E-Health Across Borders Without Boundaries: E-salus Trans Confinia Sine Finibus; Proceedings of the ERMI Special Topic Conference 14-15 April 2011 Lasko, Slovenia
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Studies in Health Technology and Informatics This book series was started in 1990 to promote research conducted under the auspices of the EC programmes’ Advanced Informatics in Medicine (AIM) and Biomedical and Health Research (BHR) bioengineering branch. A driving aspect of international health informatics is that telecommunication technology, rehabilitative technology, intelligent home technology and many other components are moving together and form one integrated world of information and communication media. The complete series has been accepted in Medline. Volumes from 2005 onwards are available online. Series Editors: Dr. O. Bodenreider, Dr. J.P. Christensen, Prof. G. de Moor, Prof. A. Famili, Dr. U. Fors, Prof. A. Hasman, Prof. E.J.S. Hovenga, Prof. L. Hunter, Dr. I. Iakovidis, Dr. Z. Kolitsi, Mr. O. Le Dour, Dr. A. Lymberis, Prof. J. Mantas, Prof. M.A. Musen, Prof. P.F. Niederer, Prof. A. Pedotti, Prof. O. Rienhoff, Prof. F.H. Roger France, Dr. N. Rossing, Prof. N. Saranummi, Dr. E.R. Siegel, Prof. T. Solomonides and Dr. P. Wilson
Volume 165 Recently published in this series Vol. 164. E.M. Borycki, J.A. Bartle-Clar, M.S. Househ, C.E. Kuziemsky and E.G. Schraa (Eds.), International Perspectives in Health Informatics Vol. 163. J.D. Westwood, S.W. Westwood, L. Felländer-Tsai, R.S. Haluck, H.M. Hoffman, R.A. Robb, S. Senger and K.G. Vosburgh (Eds.), Medicine Meets Virtual Reality 18 – NextMed Vol. 162. E. Wingender (Ed.), Biological Petri Nets Vol. 161. A.C. Smith and A.J. Maeder (Eds.), Global Telehealth – Selected Papers from Global Telehealth 2010 (GT2010) – 15th International Conference of the International Society for Telemedicine and eHealth and 1st National Conference of the Australasian Telehealth Society Vol. 160. C. Safran, S. Reti and H.F. Marin (Eds.), MEDINFO 2010 – Proceedings of the 13th World Congress on Medical Informatics Vol. 159. T. Solomonides, I. Blanquer, V. Breton, T. Glatard and Y. Legré (Eds.), Healthgrid Applications and Core Technologies – Proceedings of HealthGrid 2010 Vol. 158. C.-E. Aubin, I.A.F. Stokes, H. Labelle and A. Moreau (Eds.), Research into Spinal Deformities 7 Vol. 157. C. Nøhr and J. Aarts (Eds.), Information Technology in Health Care: Socio-Technical Approaches 2010 – From Safe Systems to Patient Safety Vol. 156. L. Bos, B. Blobel, S. Benton and D. Carroll (Eds.), Medical and Care Compunetics 6
ISSN 0926-9630 (print) ISSN 1879-8365 (online)
e-Health Across Borders Without Boundaries E-salus trans confinia sine finibus Proceedings of the EFMI Special Topic Conference 14–15 April 2011 Laško, Slovenia
Edited by
Lăcrămioara Stoicu-Tivadar University “Politehnica” Timişoara, Romania Chair of the Scientific Programme Committee
Bernd Blobel eHealth Competence Center, University Hospital Regensburg, Germany Vice-Chair of the Scientific Programme Committee
Tomaž Marčun Health Insurance Institute of Slovenia, Ljubljana, Slovenia Vice-Chair of the Scientific Programme Committee
and
Andrej Orel Marand d.o.o., Ljubljana, Slovenia Slovenian Representative to EFMI and IMIA
Preface This volume contains the proceedings of the 11-th EFMI Special Topic Conference that will be held in Laško, Slovenia on 14 and 15-th of April 2011. The EFMI STC 2011 is an important international forum for presenting results of current scientific work in health-informatics processes, systems, and technologies. Achievements in this area will be introduced to a global audience. As a major event for science, medicine, and technology, the conference provides a comprehensive overview and in-depth, first hand information on new developments, advanced systems, technologies and applications. The EFMI STC 2011 was organized by the European Federation of Medical Informatics (EFMI) in cooperation with Slovenian Medical Informatics Society (SDMI). It follows previous conferences in Bucharest, Romania (2001), Nicosia, Cyprus (2002), Rome, Italy (2003), Munich, Germany (2004), Athens, Greece (2005), Timişoara, Romania (2006), Brijuni Island, Croatia (2007), London, UK (2008), Antalya, Turkey (2009) and Reykjavik, Iceland (2010). The theme of the STC 2011 is “E-salus trans confinia sine finibus – e-Health across Borders without Boundaries”, addressing a range of important aspects of crossborder e-Health services, like Knowledge Representation, Patient Empowerment, Social Care, or Cross-border Interoperability. The EFMI Working Groups that are scientifically supporting the conference are “Healthcare Informatics for Interregional Cooperation”, “Primary Care Informatics”, “Electronic Health Records”, and “Security, Safety and Ethics”. The objective of the conference, reflected in the contents of the present volume, was to highlight health-related communication and collaboration at regional, national, and international level. Achieving and maintaining cross-border interoperability of electronic health record systems implies managing the continuous process of change and adaptation of a multitude of elements within and across electronic infrastructures in neighboring countries. The proceedings volume opens with excellent keynote and invited contributions covering cross-border e-Health from international perspective, p-Health interoperability, and education across borders. The volume contains 21 papers from specialists of the field and also PhD papers continuing the good EFMI tradition encouraging young scientists to participate and present valuable work. The papers have been selected by the Scientific Program Committee (SPC) out of 41 submissions of papers, posters, and workshop proposals sent by experts from 20 countries. Each submission has been reviewed by three reviewers which have been selected from a list of 486 internationally acknowledged domain experts from all continents. The SPC chair and vice-chairs are especially thankful to all reviewers listed in the proceedings. The scientific topics presented in the proceedings comprise interregional health information systems and applications, cross-border e-Health projects, patient and crossborder e-Health systems, multi-language and cultural issues in e-Health system, international standardization, assessment of e-Health systems, safety and security aspects of interconnected e-Health systems, e-communications in a global community, and crossborder education in health informatics.
vi
The topics presented at EFMI STC 2011 are interdisciplinary in nature and consequently of interest to a variety of professionals: medical informatics, bioinformatics, and health informatics scientists, medical computing and technology specialists, public health, health insurance and health institutional administrators, physicians, nurses, and other allied health personnel, as well as representatives of industry and consultancy in the various health fields. 74 researchers, residing in 16 countries from Europe, North America, and Asia have reported their results in this volume. The EFMI STC 2011 has been completed through 9 posters presented in a special session, and 3 workshops. The editors would like to thank all the authors for their excellent work as well as the reviewers for lending their expertise to the conference, thereby contributing to the final achievements. Furthermore, they are indebted to Slovenian Ministry of Health and HL7 International for sponsoring the print of the proceedings. Lăcrămioara Stoicu-Tivadar, Bernd Blobel, Tomaž Marčun, Andrej Orel (Editors) Acknowledgement. The editors are indebted to Ivan Eržen, SDMI President for hosting the conference, to Mojca Paulin, Danila Perhavec and Nina Dolenc for managing the conference Website, the registration process and extended communications. They thank Thomas Schabetsberger for facilitating the online congress management. Furthermore, they are grateful to Mihaela Vida from University “Politehnica” Timişoara for her careful work in formatting the proceedings on hand.
vii
Reviewers for EFMI STC 2011 The following experts contributed to the selection process of the papers: Laurence Alpay, Leiden Stig Kjaer Andersen, Aalborg Luis Antunes, Porto Kennedy Auma, Nairobi Paul Avillach, Paris Adnan Bajraktarevic, Sarajevo Laszlo Balkanyi, Stockholm Claudia Bartz, Milwaukee, WI Lejla Begic Fazlic, Sarajevo Elena Bernad, Timişoara Bernd Blobel, Regensburg Bernhard Breil, Münster Morten Bruun-Rasmussen, Copenhagen Oliver Burgert, Leipzig Ann Bygholm, Aalborg Oest James Cimino, Bethesda, MD Guillermo de la Calle, Boadilla del Monte, Madrid Kerstin Denecke, Hannover Judith Dexheimer, Nashville, TN Claudio Povo Eccher, Trento Jesualdo Fernandez-Breis, Tomas Murcia Gianluigi Fioriglio, Rome Yang Gong, Columbia, MO Dieter Hayn, Graz Harald Heinzl, Vienna Mira Hercigonja-Szekeres, Zagreb Vojtech Huser, Těrlicko
Miron Iancu, Bucharest Vassilis Koutkias, Thessaloniki Christian Lovis, Geneva Tatyana Lugovkina, Yekaterinburg Diana Lungeanu, Timişoara Tomaž Marčun, Ljubljana Mojca Paulin, Ljubljana Henning Müller, Sierre Jose Luis Oliveira, Aveiro Michael Onken, Oldenburg Andrej Orel, Ljubljana Louise Pape-Haugaard, Aalborg Niels Peek, Amsterdam Milan Petkovic, Eindhoven José Pinto, Simão de Paula Curitiba Besim Prnjavorac, Tesanj Petar Rajkovic, Nis Álvaro Rocha, Porto Samrend Saboor, Hall in Tyrol Thomas Schabetsberger, Mils Philip Scott, Portsmouth Michael Shifrin, Moscow Peter Spyns, Brussel Jürgen Stausberg, München Milton Stern, Staten Island Vasile Stoicu-Tivadar, Timişoara Lăcrămioara Stoicu-Tivadar, Timişoara Aristides Vagelatos, Athens Zhang Songmao, Beijing
This page intentionally left blank
ix
Contents Preface Lăcrămioara Stoicu-Tivadar, Bernd Blobel, Tomaž Marčun and Andrej Orel Reviewers for EFMI STC 2011
v vii
Keynotes “Initiate-Build-Operate-Transfer” – A Strategy for Establishing Sustainable Telemedicine Programs Not Only in the Developing Countries Rifat Latifi
3
Ontologies, Knowledge Representation, Artificial Intelligence – Hype or Prerequisites for International pHealth Interoperability? Bernd Blobel
11
E-Health in Graduate and Postgraduate Medical Education: Illusions, Expectations and Reality Ferenc Bari, Erzsébet Forczek and Zoltán Hantos
21
Knowledge Representation A Knowledge Management System to Study the Quality of Life in Head and Neck Oncology Patients Joaquim Gonçalves, Augusta Silveira and Álvaro Rocha
31
Detection of Nicotine Content Impact in Tobacco Manufacturing Using Computational Intelligence Lejla Begic Fazlic and Zikrija Avdagic
37
Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs Christoph Rinner, Michael Kohler, Gudrun Hübner-Bloder, Samrend Saboor, Elske Ammenwerth and Georg Duftschmid
43
Extracting Clinical Information to Support Medical Decision Based on Standards Valentin Gomoi, Mihaela Vida, Lăcrămioara Stoicu-Tivadar and Vasile Stoicu-Tivadar
49
Patient Empowerment, Social Care eHealth in Switzerland – Building Consensus, Awareness and Architecture Christian Lovis, Hansjorg Looser, Adrian Schmid, Judith Wagner and Stefan Wyss Patient Empowerment by Electronic Health Records: First Results of a Systematic Review on the Benefit of Patient Portals Elske Ammenwerth, Petra Schnell-Inderst and Alexander Hoerbst
57
63
x
Empowerment of Patients over Their Personal Health Record Implies Sharing Responsibility with the Physician Catherine Quantin, Eric Benzenine, Bertrand Auverlot, David-Olivier Jaquet-Chiffelle, Gouenou Coatrieux and François-André Allaert Personal Information Protection – Exceptional Challenges of Integrated Systems of eHealth Anka Bolka, Blaž Zadel and Martina Zorko
68
74
Research and Practice Reports XML as a Cross-Platform Representation for Medical Imaging with Fuzzy Algorithms Norbert Gal and Vasile Stoicu-Tivadar Fully Connected Emergency Intervention for the Critical Home Care System Draško Nakik, Suzana Loškovska and Vladimir Trajkovik
83 88
How to Ensure Sustainable Interoperability in Heterogeneous Distributed Systems Through Architectural Approach Louise Pape-Haugaard and Lars Frank
94
A Preliminary Study on Network Traffic Estimation in Using EPR with Thin-Client Computing over Wide Area Network Kei Teramoto, Shigeki Kuwata, Masaki Mochida and Hiroshi Kondoh
99
Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers Ksenija Lekić, Nuša Konec Juričič, Petra Tratnjek and Borut Jereb
105
E-Learning for Occupational Physicians’ CME: A Study Case M. Cristina Mazzoleni, Carla Rognoni, Enrico Finozzi, Tommaso Gri, Marco Pagani and Marcello Imbriani
111
Regional Medical Data Mining System Raul Robu and Vasile Stoicu-Tivadar
117
Assessment of Hospital Information Systems Implementation: A Case Study Dimitrios Zikos, Athanasios Mitsios and John Mantas
123
A Methodology to Assess Experiences in Implementing e-Health Solutions in Croatian Family Medicine Damir Kralj, Miroslav Končar and Stanko Tonković What Are the Barriers to Conducting International Research Using Routinely Collected Primary Care Data? Simon de Lusignan, Christopher Pearce, Nicola T. Shaw, Siaw-Teng Liaw, Georgios Michalakidis, Marília T. Vicente and Michael Bainbridge, International (IMIA) and European (EFMI), Medical Informatics Association and Federation, Primary Care Informatics Working Groups (PCI-WG)
129
135
xi
Interoperability Evaluation of Possibilities in Demographic Data Exchange Support in Czech Healthcare Miroslav Nagy, Libor Seidl and Jana Zvarova
143
Character Sets: An Invisible Pre-Requisite Towards Cross-Border Interoperability? Frank Oemig and Bernd Blobel
149
The Role of Basic Data Registers in Cross-Border Interconnection of eHealth Solutions Mirjana Kregar, Tomaž Marčun, Irma Dovžan and Lojzka Čehovin
155
Mapping the Finnish National EHR to the LOINC Kristiina Häyrinen and Juha Mykkänen
“Initiate-Build-Operate-Transfer” – A Strategy for Establishing Sustainable Telemedicine Programs Not Only in the Developing Countries Rifat LATIFI1 President and Founder, International e-Hospital Foundation Director, Telemedicine Program of Kosova and Albania Professor of Surgery, University of Arizona, Tucson, Arizona Sr. Consultant and Director, Trauma Services, Hamad General Hospital, Doha, Qatar
Abstract. Establishing sustainable telemedicine has become a goal of many developing countries around the world. Yet, despite initiatives from a select few individuals and on occasion from various governments, often these initiatives never mature to become sustainable programs. The introduction of telemedicine and e-learning in the Balkans has been a pivotal step in advancing the quality and availability of medical services in a region whose infrastructure and resources have been decimated by wars, neglect, lack of funding, and poor management. The concept and establishment of the International Virtual e-Hospital (IVeH) has significantly impacted telemedicine and e-health services in Kosova. The success of the IVeH in Kosova has led to the development of similar programs in other Balkan countries and other developing countries in the hope of modernizing and improving their healthcare infrastructure. A comprehensive, four-pronged strategy developed by IVeH “Initiate-Build-Operate-Transfer” (IBOT), may be a useful approach in establishing telemedicine and e-health educational services not only in developing countries, but in developed countries. The development strategy, IBOT, used by the IVeH to establish and develop telemedicine programs is described. IBOT includes assessment of healthcare needs of each country, the development of a curriculum and education program, the establishment of a nationwide telemedicine network, and the integration of the telemedicine program into the very core of healthcare infrastructure. The end point is the transfer of a sustainable telehealth program to the nation involved. By applying IBOT, a sustainable telemedicine program of Kosova and Albania has been established as an effective prototype for telemedicine in the Balkans. Once fully matured, the program is transitioned to the Ministry of Health, which ensures the sustainability and ownership of the program. Similar programs are being established in Macedonia, Montenegro and other countries around the world. The IBOT model has been effective in creating sustainable telemedicine and e-health integrated programs in the Balkans and may be a good model for establishing such programs in developing countries. Keywords. Sustainable telemedicine, e-learning, regions, model.
1
Corresponding Author: Rifat Latifi, MD, FACS, Professor of Surgery of Surgery, University of Arizona, Tucson, [email protected].
4
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
Introduction Obviously telemedicine is not new, however, as in the case with many innovations it taking decades for telemedicine to enter the mainstream as a healthcare delivery system. Nonetheless telemedicine and telehealth development have brought real hope to developing countries and many of the remote areas and yet has raised significant questions and anxiety among those forces hoping to retain the status quo of current medical practice. While most developments in telemedicine and other technologies in the medical field are being implemented among the developed world and prominent universities and institutions, the developing countries are the perfect examples where the telemedicine, c-health and relevant technologies are in great need. These countries need modern health information systems and advanced technology in order to become a part of the global community of medical practice. There is a need for significant medical and radical changes. So today, telemedicine is basically applied and is used from one continent to the other and one can find some elements of telemedicine in almost every part of the world. The penetration of the internet into the global scene has made this very possible. Transmission of medical information through the internet, as an attachment, enables consultation of doctors from one corner of the world is possible through these technologies. Today, the patient and the physician should not be alone anywhere in the world as long as there is some form of technology present and acceptable.
