferences. But, the distribution weights seem to be imposed externally and reflect government/ political preferences and so would not be fundamental to economics. The problem here is that this ignores the fact that individuals actually do have preferences about how much income the poor receive, just as they do about goods and services that they themselves receive. Distribution weights can then be interpreted to be expressions of WTP for redistribution. This is not a fanciful interpretation as, for example, Zarkin et al.’s (2000) survey found that people were WTP $37 per respondent for the successful drug treatment of 100 persons other than themselves or people that they know. In terms of the simple Squire and van der Tak formula, one could always ask people whether they prefer the inverse income weighting scheme to some other formulation. (I actually prefer distribution weights to be set by the square root of the inverse of a person’s income [Brent, 2006].)

SUMMARY AND CONCLUSIONS It is a weakness of traditional CBA that a person’s willingness to pay is to be recorded with no recognition of that person’s ability to pay. But, this can easily be remedied by weighting the benefits and costs in health care evaluation by a factor that reflects the social marginal utility of income. What does this weighted criterion represent? It is no more than the recognition that efficiency and distribution should be joint objectives when making health care evaluation decisions. Mainstream economics is based on WTP and so incorporates only economic efficiency. Non-economists often just focus on distribution and want to ensure that people can afford to receive interventions that are deemed necessary. What is needed is for both objectives to be incorporated simultaneously. The choice of distribution weights then represents the extent of the trade-off between objectives, that is, between distribution and efficiency. To be sure there are a large number of practical problems that have to be solved in applying distribution weights. But these problems can be manageable if the economics profession tries to solve these problems rather than pretending that distribution weights can be safely ignored when making health care evaluations. Note, that the option not to use distribution weights does not exist. The traditional criterion used throughout

190

Setting priorities for HIV/AIDS interventions

this book, that is, calculating the difference between benefits and costs is just one special case of the weighted benefits and weighted costs criterion presented in this chapter. It is the special case where both weights are set equal to 1. Why is this set thought to be “scientific” by mainstream economists while any other set is “non-scientific”? I cannot tell you precisely how much more a dollar is worth to a homeless man than to a millionaire, but, I can tell you with certainty that they are not equal, and so the traditional CBA criterion cannot be correct. To repeat what is said at the beginning of this chapter: everyone can agree that a dollar to a homeless person is worth much more than a dollar to a millionaire. It is true that this is a value judgment. But, if it is a value judgment that everyone can agree on, then we should use it in CBA practice. It is already there in CBA theory.

41.

CBA and equity II: allocating by time and other non-price methods

As a complete alternative to allocating health care resources by willingness to pay and cost, non-economists often recommended using time as the basis for rationing (that is, restricting consumption of health care). They think this is fairer than using market-type principles. Much of the National Health Service (NHS) in the United Kingdom relies (or used to rely) on this system. So if one wants to see the local physician, one needs to wait one’s turn at the doctor’s offices. Rationing here is on a “first-come, first-served basis”. How long you are willing to wait is an indicator of how important it is for you to see the doctor. Similarly, to have a nonemergency operation, there is a waiting list. If you want the operation free of charge, you must be willing to wait weeks, months, or possibly, years till your turn comes up. In this chapter we will examine the effects and implications of using time as a rationing system. This discussion is very general and relates to any type of rationing of health care by using time. To consider particular issues that are specific to HIV/AIDS rationing, we will also cover other non-price rationing methods that are applicable to the dispensing of ARV therapy in Africa, based on some of the ideas given in Rosen et al. (2005).

RATIONING HEALTH CARE USING TIME If health services are to be given free of charge, it is almost inevitable that the quantity demanded is going to exceed the quantity supplied and rationing will be necessary. When time is used to do the rationing, WTP considerations still apply, even though this is now indirect rather than direct. As everyone knows: “time is money”. The people who value time the most will pay the highest “price” (have the highest opportunity cost). Note that time is involved not just in the waiting room, but will also be relevant in making the journey to and from the doctor’s office. To a certain extent rationing by time would be rationing by income and so recognizes the ability to pay principle. This is because the rich are usually paid more for their time. So when they give up their time they give up more of their 191

192

Setting priorities for HIV/AIDS interventions

income. But, the point that will be stressed here, and in the next section, is that rationing by non-price means can have unintended effects, and hence unintended equity effects. The NHS was explicitly set up in 1944 in the United Kingdom in order to ensure that people can receive health care irrespective of their ability to pay. The system involves a fixed quantity of doctors, surgeons and treatments being supplied and financed by the government out of taxation. Waiting times function as a rationing system of this fixed quantity by increasing when there is excess demand and falling when there is excess supply. To help understand some of the equity issues of rationing by time, let us look at the study by Propper (1990), which surveyed those on the United Kingdom’s waiting lists to estimate what they thought was the value of the time that they had to wait in order to get non-urgent care. The question asked of respondents was how much they would be willing to pay for a reduction in waiting time. Note that the patient here is not actually giving up time by waiting in an office (and hence not earning a wage). Instead, being on a waiting list involves the cost of having to wait for treatment when one is ill and uncertain as to how long the wait for treatment will be. The resulting WTP for waiting list reductions for different groups are given in Table 41.1. The valuations are given separately for those who earn more than £350 per month and for those who do not. As one would expect, and what the designers of the NHS would have hoped for, those with the larger ability to pay incur the higher waiting list “price”. But, what was revealed and was unintended is that the elderly had to pay the highest price. This is easily explained by the fact that, as the elderly have the least remaining time available, they have the most to lose by having to wait for treatment. Allowing for ability to pay came at the expense of other forms of inequity, that is, age discrimination.

RATIONING HEALTH CARE USING OTHER NONPRICE METHODS Rosen et al. state that the most accepted criterion for rationing ARVs is “disease progression”. That is, access to the ARVs is to be given to those who are in the most medical need as judged by CD4 (T cell) counts. This restriction is not applied in a standard way, as the US Department of Health and Human Services sets the recommended CD4 count at 350, while the World Health Organization’s eligibility guideline specifies a CD4 count of 200 (or an AIDS-defining illness). The amount of rationing involved varies enormously by the CD4 count restriction. Auvert et al. (2004) point out that, for South Africa, using the 350 CD4 count figure as a

Allocating by time and other non-price methods

Table 41.1

Estimated value of waiting list time (£ per month)

Segment

Full-time employed Part-time employed Housewife Retired Source:

193

Weekly Household Income Below £350 per Month

Weekly Household Income Above £350 per Month

41.90 35.70 20.40 43.43

49.43 42.11 24.06 49.90

Based on Propper (1990) Table 3.

limit instead of the 200 CD4 count figure would increase the proportion of HIV positive people eligible for ARVs from 9.5 percent to 56.3 percent. Of course, defining disease progression is only the first step in the rationing sequence. If one takes the high CD4 figure for South Africa, the country probably would be unable to actually provide drugs for the 56.3 percent who need them. Rosen et al. suggest that additional “socioeconomic criteria” will also be required to ration treatment. They give doctors, nurses, teachers, judges, police officers, or post-secondary students as example target groups. The most widely applied target group is HIV positive mothers as this will reduce transmission to their babies. The downside of this rationing criterion is that it discriminates against males, and women who are not pregnant. Apart from formal rules and procedures for rationing, restricting consumption of health care also takes place using informal and often illicit arrangements that favor specific individuals or groups implicitly. There are two main concerns here in countries such as those in SSA that have weak enforcement capabilities. First, there will be queue jumping by political and economic elites to ensure that they get the drugs at the expense of others. In these cases, ability to pay enters by the back door; the rich get the drugs free. Second, some will get the low-price drugs and then, because the drugs may not be a priority to the recipients, they will sell them on the black market for higher prices. Apparently this is taking place in Zimbabwe and Swaziland. Again, ability to pay ends up paramount because it is the rich who are buying the black market ARVs.

SUMMARY AND CONCLUSIONS We have seen that non-price rationing on equity grounds often leads to other equity problems. If one is concerned with ability to pay, introduce it

194

Setting priorities for HIV/AIDS interventions

formally into the evaluation criteria by using distribution weights rather than informally through non-price rationing. When it works well, nonprice rationing can ensure that ability to pay does not dominate outcomes. But usually, ability to pay will end up being very influential. The main result of non-price rationing is to ensure that people’s preferences are not being recognized. WTP reflects these preferences. After all, if a medical expert is to decide what your CD4 count must be in order to be eligible for ARVs, it means that the patient him/herself has no say in the decision. To try to rule out WTP on ability to pay grounds leads to results that could ensure that neither efficiency nor distributional considerations are included in the determination of outcomes. As Rosen et al. (2005, pp. 355–6) conclude: “Implicit rationing will foster both inequity and inefficiency”. To this we simply add that CBA with distribution weights was especially devised to foster both equity and efficiency.

42.

Conclusions I: how not to set priorities for HIV

It is time to sum up what CBA has to offer in the setting of priorities for interventions to deal with HIV/AIDS. In this chapter we look at three main ways that priorities are set in practice that do not rely on CBA methods. We see that all sorts of mistakes, and hence a large waste of resources, follow from ignoring CBA. Then in the next, and final, chapter we show how many concerns that people have, which are thought to be essential in setting priorities, are indeed included in CBA automatically.