1. Establishing sustainable telemedicine programs Establishing sustainable telemedicine has become a goal of many developing countries around the world. Yet, despite initiatives from various organizations, only few mature to become sustainable programs. One successful program is the International Virtual eHospital (IVeH), whose mission is to create self-sustainable telemedicine and e-health programs around the world and to rebuild medical systems in the developing world, one country at a time. The organization uses telemedicine and collaboration as a platform by educating healthcare providers in the use, adoption, practice, and implementation of telemedicine, e-health and electronic libraries in order to narrow the gap created by the digital divide and healthcare imbalance. The implementation of telemedicine and e-learning in the Balkans has been pivotal in advancing the quality and availability of medical services in a region whose infrastructure and resources have been decimated by wars, neglect, lack of funding, and poor management. In the 1990s, Kosova, now an independent country, experienced an immense political and ethnic conflict, which left its medical services in a deplorable state. This severely hindered redevelopment of the healthcare system in Kosova, making telemedicine and medical distance learning the ideal solutions for the region. The idea to establish and implement telemedicine in the Balkans was presented for the first time at a 2000 G-8 telemedicine meeting in Berlin by Dr. Rifat Latifi, then assistant professor of Surgery at Virginia Commonwealth University in Richmond. The objective was to design and implement the Telemedicine Program of Kosova (TPK) as a sustainable and functional portal for information within and outside the region with a training center for telemedicine. The IVeH established the TPK in 2001 and it was inaugurated on Dec. 10, 2002. Since then, the program has connected seven hospitals
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
5
via telemedicine, provided access to an electronic medical library, and established local leadership to run the regional telemedicine centers. The TPK is world renowned as a model for telemedicine success in developing countries and has grown into the Balkan Telemedicine Program that now includes programs in Albania, Macedonia and Montenegro. 1.1. Technology Recent technological developments have made it possible to apply telemedicine in the management of trauma and emergency care, irrespective of where the patient is or where the trauma occurs. Remote areas and isolated communities have especially benefited. Today, access to trauma specialists and other highly trained caregivers during an emergency is possible from anywhere in the world, anytime, under any conditions, day or night. The physician and the patient do not have to be in the same room, the same town, the same village, or the same country, or even on the same continent. The astonishing technological developments, as well as the far-reaching vision and dreams of medical and technical leaders, simply were neither possible nor predictable just a few years ago. The technology is here, yet most of us cannot keep up with its developments. Medical and administrative communities of health care systems around the world need to become familiar with and embrace technology, so that they can act as the bridge to the digital divide between people, regions, and countries. What is variously known as telemedicine, telehealth, telepresence, teletrauma, or e-health is possible for practically every medical discipline and every condition, including dermatology, psychiatry, surgery, and radiology The IVeH telemedicine centers include video conferencing rooms, teleconsultation rooms and telemedicine training rooms, electronic libraries and electronic auditorium. Each room is equipped with Polycom’s high-definition room telpresence systems, Criticare Vital sign monitoring, and Teletrauma Real Time Monitoring and Capture Stations. A Polycom firewall traversal solution allows for seamless video collaboration with outside organizations, something that is critical to the success of the program. 1.2. Barriers and Challenges The most significant challenges in implementing telemedicine programs have been changing the mind set of physicians, and convincing governments of the need for such investments. Other challenges include administrative issues and managing projects with multiple players representing different backgrounds, cultures, education, goals, and mentalities. Creating true partnerships and transparency in the region has been the key to IVeH’s success. Many of us assume, albeit wrongly, that the business philosophy of “build-operatetransfer” of technology boom in few developing countries may be applied in e-health and telemedicine field as well. So, why does the “build-operate-transfer” philosophy not apply entirely? Well, this is far more complex than setting up “a call center” in India or in Bangladesh, thus that business philosophy does not apply entirely, although the goal is the same: “build-operate-transfer”. We should not see creating telemedicine and telehealth systems in any region, but especially in developing countries, as creating few profitable call centers, or building a factory in the middle of nowhere. The motives have to be different, deeper, and certainly, more transparent and may be more complex,
6
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
as the profitability is not the order of business and may not be visible for many years or decades. Predicting success is difficult, if not impossible. This is not the same as the Injury Severity Score (ISS) that predicts the mortality and morbidity of trauma patients. Commitment, individual, institutional and governmental is mandatory. But this is common senses, but we know that common sense may not prevail in our global business world.
2. Materials and methods An analysis of Initiate-Build-Operate-Transfer strategy related activities that led to initiation and establishment of international telemedicine programs in the Balkans is performed. Outcomes are described herein.
3. Results Application of Initiate-Build-Operate-Transfer strategy, the IVEH has succeeded in establishing world renowned program in Kosova and Albania as a model for telemedicine success in developing countries and has grown into the Balkan Telemedicine Program and is spreading to other countries in the world. Establishment of telemedicine programs can be divided into several phases. An initiation process that includes a well-publicized intensive telemedicine and e-health seminar that introduces to the physicians, nurses, and politicians the concepts and principles of telemedicine and e-health. Using the seminar as a catalyst, a technical assessment of the needs of the country and its health facilities is conducted. This technical assessment documents the details for the establishment of an integrated telemedicine and e-health program and specific needs for individual facilities.
4. Description of the strategy 4.1. The Initiation Process Balkans Intensive Telemedicine and e-Health Seminars Three advanced intensive telemedicine and e-health seminars have successfully been held to date in Kosova (2002), Albania (2007) and Macedonia (2009), and Montenegro (2010). These seminars have served as galvanizing events for the adoption of telemedicine in the Balkans. These seminars have served well as introduction of telemedicine and e-health in the region and have created a large number of health professionals armed with new capabilities. Over 1,500 physicians, IT personnel, allied health professionals, etc. from the Balkans Region have attended these seminars and other virtual lectures. The 1st Intensive Balkan Telemedicine Seminar was held in October of 2002 with over 400 people from 21 countries. This began a new chapter in the history of telemedicine and telehealth in the Balkans. The success of the first seminar paved the way for subsequent seminars which have been held in other Balkan countries. During
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
7
the next few years, the telemedicine program of Kosova grew in popularity, content, and intensity. The efforts of this first meeting were published in a book entitled, “Establishing Telemedicine in Developing Countries: From Inception to Implementation” which was published in 2004. The 3rd Intensive Balkan Telemedicine and e-Health seminar, organized by the IVeH in collaboration with the Ministry of Health of Macedonia, was held in Skopje, Macedonia, February 6-7, 2009. This seminar was dedicated to discussion of clinical applications and evidence-based outcomes of telemedicine and e-health technology as it pertains to current technologies, principles, practices and applications of telemedicine and e-health. It was very helpful for all medical and technical, as well as administrative healthcare personnel of hospitals and private practices, governmental agencies, ministries of health and other institutions and organizations dealing with healthcare issues. Hands-on experience opportunities and ample time for discussion with world-renowned experts was possible during the seminar. The introduction of telemedicine programs in countries that lack basic medical care standards is a complex process and needs to be carefully planned if one is to be successful. Introducing these countries with such seminars ensures that the many different aspects of telemedicine are discussed and the special needs for each country can be identified. The seminars solidified the initiative to bring telemedicine to the Balkans and the IVeH began working with governmental agencies to bring telemedicine to other countries around the world. 4.2. Building Phase: Create the Robust Infrastructure and Network The building process is based on the technical assessment and the goals of the project. These are divided into 1) building the network; 2) building the center with the necessary space, including auditoriums, training room, research and resource or educational rooms; 3) electronic virtual medical library; and 4) training and education of individuals to run programs independently. 4.2.1. The Network The backbone to any telemedicine program is its network infrastructure and available bandwidth. There is a need for virtual private network to serve the program. At the University Clinical Center of Kosova (UCCK) and six Regional Telemedicine Centers (RTC) this network is developed using fiber optic lines provided by Telecom of Kosova. The sites were connected through 512 kilobits per second (Kbps) dedicated link and 10 megabits per second (Mbps) dedicated link at TCK. The communication between TCK and RTC was supported by a Polycom VSX7000 view station for pointto-point and multipoint communications via a Polycom MGC-25 Multiconferencing Unit (Polycom, Pleasanton, CA). All communications were recorded and streamed live on the Internet using Polycom RSS2000 Recording and Streaming Unit. The TPK was equipped with a Medvizer clinical telemedicine consultation unit (VitelNet, McLean, VA) in each emergency room at the UCCK and each regional hospital. These units are independent from the electronic library and the video-conferencing system used for educational purposes. 4.2.2. The Main Telemedicine Center The concept of a telemedicine program requires a physical space as the center of operations from which content can be dispersed as well as acting as a central line of
8
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
communication among the regional centers. The main telemedicine center has an electronic auditorium, telemedicine training room, electronic medical library, servers, and administrative offices. Additional laboratories such as a simulation labs and smaller video-enabled conference rooms provide additional flexibility for smaller scale projects and programs. This is highly advisable and encouraged as part of the telemedicine program. This adds a special character and depth to the center and empowers the leadership to undertake new projects. Sophisticated computerized classrooms are needed in each major telemedicine and e-health center to facilitate the diffusion of the latest information to medical personnel. Often this room will be used for education of the staff from the center as well as from the ministries of health or other governmental organizations. The main control center should be connected to the operating rooms, other auditoriums and classroom of the institution. This is particularly important to create a sense of integration for telemedicine into medical operations avoiding the perception of the telemedicine center as an isolated place within the institution and the country. 4.2.3. Electronic Medical Library Universities and medical schools in developing countries cannot afford to purchase expensive medical scientific journals. Instead they rely on written manuscripts from professors and other faculty. Creation of an electronic medical library using the World Health Organization’s (WHO) Health Internetwork Access to Research Initiative (HINARI) is extremely beneficial to healthcare workers, medical students, residents and other physicians in training, and ultimately patients themselves. A modified electronic library, where physicians, nurses, students have constant access even from their homes has been created at the TPK. The electronic library is one the most important segments of the integrated telemedicine and e-health education for developing countries as it provides the latest in evidence-based practices. Each physician, student, nurse and other healthcare provider needs to be given access to remotely enter the e-library. 4.2.4. Local Leadership and Human Capacity Building Without human capacity able and willing to lead the program, it cannot be sustained. For the first three years of TPK, special attention was paid to creating the staff that can independently run the program. This included technical, managerial and electronic library staff. This concept is of utmost importance. At the initiation of the TPK we established 4 departments whose function, while separate at first glance, combined to improve function and character. These were: 1) technical; 2) educational; 3) clinical; and 4) research and development. The financial and managerial aspect of the program/center was part of the leadership of the program. While the program and/or center may be managed by a non-physician, the clinical aspect needs to be led by the clinician, one who has authority in the institution, has energy and is willing to work with others. The research and development division should serve as a grant writing group for the center in the program and work very closely with each individual involved in the program, as well as new partners. Each of the above segments of the infrastructure needs to have its own leadership and human capacity trained and able to run the program independently. Financial and other incentives are important aspects of ensuring attraction and retention of talented personnel.
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
9
4.2.5. Financial Considerations The creation and development of a telemedicine program requires financial resources. These funds allow for planning, building, initiation and initial operation. Subsequently all costs have been covered by the country recognizing the evidence that proved the program worthwhile and cost effective for national priorities in health. Table 1. Requirements for a Successful and Sustainable Telemedicine Program*
1. 2.
Plan and develop a comprehensive business plan. Identify a champion both at the government (Ministry of Health) and at the local hospital. 3. Ensure acceptance on the part of the hospitals, including its emergency department or clinic, its physicians, its nurses, and its information technology staff. 4. Ensure proper administrative support at all sites. 5. Continuously educate new personnel at all sites through innovative online and classroom techniques. 6. Use an integrated approach involving engineers and technical personnel as well as nurses and physicians, all working very closely to resolve technical and other daily concerns. 7. Seek a private and community partnership to maintain the program’s sustainability through grants and other methods. 8. Create a partnership with an existing telemedicine program to ensure the security of patient data under HIPAA** guidelines and regulations as well as to provide technical support 24/7. 9. Allow for continuous reporting, disclosure, and full transparency. 10. Ensure a credentialing process at the re ferring hospitals for physicians. ** Health Insurance Portability and Accountability Act (HIPAA) *Adopted from Latifi et al. American Journal of Surgery. December 2009.
5. Summary Over the past several years, the concept of telemedicine in the Balkans has emerged as a key element in healthcare reform in the region. In the aftermath of the Balkans war in the mid-1990s, the energy and fortitude of a number of individuals helped establish the Kosova Foundation for Medical Development and would eventually become the International Virtual e-Hospital (IVeH), a nonprofit organization focused on developing telemedicine systems for healthcare delivery in developing countries around the world (www.iveh.org). As a result of a IBOT strategy the IVeH in collaboration with the International Trust Fund and numerous governments in the Middle East, the IVeH is creating an integrated multinational Middle Eastern Telemedicine Network (METN) for the application of telemedicine and advanced technologies in the prevention, treatment and rehabilitation of victims of trauma and major injuries. Over the next two years, structured educational curriculums and exchange programs in the United States will educate physicians, nurses, and technical
10
R. Latifi / IBOT – A Strategy for Establishing Sustainable Telemedicine Programs
professionals from these countries and U.S. experts will be deployed to Middle Eastern regions for workshops, seminars and other cultural and educational activities. The IVeH will also establish the infrastructure for telemedicine and virtual medical educational networks in these countries. Advanced technologies such as telemedicine communications and virtual education modalities using the Internet will be applied to ensure sustainability of the long-term educational effects of the program. When fully implemented, this program will create a powerful international medical education network in these countries for further collaboration and development. The medical and technical leadership created during the two year period will provide a solid foundation for new progress in healthcare in these countries. In addition, the IVeH has initiated telemedicine programs in Africa (Tanzania and Nigeria) and South America (Peru and Brazil).
References [1] R. Latifi, R.C. Merrell, C.R. Doarn, G.J. Hadeed, F. Bekteshi, I. Lecaj, C. Boucha, F. Hajdari, A.Hoxha, D. Koshi, M. de Leonni Stanonik, B. Berisha, K. Novoberdaliu, A. Imeri, R.S. Weinstein, “Initiate– Build–Operate–Transfer”- A Strategy for Establishing Sustainable Telemedicine Programs in Developing Countries: Initial Lessons from the Balkans. Telemedicine and e-Health 15: 10 (2009), 114. [2] R. Latifi, Establishing Telemedicine in Developing Countries: From Inception to Implementation. IOS Press, Amsterdam, 2004. [3] R. Latifi R, (ed.) Current Principles and Practices of Telemedicine and e-Health. IOS Press, Amsterdam, 2008.
Ontologies, Knowledge Representation, Artificial Intelligence – Hype or Prerequisites for International pHealth Interoperability? Bernd BLOBEL 1 eHealth Competence Center, University Hospital Regensburg, Regensburg, Germany Abstract. Nowadays, eHealth and pHealth solutions have to meet advanced interoperability challenges. Enabling pervasive computing and even autonomic computing, pHealth system architectures cover many domains, scientifically managed by specialized disciplines using their specific ontologies. Therefore, semantic interoperability has to advance from a communication protocol to an ontology coordination challenge including semantic integration, bringing knowledge representation and artificial intelligence on the table. The resulting solutions comprehensively support multi-lingual and multi-jurisdictional environments. Keywords. Semantic interoperability, ontology, pHealth, Generic Component Model
Introduction Traditionally, interoperability has been considered from a purely technological perspective. This view seems to be rather persistent as demonstrated at HL7, even nowadays referencing the 1990 IEEE interoperability definition “Interoperability is the ability of two or more systems or components to exchange information and to use the information that has been exchanged“ [1]. Finalizing the European Interoperability Framework [2], the European Commission is meanwhile looking for a bunch of interoperability aspects such as Organizational Interoperability; Legal Interoperability; Semantic Interoperability; Technical Interoperability. From the politicians’ and administrators’ viewpoint, such classification might be understandable, not saying that this approach is scientifically correct. Interoperability happens when two or more entities (persons, organizations, devices, systems, applications, components, or single objects) communicate and collaborate – or generally interact – to achieve a common goal [3]. Architecturally, interoperability is defined by structure, behavior and interrelations of a system’s entities interacting with themselves and with the environment to create the intended structure and behavior of the system or the environment. From a business perspective, interoperability requirements define matter and quality of interactions performed internally and exter1
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
nally, i.e. environmentally, to a system of entities by running an appropriate business process to achieve common business objectives. The definition of those objectives, the reflection of structural and behavioral needs regarding system and processes, and the assessment of environmental conditions given, intended or created, objectively depend on the considered domain, but are subjectively depending on cultural background, knowledge, skills, experiences, but also additional constraints (social, political, organizational, legal, etc.). There are also other buzzwords widely introduced and thoughtlessly used such as Intelligent Agents, Ontologies, Knowledge Representation and Reasoning, Decision Support. In this paper, we will define the related concepts as well as investigate basic requirements and solutions for enabling comprehensive interoperability in pHealth, not just focusing on technical but also considering business and user-domain-specific aspects. 1. Shared Knowledge and Capabilities Define Interoperability Levels For meeting the objectives of interaction processes, knowledge about environment and context, business domain, business objectives, business process, as well as capabilities to perform communication and cooperation must be shared. Knowledge and skills are prerequisites for running the information cycle, which is composed by the following steps that must be interactively performed with iterations: (i) observation of status and interrelationships of the objects involved including the contextual and environmental conditions resulting in data, (ii) appropriate and correct interpretation (semantics) of those data, (iii) performing of the required action. Step (ii) and (iii) require the knowledge (competence) and capabilities (performance) of the domains which define the business process and impact on it. If knowledge and capabilities are not available for running the different collaboration phases by the cooperating actors, those enablers must be provided. Interoperability is therefore not first of all a matter of technical protocols for data exchange but a matter of shared knowledge. The distribution status of knowledge and capabilities defines the required interoperability level between the cooperating actors from technical interoperability at the lowest end through structural, syntactic and semantic up to service interoperability. Preparation and realization of cooperation requires knowledge management (KM) and knowledge representation (KR) [3]. Interoperability between two actors is frequently not performed directly, but supported and mediated by other actors forming an interoperability chain. The different actions or services within this interoperability chain can be tightly coupled or loosely coupled. This circumstance defines whether sharing the knowledge about the entire process or just the input/output from/to the neighbors in the chain is needed. KR started as part of Artificial Intelligence (AI) to implement computational knowledge. Based on a set of ontological commitments, its genuine task was the description of the natural world by a surrogate of the real thing to determine consequences of processes in virtual environments, i.e. by thinking rather than by acting. So, it provided a medium for human expressions and for efficient computation as well. KR is a fragmentary theory of intelligent reasoning, expressed in terms of the three components: (i) the representation’s fundamental conception of intelligent reasoning; (ii) the set of interferences the representation sanctions; (iii) the set of inferences it recommends [4].