1.

EVALUATING BY LABELING

In the HIV/AIDS field, many judgments are made about projects just by looking at the label associated with the intervention. Here we look at some of the main labels and explain why they are not helpful. (i)

“Prevention” is Better than Cure

A British saying is that “A penny’s worth of prevention is worth a pound’s worth of cure”. How can this not be true for HIV/AIDS interventions? There are two main surveys of a large number of HIV/AIDS evaluations by Creese et al. (2002) and Canning (2006). These surveys show clearly that treatment cost-effectiveness ratios are so much lower (we get so many more DALYs per dollar of expenditure) for prevention programs (such as condoms, blood safety, education and information) than for treatment projects (such as investing in ARVs). Surely, prevention is the way to go? Not necessarily. Here are some reservations: ●

Even if every prevention program were more cost-effective than every treatment (or care) program, it could be that none of the prevention programs may actually be worthwhile. As we explained in Part III (see especially Chapters 30 and 31) only a CBA can tell whether a health policy change is socially worthwhile. A CEA is not sufficient as a guide. 195

196 ●

●

(ii)

Setting priorities for HIV/AIDS interventions

Even if every prevention program were more cost-effective than every treatment program, it could be that all of the treatment programs may actually be worthwhile. Again, only a CBA can tell whether a health policy change is socially worthwhile. We saw in Chapter 31 that a combined intervention with first line and second line drugs with intensive monitoring was the least cost-effective yet, using one benefit methodology, it was found to have positive net benefits. Of course, we should expect that both prevention and treatment programs would have a wide variation in their net benefits irrespective of their labels. This is because how they are implemented in detail will alter greatly the inputs and outputs, and hence the effectiveness, benefits and costs. For example, Table 31.1 shows that combined interventions that used only first line drugs had a cost-effectiveness ratio that was around one-tenth of the least cost-effective ARV therapy. This Program Does Not Need to be Evaluated Because it Involves “Scaling Up”

The idea behind this label is that if one has first identified something that has already been shown to have “worked” for a subset of a population (that is, a sample), then scaling up that intervention to the whole population can also be thought to be worthwhile. UNAIDS (see, for example, UNAIDS, 2008, p. 7), is a great believer that scaling up is what is now required, especially for ARVs. The optimum under this approach is a target of 100 percent. Anything less than 100 percent is not good enough. Hence, although almost 1 million more were receiving ARVs between 2006 and 2008 (making the total 3 million), ARV coverage was deemed “low” as it was only 37 percent of those who needed it in 2007. The trouble with scaling up is that it is hardly ever based on what “works”, and hence what should be scaled up, by reference to an evaluation that relies on a CBA. There is a real risk of the “blind leading the blind”, that is, if an intervention has never been properly evaluated, why should cloning the intervention be any less based on guesswork? Even if the existing 3 million people on ARVs were all shown to have been evaluated and judged worthwhile by a CBA test, we have no guarantee that net benefits of extensions to ARV coverage would also be positive. This is because one should not expect benefits or costs to stay constant. As explained in Part I, it can be expected that marginal benefits will decline as the scale of operations expands because of the principle of diminishing marginal utility. Marginal costs could go up or down, as there could be economies of scale over time (if the methods of mass production can

Conclusions I

197

be utilized), or there could be rising costs (if distributing the health care services to the general population involves traveling to more remote areas of the country). But, in either case, we would not expect MC to remain constant. (iii)

“High-risk” Groups Should be Targeted

The World Bank is a great believer in the need to identify high-risk groups. Traditionally, these groups were homosexual males, injecting drug users who share needles and prostitutes. Health care evaluations are very straightforward in this belief system. If expenditures target high-risk groups they are “good” projects. Dayton (1998, p. 22) used this approach in her evaluation of the World Bank’s lending of $550 million for HIV/ AIDS interventions between 1986 and 1996. Her review “identified a weakness in the Bank’s lending program: many of the activities supported by the Bank have not been well-focused on the groups in the population most at risk for HIV infection”. However, what was much more a real weakness of the Bank’s lending program for HIV/AIDS was Dayton’s finding that only one-third of the Bank’s projects prepared adequate cost– benefit or cost-effectiveness analysis. The biggest danger with using the “high-risk” label for setting HIV/ AIDS priorities is that it is tautological. If an intervention actually does target those most at risk then it will be most effective. One only knows whether one has targeted those most at risk if one first establishes that the intervention was most effective. This tautological problem is most evident in the revelation given in the book by Pisani (2008), The Wisdom of Whores, which we referred to in Part II. She informed us that a schoolgirl in South Africa is ten times more likely to be infected with HIV than a prostitute in Beijing. In this particular case, school girls need to be targeted, and not prostitutes, as they are the high-risk group. (iv)

The Myth About “Myths”

The problem with using the myth label as a way of setting HIV/AIDS priorities is that nearly all alleged myths in this area are in some circumstances true and in other circumstances false. A CBA tells us the veracity of the alleged myth. The myth label is not useful, notwithstanding the popularity of trying to uncover myths in the HIV/AIDS literature. Myths are not helpful because they ignore the details that are vital to determining intervention success. We first discussed myths in Part I of the book in order to make clear that some things we think we know may be false. Here we wish to emphasize the other side, that is, things that we think are false

198

Setting priorities for HIV/AIDS interventions

may be true. A recent set of ten alleged myths was presented by Shelton (2007). All ten “myths” are not myths in some circumstances and all ten are myths in other circumstances. For example, is the statement “Men are the problem” a myth? Obviously, what Shelton seems to want to debunk was the label that “men” always cause HIV transmissions, while “women” do not. So the fact that a large number of females in Kenya were HIV positive when their male partners were HIV negative is evidence that not all transmissions are initiated by males in Africa. This is a point worth making, but, the statement that men are the problem is not a myth when one considers all the male violence against women through rape (especially in the Democratic Republic of Congo), which does mean that much transmission is male initiated.

2.

DENIAL

There is a saying in the United States that “If something is not broke, then you don’t need to fix it”, so if one denies that there is an HIV problem for certain specified groups, then one thinks that one does not need to carry out any expenditure on their behalf. With no expenditure taking place, there is no need to carry out an evaluation for these groups. Denial is a major strategy for dealing with HIV/AIDS in the United States. Here we will focus just on two groups where denial is prominent, African Americans and the elderly. (i)

The United States Does Not Have an HIV/AIDS Problem

The international community is adamant that countries in Africa have to have a nationally coordinated HIV/AIDS program in order to combat the disease. This is presumably because HIV/AIDS is rampant in these countries. However, in the United States there is no national program even though HIV has a high prevalence among African Americans. As we have seen, AIDS is the number one cause of death among African American females ages 25 to 34. If Washington, DC were a country, its HIV/AIDS prevalence rate of 3 percent would place it above many West African countries by degree of severity. Of the blacks in DC, 7 percent are infected (according to the blog of Mark Cichocki, Tuesday, 17 March 2009). We know that partner concurrency caused by prison incarceration is one of the main reasons why the African American community is so affected, but, are prison HIV interventions a priority in the United States? Have such interventions been shown to be ineffective or otherwise not worthwhile?

Conclusions I

(ii)

199

The Elderly Do Not Need to be Targeted

Prior to 2008, UNAIDS only tracked HIV prevalence for females between the ages of 15–49. They called this the “sexually active” age group, thus effectively denying that sex takes place over the age of 50! We should not be surprised then that in many countries of the world the elderly have experienced rising HIV rates. As we saw, CDC has reported that for the United States in 2008, 15 percent of new infections involve those who are over 50. Elderly women worldwide are especially likely to be vulnerable to HIV as many of them are post-menopausal, so they would not typically be using condoms when they have sex to protect themselves against getting pregnant. Karpiak et al.’s (2006) ROAH study (Research on Older Adults with HIV) reports that the immune system’s function declines with aging. This has two main implications. The elderly are more likely to become infected from a HIV positive partner and once infected the elderly are more likely to die (they are twice as likely to die as their younger untreated counterparts according to the ROAH study). In SSA, elderly women are even more vulnerable to HIV than in the West. According to the study related to Tanzania by HelpAge International (2004), elderly women have much lower levels of literacy (over one-third of adult rural women had no education) and therefore rely more on oral forms of communication to learn about HIV. However, prevention programs concentrate exclusively on younger persons and “highrisk groups” using modern communication methods. As a result, the elderly lacked information about the causes of HIV, prevention methods and opportunities for treatment. Not surprisingly therefore, 5.2 percent of female blood donors over the age of 55 in Tanzania tested HIV positive.

3.

TOTAL BURDEN OF DISEASE

Canning (2004, p. 134) points out that many people believe that if there is more death and suffering caused by one disease than another, then more resources should be devoted to that disease. In 2004, HIV/AIDS accounted for 12 percent of the DALYs lost to diseases in Africa. This is more than malaria (11 percent), respiratory infections (10 percent) and injuries. On this basis, HIV/AIDS would seem to have the greatest priority for health care expenditures in Africa. But, as Canning makes clear, this line of reasoning makes an elementary mistake: if nothing can be done to reduce the spread of HIV, nothing should be spent on it. Assessing the effectiveness of interventions is crucial to the setting of priorities. As we have seen, CBA

200

Setting priorities for HIV/AIDS interventions

relies on marginal analysis: changes in benefits and changes in costs. If there are no changes in benefits, there should be no changes in costs. Interestingly, the same burden of disease argument has recently been made by England (2008), but this time to make the opposite case, that too much money has been devoted to HIV/AIDS. Although HIV causes 3.7 percent of global mortality, it receives 25 percent of international health care aid (as well as large amounts of domestic expenditures). Again the point is that only by looking at the margin and seeing how effective particular evaluations are, and feeding this information into a CBA, can one tell whether HIV is more or less deserving of funds than other diseases.

SUMMARY AND CONCLUSIONS As stated at the beginning of this book, the most important message is that deciding on priorities for HIV/AIDS is too difficult a task to be left to guesswork and intuition. Health care evaluations must be made on the basis of data reflecting actual behavior, which is the approach that is fully reflected in CBA. This chapter pulled together a number of attempts that relied on guesswork and intuition and explained why they were not helpful. Labels such as “prevention”, “high risk”, and “scaling up” to mean “good” projects, and “treatment”, “low risk” and “myths” for “bad” projects, are not useful even if they were first conceived with data in mind. It is the data that relates to a particular set of circumstances, that is, a specific project based on information for a given set of inputs and outputs that needs to be the basis for the setting of priorities. Similarly, simply denying that HIV is a problem for a particular country (the United States) or demographic group (the elderly) is just another way of guessing what is likely to be worthwhile. Lastly, we made the point that “number one killer” and “number one priority” is not necessarily the same thing in the context of HIV/AIDS. The issue is whether a particular intervention can alter the course of a disease, not whether, unaltered, the course of a disease causes a lot of pain and suffering. CBA is valuable not just because it relies on data; its value lies in the way it uses this data. In the final chapter we explain why it is that CBA uses data in the best possible way.

43.

Conclusions II: using CBA to set priorities for HIV

Critics of CBA, who wish to avoid its use on the grounds that willingness to pay does not allow for ability to pay, risk the proverbial problem of “throwing out the baby with the bathwater”. So the strengths of using WTP are ignored by the critics. This is unfortunate as much of the politics of HIV/AIDS involves activists thinking that they have to get organized and petition governments in order that a particular perspective be incorporated in the HIV/AIDS decision-making process when, in fact, every perspective would automatically be included if a cost–benefit framework using WTP were used. We have throughout the book emphasized that CBA tries to include all the effects of an intervention by estimating the WTP of everyone who is affected by it. The main task in this chapter is to spell out some of the implications of this all-inclusive property of CBA. As this is the final chapter, the summary and conclusions section briefly sums up what this book has tried to achieve.

WILLINGNESS TO PAY AND SOCIAL INCLUSIVENESS WTP records individual preferences and the intensity of those preferences. When we say that the net benefits of a project are the sum of the WTP of all those affected, we are stating that, no matter whether these net benefits are positive or negative, we have included everyone’s preferences in the evaluation and these preferences will determine the decision outcome. Only if decisions are not made on the basis of WTP do activist groups need to be set up to ensure that the interests of special groups be included. We will illustrate this point with reference to women, those living with AIDS and the elderly.

WOMEN AND HIV Women make up around half of the world’s population and about half of the total number of people living with HIV and AIDS (PLWA). Gender 201

202

Setting priorities for HIV/AIDS interventions

inequality means that women’s preferences are not given full weight, especially in the context of HIV/AIDS. To use Stephen Lewis’s (2005) words, in the fourth lecture in his book Race Against Time, “Women: Half the World, Barely Represented”, he argues that the UN has never really represented women’s preferences adequately. As he says (p. 125): “The proof is in the reality: just look at the death toll that AIDS has taken on women”. What is Lewis’s solution to this problem, which he argues is the most important HIV issue? He wants there to be an international women’s organization set up as part of a multilateral UN system. This agency would be larger than any existing UN women’s organization and be an amalgam of the UN Development Fund for Women (UNIFEM), the UN Population Fund (UNFPA) and the Division for the Advancement of Women (DAW). This new agency should be funded at the same level as UNICEF, with a budget of around $2 billion a year. One does not have to be a feminist to agree with Lewis’s sentiments that there is a problem in getting the preferences of women included in HIV decision-making. However, one can question his solution in the context of women and HIV (although his proposal can still be worthwhile in other contexts). If instead of male bureaucrats at the UN agencies making HIV expenditure decisions, say UNAIDS took CBA seriously? What would this imply about how HIV priorities would be set? The answer is that women’s preferences would automatically be taken seriously. To see how this works, let us reconsider the CBA of the condom social marketing program in Chapter 27. The benefits of the subsidized condoms were measured by the area under the demand curve, where the demand curve measures WTP. Using the WTP measure, the condom subsidy program was just about worthwhile if one ignores the values of the condoms to the partners. As 50 percent of the partners were not regular partners or spouses, it is likely that the market demand curve reflecting private preferences was an underestimate of the social (total) WTP for the condoms (that is, the purchasers of the condoms plus their partners). For this reason the private benefits were increased by 50 percent to obtain the social benefits. With this added element, the condom subsidy program was highly worthwhile. What has this evaluation got to do with women and their preferences? Nowhere in the Brent (2009c) study did he mention women’s preferences. Does this mean that women’s preferences were not included in the evaluation? Not at all. It is true that it was the men who actually did most of the purchasing of the condoms. But (assuming we are mainly talking about heterosexual use of the condoms) 100 percent of the partners (that is, 50 percent of the beneficiaries) would be women. So when the external benefits of condoms were being calculated, it was the WTP of women that was

Conclusions II

203

being recorded. In other words, total benefits, that is, total WTP, were the sum of male WTP and female WTP. As long as externalities were allowed for, women’s preferences were included. It will not always be the case that external benefits and female benefits will be the same thing, But, even where there are no externalities, it will always be the case that total WTP will be the sum of male and female WTP.

PLWA AND HIV People living with AIDS are one of the most discriminated against groups of all. It is likely that none of the scientific advances that have been made in terms of AIDS treatment (such as the invention of ARVs) would have occurred in the absence of political activism by gay rights groups. The research expenditures would never have taken place unless the US government had been convinced that there was a human rights and public health issue concerning HIV/AIDS that had to be addressed. Having acknowledged this fact, it is also true that PLWA do not need to organize around every HIV/AIDS intervention to ensure that the preferences of those HIV infected are not ignored by decision-makers. What is required for the preferences of those PLWA to count is only that a CBA be carried out of all HIV/AIDS interventions using the WTP of all those affected, and for the expenditure decisions to be based on the CBA outcomes. The condom social marketing program also illustrates this point. One reason why people purchase condoms is because they are HIV infected and do not want to transmit the disease to their sexual partners. Again, the Brent (2009c) condom subsidy evaluation did not mention PLWA explicitly. Nonetheless, some of the Tanzanian respondents surveyed would have been HIV positive. What these people would be WTP for condoms would be included as part of the demand curve estimate. This means that effectively, total benefits equal the WTP of PLWA plus the WTP of those uninfected by HIV. The demand curve estimate does not discriminate; a dollar is a dollar no matter who is willing to give it up.

THE ELDERLY AND HIV As we explained in the last chapter, until recently, UNAIDS ignored the number of HIV cases that were elderly. Does there need to be a new UN agency introduced especially to look after the interests of the elderly? Again the answer may very well be yes, for many different reasons. But, for including the preferences of the elderly in health care evaluations all

204

Setting priorities for HIV/AIDS interventions

that is required is that a CBA be used to guide the decisions based on WTP. In principle, the reasoning that we applied for total WTP for women and PLWA applies to the elderly. So we have the identity that total WTP is the sum of the WTP of the elderly plus the WTP of the young. This means that the demand estimate for condoms could automatically include the preferences of the elderly. Whether this reasoning is realized in practice depends on how the data is collected for the particular demand estimate study. It turns out that for the Tanzanian condom study, the WTP of the elderly was not a part of the total WTP. This was because the PSI survey was designed so that only those between 15 and 49 years were asked to respond. This age restriction was a peculiarity of this particular study and not a weakness of the WTP concept, which should include everyone in a total WTP estimate. The restriction was not nearly as binding in the estimate of the WTP for an HIV vaccine covered in Chapter 25 where those surveyed for their WTP had to be in the age range 18 to 60 years. To sum up: if, which may in some cases be a “big if”, benefit estimates actually do represent total WTP, then automatically everyone’s preferences are incorporated in the CBA outcome. One does not need to set up an organization to represent each and every group who make up the total. If hospital staff have a bias against women, PLWAs and the elderly these people will go to clinics where there is less bias. If the hospital treatment is then being compared to the clinic treatment on the basis of a CBA using WTP, the clinic will have higher net benefits and the hospital treatment will be closed and the clinic will be kept open. The stigmatized groups do not need special representation; their WTP would have “spoken up” for them.