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
13
2. pHealth – Basic Principles and Requirements pHealth, i.e. personal health or personalized health, sometimes also connected to the pervasive health paradigm, addresses health services delivery independent of time and location of actors and resources. Such a paradigm has to be technically enabled by mobile (anywhere accessibility), pervasive (anywhere multi-disciplinary service provision), and for the sake of individualization even autonomous technology2, guaranteeing fully distributed, flexible, scalable, intelligent, user-friendly, lawful, trustworthy, and standardized solutions [3]. For meeting the aforementioned requirements, the intended system has to be component-oriented, knowledge-based, ontology-driven, formally modeled as well as designed and implemented following a unified process to provide interoperability at the appropriate level. Despite the fact that ICT support sometimes initiates a re-engineering of business domains and their processes, the business domain and its processes define the requirements for ICT solutions. Therefore, the design, implementation and deployment process must be led by user domain experts, but not by IT geeks. In organization-centered, and partially in process-controlled, care settings, business processes and the related workflow with its actions (abstracted as steps), but also structure and function of the system regarding the components involved, their properties, functions (services) and interaction details as well as environmental and contextual conditions can be widely predicted, negotiated, harmonized and enforced. In personal health settings, the status of the patient, her expectations, intentions, conditions define delivery system structure and business process in an unpredictable way. This applies even more to pHealth settings in a regional, national or even international environment. Here, not just the multi-disciplinary characteristics of the pervasive components set to be integrated, but also their modifications and expression style ruled by changing regulatory and cultural environments must be mastered.
3. Modeling Personal Health Systems Communication and collaboration is always bound to the challenge of properly describing the reality. The outcome of this description process independent of the expressivity of the description means is model as a partial reflection of reality focusing on aspects the modeler (or her audience) is interested in [5]. 3.1. The Need for an Open Systems Approach To manage cooperatively multi-disciplinary systems like health services, we have to abstract from the real system’s diversity just focusing on the current business objectives. The focus of interest defines the considered characteristics (components, processes, etc.) as well as the abstract system's boundaries to the environment. By that way, components can be externalized if their function can be sufficiently summarized as environmental effect. On the other hand, influential components can be internalized. The system design results in a simplified model of reality – the system architecture – describing structure, function as well as internal and external interrelationships of the sys2
More details on the impact of those technical paradigms on the care paradigm change towards pHealth can be found in [3] or [6].
14
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
tem's components relevant for the business objectives. Here, we have to have in mind that the business objective in eHealth is first of all defined by the specialties involved such as medicine, biomedical engineering, biology, genomics, legal affairs, administration, etc., and just supported by information and communication technology (ICT). More details about this traditional approach of system theory and system engineering in the papers context can be found in [7]. 3.2. Reference Models Not really facilitating interoperability, there are many even standard definitions for reference models. The Organization for the Advancement of Structured Information Standards (OASIS) defines a reference model as an abstract framework for understanding significant relationships among the entities of some environment by defining a minimal set of unifying concepts, axioms, and relationships within a particular problem domain [8]. Consequently, a reference architecture describes the abstract architectural elements in a domain [9]. Reference models are by definition independent of specific standards, technologies, implementations, or other concrete details [7]. Depending on the level of abstraction ranging from foundation architectures to common systems architectures, and industry and organization-specific architectures, it enables the development of specific reference or concrete architectures using consistent standards or specifications supporting that environment. All reference models defined by OASIS [10], The Open Group (TOG) [11] or the Object Management Group (OMG) [12] deal with IT-specific models, not tackling biology, life sciences, etc., but relate to systems for distributed processing as described in ISO 10746 [13], large software systems as described in IEEE 1471 [14], or the Service Oriented Architecture (SOA), meanwhile jointly standardized by OASIS, The Open Group and OMG. Another, more generic reference model is the Generic Component Model (GCM) [3], shortly introduced in the next section. 3.3. The Generic Component Model The development of the generic system description or architecture framework used has started in the early nineties in the OMG architecture-centric standards development environment. The resulting and by the Magdeburg Medical Informatics Department enhanced Generic Component Model (GCM) [7, 15] has been developed using system engineering, control theory, cybernetics and their application on technical as well as living systems, the approach describes any kind of system beyond ICT ones. The GCM defines three dimensions comprehensively characterizing any system: the architectural components perspective describing composition/decomposition including concepts and interrelationships representation; the domain perspective separating system’s aspects described by domain ontologies; the views on the system describing the development process in consistence with ISO 10746 (Figure 1). For connecting different instances within and between the aforementioned dimensions, reference models or meta-models are needed, including those for engineering ontologies and terminologies used for representing concepts as well as mapping between different domain languages and their concepts. For mapping an ontology, many dimensions of an ontology have to be considered like the concepts hierarchy, the semantic relationships, the instances and the possible axioms defining a given knowledge domain.
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
15
Business Concepts Relations Networks
Development Process Perspective
Aggregations
Technology View
Engineering View
Computational View
Information View
Enterprise View
Details
System Component Composition
Domain n Domain 2 Domain 1
System Viewpoint
Figure 1. The Generic Component Model
4. The Ontology Challenge in Architectural Modeling Some years ago, ontologies have evolved in computer science as computational artefacts providing computer systems with a conceptual and nevertheless a computational model of a particular domain of discourse. This facilitates computer systems to draw conclusions (perform interferencing) from domain knowledge like humans do. There are two distinguishable approaches to ontologies: a) the philosophical one started in the ancient time already as the study or concern about what kinds of things exist - what entities or `things' there are in the universe [16] and b) the practical definition in the context of information processing. As important representative starting from the first approach, Barry Smith defines an ontology as representation of universals or classes of reality and the relations existing between them. Hereby, universals are “the real invariants or patterns in the world apprehended by the specific sciences” [17]. Inaugurating the second approach, Gruber defines “An ontology is the specification of conceptualizations, used to help programs and humans share knowledge” [18]. In that context, ontologies are the structural frameworks for organizing information on a certain domain by defining the basic terms and relations in a special domain. Among others, they are being used as reference for the related community. In technical environments, ontologies are used as basis for interoperability between systems as well as for search, integration and exchange of domain-specific data. We will inter-relate the two basic approaches to ontologies using the GCM in Section 4.1. From a knowledge representation point of view ontologies can have the following components: Concepts representing sets or classes of entities in a domain frequently organized in taxonomies; Instances representing the actual entities; relations types such as specialization or partitive relationships; and axioms representing facts that are always true in the topic area of the ontology such as domain restrictions [19].
16
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
4.1. Ontologies as a System
Business Concepts
General Ontology
Relations Networks
Top-Level Ontology
Aggregations (Basic Services / Functions)
Domain Ontology
Technology View
Engineering View
Computational View
Information View
Application Ontology
Enterprise View
Details (Basic Concepts)
Development Process Perspective
System’s Architectural Perspective
According to the architectural perspective of the GCM, systems components and their interrelationships domain-specifically described through the domains concepts and relations (component aggregation rules or logics) are represented by business case specific application ontologies derived from the domain’s ontology. The integration of the different domains included is controlled by meta-ontologies (upper level ontologies), and modeled in an ICT environment using IT-specific ontologies. Being just another system (not to be merged with the pHealth system, where the ontology system must be developed for modeling it), the system of ontologies has to be managed within an architectural framework like GCM as well, as shown in Figure 2.
Figure 2: The GCM representation of the system of ontologies
An ontology and its components can be represented in a spectrum of representation formalisms ranging from very informal to strictly formal [20]. Collaboration in a system of concepts as any interoperability depends on the knowledge shared between the principals involved. As less business process related as well as contextual knowledge (including common knowledge) is available at the actors’ side, as more knowledge must be communicated by making it explicit. This implies that the level of formalization and expressivity has to be increased providing another ontology type (Figure 3). In general, the more formal the used representation language, the less ambiguity there is in the ontology [19]. Formal languages are also more likely to implement correct functionality. The aforementioned different ontology types for concept representation in the medical domain have been partially standardized. Here, frames to represent Arden Syntax Medical Logic Modules or openEHR Archetypes, ULM concepts for Detailed Clinical Models, but also XML-based specifications for CDA documents have to be referenced as examples. Analog to the limitation of the interoperability level to the enhancement needed, also the formalization of ontologies should be restricted to level required for enabling the principal’s involvement.
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
Formal taxonomies Data models XML Schema Database schemas
Principled, informational hierarchies XML DTD Structured glossaries Thesauri
Data dictionaries Ad hoc hierarchies “ordinary” glossaries Terms
&ŽƌŵĂůŝnjĂƚŝŽŶ Figure 3: Types of ontologies (after [21], changed)
5. Why do we Need Ontology-Driven Approach to Practical pHealth? As status of the subject of care with her requirements, needs, expectations, wishes, environmental as well as contextual conditions define the aggregation of needed components, their constraint functionality as well as the domains being involved. In other words, the systems must be able to make intelligent design decisions at runtime, thereby finding the appropriate component offering the best-fitting service, possibly constraining properties, defining the invocation context and processing it. Architectures like SOA do not provide such capability. Here, an ontology-driven middleware managing the inference decision and realization based on formal representations of business process models, services capabilities and the invocation context provide the solution [22]. The challenge of finding semantic similarities between service request and service delivery through ontology mapping and selecting the right service can be done by a context sensitive invoker. In that context, assertions as mappings of context-sensitive variables are bound to policies as subset of those assertions ontologies containing the required logic for realizing the appropriate invocation of the selected component [22]. As the architectural design of the pHealth system requires the intelligent aggregation of appropriately adapted (constrained) services within domains (service aggregation) and between domains (service constraints), ontologies are needed at the level of components for defining and specializing them, but also at the level of relationships within the domains, between them, and through the development process, as shown in Figure 4. Mathematically, the abstract model of the GCM approach forms a threedimensional type representation, which corresponds to the improved Barendregt’s Lambda Cube consisting of a set of Pure Type Systems (PTSs), additionally refined through parameters, constraints, context, etc. By that way deploying universal logic, the three dimensions’ construction rules are created [23, 24], representing knowledge
18
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
or more generally ontologies of a multi-domain, multi-disciplinary pHealth system in an open representation space.
Figure 4. Ontology –driven relationships in the GCM
6. Discussion Semantic interoperability deals with the challenge of common understanding and use of shared information to meet common business objectives. As the different disciplines involved in pHealth manage their specific knowledge representations based on their evolution and environmental constraints, the mutual understanding of domain-specific, to the partner foreign concepts, relations and joint use is unlikely. So, interoperability is not a technical challenge as traditionally managed in standardization and system implementation. This holds even more in an international environment, where domains impacting the care delivery process compared with local settings have to be considered such as different jurisdictions and language domains (also the translation between languages must be based on concept similarities and cannot be managed through word by word translation). Engineering and coordinating the different ontologies of a pHealth solution architecturally integrates the different disciplines involved in personalized care settings, thereby allowing stakeholders of using their domain-specific languages, “translating” semantically between them, and thereby providing comprehensive interoperability amongst them. This aspect of a formal architectural approach is especially important for the empowerment of patients (or citizens in general) as a characteristic of personal health [3]. Only if the subject of care is enabled to reflect his/her concerns, wishes, expectations, etc., as well as to perceive the recommendations, regulations and procedures defined by care professionals properly, he/she can manage the care process safely and responsibly. The semantic challenges described in the paper are not limited to human understanding. They also concern semantically-enriched applications which are autono-
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
19
mously designed at runtime. This requires dynamic interaction between dynamically aggregated (find, bind, execute) components (services). The system must enable the intelligent, policy- or generally rule-conform, event-sensitive selection understanding concepts and interrelations like humans do. So, the entire ontology-based architectural framework represented by the GCM must be ontologically refined within the RM-ODP process. For that purpose, education on the one hand and intelligent support as presented here on the other hand has to be provided.
7. Conclusions Traditionally, semantic interoperability between different principals has been considered and standardized from a technology perspective. eHealth and pHealth solutions are characterized by the integration of different disciplines representing different domains based on different ontologies, policies, etc. Therefore, semantic interoperability has to advance from a communication protocol to an ontology coordination challenge. In that context, eHealth and pHealth solutions have to be based on a formally modeled architecture comprehensively meeting the GCM architecture framework and the underlying principles the paper addressed. For meeting the interoperability challenge, the information cycle has to be completed, based on shared knowledge and capabilities. System properties and environmental conditions define the interoperability level needed. The solution should be restricted to that interoperability level. The management of multi-disciplinary environments requires a system-theoretical or cybernetic approach to model real-world systems within a clearly defined architecture framework is inevitable. The GCM supports analysis and design of multi-disciplinary real-world systems. Being based on a formal type theory for expressing any representation style, the GCM manages the systems architecture for multiple, interconnected domains as well as the ICT development process. Representing any system, the GCM also allows for managing language systems and ontologies. By that way, the motto of this conference is realistically. Existing standards such as ISO 10746 [13] or IEEE 1471 [14], they can be easily and comprehensively represented by the GCM as well. The integration of different domains as well as adaptive and intelligent behavior of ubiquitous and personalized health services require ontology management and coordination as well as autonomic computing based on artificial intelligence (AI) approaches. Interoperability of pHealth solutions is described by a system of interrelated ontologies.
Acknowledgement The author is indebted to his colleagues from ISO, CEN, HL7, and IHTSDO for kind collaboration.
20
B. Blobel / Ontologies, Knowledge Representation, Artificial Intelligence
D. Kaminker, Standards for Interoperability, Tutorial at HL7 January Working Group Meeting 2011. European Commission, Directorate-General for Informatics, DIGIT/01 - European eGovernment services (IDABC) (2004) Final European Interoperability Framework - November 2004 (EN) B. Blobel, Architectural approach to eHealth for enabling paradigm changes in health, Methods Inf Med 49,2 (2010), 123-134. R. Davis, H. Shrobe, and P. Szolovits. What is Knowledge Representation? AI Magazine 14,1 (1993) 17-33 B. Blobel, Concept Representation in Health Informatics for Enabling Intelligent Architectures. In: A. Hasman, R. Haux, J. van der Lei, E. De Clercq, F. Roger-France (Edrs.) Ubiquity: Technology for Better Health in Aging Societies, pp. 285-291. Series Studies in Health Technology and Informatics, Vol. 124. IOS Press, Amsterdam, 2006. B. Blobel, C. Gonzalez, F. Oemig, D.M. Lopez, P. Nykänen, P. Ruotsalainen, The Role of Architecture and Ontology for Interoperability. In: B. Blobel, E.P. Hvannberg, V. Gunnarsdóttir (Edrs.) Seamless Care – Safe Care: The Challenges of Interoperability and Patient Safety in Health Care, p. 33-39. Series Studies in Health Technology and Informatics, Vol. 155. IOS Press, Amsterdam, Berlin, Oxford, Tokyo, Washington, 2010. B. Blobel, Analysis, design and implementation of secure and interoperable distributed health information system. Stud Health Technol Inform Vol. 89, IOS Press, Amsterdam, 2002. OASIS, OASIS Reference Model for SOA, Version 1.0, OASIS Standard, October 2006: docs.oasisopen.org/soa-rm/v1.0/soa-rm.pdf OASIS Reference Architecture for SOA Foundation, Version 1.0, OASIS Public Review Draft 1, April 2008: docs.oasis-open.org/soa-rm/soa-ra/v1.0/soa-ra-pr-01.pdf Organization for the Advancement of Structured Information Standards (OASIS), www.oasis.org The Open Group, www.opengroup.org Object Management Group, Inc., www.omg.org ISO/IEC 10746:1996 Information technology – Reference Model for Open Distributed Processing. IEEE 1471:2000 Recommended Practice for Architectural Description of Software-Intensive Systems. B. Blobel, Assessment of Middleware Concepts Using a Generic Component Model. Proceedings of the Conference “Toward An Electronic Health Record Europe ’97”, pp. 221-228. London, 1997. S. Blackburn. The Oxford Dictionary of Philosophy. Oxford University Press, 1996. B. Smith, M. Brochhausen, Establishing and Harmonizing Ontologies in an Interdisciplinary Health Care and Clinical Research Environment. In. B. Blobel et al. (Edrs.) eHealth: Combining Health Telematics, Telemedicine, Biomedical Engineering and Bioinformatics to the Edge, 219-33. Studies in Health Technology and Informatics, Vol 134. IOS Press, Amsterdam, 2008. T.R. Gruber, Towards Principles for the Design of Ontologies Used for Knowledge Sharing. In: R. Poli, N. Guarino (Edrs.) International Workshop on Formal Ontology, Padova, Italy, 1993. Available as technical report KSL-93-04, Knowledge Systems Laboratory, Stanford University: ftp.ksl.ftanford.edu/pub/KSL_Reports/KSL-983-04.ps. P. Lambrix, H. Tan, V. Jakoniene, and L. Strömbäck, Biological Ontologies. In: C.J.O. Baker and K.-H. Cheung (Edrs.) Semantic Web – Revolutionizing Knowledge Discovery in the Life Sciences, SpringerVerlag, Berlin-Heidelberg-New York, 2007 Jasper R. and Uschold M. A Framework for Understanding and Classifying Ontology Applications, in: Proceedings of the IJCAI-99 Workshop on Ontologies and Problem- Solving Methods: Lessons Learned and Future Trends, 1999. M. Rebstock, J. Fengel, H. Paulheim, Ontologies-Based Business Integration, Springer-Verlag, BerlinHeidelberg, 2008. R. Krishnamurthy, V. Ranganathan, B. Senguttuvan, Drug Discovery: An Innovative Approach to Application Integration using SOA & Ontology. White Paper, Cognizant, 2009. Bloe R, Kamareddine F, Nederpelt R (1996) The Barendregt Cube with Definitions and Generalized Reduction. Information and Computation 126 (2), 123-143. F. Kamareddine, T. Laan, R. Nederpelt, A Modern Perspective on Type Theory, Kluwer Academic Publishers, New York, 2004.