SUMMARY AND CONCLUSIONS Some are going to be disappointed with the main findings of this book. The conclusion is not going to be that investing in X has the highest priority while intervening through Y should be discontinued. This is not to say that at this point in time there are not promising generic categories of intervention. Male circumcision and information and education programs have been shown to be effective in a large number of different situations. However, even when generically promising interventions do exist, whether the promise will be actually realized will depend on the details of the intervention. If local conditions are such that inputs are freely available and costs are low, and outputs are highly valued by certain communities producing large benefits, then worthwhile interventions will have been

Conclusions II

205

identified. But, the local conditions may not be so supportive in other contexts. For example, if male circumcision is carried out by trained surgeons or doctors who know what they are doing, then widespread MC can be worthwhile. If, on the other hand, there are tribal doctors who carry out the male circumcision by using the same blunt knife for all patients, then MC can go horribly wrong as it could actually spread HIV transmission. If one is not going to list specific intervention priorities, what then is the main contribution of this book? The answer can be understood by an analogy. When economists are asked to help people on an island who are starving they respond by saying that to give the islanders fish will save them today, but to give them a fishing rod the islanders will be helped by having fish over their lifetimes. In the same way, we are suggesting a “fishing rod” solution to HIV/AIDS problems, where the rod this time is cost–benefit analysis. Interventions that have been suggested in the past (for example, condoms), or at present (male circumcision), or will appear in the future (for example, microbicides or vaccines) can all be evaluated by a common CBA methodology that does allow decision-makers to separate the wheat from the chaff. And what does one get when one uses CBA? One gets a comprehensive methodology that has been applied to many different policy decisions (for example, transport, the environment, education, agriculture, as well as all areas of health and mental health) that can deal with many different considerations (airport noise, road congestion, global warming, pain and suffering from injuries, and so on). As a result, one already has available a framework that can deal with all the ingredients that make HIV/AIDS interventions impossible to evaluate by guesswork and common sense (for example, information programs seeking sexual behavior change and HIV testing where people do not want to know the outcomes). One does not need to reinvent the wheel and pretend that the CBA methodology does not exist by going back to first principles and asking basic questions, such as, “What do we want to achieve with HIV/AIDS interventions?”, and “How can we best go about securing what we want to achieve?”. Policy economists have been asking these sorts of questions for decades and have come up with an extensive CBA theoretical framework linked to an extensive body of real world applications. The reader can find out more about CBA by consulting the many text books for the theoretical and practical details. The purpose of this book is to motivate the reader to consult this literature and to try to convince the reader that all this hard work in finding out more about CBA is likely to be necessary and worthwhile, as there is no workable alternative to CBA for setting priorities for HIV/AIDS. To be sure, CBA does have its problems. But, these problems are basically of the nature of trying to get meaningful data and dealing

206

Setting priorities for HIV/AIDS interventions

with uncertainties about the effects of inputs on outputs. Methods to value the inputs and outputs once they are known are well established and there are a wide range of valuation methods available. The reader just needs to adopt the CBA valuation method he or she feels comfortable with and accept the resulting interventions that using that method endorses. Lastly, the reader needs to go out and advocate the use of CBA to all those who are involved in deciding on HIV/AIDS interventions. UNAIDS and the World Bank should be the first targets for this advocacy.

References Adimora, A.A., Schoenbach, V.J. and Doherty, I.A. (2006), “Concurrent Sexual Partnerships among African Americans in the Rural South”, Annals of Epidemiology, 14(3), 155–60. Adimora, A.A., Schoenbach, V.J., Martinson, M., Donaldson, K.H., Stancil, T.R. and Fullilove, R.E. (2004), “HIV and African Americans in the Southern United States: Sexual Networks and Social Context”, Sexually Transmitted Diseases, 33(7), S39–S45. Aliber, M. and Walker, C. (2006), “The Impact of HIV/AIDS on Land Rights: Perspectives from Kenya”, World Development, 34(4), 704–27. Arnesen, T. and Nord, E. (2006), “The Value of DALY Life: Problems with Ethics and Validity of Disability Adjusted Life Years”, British Medical Journal, 319(7222), 1423–5. Auvert, B., Males, S., Puren, A., Taljaard, D., Carael, M. and Williams, B. (2004), “Can Highly Active Antiretroviral Therapy Re-educate the Spread of HIV?”, Journal of the Acquired Immune Deficiency Syndrome, 36(1), 613–21. Badru, A. (2000), Tanzania Male Condom Consumer Profile Survey (TMCCPS-99), First Report, PSI/Tanzania Research Department Report No.1, Dar Es Salaam, Tanzania. Bailey, R.C., Egesah, O. and Rosenberg, S. (2008), “Male Circumcision for HIV Prevention: A Prospective Study of Complications in Clinical and Traditional Settings in Bungoma, Kenya”, Bulletin of the World Health Organization, 86(9), 669–77. Barnett, T. and Whiteside, A. (2006), AIDS in the Twenty-first Century: Disease and Globalization (2nd edition), New York: Palgrave Macmillan. Beegle, K. (2005), “Labor Effects of Adult Mortality in Tanzanian Households”, Economic Development and Cultural Change, 54(3), 655–83. Beegle, K. and Krutikova, S. (2006), “Impact of HIV/AIDS on Transition into Marriage”, World Bank Policy Research Working Paper No. 4139, Washington, DC: World Bank. Beegle, K., De Weert, J. and Dercon, S. (2008), “Adult Mortality and Consumption Growth in the Age of HIV/AIDS”, Economic Development and Cultural Change, 57(2), 299–326. 207

208

Setting priorities for HIV/AIDS interventions

Bloom, D.E. and Glied, S. (1991), “Benefits and Costs of HIV Testing”, Science, 252(5014), 1798–1804. Boily, M.-C., Desai, K., Masse, B. and Gumel, A. (2008), “Incremental Role of Male Circumcision on a Generalized HIV Epidemic through its Protective Effect against Other Sexually Transmitted Infections: from Efficacy to Effectiveness to Population-level Impact”, Sexually Transmitted Infections, 84(2), ii28–ii34. Bollinger, L.A. (2008), “How Can We Calculate the ‘E’ in CEA?”, AIDS, 22(1), S51–S57. Boozer, M.A. and Philipson, T.J. (2000), “The Impact of Public Testing for Human Immunodeficiency Virus”, The Journal of Human Resources, 35(3), 419–46. Brent, R.J. (1991), “A New Approach to Valuing a Life”, Journal of Public Economics, 44(3), 165–71. Brent, R.J. (1996), Applied Cost–Benefit Analysis, Aldershot, UK and Brookfield, VT, USA: Edward Elgar. Brent, R.J. (1998), Cost–Benefit Analysis for Developing Countries, Cheltenham, UK and Lyme, NH, USA: Edward Elgar. Brent, R.J. (2003), Cost–Benefit Analysis and Health Care Evaluations, Cheltenham, UK and Northampton, MA: Edward Elgar. Brent, R.J. (2006), “Does Female Education Prevent the Spread of HIVAIDS in Sub-Saharan Africa?”, Applied Economics, 38(5), 491–503. Brent, R.J. (2009a), “A Social Cost–Benefit Criterion for Evaluating Voluntary Counseling and Testing with an Application to Tanzania”, Health Economics (forthcoming). Brent, R.J. (2009b), “An Implicit Price of a DALY for Use in a Cost– Benefit Analysis of ARVs”, Applied Economics (forthcoming). Brent, R.J. (2009c), “A Cost–Benefit Analysis of a Condom Social Marketing Program in Tanzania”, Applied Economics, 41(4), 497–509. Brent, R.J. (2009d), “A Cost–Benefit Analysis of Female Primary Education as a Means of Reducing HIV-AIDS in Tanzania”, Applied Economics, 41(14), 1731–43. Brent, R.J. (2009e), “Why Countries have Both Subsidized and Free Condoms to Prevent HIV/AIDS: The Role of Stigma Mitigation”, Department of Economics, Fordham University. Brent, R.J. (2009f), “Female Education and HIV/AIDS: The Condom Piece of the Puzzle”, Department of Economics, Fordham University. Canning, D. (2006), “The Economics of HIV/AIDS in Low Countries: The Case For Prevention”, Journal of Economic Perspectives, 20(3), 121–42. Caraël, M. (1995), “Sexual Behaviour”, in J. Cleland and B. Ferry (eds), Sexual Behaviour and AIDS in the Developing World, London: Taylor and Francis for the World Health Organization, pp. 75–123.