E-Health in Graduate and Postgraduate Medical Education: Illusions, Expectations and Reality a
Ferenc BARI a,1, Erzsébet FORCZEK, and Zoltán HANTOS b Department of Medical Physics and Informatics, Faculty of Medicine, University of Szeged, Szeged, Hungary
Abstract. With the overall growth of informatics, the medical education system should also provide programs at both graduate and post-graduate levels. While there is a wide consensus as to the importance of this urgent need, several factors slow down the construction and operation of effective education programs in medical and nursing schools. The increasing need for better and more comprehensive training in informatics is strongly limited by several factors including undefined output skills, tight time frame etc. An efficient development of partnerships within the health care system assumes that all professionals involved must possess strong informatics and interpersonal knowledge, and skills reaching beyond their own individual fields. There is an emerging need to define the basic skills and knowledge for each level of the health care education. Transborder cooperation offers a unique opportunity for the establishment of common criteria for basic skills and knowledge, via joint discussions, collaborative thinking and concerted action Keywords. medical informatics, education, graduate- and post-graduate studies, cross-border collaboration
Introduction The use of e-Health approaches and systems in the healthcare sector is becoming a daily practice, as almost all medical practitioners use computers increasingly to store patients' data, assist the patient, make decisions by support systems, check data bases, etc. According to E-Health indicators, almost 100% of Hungarian doctors use a computer in office; whereas, permanent internet access and the use of professional web sites are still limited and personal web presentation is below the European average. Although medical professionals and institutions emphasize the need of the continuing medical education (CME) for maintaining and improving knowledge, skills, competence and performance, there is no concept in CME that cover certain specific fields, such as medical informatics. With the overall growth of informatics, the medical education system should also provide programs at both graduate and post-graduate 1
Corresponding author: Ferenc Bari, Department of Medical Physics and Informatics, Faculty of Medicine, University of Szeged, Korányi fasor 9, 6720 Szeged, Hungary, e-mail: [email protected]
22
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
levels. While there is a wide consensus as to the importance of this urgent need, several factors hamper the construction and operation of a CME system in medical informatics. In particular, x lack of involvement of health informatics in the core curriculum of most medical schools, x limited resources of educational institutions, including medical and nursing schools, x competence and skill level are not defined and measured, x there are no comprehensive education programs available. The lack of formal health informatics training begs the questions as to whether future doctors will understand the full capabilities of electronic platforms and technologies, and whether they will know to evaluate and integrate them into their practices. E-Health training problems affect almost all countries. Establishing formal crossborder networks would help in constructing and updating curricula, and setting acceptable competence and skill levels. Therefore, there is an urgent need for a harmonized international program to be launched in the near future. This paper focuses on demonstrating the current situation in Hungary and neighboring countries, and on project proposals for future solutions in creating crossborder network(s) for effective e-Health education.
1. Changing Mission of Medical Informatics The challenges faced by medicine today are enormous. Setting and continuously updating curricula is an ongoing task of all medical schools. Medicine has always welcomed new technology, therefore informatics was incorporated into everyday medical practice decades ago. In fact, the development of many fields in informatics was driven by the demands of health care and occasionally stimulated by the medical educational system. Although there are obvious variations among different specialties; physicians spend 25-30% of their time with administrative tasks executed with the help of computers and other information technology devices. Nowadays, with universal accessibility, internet databases have been considered a major avenue in making medical knowledge available to almost all medical professionals. The booming field of medical informatics and e-Health assists physicians and healthcare specialists in diagnosis, follow-up, and treatment of patients through electronic means over wide geographic distances. Aspects such as more accurate diagnosis, better treatment solutions and patient relationships are increasingly more emphasized when using eHealth facilities. It has been accepted widely that medical informatics forms a hub for medical database management, and it considerably contributes to the evolution of diagnostic and treatment technologies. As a result, much is expected of medical informatics with respect to promoting improvement of health conditions throughout the world, both by contributing to the quality and efficacy of health care, and accomplishing research in the fields of innovative biomedical as well as computer, health, and information sciences.
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
23
2. Factors Influencing the Involvement of Medical Informatics in the Curriculum When we consider effective cross-border collaborations in education programs of our discipline, it is worth mentioning that the terms medical informatics and biomedical informatics have diverse meanings between different medical schools in distinct geographical regions and even within the same country. Therefore intensive discussions and careful exploration of the various definitions/descriptions in the neighboring countries would help in reaching consensus and in finding common terminology. The artificial political, cultural and economical separation during the second half of the previous century also has forced Hungary to find individual solutions for problems in medical informatics training, and in almost all fields of health care [1]. When globalization incorporated our countries into Western culture, countries of the former Eastern block had no choice but to follow international trends, by taking on widely accepted operation systems, adapting basic documentation methods and implementing western-type education systems. Even though adaptation became a universal routine, various backgrounds and former goals resulted in information systems in the former socialist countries becoming disparate and confusing [2]. Now, the necessity of more intensive regional dialogues and cooperation urges experts in biomedical informatics to engage in more intensive collaboration, in order to establish joint concepts and practice for improving the health conditions of our countries. Consequently, we have to list and voice our individual problems, and analyze them in order to find common solutions. It has been emphasized several times that human resources are critical in determining the level of performance of the health care delivery and for the attainment of national health goals in all countries. Lack of involvement of health informatics in core curriculum in most medical schools creates artificial barriers in medical education and slows down information transfer. The question asked many years ago, "To what degree are medical schools teaching medical informatics and how?" has still only partially been answered. Medical Informatics is a constantly developing field. Some years ago, topics like literature searching, internet use, computer assisted diagnosis, hospital information systems and electronic patient records constituted the maximum requirement for a medical graduate. Today telemedicine, telediagnosis, and patient-support systems also require more and more attention in a curriculum. As there are no standards on what makes up an informatics curriculum, medical schools teach what they consider important, what can be financed and what can be forced into the tight time-frame. There are many reasons why medical schools are reluctant to incorporate medical informatics in their core curriculum. First of all, medical school professors do not all fully estimate the importance and meaning of medical informatics. Secondly, the densely packed medical school curriculum does not leave a suitable time slice for the proper training of medical informatics [6]. The increasing need for better and more comprehensive training in informatics is strongly limited by several factors including teachers. The most sensitive point is how to recruit and maintain staff for teaching medical (health) informatics. Since there is a general shortage in well-trained experts, medical schools have to compete for teachers with informatics companies, which offer far higher salaries and better career opportunities. Consequently, it is almost impossible to offer such income or/and promotion opportunities that would attract young professionals to join informatics departments. Common cross-border projects combining the efforts on specific issues
24
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
should successfully aim at finding salary supplements and facilitating common research-development activity. Medical informatics should not be solely the subject of specific courses or specific departments. It should be a general approach throughout an entire training program. Therefore it seems desirable that education professionals undergo special training to learn new teaching methods and utilization of educational aids including information technology, newer media, interactive learning, group education and individualized teaching. There are obvious questions to be answered: who trains the trainers, who sponsors such trainings and how could it be fitted into the program of a Faculty? This is another area where cross-border projects are desired. With the formation of special interest groups and organization of intensive courses a certain region could become an engine in promoting the better knowledge of informatics among medical teachers [3, 4]. The educator cannot avoid e-learning, virtual environments or distance teaching, especially if the trends towards more international and transnational students, teachers and curricula strengthen. Even if medical informatics modules obtain some limited niche, they must be adapted to the rigid, weekly-based, formal and traditional lecture–practice structures. Innovations like intensive, course block structures, in which students learn about problem-driven medical informatics as a component of diagnosis are very rare. Such structural changes in the educational system disturb the traditional composition of a semester and may generate undesired, contradicting opinions and tension among professors of other disciplines. Additionally, even with the implementation of the potential inventions proposed above, very limited skills could be taught within a short period of a week or so. It is important to recognize medical informatics as a tool and as a skill. Any skill is learned and imprinted the most effectively when the education is problem-driven and put into the context of the environment it is to be used within. What are the major motivations that could attract medical students to take up medical informatics as an elective subject? How can these students transform informatics knowledge and computer skills gained during the module into skills that aid in learning fundamental medical subjects such as anatomy, physiology, surgery or general medicine more effectively? For these reasons, we have to be aware of the demands from other departments and the specific electronic teaching material they offer when we assemble the material of our medical informatics course. A wide knowledge about the nature of medical information and how it can be obtained from the internet also prompts the departments of medical informatics to develop their general curriculum accordingly. A further question is the phase of medical training in which medical informatics should be taught. At the Faculty of Medicine of the University of Szeged, medical informatics is taught in the second semester of the first academic year as an elective subject. Currently, about 85-90 % of the students choose this course. The lectures (1 hr/week for 14-15 weeks) cover the basic concepts of informatics, its development, the fundamentals of computer architecture, principles and functions of operating systems and computer networks. Special emphasis is given to the building up of electronic documents, the characteristics of textual, tabular, graphic and other components of documents and their unification. The lectures deal with the potential possibilities inherent in the Internet, the properties of databases and data warehouses, and the most significant medical, biological and bibliographic databases available. The aim of the practical course (2hr/week in small groups [less than 16 students]) is to provide the students with a
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
25
basic practical knowledge in electronic communication and evaluation of biomedical data, and with the tools and knowledge to create electronic documents. For students who are interested in extra-curricular research activity there are elective courses in the 4th-5th years dealing with specific problems like biomedical signal processing or advanced statistics. The use of statistical packages assumes a basic knowledge in calculus and applied statistics but the proper use requires also advanced knowledge in informatics (Fig.1).
Figure 1. Schematic representation of medical informatics education-program at the University of Szeged, Hungary.
Apart from the proper establishment of a medical informatics module, it would be very useful for medical students if they are provided with a detailed picture on the specific information infrastructure available at their medical school (mailing system, security, e-library, administrative supports like registration etc.), upon their admission. This introduction might also promote the familiarity of medical students with information technology and electronic facilities at the medical school. At the same time, this may not necessarily be the task of the medical informatics departments exclusively, but could be a joint effort between various departments. Effective practice of medical information systems requires harmonized educational programs in medicine, nursing, health care management, dentistry, pharmacy, public health, health record administration, and informatics/computer science. Because of the diverse structure of the graduate and postgraduate education, it is almost impossible to establish dedicated complementary programs [5, 7]. At the University of Szeged, our department is responsible for the teaching of medical informatics for students in medicine, dentistry, pharmacy and nursing. It has always been a complex task to create profession-specific education programs. There is an agreement that all curricula should include some common components (basic concepts and common software available for
26
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
higher education in Hungary) but there is an ongoing debate on what kind of additional specific tools and resources should be involved. Our approach is very practical. With close cooperation with faculty members of the schools of dentistry, pharmacy and nursing, and via the specialization of our teaching staff we try to define the professionspecific targets. Specifically, nurses have commonly been regarded to have poor IT skills (and are not motivated for the introduction of IT); it has also been observed that nurses are more frequently reluctant to use computers than other healthcare staff groups, and have made more statements against curriculum development in all disciplines. We have to make it clear that computers will not disappear from health care; on the contrary, their increasing use is unequivocally predicted. Changing the views of these nurses is a particular challenge for our department. Mandatory courses in informatics and statistics are an essential part of postgraduate university education programs in Hungary. In our faculty, informatics and statistics are taught in two semesters (30 hr each). In contrast to undergraduate courses that concentrate on basic skills and everyday medical practice, PhD courses focus on specific aspects of info-communication involved in all phases of research: the efficient use of scientific databases, data organization, processing and presentation. An important feature of these courses is that medical PhD programs involve a substantial number of non-MD participants, such as biologists, pharmacists, chemists, physicists, etc.; this necessarily broadens the scope of the training. Although it is desirable for all kinds of medical professionals to graduate with a certain set of computer and information management skills, there are no general computer skill assessment systems which allow national or international comparisons and evaluation of such skill-sets. It is imperative to define the basic skills and knowledge requirements for various health care professionals [7]. A challenging project would be to construct an international evaluation system which would be accepted by the various faculties teaching health sciences. Development of an ongoing training program for employees including the educational staff is a task closely related to the graduate and postgraduate courses organized for students.
3. Conclusions In summary, an efficient development of partnerships within the health care system assumes that all professionals involved must possess strong informatics and interpersonal knowledge, and skills reaching beyond their own individual fields. There is an emerging need to define the basic skills and knowledge for each level of the health care education. Trans-border cooperation offers a unique opportunity for the establishment of common criteria for basic skills and knowledge, via joint discussions, collaborative thinking and concerted action.
Acknowledgement This work has been supported by Hungarian Research Funds (OTKA K81266 and ETT).
F. Bari et al. / E-Health in Graduate and Postgraduate Medical Education
27
References [1] [2] [3]
[4]
[5] [6] [7]
N. Balogh, The role of XML in medical informatics in Hungary, Stud Health Technol Inform. 90 (2002), 168-73. M. Duplaga, E-health development policies in new member states in Central Europe, World Hosp Health Serv. 43 (2007), 34-38. M.W. Jaspers, R.M. Gardner, L.C Gatewood, R. Haux, R.S Evans, An international summer school on health informatics: a collaborative effort of the Amsterdam Medical Informatics Program and IPhiE-the International Partnership for Health Informatics Education, Int J Med Inform. 76 (2007), 538-546. M.W. Jaspers, R.M Gardner, L.C Gatewood, R. Haux, D. Schmidt, T Wetter, The International Partnership for Health Informatics Education: lessons learned from six years of experience. Methods Inf Med. 44 (2005), 25-31. M.C Shanahan, Transforming information search and evaluation practices of undergraduate students, Int J Med Inform, 77 (2008), 518-526. S. Strauss, Canadian medical schools slow to integrate health informatics into curriculum, CMAJ 182 (2010), E551-2. J.H. Wu, Y.C. Chen, R.A. Greenes, Healthcare technology management competency and its impacts on IT-healthcare partnerships development, Int J Med Inform. 78 (2009), 71-82.
A Knowledge Management System to Study the Quality of Life in Head and Neck Oncology Patients Joaquim GONÇALVESa, Augusta SILVEIRAb and Álvaro ROCHA c, 1 a Instituto Politécnico do Cávado e do Ave, Barcelos, Portugal b Universidade Fernando Pessoa, Porto, Portugal c Universidade Fernando Pessoa, GIMED, Porto, Portugal
Abstract. The perception that an individual holds about his place in life, which depends upon his culture and values, defines this individual’s Quality of Life (QoL). When applied in a health context this known as: Health-Related Quality of Life (HRQoL). The assessment of HRQoL is a Medical goal; it is used in clinical research, medical practice, health-related economic studies and in planning health management measures and strategies. Obtaining a patient self-assessment with QoL measuring instruments on the platform developed in this project, through user-friendly software, aids the study, promotes the creation of databases, and accelerates its statistical treatment. The possibility of graphically representing results that physician needs to analyze, immediately after the answer collection, makes this assessment a diagnosis instrument ready to be used routinely in clinical practice. Knowledge Management Systems (KMS) applied to this context enable knowledge creation and storage, and guide therapeutic decisions.
1. Introduction The concept of “Quality of Life – QoL” is used in different contexts and situations, reaching practically all sectors of society. The perception an individual holds about his place in life, which depends upon his culture and values, defines this individual’s Quality of Life (QoL). When applied in a health context this is known as: HealthRelated Quality of Life (HRQoL) [1]. Nowadays, indicators of HRQoL are used in health management strategies, managers, economists, political analysts and pharmaceutical companies use QoL measures from the World Health Organization (WHO) in some of their departments [2]. Today, HRQoL is a medical goal, being used in epidemiological studies, clinical essays, medical practice, health-related economic studies, and in planning and comparing measures and strategies [3]. In the present paper we intend to demonstrate the importance of HRQoL assessment in oncologic patients, and the relevance of Knowledge Management Systems (KMS) as decision 1
Álvaro Rocha, Universidade Fernando Pessoa, GIMED, Praça 9 de Abril 349, 4249-004 Porto, Portugal. E-mail: [email protected]
32
J. Gonçalves et al. / A Knowledge Management System to Study the Quality of Life
making aids. We analyze this problem and show the results obtained with a platform developed for the self-evaluation questionnaire that measures patients QoL in this project.