References

209

Caulfield, L.E., Richard, S.A., Rivera, J.A., Musgrove, P. and Black, R.E. (2006), “Stunting, Wasting, and Micronutrient Deficiency Disorders”, in Disease Control Priorities in Developing Countries (2nd edition), New York: Oxford University Press, Ch. 28, pp. 551–68. CDC, Centers for Disease Control and Prevention (2008), Morbidity and Mortality Weekly Report, 3 October, available at http://www.cdc.gov/ hiv (accessed 23 September 2009). Cohen, J. (2004), “Why is There Such a High Percentage of HIV and AIDS among Black Women?”, Slate Magazine, posted 27 October 2004, available at http://www.slate.com/id/2108724/ (accessed 26 September 2009). Crease, A., Floyd, K. and Guinness, L. (2002), “Cost-effectiveness of HIV/ AIDS Interventions in Africa: A Systematic Review of the Evidence”, The Lancet, 359(9318), 1635–42. Dayton, J. (1998), “World Bank HIV/AIDS Interventions: Ex-ante and Expost Evaluation”, World Bank Discussion Paper No. 389, Washington, DC: World Bank. De Walque, D. (2006), “Who Gets AIDS and How? The Determinants of HIV Infection and Sexual Behaviors in Burkina Faso, Cameroon, Ghana, Kenya and Tanzania”, World Bank Policy Research Working Paper No. 3844, Washington, DC: World Bank. De Walque, D. (2007), “How Does the Impact of an HIV/AIDS Information Campaign Vary with Educational Attainment? Evidence from Rural Uganda”, Journal of Development Economics, 84(2), 686–714. Drain, P.K., Smith, J.S., Halperin, D.T. and Holmes, K.K. (2004), “Correlates of National HIV Seroprevalence: An Ecological Analysis of 122 Developing Countries”, Epidemiology and Social Science, 35(4), 407–20. Drain, P.K., Halperin, D.T., Hughes, J.P., Klausner, J.D and Bailey, R.C. (2006), “Male Circumcision, Religion, and Infectious Diseases: An Ecological Analysis of 118 Developing Countries”, BMC Infectious Diseases, 6(1), 172 (electronic publication 30 November 2006, doi: 10.1186/1471-2334-6-172). Eley, B.S. et al. (2006), “Antiretroviral Treatment for Children”, South African Medical Journal, 96(9), 988–93. England, R. (2008), “The Writing is on the Wall for UNAIDS”, British Medical Journal, 336(7652), 1072. Epstein, H. (2007), The Invisible Cure: Africa, the West and the Fight Against AIDS, New York: Farrar, Straus and Giroux. Fawzi, W.W. et al. (1998), “Randomized Trial of Effects of Vitamin Supplements on Pregnancy Outcomes and T Cell Counts in HIV-1Infected Women in Tanzania, The Lancet, 351(9114), 1477–82.

210

Setting priorities for HIV/AIDS interventions

Fawzi, W.W. et al. (2004), “Randomized Trial of Multivitamin Supplements and HIV Disease Progression and Mortality, New England Journal of Medicine, 351(1), 1 July, 23–32. Gillespie, S. and Haddad, L. (2002), “Food Security as a Response to AIDS”, International Food Policy Research Institute’s 2001–2002 Annual Report: AIDS and Food Security, pp. 10–16, available at: www. ifpri.org (accessed 24 September 2009). Gillespie, S. and Kadiyala, S. (2005), “HIV/AIDS and Food and Nutrition Security: From Evidence to Action”, Food Policy Review No.7, Washington, DC: International Food Policy Institute. Gillespie, S., Haddad, L. and Jackson, R. (2001), HIV/AIDS. Food and Nutrition Security: Impacts and Actions, Washington, DC: International Food Policy Institute. Global Campaign for Education (2004), Learning to Survive: How Education for all would Save Millions of Young People from HIV/AIDS, available at http://www.schoolsandhealth.org/Documents/Learning%20 to%20Survive.pdf (accessed 6 October 2009). Glynn, J.R., Carae, M., Buvé, A., Anagonou, S., Zekeng, L., Kahindo, M. and Musonda, R. (2004), “Does Increased General Schooling Protect against HIV Infection? A Study in Four African Cities”, Tropical Medicine and International Health, 9(1), 4–14. Gray, P.G. (2004), “HIV and Islam: Is HIV Prevalence Lower Among Muslims?”, Social Science and Medicine, 58(9), 1751–6. Grüne-Yanoff, T. (ND), “Mismeasuring the Value of Statistical Life”, available at http://www.infra.kth.se/~gryne/VLS061120.pdf (accessed 11 May 2009). Guinness, O. (2001), Doing Well and Doing Good, The Trinity Forum Study Series, Colorado Springs: NavPress. Haacker, M. (2006), “When You are Talking a Possible $1,000 a Day TaxFree, It’s Real Attractive”, seminar paper, International Monetary Fund. Halperin, D.T. and Epstein, H. (2004), “Concurrent Sexual Partnerships Help to Explain Africa’s High HIV Prevalence: Implications for Prevention”, The Lancet, 364(9428), 4–6. Hargreaves, J.R. and Glynn J.R. (2002), “Educational Attainment and HIV-1 Infection in Developing Countries: A Systematic Review”, Tropical Medicine and International Health, 7(6), 489–98. Hargreaves, J.R. et al. (2008), “Systematic Review Exploring Time Trends in the Association Between Educational Attainment and Risk of HIV Infection in Sub-Saharan Africa”, AIDS, 22(3), 403–14. HelpAge International (2004), “The Cost of Love: Older People in the Fight Against AIDS in Tanzania”, available at www.helpage.org/ Resources/Researchreports (accessed 1 October 2009).

References

211

Holtgrave, D.R. and Qualls, N.L. (1995), “Threshold Analysis and Programs for Prevention of HIV Infection”, Medical Decision Making, 15(4), 311–17. Irwin, A., Millen, J. and Fallows, D. (2003), Global AIDS: Myths and Facts, Cambridge, MA: South End Press. Karpiak, S.E., Shippy, R.A. and Cantor, M.H. (2006), Research on Older Adults with HIV, New York: AIDS Community Research Initiative of America. Knodel, J. (2008), “Poverty and the Impact of AIDS on Older Persons: Evidence from Cambodia and Thailand”, Economic Development and Cultural Change, 57(2), 441–75. Kremer, M. (1994), “Can Having Fewer Partners Increase Prevalence of AIDS?”, NBER Working Paper No. 4942, Cambridge, MA: National Bureau of Economic Research. Lagarde, E., Auvert, B., Caraël, M. et al. (2001), “Concurrent Sexual Partnerships and HIV Prevalence in Five Urban Communities of SubSaharan Africa”, AIDS, 15(7), 877–84. Laumann, E.O. and Youm, Y. (1999), “Racial/Ethnic Group Differences in the Prevalence of Sexually Transmitted Diseases in the United States: A Network Explanation”, Sexually Transmitted Diseases, 28(9), 250–61. Lewis, S. (2005), Race Against Time, Toronto: House of Anansi Press. Loevinsohn, M. and Gillespie, S. (2003), “Food Security, and Rural Livelihoods”, FCND Discussion Paper No. 157, Washington, DC: International Food Policy Institute. Mackay, J.L. (2001), “Why Have Sex?”, British Medical Journal, 322(7289), 666. MEASURE (2001), AIDS in Africa During the Nineties: Tanzania, National AIDS Control Programme, Tanzania; Bureau of Statistics, Tanzania; Carolina Population Center, University of North Carolina at Chapel Hill. Miller, T.R. (2000), “Variations between Countries in Values of Statistical Life”, Journal of Transport Economics and Policy, 34(2), 169–88. Moore, M.L. and Viscusi, W.K. (1988), “Doubling the Estimated Value of Life: Results using New Occupational Fatality Data”, Journal of Policy Analysis and Management, 7(3), 476–90. Morris, M. and Kretzschmar, M. (1997), “Concurrent Partnerships and the Spread of HIV”, AIDS, 11(5), 641–8. National Bureau of Statistics Tanzania (2002), Household Budget Survey 2000/2001, Final Report, Dar Es Salaam: United Republic of Tanzania. Nattrass, N. (2006), “AIDS, Gender and Access to Antiretroviral

212

Setting priorities for HIV/AIDS interventions

Treatment in South Africa”, Working Paper No. 178, University of Cape Town, Centre for Social Science Research (CSSR). Ngamvithayapong, J. (2007), HIV Testing for Life: HIV Testing for all Tuberculosis Patients, Chiang Rai, Thailand: Chiang Rai Rung Roje, Co. Ltd. Norton, E.C., Martin, R.F. and Wechsberg, W.M. (1998), “Threshold Analysis of AIDS Outreach and Intervention”, in D.R. Holtgrave (ed.), Handbook of Economic Evaluation of HIV Prevention Programs, New York: Plenum Press, Ch. 14. Oster, E. (2005), “Sexually Transmitted Infections, Sexual Behavior, and the HIV/AIDS Epidemic”, Quarterly Journal of Economics, 120(2), 467–515. Over, M. (1998), “The Effects of Societal Variables on Urban Rates of HIV infection in Developing Countries”, in M. Ainsworth, L. Fransen and M. Over (eds), Confronting AIDS: Evidence from the Developing World, Luxembourg: European Communities. Owens, D.K., Nease, R.F., Harris, R.A (1993), “Use of Cost-effectiveness and Value of Information Analyses to Customize Guidelines for Specific Clinical Practice Settings”, Medical Decision Making, 13, 395. Philipson, T. and Posner, R. (1995), “A Theoretical and Empirical Investigation of the Effects of Public Health Subsidies for STD Testing”, Quarterly Journal of Economics, 110(2), 445–74. Pisani, E. (2008), The Wisdom of Whores: Bureaucrats, Brothels and the Business of AIDS, New York: W.W. Norton & Company. Potts, M., Halperin, D.T., Kirby, D., Swidler, A., Marseille, E., Klausner, J.D., Hearst, N., Wamai, R.G., Kahn, J.G. and Walsh, J. (2008), “Reassessing HIV Prevention”, Science, 320(5877), 749–50. Propper, C. (1990), “Contingent Valuation of Time Spent on NHS Waiting Lists”, Economic Journal, 100(400), 193–9. Rispel, L. and Setswe, G. (2007), “Stewardship: Protecting the Public’s Health”, in S. Harrison, R. Bhana and A. Ntuli (eds), South African Health Review 2007, Durban: Health Systems Trust, available at http:// www.hst.org.za/publications/711 (accessed 24 September 2009). Rosen, S., Sanne, I., Collier, A. and Simon, J.L. (2005), “Hard Choices: Rationing Antiretroviral Therapy for HIV/AIDS in Africa”, Lancet, 365(9456), 354–6. Sachs, J. (2005), The End of Poverty, New York: The Penguin Press. Schelling, T.C. (1968), “The Life You Save May Be Your Own”, in S.B. Chase Jr (ed.), Problems in Public Expenditures Analysis, Washington, DC: Brookings Institution, pp. 127–61. Semba, R.D. and Tang, A.M. (1999), “Micronutrients and the Pathogenesis of Human Immunodeficiency Virus Infection”, British Journal of Nutrition, 81(3), 181–9.