2. Evaluation of HRQoL in Oncologic Patients Malignant tumors are the second leading cause of death in Portugal. Their relevance as a morbidity and mortality factor is growing and their social impact is being recognized [1]. The global weight of oncologic disease is growing, given the economic and social costs involved in the prevention, treatment and rehabilitation of this disease [6]. Research methods used in oncology enable us to analyze the oncologic process in its physiopathologic and clinical aspects, penetrating wide domains like the psychological, social, economic and organizational domains [1]. Epidemiology and statistics are significant aspects of this study, since oncologic care can only be programmed based in safe databases [7]. Assessing the implementation of these diseases in our community helps to recognize the global impact of tumors and to evaluate the effectiveness of the adopted control measures [2]. The time where therapeutic decisions were not discussed with the patient and the family, and treatment options were not even considered, has long since passed. Oncologic patients were frequently not informed of their diagnosis after their families were. This reality has changed and, today, patients participate, or should participate, in the several stages of their treatment [1]. In fact, patients motivated to participate in their treatment and rehabilitation plan often show a better QoL, and should therefore be involved in the strategies developed to fight their disease. Furthermore, evidence shows that a global patient QoL optimization can lead to a higher survival rate and to a higher quality of life [1]. Promoting the integration of QoL assessment in clinical practice can result in the optimization of infrastructures and methods capable of improving patients QoL [8]. A validated, safe and scientificallybased measuring instrument must be made available in a simple format, understood both by the patient and the physician, and it should be completed in less than 10 minutes [9]. Although being a subjective concept, HRQoL is quantified objectively and does not merely represent the inexistence of disease [10]. The multidimensional conception of HRQoL comprises a wide range of physical, psychological, functional, emotional and social variables, and these, as a whole, define welfare [11]. These domains vary individually according to religion and beliefs, culture, expectations, perceptions, education, knowledge, etc. [11]. Table 1 represents schematically some of the HRQoL dimensions and elements, proposed by the WHO [12]: QoL Domains Physical Health
Self-Esteem Spirituality Body Image Thoughts Negative Feelings Positive Feelings
Sexual Activity Social Support Family Personal relationships
Economy Information Means of transportation Security Services Free time
Table 1 – Dimensions and items for HRQoL assessment
J. Gonçalves et al. / A Knowledge Management System to Study the Quality of Life
33
3. KMS in Routine HRQoL Assessment Preliminary studies on oncologic patients conclude that the use of an adequate software for the HRQoL assessment, data collection and processing, allows us to obtain selfanswered questionnaires from patients, an automatic quotation of these questionnaires, the creation of a database and the statistic analysis of the results, favoring a routine HRQoL clinical assessment [13]. Moreover, the graphical representation of results enables a fast patient HRQoL assessment by the physician, and this evaluation becomes a diagnosis instrument to be used in routine clinical practice [13]. HRQoL assessment is dynamic and requires periodic reevaluations [14], it should be done objectively and quantitatively on a routine basis. And to do so, the selection of a measuring instrument with good psychometric characteristics, easy to administrate and to quantify, that doesn’t increase the appointment time and with a multidimensional character is most important. It must be answered and quoted before appointments. The results should remain confidential and anonymous, and when graphically represented they allow an easy reading of the patient’s self-perception. Thus, HRQoL assessment becomes a diagnosis instrument that identifies patient’s problems, highlights certain signs and symptoms that could otherwise go unnoticed, improves the physician-patient communication and assists therapeutic decisions; in other words, it renders the appointments easier. By analogy, the physician can evaluate the evolution of his patient’s state comparing two or more assessments obtained in different periods [15]. However, a routine assessment implies the design of a new appointment protocol. The analysis and specification of the information system requirements, as well as the specification of necessary activities for the process, defines the Knowledge Management System which supports the clinical decision aid system, based on the HRQoL assessment.
4. Friendly Software Design It cannot be denied that the health domain is extremely sensitive, and every aspect that interferes with traditional processes is potentiated in terms of impact. Thus, we started this project assessing the influence that a technological environment would have in the patient’s behavior. Knowledge management systems can and should be used in order to optimize certain procedures, but the type of organization they are introduced in must be kept in mind. A model of knowledge management is described in Table 1. The purpose of this paper involved the development of a platform that would not interfere with patient’s answers when used, and that could be used and applied by health professionals. This software should run through a browser working in the health unit’s intranet, or even in the internet. The main requirement in the creation of this software was building an interface as close to a traditional paper form as possible. Using keywords like usability, accessibility and confidentiality, the intention was to build a simple interface with an intuitive use, where the correction to an answer could be done in a clear, objective way, where the patient could clearly understand the confidentiality of his answers, and accessible to all types of patients. Blind, illiterate and physically challenged individuals are recurrent amongst the frequent oncology service patients of the IPO (Portuguese Institute of Oncology) in Oporto. Sound or touch screens are presently the two used interface solutions, but we are still investigating the use of other
34
J. Gonçalves et al. / A Knowledge Management System to Study the Quality of Life
communication devices. Fig. 1 shows a view of the patient’s screen when he answers the questionnaire.
Fig. 1 – View of the patient when answering the questionnaire
5. Methodology and Results In order to assess the impact created by the application in the given answers we randomly selected patients from the otorhinolaryngology service in Oporto’s IPO. We selected fifteen days from May, June and July, and all patients attending consultations on those days were invited to participate in this study. All of them accepted the invitation. We obtained a sample of 54 individuals. These patients answered the same questionnaire twice, one in paper form – the traditional model – and the other on the computer using the software developed for that purpose, with 40 minutes temporal gap. Half of the patients answered first on the paper form and the other half answered first on the computer platform, the minimum time between answers was 40 minutes. In both cases the answer time was measured and the patient’s preference between the paper and the computer was registered. Information regarding patient’s affinity with computer use was also registered. In order to understand if the computer-based environment influenced or not the answers we analyzed the obtained values for each given answer, in both of the assessment moments, using a collection of statistical models and tests. Answers obtained in paper format and through the computer-based platform were matched. To understand if the computer-based platform did not influence the patients answers we hypothesized that distributions, for each variable in study, were identical. We first tested the entire set of answers and then two subsets, which divided patients that answered firstly on paper and patients that answered firstly on the computer. In the validation process two questionnaires were used, both from EORTC (European Organisation for Research and Treatment of Cancer): QLQ-C30 and QLQ-H&N35. The first one is a global questionnaire developed for all type oncologic patients. It has thirty questions grouped in five domains (physical, social, emotional, functional and cognitive). The second is a specific questionnaire for Head and Neck oncology patients, with thirty five questions. The two statistical hypotheses for a bilateral test in each situation were written: Hypothesis H0 F(X0)=F(X1); Hypothesis H1: F(X0)<>F(X1); We used the Wilcoxon test, the most appropriate when the dependent variable is measured in an ordinal scale [16]. In both of the questionnaires (QLQ-C30 and QLQH&N35) adopted to evaluate the QoL the test results did not allow to conclude if there
J. Gonçalves et al. / A Knowledge Management System to Study the Quality of Life
35
were significant differences between distributions, for the two samples and the three mentioned situations. A high level of significance was always attained, independently of the global or the partial analysis of the sample, divided between those who firstly answered on paper and those who firstly answered on the computer, so the hypothesis of a significant difference between answers not existing was accepted. We can thus state that the software use does not bias patient’s answers.
6. Data Analysis After the information registration stage, concerning the patient’s QoL, it is important to forward this information in a clear and objective way to the physician, to enable an improved decision making. This project is presently in that stage, and the participation of physicians has been fundamental not only to define which variables to investigate but also in the design of information output. Measures verified in clinical analysis differentiate patients from each other, but we understand that QoL measuring should also be considered in patient standardization. We are currently working in this area, but we still have very few results since we are only beginning. Notwithstanding, a first version of the output the physician will receive is shown next. Fig. 2 shows a patient’s answers: signaled in yellow are the answers that are bellow the expected for this type of patient, and signaled in blue are the ones above the expected.
Fig. 2 – Graphic with patient’s answers
7. Conclusions In this paper we defined the concept of QoL, in different contexts and situations, which reaches almost every sector of society. The main focus, however, was on the Health context. Some studies have suggested that the implementation of a patient HRQoL assessment in Portugal is challenged and questioned for several factors involving health institutions, health professionals and patients [4]. The reasons include: a lack of familiarity with relevant studies in this area; the absence of sensitivity; lack of time; reluctance in accepting that the patient’s perceptions regarding their own outcomes are as important as the physicians [5]; difficulty in quantifying subjective parameters; difficulty in converting tactic knowledge in explicit knowledge; inexistence of friendly computer-based applications; inexistence of health care service infrastructures that enable a routine HRQoL assessment. The purpose of this project was to give the
36
J. Gonçalves et al. / A Knowledge Management System to Study the Quality of Life
physician the opportunity to use patient’s QoL measurement as clinical decision support elements. We observed that a timely knowledge of the patient’s QoL-related elements constitutes another factor that may, in certain circumstances, contribute to a better decision making, and that a systematic patient QoL data collection allows the typification of this information and to infer therapeutic strategies for a specific patient. Moreover, therapeutic alternatives can help the physician by giving him important data from which he can infer the patient’s future QoL. We proved the validity of the developed platform in the acquisition of data required for the QoL assessment, and in allowing a routine QoL assessment to become a part of the appointment. We are now developing a platform for the collection of clinical information, in order to typify patients and therapies according to a specific patient’s QoL and a class of patients. The need to develop this platform underlines the importance of KMS as decision making aids.
References [1] [2] [3] [4] [5]
[6]
[7] [8]
[9]
[10] [11] [12] [13] [14]
[15]
[16]
F.L. Pimentel, Qualidade de Vida do Doente Oncológico, Edição: De autor, 2003. D. Schottenfeld, J.F. Fraumeni, Cancer epidemiology and Prevention, New York: Oxford University Press, 1996. B.W. Stewart and P. Kleihues, World Cancer Report, IARC Press, 2003. V.C. Rodrigues, Oral cancer in the UK: to screen or not to screen, Oral Oncology, 34 (1998), 454-465. M. Caçador, C. Adónis, R. Pacheco, H. Estibeiro, and J. Olias, Quality of life on the laryngeal preservation protocol-methods of evaluation, Revista Portuguesa de Otorrinolaringologia, 42, 4 (2001), 305-313. D.M. Hecker, J.P. Wiens, T.R. Cowper, S.E. Ekert, C.A. Gitto, R.F. Jacob, G.K. Mahanna, G.E. Turner, A. Potts, W. Logan, and R.L. Wiens, Can we assess quality of life in patients with head and neck cancer? A preliminary report from the American Academy of Maxillofacial Prosthetics, J Prosthet Dent, 88 (2002), 344-351. IPO, Roreno - Registo Oncológico do Norte, Instituto Português de Oncologia - Centro do Porto, 1996. G. Varzim, E. Monteiro, R. Silva, C. Pinheiro, M. Aires, and C. Lopes, Factores de predisposição genética para carcinoma espinocelular da laringe: a importância dos polimorfismos metabólico, Revista Portuguesa de Otorrinolaringologia, 39, 2 (2001), 131-139. J.L. Colli, and A. Colli, International comparisions of prostate cancer mortality rates eith dietary practices and sunlight levels”, Urologic Oncology - Seminars and Originals Investigations, 24, 3 (2006), 184-194. Pinheiro, P.S., Tyczynski, J.E., Bray, F., Amado, J., Matos, E., Miranda, A.C. e Limbert E., “Cancer in Portugal”, IARC Technical Publication, 38 (2002), 9-65. Pillai, M.R., Phanidhara, A., Kesari, A.L., Nair, P. e Nair, M.K., “Cellular manifestations of Human Papillomavirus infection in the oral mucosa”, Journal of Surgical Oncology, 71 (1999), 10–15. Meneses, R.F., Promoção da Qualidade de vida de doentes crónicos: Contributos no contexto das Epilepsias Focais, Edições Universidade Fernando Pessoa, 2005. A. Silveira, Qualidade de vida do Doente Oncológico da Cabeça e Pescoço do Instituto Português e Oncologia do Porto Francisco Gentil, Tese de Mestrado, Universidade do Porto, 2007. E.A. Sobrinho, M.B. Carvalho, and S.A. Franzi, Aspectos e tendências da avaliação da qualidade de vida de pacientes com câncer da cabeça e pescoço, Revista da Sociedade Brasileira de Cancerologia, 15 (2001.), 8–10. A. Bottomleya, S.E. Vachaleca, and K. Bjordalb, The development and utilisation of the European Organisation for research and treatment of cancer quality of life group item bank, European Journal of Cancer, 38 (2002), 1611–1614. J. Maroco, Análise Estatística com Utilização do SPSS, Edições Sílabo, 2007.
Detection of Nicotine Content Impact in Tobacco Manufacturing Using Computational Intelligence Lejla BEGIC FAZLIC1 and Zikrija AVDAGIC Faculty of Electrical Engineering of University of Sarajevo
Abstract. A study is presented for the detection of nicotine impact in different cigarette type, using recorded data and Computational Intelligence techniques. Recorded puffs are processed using Continuous Wavelet Transform and used to extract time–frequency features for normal and abnormal puffs conditions. The wavelet energy distributions are used as inputs to classifiers based on Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Genetic Algorithms (GAs). The number and the parameters of Membership Functions are used in ANFIS along with the features from wavelet energy distributionare selected using GAs, maximising the diagnosis success. GA with ANFIS (GANFIS) are trained with a subset of data with known nicotine conditions. The trained GANFIS are tested using the other set of data (testing data). A classical method by High-Performance Liquid Chromatography is also introduced to solve this problem, respectively. The results as well as the performances of these two approaches are compared. A combination of these two algorithms is also suggested to improve the efficiency of this solution procedure. Computational results show that this combined algorithm is promising. Keywords. Adaptive Neuro-Fuzzy Inference System (ANFIS); Genetic Algorithm (GA), fuzzy logic; nicotine detection, HPLC (High-Performance Liquid Chromatography).
Introduction Tobacco use is the world’s leading cause of death. Tobacco is addictive in all forms and it increases the risk of many cancers, heart attack, stroke, etc. Even second-hand smoke adversely affects pregnancy outcomes and causes lung cancer and heart disease. All cigarettes share the same features like nicotine, CO, length, filter type and tar concentration with very few differences . We used an approximately typical smoking pattern that consists of one 35 cm³ "puff" of 2 seconds duration once per minute. For successful applications of these Adaptive Interference (AI) techniques, several parameters needed to be selected mostly on trial and error basis. Combination of Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Genetic Algorithms (GAs) – GANFIS – was trained with a subset of cigarette nicotine data. The trained GANFIS are tested using the other set of data (testing data), not used in training. The results have been compared with ANFIS and High-Performance Liquid Chromatography 1
L. Begic Fazlic and Z. Avdagic / Detection of Nicotine Content Impact in Tobacco Manufacturing
(HPLC), and the results show the effectiveness of the proposed approach. Our goal was to predict nicotine impact using proposed algorithm and minimizing the complexities as much as possible.
1. ANFIS Architecture Based on the Sugeno Fuzzy Model [2], ANFIS was proposed by Roger Jang in 1992 [3, 4]. The architecture of a two-input two-rule ANFIS is shown in Figure 1.
Figure 1. ANFIS architecture
ANFIS has five layers, in which node functions of the same layer have the same function type as described below [2, 3]: Layer 1: Every node i in this layer is an adaptive node with node function: O j ,i
premise parameters updated through Back Propagation
Learning Algorithm. Layer 2: Every node i in this layer is a fixed node labeled П, whose output is the product of all the incoming signals: O2,i
Yi
P A ( x) u P B ( x) i
i
Layer 3: Every node i in this layer is a fixed node labeled N. The ith node calculates the ratio of the ith rule’s firing strength to the sum of all rules’ strengths:
L. Begic Fazlic and Z. Avdagic / Detection of Nicotine Content Impact in Tobacco Manufacturing
39
Yi Y 1 Y 2
O3,i Y i
Layer 4: Every node i in this layer is adaptive node with node function:
Y i fi
O4, j
Y n ( pi ( x) qi ( y) ri )
Where ^pi , qi , ri `, are consequent parameters [3, 4], updated through Recursive LeastSquares Estimation. Layer 5: The single node in this layer is a fixed node labeled Σ, which computes the overall out-put as the summation of all the incoming signals:
O5
¦Y
¦Y f ¦Y i
i
i
fi
i
i
i
i
Y 1 f1 Y 2 f 2 Y1 Y 2
i
2. Genetic Algorithms Genetic algorithms are one of the best ways to solve a problem for which little is known. They are very general algorithms and so, they will work well in any search space. Genetic algorithms use the principles of selection and evolution to produce several solutions to a given problem. The use of GA needs the consideration of six basic issues: chromosome (genome) representation, creation of initial population, selection function, genetic operators such as mutation and crossover for reproduction function, termination criteria, and the evaluation (fitness) function [7, 8]. A population of 240 individuals was used, starting with randomly generated genomes. GA was used to select the most suitable features and one variable parameter related to the particular classifier: the number of membership function for ANFIS. For a training run needing N different inputs to be selected from a set of Q possible inputs, the genome string would in the genome ( ) consist of N + 1 real numbers. The first N integers are constrained to be in the range 1≤N≤Q
(1)
(2) The last number XN + 1 has to be within the range Bmin ≤ xN + 1 ≤ Bmax. The parameters Bmin and Bmax represent the lower and the upper bounds on the classifier parameter. A probabilistic selection function [7], namely normalised geometric ranking, was so used that the better individuals, based on the fitness criterion in the evaluation function, had a higher chance of being selected. The maximum number of generations (240) was adopted as the termination criterion for the solution process. The
40
L. Begic Fazlic and Z. Avdagic / Detection of Nicotine Content Impact in Tobacco Manufacturing
prediction success for the training data was used as the fitness criterion in the evaluation function [5, 6]. For each puff signal, wavelet detail coefficients at the first four levels (four sets of coefficients), and wavelet approximation coefficients at the fifth level (one set of coefficients) were computed. A total of four values for the five wavelet coefficient sets were computed for using them to form the feature vectors. The best features are determined by applying divergence analysis [5] to the training sets of 240 vectors. Divergence analysis enables to determine the best features which increase the classification performance, and also it gives information to determine the dimension of the feature vectors.
3. GANFIS Architecture There is also a need to develop and automated detection and diagnostics system combining advanced signal processing techniques with Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Genetic Algorithms (GA). In this paper, the new approach is extended to combine GA with ANFIS [6], termed here as GANFIS, for automatic selection of inputs and the number of Membership Functions (MFs) in the detection of nicotine in cigarettes. The process flow chart is shown in Figure 2.