References

213

Shelton, J.D. (2007), “Ten Myths and One Truth about Generalized HIV Epidemics”, Lancet, 370(9602), 1809–11. Shelton, J.D., Cassell, M.M. and Adetunji, J. (2005), “Is Poverty or Wealth at the Root of HIV?”, The Lancet, 366(9491), 1057–8. Sinanovic, E. and Kumaranayake, L. (2006), “Sharing the Burden of TB/HIV? Costs and Financing of Public–Private Partnerships for Tuberculosis Treatment in South Africa”, Tropical Medicine and International Health, 11(9), 1466–74. Squire, L. and van der Tak, H. (1975), Economic Analysis of Projects, Baltimore: Johns Hopkins University Press. Stillwaggon, E. (2002), “HIV/AIDS in Africa: Fertile Terrain”, Journal of Development Studies, 48(6), 1–22. Stillwaggon, E. (2006), AIDS and the Ecology of Poverty, New York: Oxford University Press. THIS (2005), Tanzania HIV/AIDS Indicator Survey 2003-04, Calverton, Maryland: Tanzania Commission for AIDS (TACAIDS), National Bureau of Statistics (NBS) and ORC Macro. Thomas, J.C. and Sampson, L.A. (2005), “High Rates of Incarceration as a Social Force Associated with Community Rates of Sexually Transmitted Infection”, Journal of Infectious Diseases, 191(Supplement), S55–S60. UNAIDS, Joint United Nations Programme on HIV/AIDS (2004), Epidemic Update 2004, available at www.unaids.org (accessed 21 September 2009). UNAIDS, Joint United Nations Programme on HIV/AIDS (2006a), Report on the Global AIDS Epidemic, A UNAIDS 10th Anniversary Special Edition, Geneva: available at www.unaids.org (accessed 21 September 2009). UNAIDS, Joint United Nations Programme on HIV/AIDS (2006b), Epidemic Update 2006, available at www.unaids.org (accessed 21 September 2009). UNAIDS, Joint United Nations Programme on HIV/AIDS (2007), Epidemic Update 2007, available at www.unaids.org (accessed 21 September 2009). UNAIDS, Joint United Nations Programme on HIV/AIDS (2008), Report on the Global HIV/AIDS Epidemic 2008, available at www.unaids.org (accessed 21 September 2009). University of California San Francisco (2003), “How do Sexual Networks Affect HIV/STD Prevention?”, a publication of the Center for AIDS Prevention Studies (CAPS) and the AIDS Research Institute, posted April 2003, available at http://www.caps.ucsf.edu/pubs/FS/pdf/ networksFS.pdf (accessed 26 September 2009). UN Millennium Project (2005), Combating AIDS in the Developing World,

214

Setting priorities for HIV/AIDS interventions

Task Force on HIV/AIDS, Malaria, TB, and Access to Essential Medicines, Working Group on HIV/AIDS. Viscusi, N.K. and Aldy, J.E. (2003), “The Value of a Statistical Life: A Critical Review of Market Estimates throughout the World”, Journal of Risk and Uncertainty, 27(1), 5–76. Voluntary HIV-1 Counseling and Testing Efficacy Study Group (2002), “Efficacy of Voluntary HIV-1 Counseling and Testing in Individuals and Couples in Kenya, Tanzania and Trinidad: A Randomised Trial”, Lancet, 356(9244), 103–112. Whiteside, A. (2008), HIV/AIDS: A Very Short Introduction, New York: Oxford University Press. Whittington, D., Matsui-Santana, O., Freiberger, J.J., Van Houtven, G. and Subhrendu, P. (2002), “Private Demand for a HIV/AIDS Vaccine: Evidence from Guadalajara, Mexico”, Vaccine, 20(19–20), 2585–91. World Bank (1993), World Development Report 1993: Investing in Health, New York: Oxford University Press. World Bank (1999), Confronting AIDS: Public Priorities in a Global Epidemic, revised edition, New York: Oxford University Press. World Bank (2002), HIV/AIDS and Education, Washington, DC: World Bank. Yamano, T. and Jayne, T.S. (2004), “Measuring the Impacts of Primeage Adult Death on Rural Households in Kenya”, World Development, 32(1), 91–119. Zarkin, G.A., Cates, S.C. and Gala, M.V. (2000), “Estimating the Willingness to Pay for Drug Abuse Treatment”, Journal of Substance Abuse Treatment, 18, 149–59.

Index ability to pay 185, 186, 191, 192, 193–4 abstinence 17–18, 19 acquired immunodeficiency syndrome see AIDS Adimora, A.A. 86, 90 Africa condom use 85 HIV prevalence rates 37 see also South Africa; Sub-Saharan Africa age discrimination 192, 199 age of first sex 95, 96, 98 agricultural households, impact of HIV adult death 68–74 consumption, long run 71–2 household’s downward spiral 68–9 labor supply, short run 70–71 parents’ finances 73 small-scale farm 69–70 women’s land rights 72–3 agricultural policies 75–7 AIDS 3, 11 complexity of 11 deaths worldwide 15 alcohol prohibition 65 Aldy, J.E. 154, 156 Aliber, M. 72 antioxidants 41–2 antiretroviral drugs (ARVs) 13, 15, 67, 75, 77, 196 black market 193 cost-effectiveness analysis 141, 144–5 rationing, criterion for 192 South Africa, use of 47 toxicity 75 Arnesen, T. 140 ARVs see antiretroviral drugs (ARVs) Asia, HIV prevalence rates 37 Auvert, B. 192

Badru, A. 188 Bailey, R.C. 66, 67 Barnett, T. 56, 57 Beegle, K. 70, 71, 72 behavioral disease, HIV/AIDS as 175 see also risky behavior black market antiretroviral drugs 193 blood donors with HIV 199 blood transfusions 4 Bloom, D.E. 180, 181, 182, 183, 184 Boily, M.-C. 65 Bollinger, L.A. 95 Boozer, M.A. 20 Botswana, income inequality 56 Brazil 60, 84, 85 bread program cost–benefit analysis 27–9 breastfeeding 76 Brent, R.J. 53, 55, 57, 58, 59, 60, 62, 64, 82, 117, 118, 122, 125, 126, 142, 143, 144, 147, 149, 150, 152, 158, 159, 161, 162, 169, 186, 189, 202, 203 budgets, variable 169–70 burden of disease 143, 200–201 Burkino Faso, HIV transmission knowledge 12–13 calorie increase, effect on HIV/AIDS 51–2 Cambodia, impact of HIV on agricultural households 73 Cameroon, HIV transmission knowledge 13 Canning, D. 195 capital punishment 66 Caraël, M. 84 casual partners 81, 82, 84 Caulfield, L.E. 49 CBA see cost–benefit analysis (CBA) CD4 T cells 45, 66, 192–3

215

216

Setting priorities for HIV/AIDS interventions

CEA see cost-effectiveness analysis (CEA) Centers for Disease Control and Prevention (CDC) budget 112 choices in policy-setting, need for 7–10 circumcision 65–7, 178, 205 Cohen, J. 86 commercial sex workers 4, 56 , 65, 122, 163 common sense 3, 171, 205 commodification 172–5 competitive markets 117–19, 126 concurrency 84–8, 91, 198 African Americans 86–7 Sub-Saharan Africa 85–6 condom promotion 16, 85, 122, 169 condom social marketing program 117–18, 122–3, 127, 169, 202–3 distribution weights 187–8 private benefits and costs 123–5 social benefits and costs 125–6, 169 condom use 81–2, 95–8 and education 62 and low income 60 in marriage 62 consumer sovereignty 177–8, 179 consumer surplus 124 contingency valuation 114 cost minimization 99, 128–31 as a cost–benefit analysis 128–9 costs, what to include 129–30 strengths and weaknesses 130–31 TB treatment in South Africa 132–3 conclusions 135–6 costs, estimating 133–5 drug costs 135 patient costs 134 cost–benefit analysis (CBA) 4, 8, 164, 201, 205–6 distribution weights 186–90, 194 economic efficiency 176–9 equity 185–90 first level (101) 27–9 HIV testing, evaluation 180–84 private and social perspectives, comparison 182–3 private perspective 180–82 social perspective 182 second level (201) 30–32

social inclusiveness 201–5 value judgements 176, 177, 190 cost-effectiveness analysis (CEA) 99, 137–8, 195 antiretroviral drugs 141, 144–5 as cost–benefit analysis 138–9 revealed preference approach 141–2, 145 strengths 139 weaknesses 139–40, 169–70 costs 99–100 cost-utility analysis 137 Côte d’Ivoire, sexual partners 85 counterintuitive results abstinence 17–18, 19 marriage 18–19 testing 19–20 Creese, A. 138, 195 crop selection 76 cross-section analysis 55 DALYs see disability adjusted life years (DALYs) Dayton, J. 197 De Walque, D. 12, 13, 61, 66 death penalty 66 Democratic Republic of Congo as origin of epidemic 82 denial 11, 46–7, 198–9 developing countries, HIV prevalence rates comparison 54–5 diminishing marginal utility, law of 119, 157 disability adjusted life years (DALYs) 50, 110–12, 137, 140, 142, 143, 199 discordant couples 158 discrimination 192, 193 disease progression 192–3 distribution weights 186–7, 189–90, 194 condom social marketing project 187–8 criticisms of 188–9 Drain, P.K. 65 economic efficiency 176–9 consumer sovereignty 177–8, 179 willingness to pay 176 education and HIV 58–63 African Americans 87

References Index and condom use 62 education/income positive relation 62–3 female education evaluation 150–52 HIV positive link 53, 58–9 intervention types 150 neutral relation 61 positive/negative link paradox 59–61, 62–3 universal primary education 61 education programs 105, 150, 178 threshold analysis 105–8 effects, threshold estimate 106–7 intervention 105–6 threshold value 107–8 education vaccine 61, 62 effectiveness of interventions 95–100 elderly people discrimination 192 and HIV 199, 203–4 Eley, B.S. 47 The End of Poverty (Sachs) 4–5 England, R. 200 epidemics, types of 3–4, 132 Epstein, H. 56, 85, 87, 122 equity ability to pay 185–90 distribution weights 186–7 non-price rationing 191–4 time as rationing system 191–2 willingness to pay 120, 185–6 Europe, HIV prevalence rates 37 external costs 130 Fawzi, W.W. 44, 46 female education 58, 63 effectiveness evaluation 150–52 and HIV positive relation 53, 58–9 see also women fixed budgets 169–70 free radicals 41–2 gender and HIV transmission knowledge 12–13 and HIV transmission rates 81, 82 gender inequality 201–2 generalized HIV epidemics 4, 36 genital ulcers 80

217

GFATM 15, 143, 145 Ghana, HIV transmission knowledge 13 Gillespie, S. 4, 41, 68, 75–6 Glied, S. 180, 181, 182, 183, 184 Global Fund to Fight AIDS, Tuberculosis and Malaria see GFATM Glynn, J.R. 60, 61 Gray, P.G. 64 Grüne-Yanoff, T. 157 Guinness, O. 173 Haacker, M. 160 Haddad, L. 41 Halperin, D.T. 85 Hargreaves, J.R. 60, 61 harm reduction interventions 175, 178 health care firms’ packages 181 non-price rationing 191–4 healthy-looking people 12–13, 16 HelpAge International 199 heterosexual transmission of HIV 5, 38, 39, 78–9 high-risk groups 4, 17, 39, 86, 90 interventions, effectiveness of 95–8, 122 targeted for intervention 197 testing 20, 182 see also risky behavior HIV 3–4 HIV education, evaluation 105–8 effects, threshold estimate 106–7 intervention 105–6 threshold value 107–8 HIV prevalence rates developing countries, comparison 54–5 Muslim countries 64 United States 38–9 worldwide 36–8 HIV testing 183–4 counterintuitive results 19–20 and the elderly 203–4 private evaluation 180–82 private and social perspectives, comparison 182–4 social evaluation 182

218

Setting priorities for HIV/AIDS interventions

see also voluntary counseling and testing (VCT) HIV transmission 4 and donor instructions 178 and education 105–8 healthy-looking people 12–13 knowledge of 11–13, 16 malnutrition, direct role of 41–2 malnutrition, indirect role of 42–3 mother-to-child 4, 105, 138 sexual behavior 5, 78–83 sexually transmitted diseases, role in 80 and testing 20 HIV/TB dual epidemic 132 Holtgrave, D.R. 106, 109, 110, 112 home-making services, value of 174 homosexual men, micro-nutrient deficiencies 45 see also MSMs housework, value of 173–4 human capital approach 146–9, 168 female primary education evaluation 150–52 strengths and weaknesses 149 voluntary counseling and testing 160, 162 human immunodeficiency virus see HIV hunger see malnutrition illegal activities 174, 175 immune system 11, 41, 44, 79, 132, 199 incarceration 86–7, 90, 198 income and education relation 62–3 and HIV 53–7, 60, 86 negative relation 54, 57 positive relation 54, 55–6, 57, 60 income inequality 56–7 India, HIV cases 14 information programs 150, 178 interracial sex 90 interventions 4, 74, 85 and agricultural policy 75–7 as commodities 173–4 details, importance of 168–9 education, types of 150 effectiveness of 95–100 priority-setting

burden of disease 199–200 denial 198–9 difficulties of 4 labeling 195–8, 200 social inclusion 201–4 intravenous/injecting drug users (IDUs) 4 HIV education 105–6, 107–8 HIV transmission 39, 90 micro-nutrient deficiencies 44, 45 Irwin, A. 13, 14 Islam and HIV rates 53, 64–7 Jayne, T.S. 69 Kadiyala, S. 4 Kagera see Tanzania: agricultural households, impact of HIV Karpiak, S.E. 199 Kenya agricultural households, impact of HIV on small-scale farm 69–70 on women’s land rights 72–3 HIV transmission knowledge 12–13 land ownership 72–3 male circumcision 66 marriage and HIV levels 19 sexual partners 84, 85 women’s land rights 72–3 Knodel, J. 73 knowledge of HIV transmission 11–13, 16 Kremer, M. 17–18 Kretzschmar, M. 85 Krutikova, S. 72 Kumaranayake, L. 132, 134 labeling, evaluation by 195–8, 200 Lagarde, E. 87 Latin America, HIV prevalence rates 37 Laumann, E.O. 89 Lesotho, sexual partners 84, 85 Lewis, S. 9–10, 15, 26, 202 localized HIV epidemics 4, 36 Loevinsohn, M. 75–6 Mackay, J.L. 66 macro-malnutrition 41, 42, 51–2

References Index malaria 8, 79 intervention analysis 30–32 Malawi, agriculture and HIV link 75 male circumcision 65–7, 178, 205 males having sex with males (MSMs) 4, 39, 67, 86, 90, 97 malnutrition 41, 57, 79 and antiretroviral drugs 75 calorie increase, effectiveness 51–2 cause of AIDS claim 46–7 as consequence of HIV/AIDS 68–9 at country level 50–51 and HIV, vicious circle 41–2 HIV transmission, direct role in 41–2 HIV transmission, indirect role in 42 measures of 47, 49–50 see also micro-nutrient deficiencies Manila, concurrent partners 85 marginal benefits 27–8, 30–31, 117–19 marginal costs 27–8, 30–31, 117–18 markets 117–19, 126 marriage and concurrency 87 condom use in 62 as HIV risk factor 18–19 Mbeki, Thabo 46 MEASURE 79 medication see antiretroviral drugs (ARVs) Mexico, willingness to pay for vaccine study 113–16 micro-malnutrition 41 see also malnutrition micro-nutrient deficiencies 42, 44, 45 micro-nutrients 41, 44 multivitamin supplementation 44–7 vitamin A 44, 46, 47–8 Millennium Development Goals (MDGs) 5, 21, 24 feasibility as economic issue 24–5 feasibility as political issue 25–6 Miller, T.R. 155, 157 mining industry 56, 86, 133, 154, 157 mitigation 77, 169, 178 monogamy 84, 85 Moore, M.L. 154 Morris, M. 85 mosquitoes 79 mothers with HIV 4, 46, 47, 193

219

mother-to-child transmission of HIV 4, 105, 138 MSMs see males having sex with males (MSMs) multivitamin supplementation 44–6 South Africa, use of 46–7 Muslim countries and HIV negative relationship alcohol prohibition 65 capital punishment 66 male circumcision 65–7 Muslim population and HIV rates 53, 64 myths 13–15, 197–8 National Health Service (NHS) 191, 192 National Institute on Drug Abuse (NIDA) 105 nations’ values 174 Nattrass, N. 15, 47 network analysis 88–9 African American HIV rates 89–91 Ngamvithayapong, J. 132 NHS see National Health Service (NHS) Niger, religion and HIV rates 53–4 Nigeria, HIV cases 14 non-governmental organizations (NGOs) 7, 133 non-price rationing of health care 191–4 Nord, E. 140 North America, HIV prevalence rates 37 North Carolina Cooperative Agreement Program (NC CoOP) 105 Norton, E.C. 105, 106, 107, 108 nutrition see malnutrition; micronutrient deficiencies; micronutrients nutrition supplements 77 cost–benefit analysis 27–9 obesity 56–7 occupational fatality rates 154 Oceania, HIV prevalence rates 37 opportunity costs 8, 130, 133 Oster, E. 80, 82, 84