GA Parameter initialization
Training data Recordered puff sounds
Extraction of feature
Population initialization
Evaluation of fitness function
ANFIS TRAINING
Testing data Crossover
Selection
Mutation
Reachead termination criteria
Detection of nicotine impact
Figure 2: Process algorithm
In the first stage, we identified the relationship between responses and input variables. In the second stage, we used optimization techniques to obtain a setting of input variables, which give system the most desirable responses. First, puff recorded signals are processed through advanced signal processing techniques for feature extraction. A portion of the extracted dataset is then used for training the CI classifiers, and the rest for testing.
L. Begic Fazlic and Z. Avdagic / Detection of Nicotine Content Impact in Tobacco Manufacturing
41
During the training phase, GA helps in getting the optimized parameters of the classifiers (ANFIS) and the most suitable features for best classification performance. Once the set of features and the classifier parameters have been automatically selected by GA satisfying the termination criteria, the effectiveness of the process was verified using the test data for diagnosing the patient’s lung condition. The puffs data had been processed extracting wavelet coefficients which were then used to get the wavelet energy distribution.
4. Problem Description Nicotine and tar content is measured by a machine that smokes every cigarette in exactly the same status. We used different classes of cigarettes and different classes of filter types. Studies show that smoking of low-nicotine has same impact on lung as smoking of classical nicotine cigarettes. So although the cigarette might contain less nicotine, the impact on lung diseases is actually at the same amount.
5. Experimental Results Prediction of nicotine impact by different types of filter cigarettes is a typical problem. Several attributes of the cigarette's profile information are used to predict another continuous attribute, that is, the nicotine contentration. We used data collected from different tobacco manufactures. The spectrum of the puff sound signal was divided into sub-bands to extract the discrepancies between abnormal and normal subjects. Because puff (respiratory) signals are non-stationary, the discrete wavelet transform was involved for the sub-band analysis. For each signal, wavelet detail coefficients at the first four levels (four sets of coefficients), and wavelet approximation coefficients at the fifth level (one set of coefficients) were computed. From the calculated five RMS values, a subset of the best coefficients was searched by ANFIS. The data set was obtained from the original data file 'cig.dat' [8]. The variation of the wavelet energy distribution for different puff sound data can be used for detection nicotin impact. A total of four values for the five wavelet coefficient sets were computed, and they have been used to form the feature vectors. The best features were determined by applying divergence analysis [6] to the training sets of 240 vectors. Divergence analysis enables to determine the best features which increase the classification performance. Furthermore, it gives information to determine the dimension of the feature vectors. This paper presented a new application of the combination GA and ANFIS model for the detection of nicotine impact on puffs in different types of cigarettes. The predictions of the ANFIS classifiers were combined with the GA classifier. For GANFIS, the number of input features was varied from 1 to 5, and the number of MFs was in the range of 2–4. We used generalised bell curve MFs. The initial input MFs were generated spanning uniformly over the range of each input. The output MFs were taken as constants for the zero-order Sugeno fuzzy model. The number of MFs for all inputs was taken equal. In GA, the genome consisted of the most suitable features and the classifier parameter. A population size of 240 individuals was used starting with randomly generated genomes. The prediction accuracy (RMSE) for the training data was used as the fitness criterion in the evaluation function. Table 1 shows the results of the test success of GANFIS for different numbers of selected inputs (1–5) from the
42
L. Begic Fazlic and Z. Avdagic / Detection of Nicotine Content Impact in Tobacco Manufacturing
input feature and the different puff states corresponding to one test dataset. The target outputs for three states were assigned as LessImpact(0.0), Tolerant(0.5) and BigImpact (1.0). The number of membership function selected by GA was 3. Table 1: Test success of GANFIS Inputs number
MFs number
Training time
Output
Output
Output
(0)
(0.5)
(1)
RMSE
1
3
5,34
0.0322
0,48
1
0.023
2
3
8,76
0.0162
0,493
1
0.020
3
3
13,23
0.0004
0,5
1
0.003
4
3
156,2
0.0001
0,5
1
0.002
5
3
1328,3
0.0000
0.5
1
0.000
6. Conclusion This paper presented a new application of the GANFIS model for the detection of nicotine impact in different types of cigarettes, using recorded data and Computational Intelligence techniques. Recorded puffs are processed using Continuous Wavelet Transform and used to extract time–frequency features for normal and abnormal puffs conditions. GANFIS are trained with a subset of cigarette nicotine data. The trained GANFIS are tested using the other set of data, not used in training. The results are compared with HPLC and ANFIS approach [5]. The classification success of HPLC was found to be better than GANFIS, but the training time for the former was much higher than the latter for a small number of inputs. We incorporate smoker behavior into the new artificial ‘smoking’ method to provide a more accurate reflection of in vivo smoker exposure to toxic compounds than indicated by package labeling.
References [1] [2] [3] [4] [5] [6]
[7] [8]
Z. Avdagic, Lejla Begic Fazlic, Huse Fatkic „Automatic detection of Co Content in Smoke Condensate using Adaptive Neuro Fuzzy Inference Systems ”, Inista 2007. D. Dubois and H. Prade, “An introduction to fuzzy systems”, Clinica Chimica Acta 270, Elsevier Science Inc. New York, NY, USA, 1998, pp. 3–29. J.-S.R. Jang, “ANFIS: Adaptive-network-based fuzzy inference system”, IEEE Trans. Syst. Man Cybern. 23 (3) (1993), 665–685. J.-S.R. Jang, “Self-learning fuzzy controllers based on temporal backpropagation”, IEEE Trans. Neural Networks 3 (5) (1992), 714–723. L. McIntyre, Using Cigarette Data for an Introduction to Multiple Regression, Journal of Statistics Education, Vol.2, No. 1, 1994. http://www.amstat.org/publications/jse/v2n1/datasets.mcintyre.html B. Samantha, C. Nataraj, Automated diagnosis of cardiac state in healthcare systems using computational intelligence, International Journal of Services Operations and Informatics, Vol 3., No.2 (2008), 162-177. Z. Michalewicz, Genetic algorithms + Data Structures = Evolution Programs, Springer-Verlag, New York, 1999. L. Begic Fazlic, Z. Avdagic, Optimized Detection of Tar Content in the Manufacturing Process Using Adaptive Neuro-Fuzzy Inference Systems, In: K.-P. Adlassnig, B. Blobel, J. Mantas, I. Masic (Edrs.) Proceedings of MIE 2009: Medical Informatics in a United and Healthy Europe, pp 615-619, SHTI Vol. 150, IOS Press, Amsterdam, 2009.
Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs Christoph RINNERa,1, Michael KOHLERa, Gudrun HÜBNER-BLODERb, Samrend SABOORb, Elske AMMENWERTHb and Georg DUFTSCHMIDa a Working Group Medical Information and Retrieval Systems – MIAS of the Section for Medical Information Management and Imaging, Medical University Vienna, Austria b Institute for Health Information Systems, UMIT – University for Health Sciences, Medical Informatics and Technology, Hall/Tyrol, Austria
Abstract. Archetypes model individual EHR contents and build the basis of the dual-model approach used in the ISO/EN 13606 EHR architecture. We present an approach to create archetypes using an iterative development process. It includes automated generation of electronic case report forms from archetypes. We evaluated our approach by developing 128 archetypes which represent 446 clinical information items from the diabetes domain. Keywords. EHR, ISO/EN 13606, Archetypes, semantic interoperability
Introduction Cross-organizational electronic health record (EHR) communication, which enables authorized healthcare providers to access all relevant patient data regardless of where the data were created, will constitute a key component of future health care. A key requirement for EHR communication is semantic interoperability. Semantic interoperability communicates meaning and hereby aims to ensure that the communicated information is understood in exactly the same way by the sender and the recipient [1]. The ISO/EN 13606 EHR architecture standard [2, 3] is a framework for achieving semantic interoperability. In February 2010, part 5 of 13606 was ratified by CEN and ISO as the last of 5 parts. Since then the complete 13606 has the status of an official European and International standard. ISO/EN 13606 is based on the dual-model approach which combines a static reference model and archetypes to represent EHR contents. Archetypes are used to model individual EHR contents by constraining the reference model. They form an important layer for semantic interoperability between communicating EHR systems. When supported by suitable tools, the archetype concept allows medical domain experts and computer scientists to jointly and efficiently model EHR contents in a computer processable form.
1 Corresponding Author: Christoph Rinner, Dipl.-Ing.; Medical University of Vienna, Center for Medical Statistics, Informatics, and Intelligent Systems, Spitalgasse 23, 1090 Vienna, Austria; Email: [email protected], Phone: (+43)1 40400-6693; Fax: (+43)1 40400-6697; Url: http://www.meduniwien.ac.at/user/christoph.rinner/
44
C. Rinner et al. / Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs
In the following we present an approach to create ISO/EN 13606 archetypes using an iterative development process. It includes an automated generation of electronic case report forms (eCRFs) from archetypes which enables immediate testing and creation of EHR data compliant to the reference model and archetypes. We evaluated our approach by developing 128 archetypes which represent 446 clinical information items. These items were collected in an analysis of the information needs of physicians when treating diabetes patients [4]. The development of archetypes as well as the analysis of information needs in the treatment of diabetes are parts of the EHR-ARCHE2 project. In this project we examine to what extent dual-model based EHR architectures can support EHR users in selectively retrieving information that is relevant in their respective search context and thus avoid information overload.
1. Method For the development of the archetypes we propose an approach which adapts an iterative software development model [5] to the archetype domain (see Figure 1). It extends an existing approach described in [6] in a way that adds an iterative archetype refinement and especially supports the participation of medical domain experts in the development of archetypes.
Figure 1. Iterative archetype development model.
Based on the required clinical information items and existing archetypes the initial planning is done. Using the results of the planning step, archetypes are created and adapted. From these archetypes eCRFs are automatically generated that can be used to test archetypes and create test EHR documents. Based on the evaluation of the results, the required changes are planned and the cycle continues until the archetypes can finally be released. 1.1. Initial planning The initial design of archetypes combines the first two steps of [6]. In the first step “Define high level archetype concepts”, the medical concepts that should be described with archetypes have to be structured and organized. This organization can be done as a mind map or a table representing the hierarchical order. During this phase the medical domain expert and the computer scientist are working together very closely, hence a format that is understood by both parties should be chosen. 2
http://www.meduniwien.ac.at/msi/arche/
C. Rinner et al. / Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs
45
In the second step “Check for existing reference archetypes”, archetypes that match the medical concepts are searched. By specializing existing archetypes uncontrolled growth of archetypes can be prevented and semantic interoperability with EHR systems which already use these archetypes is made possible. Furthermore existing archetypes already have numerous refinement and revision cycles, which make them a valuable source. 1.2. Creation and adaptation of archetypes Based on the results of the planning new archetypes are created or existing archetypes are specialized and refined. The archetype coding is done by computer scientists. Even though we employ visual archetype editors in this step as recommended in [6], several properties of archetypes (e.g. inheritance, cardinalities) still remain rather unfamiliar for medical domain experts. 1.3. Generation of eCRFs and creation of documents Extending the approach in [6], we apply a tool that allows on-the-fly eCRF generation from archetypes and thus the creation of archetype conformant EHR data. Using the well known concepts of forms and documents, medical domain experts can hereby more easily evaluate the archetypes, e.g. spot missing elements or wrong cardinalities. Further the generation of an eCRF from an archetype ensures a formal correctness of the archetype in various facets. The correctness of the archetype definition language (ADL) specification [2, 3] of the archetype and the conformance to the reference model and data types are typically already checked in the archetype editor. Our tool further checks whether external archetypes referred to in slots and specializations are present, and whether archetype node names, translations, and term bindings are correctly defined. Both, the testing of archetypes via eCRFs and their formal verification are means to improve the quality of the archetypes. 1.4. Evaluation and planning The evaluation step may lead to further planning and refinement of the archetypes. The evaluation relies on the information obtained during the formal verification of the archetype and the information obtained by the medical domain expert when applying the eCRFs. The results of the evaluation flow into the next development cycle. The latter starts with planning the required changes. This step needs a close collaboration between the computer scientist and the medical domain expert. 1.5. Release Archetypes After the last adaptations have been made, the archetype can be released. OpenEHR archetypes should be submitted to the Clinical Review Board of openEHR. For ISO/EN 13606 no such infrastructure exists, however it is important that all partners exchanging information based on the ISO/EN 13606 standard have access to the archetypes to enable semantic interoperability.
46
C. Rinner et al. / Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs
2. Results The initial planning was based on 446 clinical information items which were collected in an analysis of the information needs of physicians when treating diabetes patients [4]. These information items were structured hierarchically, i.e. higher level groupings of related finer-grained items were already identified in this step. Based on this hierarchical list, existing archetypes were searched. For ISO/EN 13606 no archetypes are publicly available, hence we could only rely on openEHR archetypes. Available resources included the openEHR Clinical Knowledge Manager 3, the openEHR SVN repository4 and the NHS SVN repository5. For roughly one quarter of the clinical information items existing openEHR archetypes were found. Originally we planned to convert the existing openEHR reference archetypes into ISO/EN 13606 archetypes using [7] and specialize these archetypes according to our requirements. However, we finally decided to refrain from this idea. The main reason was that the openEHR approach includes the concept of templates, which allows existing archetypes to be further constrained and thus customized for local purposes. As a result openEHR archetypes are designed to include the maximum set of potentially relevant items for a particular medical concept. The ISO/EN 13606 standard however does not support the template mechanism. Therefore we would have had to reduce the typically extensive openEHR archetypes to those few items relevant for our context by specializing the converted openEHR archetype and setting the cardinality of most items to zero. As this seemed as a laborious and also not really convincing approach, we decided to create our own set of archetypes and replicate those structures of existing archetypes that were relevant in our context. Still, the openEHR archetypes found were a valuable source during the archetype creation and complemented our clinical information items. Archetypes were imported into a tool developed in the course of the EHR-ARCHE project which allowed the automatic generation of eCRFs (compare section 1.3). The documents collected via the eCRFs can either be stored locally during testing or synchronized with an IHE XDS [8] repository using the SENSE [9] framework. The creation of the test documents and the evaluation using the eCRFs was done by a medical domain expert located in Hall/Tyrol. The results were documented using a text file. Computer scientists in Vienna created and adapted the archetypes for the next iteration cycle based on this information. Our approach was tested with 128 ISO/EN 13606 archetypes that represent 446 information needs. The archetypes were created using the linkEHR archetype editor [10] and translated to German and English. The composition of archetypes (350 times) using the ADL slot mechanism and the specialization of archetypes (40 times) were intensively employed. In Figure 2 this is illustrated for the “family history” archetype.
C. Rinner et al. / Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs
47
Figure 2. Example of “Family History” archetype.
This archetype occurs in 4 COMPOSITION archetypes as slot and has 4 ENTRY archetypes referenced via slots (black line and dark arrow). Two of these ENTRY archetypes are specializations of other ENTRY archetypes (dashed line and white arrow). In our test setting, the medical domain expert and the computer scientist were geographically separated. After the initial planning, where a close collaboration between the two was needed, all other steps were done asynchronously which allowed free time management for both parties. At the average each archetype underwent 2 – 3 development cycles.
3. Discussion and Further Steps In our method for archetype development we emanate from an approach described in [6]. In particular, our steps Initial planning, Creation and adaptation of archetypes, and Release Archetypes are conformant to [6]. We extend their approach by adding the concept of an iterative archetype refinement which is derived from general software development models. Further we added the steps Generation of eCRFs and creation of documents as well as Evaluation and planning to the development process. By automatically generating eCRFs from archetypes and creating EHR documents via the eCRFs, medical domain experts could easily evaluate the archetypes. The iterative archetype refinement then provided for the integration of changes identified in the evaluation. These additions of the approach proved useful for the integration of medical domain experts in the development of archetypes and helped to raise the archetypes’ quality. Several tools have been published for the automatic creation of eCRFs from archetypes [11-13]. We decided to develop our own tool, as we needed a 13606 environment instead of an openEHR one [12]. We further did not want to be limited by the conditions of a particular EHR system in which the eCRFs are created [11]. The system described in [13] came closest to our needs. However, we desired a direct integration of the tool in our IHE XDS environment to enable immediate processing of the EHR documents created via the eCRFs. We identified a problem in our method when trying to reuse openEHR archetypes in the context of the ISO/EN 13606 standard. As explained in section 2, openEHR archetypes typically cover the maximum set of potentially relevant information items for a medical concept. They are then “reduced” to the needs of a particular setting via openEHR templates. As the template concept does not exist in the ISO/EN 13606 standard, openEHR archetypes are frequently too “overloaded” for an efficient reuse in
48
C. Rinner et al. / Creating ISO/EN 13606 Archetypes Based on Clinical Information Needs
a 13606-based domain. Although we replicated relevant parts of the openEHR archetypes in our ISO/EN 13606 archetypes, we hereby lost the formal relation to the source archetypes. We developed an archetype repository in which we stored all 128 archetypes. During further progress of the EHR-ARCHE project we plan to offer public access to the archetype repository via the project homepage. In the next step of the project we will create a set of EHR documents based on our archetypes. These EHR documents will later form the test set for selectively retrieving relevant information in a particular search context by means of an archetype-based query mechanism.
Acknowledgement The project EHR-ARCHE is funded by the Austrian Science Fund, project number P21396.