220

Setting priorities for HIV/AIDS interventions

Oster simulation model 80–83 Over, M. 56, 58 Owens, D.K. 112 oxidative stress 42 panel data 55 parasitic diseases 42, 43, 79 partnerships see sexual partners patient costs 174 people living with HIV/AIDS (PLWA) 36–7, 201, 203 Philipson, T.J. 19, 20 Pisani, E. 39, 75, 77, 197 policy-setting, difficulties 4–5 Posner, R. 19 Potts, M. 122 poverty 5, 54, 77, 80, 91 preferences 141–2, 164, 178 non-price rationing 194 of people living with AIDS 203 willingness to pay 185, 189, 201 of women 202–3 pregnant women with HIV, decline in 15 micro-nutrient deficiencies 44, 45, 72 multivitamin supplementation 44–6 President’s Emergency Plan for Aids Relief (PEPFAR) 15 prevention programs 14, 95–8, 195–6, 199 priority-setting burden of disease 199–200 denial 198–9 difficulties of 4 labeling 195–8, 200 social inclusion 201–4 “prison industry” 86 see also incarceration private costs 130 Propper, C. 192, 193 public–private partnership (PPP) and cost minimization 132–6 Qualls, N.L. 106, 109, 110, 112 Race Against Time (Lewis) 9–10, 202 Rath Health Foundation 47 rationing of health care non-price methods 191–4

socio-economic criteria 193 by time 191–2, 193 religion as determinant of HIV see Islam and HIV rates resource constraints 9–10 revealed preference approach 141–2, 145 Rio de Janeiro, sexual partners 84, 85 risky behavior 11, 19–20, 54, 65, 175, 178 see also high-risk groups Rispel, L. 56 Rosen, S. 191, 192, 193, 194 rural–urban migration 56 Sachs, J. 4–5 Sampson, L.A. 90 scaling up 196–7 Schelling, T.C. 9, 153 schistosomiasis 43, 79 Semba, R.D. 41, 42, 44, 45 serial monogamy 84, 85 Setswe, G. 56 sexual behavior concurrency 84–7 HIV transmission 5, 78–83 networks 88–91 Oster simulation model 80–83 sexual networks 88–9 sexual partners casual 81, 82, 84 concurrency, African Americans 86–7 concurrency, Sub-Saharan Africa 85–6 serial monogamy 84, 85 sexually transmitted diseases (STDs) 43, 60, 80 sexually transmitted infections (STIs) treatment 95, 97, 98 Shelton, J.D. 85, 198 Sinanovic, E. 132, 134 Singapore, sexual partners 85 social benefits of condom social marketing program 125–6, 169 of HIV testing 182–3 social complexity of AIDS 11 social costs 130, 133–4, 135, 183

References Index social evaluation 180, 182–4 social inclusiveness elderly with HIV 203–4 people living with AIDS 203 women and HIV 201–3 socialism 183 South Africa antiretroviral drugs (ARVs) 15, 47 denial of HIV as cause of AIDS 46–7 HIV cases 14, 37 income inequality 56 multivitamins as AIDS cure 46–7 obesity 56–7 TB treatment, cost minimization 132–3 costs, estimating 133–5 drug costs 135 patient costs 134 vitamin A in HIV prevention 47–8 Squire, L. 186 Sri Lanka, concurrent partners 85 STDs see sexually transmitted diseases (STDs) stigma 11, 91, 169 Stillwaggon, E. 42, 43, 44, 51, 52, 55, 56, 78, 79, 80 STIs see sexually transmitted infections (STIs) treatment stunting 47, 49–50 Sub-Saharan Africa concurrency 85–6 condom use 82 elderly women, vulnerability to HIV 199 female education and HIV 53, 59, 61 HIV determinants 53–4 HIV prevalence rates 37, 80–83 HIV transmission 78–9, 80, 81, 82 hunger as consequence of HIV/ AIDS 68–9 malnutrition 50–51 marriage and HIV 19 parasitic diseases 43 sexual behavior and HIV transmission 81, 82 and United States, HIV comparison 90–91 switching values 101

221

Tang, A.M. 41, 42, 44, 45 Tanzania agricultural households, impact of HIV on consumption, long run 71–2 on labor supply, short run 70–71 agriculture and HIV link 75 condom social marketing program 117–18, 122–3, 127, 169, 202–3 distribution weights 187–8 private benefits and costs 123–5 social benefits and costs 125–6, 169 condom use 85 education of females 60, 147–9 and HIV 60, 62–3 primary 147–9 elderly women, vulnerability to HIV 199 HIV transmission knowledge 12, 13 income 55, 57, 188 multivitamin supplementation 44–6 sexual partners 84, 85 voluntary counseling and testing 158–64 targets 21–3 desirability of 21–3 feasibility 21 as economic issue 24–5 as political issue 25–6 TB treatment, cost minimization 128–9, 132–6 TB/HIV dual epidemic 132 T-cell counts 45, 46, 66, 192–3 testing see HIV testing; voluntary counseling and testing (VCT) Thailand agricultural households, impact of HIV 73 condom use in brothels 122 sexual partners 84, 85 THIS 12 Thomas, J.C. 90 “3 by 5 Initiative” 21, 26 threshold analysis 101–2 HIV avoidence, benefits 109–12 benefits, calculation 112

222

Setting priorities for HIV/AIDS interventions

benefits, threshold estimate 109–10 DALY, estimating 110–12 HIV education, evaluation 105–8 effects, threshold estimate 106–7 intervention 105–6 threshold value 107–8 HIV/AIDS vaccine, cost 113–16 benefits, contingency valuation 114 costs, threshold estimate 114 willingness to pay, estimating 114–16 strengths 103 weaknesses 103–4 time as rationing system 191–2 time series studies 55 trade-offs 8–10, 76 transmission see HIV transmission transport infrastructure 76 Tshabalala-Msimang, Dr. 46–7 Uganda 19, 61, 122 UNAIDS 15, 16, 19, 196 HIV/AIDS data 35–6 underweight 47, 50 United Kingdom 155 HIV contracted abroad 14 HIV transmission knowledge 16 National Health Service (NHS) 191, 192 United States and Africa, HIV comparison 90–91 African Americans concurrency 86–7 education of 87 HIV prevalence 87, 89–90, 198 sexual networks 89–90 antiretroviral drugs (ARVs) 67 condom use 81–2 denial of HIV/AIDs problem 198 female HIV/AIDS rates 14 HIV prevalence rates 38–9, 80–83, 87, 198 HIV testing 19 HIV transmission 39, 81–2 incarceration 86–7, 90, 198 malnutrition, measures of 49–50 micro-nutrient deficiencies 44 occupational fatality rates 154

sexual partners 81, 84 value of a statistical life 155 unprotected sex 11, 19, 105, 160, 163 value judgments 176, 177, 190 value of a statistical life approach 153–5 foregone benefits 160, 163 strengths and weaknesses 156–7 voluntary counseling and testing 158–64 cost–benefit analysis 160–64 cost–benefit results 161–2, 163–4 costs 160, 163 effectiveness, estimating 158–9 van der Tak, H. 186 Viscusi, N.K. 154, 156 vitamin A 44, 45, 46, 47–8 vitamins shortage of 41, 42, 47–8 South Africa, HIV policy 46–7 supplementation 43, 44–8 voluntary counseling and testing (VCT) 158–64 cost–benefit analysis of 160–64 benefits 160 cost–benefit results 161–2, 163 costs 160, 163 effectiveness, estimating 158–9 human capital method 162 see also HIV testing Voluntary HIV-1 Counseling and Testing Efficacy Study Group 159 Walker, C. 72 wasting 47, 49 Whiteside, A. 56, 57, 113 Whittington, D. 113, 114, 115 widows’ land ownership 72–3 willingness to pay (WTP) 117–21, 172 condom social marketing program 122–3, 127 private benefits and costs 123–5 social benefits and costs 125–6, 169 and economic efficiency 176 elderly and HIV 204 equity concerns 120–21, 185, 186

References Index and non-price rationing 191–2 people living with AIDS 203 social inclusion 201 strengths 119–20 vaccine 113, 114–16 weaknesses 120–21, 185, 186 women and HIV 202–3 Wilson, Phill 86 The Wisdom of Whores (Pisani) 197 women education 58, 63 effectiveness evaluation 150–52 and HIV positive relation 53, 58–9 elderly, vulnerability to HIV 199 and HIV 201–3 land ownership 72–3 mothers with HIV 4, 46, 47, 193

223

preferences 202–3 see also gender workplace programs 96, 98 World Bank 9, 58, 62, 82, 150 lending program 197 World Development Report (1993) 140 World Food Program (WFP) 10 World Health Organization (WHO) “3 by 5 Initiative” 21, 26 WTP see willingness to pay (WTP) Yamano, T. 69 Youm, Y. 89 Zambia 19, 84 zero price 173, 175