References [1] [2] [3] [4]
[5] [6] [7] [8] [9] [10]
[11]
[12] [13]
P. Gibbons, et al., Coming to Terms: Scoping Interoperability for Health Care. 2007, Health Level Seven EHR Interoperability Work Group. European Committee for Standardization, EN 13606 Electronic healthcare record communication. 2007. International Organization for Standardization, ISO 13606 Electronic health record communication. 2008. G. Hübner-Bloder, G. Duftschmid, M. Kohler, C. Rinner, S. Saboor, and E. Ammenwerth. Systematische Erhebung der Informationsbedürfnisse von Ärzten bei der Behandlung von Diabetes mellitus Patienten. in 55. GMDS-Jahrestagung. 2010. Mannheim. C. Larman and V.R. Basili, Iterative and incremental development: A brief history. Computer 36(6) (2003), 47-56. S. Garde, E. Hovenga, J. Buck, and P. Knaup, Expressing clinical data sets with openEHR archetypes: A solid basis for ubiquitous computing. Int J Med Inform 76(5-6) (2007), 334-341. C. Martinez-Costa, M. Menarguez-Tortosa, and J.T. Fernandez-Breis, An approach for the semantic interoperability of ISO EN 13606 and OpenEHR archetypes. J Biomed Inform 43(5) (2010), 736-46. Integrating the Healthcare Enterprise (IHE), IT Infrastructure Technical Framework, vol. 1 (ITI TF-1, chapter 10), vol. 2 (ITI TF-2, chapter 3.14 and Appendix L), I.t.H.E. (IHE), Editor. 2007. ITH icoserve. sense - smart eHealth solutions. 2010; Available from: http://www.ithicoserve.com/loesungen/sense-smart-ehealth-solutions/uebersicht/. J.A. Maldonado, D. Moner, D. Boscá, J.T. Fernández-Breis, C. Angulo, and M. Robles, LinkEHR-Ed: A multi-reference model archetype editor based on formal semantics. Int J Med Inform 78(8) (2009), 559-70. J. Chaloupka, Automated integration of archetypes into electronic health record systems based on the Entity-Attribute-Value model, in Section for Medical Information Management and Imaging. Diploma Thesis 2009, Medical University of Vienna: Vienna. S. Arikan, T. Shannon, and D. Ingram. Opereffa. 2009; Available from: http://opereffa.chime.ucl.ac.uk/introduction.jsf. A. Brass, D. Moner, C. Hildebrand, and M. Robles, Standardized and flexible health data management with an archetype driven EHR system (EHRflex). Stud Health Technol Inform Vol. 155, 212-8, IOS Press, Amsterdam, 2010.
Extracting Clinical Information to Support Medical Decision Based on Standards Valentin GOMOIa,1,Mihaela VIDAa, Lăcrămioara STOICU-TIVADARa and Vasile STOICU-TIVADARa a University “Politehnica” Timişoara, România
Abstract. The paper presents a method connecting medical databases to a medical decision system, and describes a service created to extract the necessary information that is transferred based on standards. The medical decision can be improved based on many inputs from different medical locations. The developed solution is described for a concrete case concerning the management for chronic pelvic pain, based on the information retrieved from diverse healthcare databases. Keywords. Guidelines, protocols, standards, medical decision, syntax.
Introduction Using medical guidelines and protocols resulted in the improvement of patient care. A guideline is a document with recommendations and instructions to assist the medical professional and the patient in decision making, based on results of scientific research followed by discussion and expression of expert-opinions, to make effective and efficient medical practice explicit [1]. A clinical protocol is defined as a local adaptation of one or more clinical guidelines. A protocol provides information on duration, dosages, procedures which are not present in the guideline [2]. Guidelines and protocols can be generated, edited, verified, executed and visualized with the help of computers. The computerized implementation of guidelines has as advantages: faster implementation of new medical knowledge, integration of many local databases, decreasing in costs, better protocol visualization, in a graphical manner, avoids reading vast amounts of data regarding each step of a narrative medical protocol [3]. The integration of many local databases represents an advantage for clinical practice, gathering more relevant clinical information. Putting together information from different sources provides the inconvenience that data is stored in many formats. A solution created and implemented in a certain medical unit is very difficult to be deployed in another unit. To enable communication between different medical units, some standards have to be applied, e.g. HL7 CDA (Health Level 7 Clinical Document Architecture) [4, 5]. Consequently, the medical decision system has to implement the same standard. To solve the above mentioned issues, Semantic Web technologies are used to create data stores on the Web, build vocabularies, and write rules for data handling. XML (Extensible Markup Language) and ontology (e.g. Web Ontology Language) are 1
Corresponding Author: Valentin Gomoi, Faculty of Automation and Computers, Bd. V. Parvan 2, 300223, Timisoara, Romania, email: [email protected]
50
V. Gomoi et al. / Extracting Clinical Information to Support Medical Decision Based on Standards
two components of the Semantic Web. In our case the XML is a HL7 CDA message in XML format [6]. The HL7 Clinical Document Architecture (CDA) is a document markup standard that specifies the structure and semantics of clinical documents for the purpose of exchange. A CDA document is a defined and complete information object that can include text, images, sounds, and other multimedia content. CDA documents are encoded in Extensible Markup Language (XML). The clinical content of CDA documents is defined in the RIM (Reference Information Model) [7, 8]. The paper presents a method for connecting medical databases and a medical decision system - Egadss, and a service created to extract the necessary information that is transferred based on HL7 CDA standard. This results in improved guidelines and protocols for clinical diagnosis.
1. System Components In this section the two main components of the system are described: the medical decision support used to generate recommendations for a certain patient, and the service created to retrieve information from multiple medical databases. The communication between databases and Egadss application [9, 10 and 11] is realized based on HL7 CDA [8]. The Egadss application supports only HL7 CDA messages as data inputs [9]. Figure 1 presents the proposed system architecture. The application sends a data request.
Figure 1. System architecture
V. Gomoi et al. / Extracting Clinical Information to Support Medical Decision Based on Standards
51
1.1. HL7 CDA Component The HL 7 CDA Component extracts the data from a local database and presents it as HL7 CDA in XML format and provides it to Egadss. The HL7 CDA Component implementation is made in Visual Studio .net 2008, in C# language. The databases for the current activity are on SQL Server 2008, but the solution is the same for Oracle or MySQL. The HL7 CDA message is implemented currently in C#, but it can be implemented also in Java, the only requirement being the XML format. Egadss reads an XML file. The HL7 component represents a major improvement of the existing solutions, allowing the decision making part to have access to more complex information, besides the one that can be found only in local databases. An example of an XML file is presented in Figure 2, containing the data concerning the laboratory results in HL7 CDA format. In the Laboratory section the value of potassium test is represented and for this an ICD-9 (International Classification of Diseases) code is associated [12]. This information is used by Egadss.
Figure 2. Representative data from the CDA message.
1.2. Egadss Egadss is a clinical decision support system that has as inputs HL7 CDA standard messages as XML files. This results in a standardized communication interface between databases and “recommendation generator”. Egadss extracts the patient data from the XML and uses this information to generate new clinical recommendations, making the system very flexible and easy to be implemented in any medical unit. The recommendations resulted from Egadss are also structured as a CDA level 2 document. Another advantage of Egadss is the use of Arden Syntax for the representation of the guidelines. The Arden Syntax is a clinical guideline formalism accepted as an official standard by HL7 [10]; it is a textual language, intuitive and without room for ambiguity. It is freely available, a mature and actively maintained open standard. This is the reason why Arden Syntax is used instead of other guideline formalisms as Proforma, GLIF (GELLO), Asbru, etc. [9, 10, 11, 13]. Egadss is an already developed system and value is added using the HL7 CDA Component.
2. Guideline Representation and Computer Implementation Based on the new HL7 CDA Component the decision making system has access to data from different clinical institutions offering more precise medical recommendations based on this information.
52
V. Gomoi et al. / Extracting Clinical Information to Support Medical Decision Based on Standards
In order to obtain medical recommendations based on the data received from the HL7 CDA Component, Egadss uses guidelines represented with Arden Syntax, as a knowledge source for its expert system [9]. Currently Egadss uses ICD9, and in Romania is used ICD10 [12]. For the current study, the HL7 CDA Component and Egadss were tested for the management of chronic pelvic pain (chosen as the model guideline used to test the data) adapted to the Romanian situation. A branch of the protocol for the “Management of chronic pelvic pain” is presented in Figure 3 [14]. This branch represents the recommendation that should be made in case that the patient had voiding symptoms and he made a urinalysis. In Figure 4, the Knowledge part of the MLM (Medical Logic Module) used to represent this part of the guideline can be seen. A MLM is the Arden Syntax representation of a single medical decision. Each guideline can by defined with the help of multiple MLMs. Each module contains 3 slots: Maintenance (contains: MLM name, institution, version, author, specialist), Library (contains: keywords, explanations, purpose, citations) and Knowledge which contains the medical knowledge grouped in 5 mandatory slots (type, data, evoke, logic, action) and other two optional slots (priority, urgency) [15].
Figure 3. Management of chronic pelvic pain
The next step in implementing the guideline is to obtain the necessary rules for clinical recommendations, based on the MLM file read by Egadss, which translates the medical rules from the MLM into CLIPS rules. These are then used by the expert system from Egadss to generate the medical recommendations. CLIPS is the basis for the guideline execution mechanism in Egadss [11]. After running the CLIPS rules on the patient data, new recommendations were produced. The patient data are extracted from the XML (section 1.2) received from the HL7 CDA Component. This implementation of the HL7 CDA Component used in con-
V. Gomoi et al. / Extracting Clinical Information to Support Medical Decision Based on Standards
53
junction with Egadss and implementing the HL7 CDA had shown that a future work in developing an automatic generator of computer based-guidelines and protocols can be based on these components.
Figure 4.Medical Logic Module for chronic pelvic pain
3. Evaluation and Future Developments Developing a system which implements standards for the main inputs and outputs, makes easier the integration of different medical institution in order to improve medical decision and patient care. Implementing the HL7 CDA Component for retrieving data improves the access to information from different institutions from different countries where the different components of the system are installed. In this manner the recommendations which are generated are more precise. Based on the already developed components that were presented in this paper, a future improvement of the system will be the integration of data mining techniques in retrieving medical information concerning a large number of patients and medical unit components. A further major step in the development of this system is the introduction of a module that is able to automatically generate medical rules based on the information obtained from medical databases concerning a certain situation or symptoms. This should lead to a number of steps (a protocol) for the treatment of patients. This also implies the implementation of a powerful tool for the extraction and analyses of large amounts of information. The module can be easy turned into a validation machine for the existing guidelines. This can be achieved by comparing the results of the steps presented in the guideline and the same steps which are applied for the same symptoms or the same medical events and represented in medical databases.
4. Conclusions In this paper is presented a methodology and a system supporting medical recommendations generation. The implementation of HL7 CDA standard resulted in a better communication between different components of the system and enhanced its complexity. Using standards helps integrating the system in any medical unit. In this way, one of the major problems, the local adaptation of clinical decision support systems can be solved.
54
V. Gomoi et al. / Extracting Clinical Information to Support Medical Decision Based on Standards
Implementing HL7 CDA Component and Egadss will help the doctors to increase the quality of medical care, reduce the variation in medical practice, give more efficient treatments, and use new medical knowledge in their current clinical practice. It is a good solution for cross-border cooperation, for studies in border areas from different countries that have the same medical issues, as a support for reducing costs and improving clinical practice. The presented system can be further developed to support the automatic generation of medical guidelines and protocols.
Acknowledgment This work was partially supported by the strategic grant POSDRU/88/1.5/S/50783, Project ID50783 (2009), co-financed by the European Social Fund – Investing in People, within the Sectorial Operational Programme Human Resources Development 2007-2013 and National grant 11-066 / 18.09.2007 TELEASIS.
References [1]
[2]
[3]
[4]
[5]
[6] [7]
[8] [9]
[10]
[11] [12] [13] [14] [15]
K. Rosenbrand, J. van Croonenborg, J. Wittenberg, Guideline Development, in Computer-based Medical Guidelines and Protocols: A Primer and Current Trend, A. tenTeije, S. Miksch and P.J. Lucas, vol. 139 (2008), 3 -21. P. Groot, A. Hommersom, P. Lucas, Adaptation of Clinical Practice Guidelines, in Computer-based Medical Guidelines and Protocols: A Primer and Current Trend, A. tenTeije, S. Miksch and P.J. Lucas, vol. 139 (2008), 121-139. K. Rosenbrand, J. van Croonenborg, J. Wittenberg, Guideline Development, in Computer-based Medical Guidelines and Protocols: A Primer and Current Trend, A. tenTeije, S. Miksch and P.J. Lucas, vol. 139 (2008), 3 -21. M. Vida, M, V. Gomoi, L. Stoicu-Tivadar, V. Stoicu-Tivadar, Using HL7 CDA Standard to Connect Medical Databases as a Support for Automatic Generation of Computer Interpretable Guidelines, Proceedings of the 31st National Conference on Medical Informatics, (2010), 69-74. M. Vida, M, V. Gomoi,L. Stoicu-Tivadar, V. Stoicu-Tivadar, Generating medical computer-based protocols using standardized data transmission, Soft Computing Applications (SOFA), 2010 4th International Workshop, (2010), 155 – 158. XML and OWL, http://www.w3.org/, 17.08.2010 B. Blobel , K. Engel, P. Pharow , HL7 Version 3 Compared to Advanced Architecture Standards, Methods of Information in Medicine 45 (2006) 343–53. Available at: http://hl7t3f.org/documents/schattauer_9_2006_4_343.pdf. Accessed in 15.10.2010 HL7 Clinical Document Architecture, Release 2.0, HL7 version 3 Interoperability Standards, Normative Edition 2009, Disk 1 – Standards Publication. I. Bilykh, J. H. Jahnke, G. McCallum, M. Price, Using the clinical document architecture as open data exchange format for interfacing EMRs with clinical decision support systems, Proceedings of the 19th Symposium on Computer-Based Medical Systems (CBMS’06), (2006), 855-560. J. H. Weber-Jahnke, G. McCallum, A light-weight component for adding decision support to electronic medical records, Hawaii International Conference on System Sciences, Proceedings of the 41st Annual, (2008), 251 – 251. Documentation about medical decision support systems, http://www.egadss.org/, 15.08.2010 International Classification of Disease (ICD), http://www.who.int, Accessed in 17.09.2010 G. McCallum. EGADSS: A Clinical Decision Support System for use in a Service-oriented Architecture, MSc. Thesis, Computer Science, University of Victoria, BC, Canada, (2006) J. E. Turrentine, Clinical protocols in obstetrics and gynecology, third edition, (2008), pp. 82-85. P.Clercq, K. Kaiser, A. Hasman, Computer-Interpretable Guideline Formalisms, in Computer-based Medical Guidelines and Protocols: A Primer and Current Trend, A. ten Teije, S. Miksch and P.J. Lucas, vol. 139, (2008), 22-43.
eHealth in Switzerland – Building Consensus, Awareness and Architecture Christian LOVIS a,1, Hansjorg LOOSER b, Adrian SCHMID c, Judith WAGNER d and Stefan WYSS c a University Hospitals of Geneva b Ministry of Health, Canton St.Gallen c Swiss ehealth Coordination Office d Swiss Medical Association
Abstract. This paper reports on the process of the Swiss national strategy to define and implement eHealth. Switzerland is a federal political organization with 26 cantons that are autonomous for the health legal framework. Switzerland must also provide support for four national languages. Thus, this experience addresses many challenges that are experienced at the European level in a much larger scale. Also, Switzerland benefits from the major projects ongoing in Europe, such as epSOS, to define its own strategy. Keywords. eHealth, national infrastructure, national strategy, architecture
Introduction Background Switzerland is a federal republic consisting of 26 cantons, with Bern as the seat of the federal authorities, situated in Western Europe and bordered by Germany to the north, France to the west, Italy to the south, and Austria and Liechtenstein to the east. The country is geographically situated between the Alps, the Central Plateau and the Jura and has a population of approximately 7.8 million people concentrated mostly on the Plateau. It is one of the richest countries in the world regarding the per capita gross domestic product. Switzerland comprises three main linguistic and cultural regions: German, French, and Italian, to which the Romansh-speaking valleys are added. The Swiss therefore do not form a nation in the sense of a common ethnic or linguistic identity. The strong sense of belonging to the country is founded on the common historical background, shared values, such as federalism and direct democracy. The 26 cantons of Switzerland are the member states of the federal state of Switzerland. Each canton has its own constitution, legislature, government and courts. Most of the cantons' legislatures are unicameral parliaments. The Swiss Federal Constitution declares the cantons to be sovereign to the extent their sovereignty is not limited by federal law. The cantons also retain all power and competencies not explicitly delegated to the Confederation by the Constitution. Most significantly, the 1
Corresponding Author: Christian Lovis, MD, PhD, Professor; University Hospitals of Geneva, 24, rue Micheli-du-Crest, CH-1211 Geneva 14; Email: [email protected]; Phone: +41 22 372-6180
58
C. Lovis et al. / eHealth in Switzerland – Building Consensus, Awareness and Architecture
cantons are responsible for healthcare, welfare, law enforcement and public education; they also retain the power of taxation. Healthcare Governance in Switzerland The healthcare in Switzerland is regulated by several legal frameworks. On one side, there is the Federal Health Insurance Act of 1994. The federal level does only set the legal framework for health insurance, access, equity and reimbursement. Health insurance is compulsory for all persons residing in Switzerland. The Swiss healthcare system is a combination of public, subsidized private and totally private systems. The insured person has full freedom of choice among the recognized healthcare providers competent to treat their condition (within his canton) on the understanding that the costs are covered by the insurance up to the level of the official tariff. There is freedom of choice when selecting an insurance company officially registered to which one pays a premium, usually on a monthly basis. On the other side, there are 26 ministries of health and legal frameworks that set the health laws, for each canton. For example, while a doctor requires a Federal diploma of medicine, one has actually to receive the formal authorization of each canton to have the authorization to practice. Thus, the competence for organizing the healthcare system is at the cantonal level. There is a Federal Office of Public Health (FOPH) that is in charge of health promotion, disease prevention and health protection campaigns mainly. There is also a coordination group of representative of all 26 cantons, the “Swiss Conference of cantonal directors of health” (GDK) which elaborates recommendations in order to improve collaboration between cantons, and between cantons and the federal government. Finally, there are the same professional groups as in most European countries, which play very important roles, such as the Swiss Federation of physicians, of nurses, of pharmacists, etc. as well as the academy of medical sciences, the university and high schools, amongst others. 1. Building eHealth – Building a New Governance Building an eHealth framework for Switzerland is confronted to many challenges that are similar to building a eHealth framework for Europe: many languages, many legal frameworks and political organizations, different cultures and understandings. Thus, one of the first important actions has been to define a Swiss coordination organism, in charge of organizing the process, provide sustainability and coherence. In order to succeed, the eHealth strategy must be nationally planned and coordinated while respecting the needs, requirements and autonomy of each canton. The Confederation and the cantons have therefore concluded a framework agreement and created a coordination body. The financing is provided by the federal state and the GDK. The coordination organ represents political governance in the steering committee and has representative of numerous stakeholders, including patients, in the advisory board. It has a project management team which coordinates the work of four working groups devoted to a) standards and architecture; b) pilots and implementation; c) the Swiss patient portal and d) education. In addition to this coordination organ, Switzerland is actually working at proposing a comprehensive and federal legal framework for eHealth.
C. Lovis et al. / eHealth in Switzerland – Building Consensus, Awareness and Architecture
59
Figure 1. Governance of the Swiss eHealth project
2. Moving from Vision to Action As illustrated in the Fig1, the project is made of four working groups addressing the standards and architecture (S&A WG), the pilots (P WG), the patient’s portal (PP WG) and education (E WG) respectively, while the overall legal federal framework is addressed by another team. 2.1. Standards and Architecture In order to respect the political organization of Switzerland and the legal framework that addresses data processing, personal data and privacy, a set of major principles have driven the work of the S&A WG work. The following points summarize the major principles and guidelines: - No central patient registry - No central document registry - Patients data resides where it created - Patients self-determination at which information is available - Consent management - Use of standards wherever they exist - Use of European standards wherever they exists This architecture is presented in Figure 2. While all elements are distributed, the standards, definitions, including for example roles definition, are shared. Adopting a concept that is similar to the National Contacts Points (NCP) defined in the epSOS European project, the S&A WG defined Switzerland as being a group of interconnected “communities”. These communities can have any size and cover heterogeneous groups of care providers, such as groups of physicians, hospitals, or
60
C. Lovis et al. / eHealth in Switzerland – Building Consensus, Awareness and Architecture
even states. They have their own internal organization, but communicate with other communities by the mean of “gateways”. The S&A WG has defined and published the specifications of the gateways, following IHE cross-community exchanges profiles. In addition, the S&A WG defined a role-based access management system as well as sets of global rules to manage new roles, new care providers, etc. There are no recommendations on the internal organization, regulation, technical, etc. of a community, which thus allow supporting cantonal differences, especially at the legal level. The global processes will follow the IHE principles, with profiles clarifying the use of standards for specific use cases and IHE Connectathon™" and will mostly, in a first step, be based on document exchanges. This has been found sound to allow a smooth transition from unstructured information to structured one. After consultation, HL7-CDA-CH has been selected. The CH allows having some specific adaptations, which might hinder international interoperability, but ease adoption in Switzerland. Furthermore there is a recommendation for a nationwide set of metadata, in order to describe patient documents with variables, contents and nomenclatures. There is ongoing work for semantic standards, for drugs, lab, clinical information, etc.
Figure 2. Overall architecture
2.2. Pilots One of the big challenges is to have running pilots, that are getting concrete with large scale implementation, professional groups, care providers, patients, getting involved in the strategy and using the tools in their daily practice. Another important challenge is
C. Lovis et al. / eHealth in Switzerland – Building Consensus, Awareness and Architecture
61
to find the funding for these pilots, taken into account that there is no legal basis for federal financing and that canton’s authorities have usually little means for such projects. Thus, a private-public partnership is heavily promoted. Several pilot projects are ongoing currently in various cantons in Switzerland. The most advanced project is the e-toile project in Geneva. Geneva canton is the first canton in Switzerland that achieved to have a legal framework adopted for the eHealth and a large and successful public-private partnership for the implementation. The pilot shall start in late 2010. 2.3. Evaluation Evaluation is an important aspect of the project. It is the major way of getting experience and learning from the first phases in order to improve as fast as possible the global approach. Therefore, an evaluation framework has been developed from the early beginning of the project. However, the success of eHealth projects depend on many factors besides purely technical aspects, such as legal framework, culture, societal aspects, political will, etc. Evaluating these non-technical aspects can have very negative effects on adoption by leveraging fear in some circles. Thus, the evaluation framework has been divided in two very clear pillars. The first pillar addresses the global context, but not the technical implementation. This pillar is under the idea of “readiness”. The second pillar evaluates the technical implementation, it is very similar to an "IHE Connectathon™" and it addresses conformance and certification issues. Some of the elements are summarized in the following table: Table 1. Evaluation framework
eHealth Swiss Readiness Domain
Conformance
Societal, political, legal, organisational - Use of eHealth is recognized
Technical and content - Implementation is conform to the S&A WG
- Political will is improved Goals
- Swiss connectathon passed - Professional organisation support - Process is transparent and can be - Legal framework is achieved
certified
- Dissemination: all actors receive targeted information - Political: ehealth is included in the strategic planning Dimension
- Legal: there is a legal framework
s
- Cooperation: all stakeholders are included
-
Conformance: How are the political, legal, etc requirements implemented.
-
Certification: How are the standards and architecture defined by the S&A
- Organisation: enabling rules and context - Education: programs for pre-postgraduation and continuous education - Semantic: Consensus is reached
WG implemented, including semantics.
62
C. Lovis et al. / eHealth in Switzerland – Building Consensus, Awareness and Architecture
3. Conclusion eHealth addresses numerous challenges, societal, cultural, educational, financial legal and technical to quote only some of them. Building a global consensus amongst all stakeholders, creating awareness and sufficient educational level to avoid rising fear and privacy concerns are major factor of successes. Technical solutions and commercial pressure must not impede this process. The transition and readiness can seem slow to obtain, but this is the real price of success. Interoperability has become one of the major targets in this field. However, one must not expect technical solutions to solve misunderstanding and lack of collaboration between people. Thus, addressing this aspect is a major prerequisite in order to have eHealth strategies succeed. Facilitators and motivators are important, this include availability of technical specification, semantic content, but also clear and understandable factsheets demonstrating the benefits, promoting education and incentives for industries and users.
References [1]
[2]
The website of the coordination organ, where all documents are available: www.e-health-suisse.ch Important papers, available in French and German: Strategie eHealth Schweiz, The global eHealth strategy Empfehlungen I "Standards und Architektur", The target architecture version 1 The legal framework discussed, http://www.e-health-suisse.ch/umsetzung/00146/00157/index.html?lang=de
Patient Empowerment by Electronic Health Records: First Results of a Systematic Review on the Benefit of Patient Portals Elske AMMENWERTHa,1, Petra SCHNELL-INDERSTb and Alexander HOERBSTa a Institute for Health Information Systems, UMIT – University for Health Sciences, Medical Informatics and Technology, Hall in Tyrol, Austria b Institute for Public Health, Medical Decision Making and HTA, UMIT – University for Health Sciences, Medical Informatics and Technology, Hall in Tyrol, Austria
Abstract. Patient portals provide patients with access to a provide-managed electronic health record (EHR). They may provide an interesting approach to increase patient empowerment. The objective of this paper is to provide a first overview of the state-of-the-art and the impact of patient portals. Based on a systematic literature search, we identified five evaluation studies on patient portals. These studies demonstrate only little effect of patient portals on patient empowerment. Keywords. Patient portal, electronic health record, review, impact evaluation
Introduction
The emergence of the Internet and of the Electronic Health Record (EHR) has also brought new opportunities for a new and more active role of the patient [1. 2] in his/her care. This is often denoted as patient empowerment, describing a situation where the patients’ role is changing from a patronized patient to an informed patient and further to a responsible, autonomous and competent partner in his or her own care [3]. One important approach here is the concept of “patient portals”. Patient portals can be defined as provider-tethered applications that allow patients to access health information that is documented and managed by a health care institution [4]. As part of a patient portal, institutions may allow patients a typically web-based access to selected clinical data which is governed by the respective institutions as part of a person’s EHR. The patients can then access clinical data, read and print it or integrate them into a PHR or any other (electronic or paper-based) type of patient-owned record. Besides providing sole access to EHR data, patient portals may also offer additional services such as medication refills, appointment scheduling, access to general 1
Corresponding Author: Elske Ammenwerth, PhD, Professor; Institute for Health Information Systems, Eduard Wallnöfer Zentrum 1, 6060 Hall in Tyrol, Austria; E-Mail: [email protected]
64
E. Ammenwerth et al. / Patient Empowerment by Electronic Health Records
medical information such as guidelines, or secure messaging between a patient and an institution [4]. At the moment, there seem to be little evidence whether patient portals can really increase patient empowerment. Objectives
The objective of this paper is to review the impact of electronic patient portals on patient empowerment.
1. Methods We systematically searched for evaluation studies on patient portals in scientific databases and journals such as PubMed, Cochrane Library, CINAHL, EMBASE,
ACM Digital Library and the Evaluation Database (http://evaldb.umit.at). We included all kinds of study designs (RCT, non-RCT). We focused on studies that measured the impact of a patient portal on the outcome criteria such as patient satisfaction with the provided care, patient empowerment, costs and resource consumption, mortality or other relevant clinical parameters. We limited the search to papers after 1990. We basically used Mesh terms to retrieve papers, using a combination of the terms "Medical Records Systems, Computerized", "Health Records, Personal", “Access to Information" or "Patient participation". All papers that seemed eligible were read in full text by two researchers. Each study was systematically described addressing clinical setting, type of intervention, type of study as well as outcome. 2. Results
We identified 603 papers, 13 of them comprised an experimental or quasiexperimental study design. Of those 13 papers, 5 studies [5-9] were finally eligible and then analyzed in detail (see Table 1). Table 1. Details of retrieved studies on patient portals. Author/Year Tuil, 2007 [5]
Users Patients undergoing IVF treatment
Selected functionalities of the patient portal Personal health record, offering access to own medical record with all available information concerning the patient’s IVF or ICSI treatment, and tailored, context-sensitive clarification of clinical information.
Zhou, 2007 [6]
Patients that used KP Health Connect Online for longer than 13 months Patients with
KP HealthConnect, offering access of parts of their individual health record; health summary with problem list, medications, allergies; and health record with immunizations.
Grant, 2008
“Diabetes-Mellitus-specific PHR” offering medication module to
E. Ammenwerth et al. / Patient Empowerment by Electronic Health Records
65
Table 1. Continued. [7]
Diabetes Mellitus Type 2
review medications; view most recent results and current treatments; generate a DiabetesCarePlan based on patients’ responses to the questions, to be used at the next clinical visit.
Earnest, 2004 [8]
Patients with congestive heart failure Patients with congestive heart failure
SPPARO (“System Providing Patients Access to Records Online”) offering online access to clinical notes, lab reports, test results SPPARO (“System Providing Patients Access to Records Online”) offering online access to clinical notes, lab reports, test results
Ross, 2004 [9]
Four of the found studies were randomized controlled trials (RCT), the remaining study combined a matched-control and a cohort study. The number of participants in each study ranged from 81 to 6.402 patients. The found studies evaluated the impact on a variety of outcome criteria. One study [7] focused on changes in clinical outcome parameters, including HbA1c, blood pressure, LDL, and medication adjustments. One study [6] focused on changes of resource consumption, including office visit rates and telephone contacts. Two studies [5, 8] focused on changes of more subjective parameters such as patient satisfaction, patient knowledge, and patient anxiety; these were measured by validated questionnaires. The fifth study [9] combined several criteria and included changes of mortality, of treatment adherence, of resource consumption (message number) and of subjective parameters (subjective health status, patient empowerment, medication adherence). Significant changes in the patient portal group, compared to a control group, could only be observed for the following parameters: decrease in office visit rates and telephone contacts [6]; increase in number of messages sent [9]; changes of the medication regimen [6]; and better adherence to treatment [9]. For the other parameters, studies did not find significant changes between intervention and control group. 3. Discussion 3.1. Answers to Study Questions We systematically searched the literature and found five controlled studies focusing on the impact of patient portals. The studies were quite heterogeneous with regard to clinical setting, type of intervention and measured outcome. Basically, most of the measured parameters did not show a significant difference between intervention and control group. In particular, no significant changes could be observed for parameters related to patient empowerment. Patient portals basically present clinical information to the patients. Can we expect that giving patient access to clinical information can have an impact? Ross et al. [10] reviewed the outcome of 29 descriptive or controlled studies on adult’s patient access to (paper-based) medical records, published between 1970 and 2002. The review found that several studies showed an improvement of doctor-patient communication by patient-accessible medical records. There were, however, only conflicting findings on improvements in adherence (such as adherence to treatment, smoking behavior), patient education (such as understanding or recall of medical information; feeling of being well informed), and patient empowerment (such as sense of autonomy); in these cases, some controlled studies showed an improvement, while others did not. No changes could be
66
E. Ammenwerth et al. / Patient Empowerment by Electronic Health Records
observed with regard to patient anxiety, patient confusion, patient depression, or patient satisfaction with care. Ross et al. [10] summarized that studies suggest potential for modest benefits for example in enhancing doctor-patient communication, but that studies were of limited quality, and that more research is necessary. Compared to paper-based access to records, online patient portals allow a patient independently and repeatedly accessing the information; the information is better legible; and the user can link the information to further sources of medical information available on the Internet [10]. Also, patient portals can be adapted to the patient’ wishes and knowledge level [5]. They can also be completed by secure communication links with healthcare providers or other functions. Overall, we could expect a higher impact of online portals compare to paper-based access. However, as our results show, only little impact could be found. 3.2. Study limitation We conducted a systematic literature search, and carefully chose the search terms; however, due to the variety of terms that may be used for patient portals, we cannot be sure to have retrieved all studies. We did not search for grey literature. The review was conducted by two researchers; any differences in judgment were solved by discussion. 4. Conclusion Portals basically provide better information from the medical record to patients. However, better-informed patients are not necessarily healthier patients. Descriptive evidence from a large number of studies suggests that patients interested in access to their patient records, and that they find it helpful and useful. These findings, however, do not guarantee that there is in fact a measurable impact.
References [1] [2] [3] [4]
[5]
[6]
[7]
S. Munir, R. Boaden, Patient empowerment and the electronic health record, Stud Health Technol Inform., 84 (Pt 1) (2001), 663-665. M. Joubert, A. Gaudinat, C. Boyer, A. Geissbuhler,M. Fieschi, Hon Foundation Council M. WRAPIN: a tool for patient empowerment within EHR,Stud Health Technol Inform.,129(Pt 1)(2007), 147-151. G. Nagel, Patientenkompetenz. Krankenhauspharmazie, 26 (2005), 128-133. F.C. Bourgeois,K.D. Mandl,D. Shaw, D. Flemming, D.J. Nigrin, Mychildren's: integration of a personally controlled health record with a tethered patient portal for a pediatric and adolescent population, AMIA Annu Symp Proc. (2009), 65-69. W.S. Tuil, C.M. Verhaak, D.D. Braat, P.F. de Vries Robbe, J.A. Kremer, Empowering patients undergoing in vitro fertilization by providing Internet access to medical data, Fertil Steril. 88(2) (2007), 361-368. Y.Y. Zhou, T. Garrido, H.L. Chin, A.M. Wiesenthal, L.L. Liang, Patient access to an electronic health record with secure messaging: impact on primary care utilization., Am J Manag Care. 13 (7) (2007), 418-424. R.W. Grant, J.S. Wald, J.L. Schnipper, T.K. Gandhi, E.G. Poon, E.J. Orav, et al., Practice-linked online personal health records for type 2 diabetes mellitus: a randomized controlled trial, Arch Intern Med. 168 (16) (2008), 1776-1782.
E. Ammenwerth et al. / Patient Empowerment by Electronic Health Records
[8]
67
M.A. Earnest, S.E. Ross, L. Wittevrongel, L.A. Moore, C.T. Lin, Use of a patient-accessible electronic medical record in a practice for congestive heart failure: patient and physician experiences, J Am Med Inform Assoc. 11 (5) (2004), 410-417. [9] S.E. Ross, L.A. Moore, M.A. Earnest, L. Wittevrongel, C.T. Li, Providing a web-based online medical record with electronic communication capabilities to patients with congestive heart failure: randomized trial, J Med Internet Res. 6 (2) (2004), e12. [10] S.E. Ross, C.T. Lin, The effects of promoting patient access to medical records: a review, J Am Med Inform Assoc. 10 (2) (2003), 129-138.
Empowerment of Patients over their Personal Health Record Implies Sharing Responsibility with the Physician Catherine QUANTINab 1, Eric BENZENINEb, Bertrand AUVERLOTb, David-Olivier JAQUET-CHIFFELLEc, Gouenou COATRIEUXd and François-André ALLAERTbe a Inserm, U866, Univ de Bourgogne, Dijon, France b CHRU, Service de Biostatistique et d’Informatique Médicale, Dijon, France c Bern University of Applied Sciences et Université de Lausanne, Suisse d Institut TELECOM, TELECOM Bretagne, Unité INSERM 650 LaTIM e Ceren Esc Dijon & Dpt biostat Ecole de Santé Publique, Liège, Belgique
Abstract: Through this article, we point out the unavoidable empowerment of patients with regard to their personal health record and propose the mixed management of patients’ medical records. This mixed management implies sharing responsibilities between the patient and the Medical Practitioner (MP) by making patients responsible for the validation of their administrative information, and MPs responsible for the validation of their patients’ medical information. We propose a solution to gather and update patients’ administrative and medical data in order to reconstitute patients’ medical histories accurately. This method is based on two processes. The aim of the first process is to provide patients administrative data, in order to know where and when they received care (name of the health structure or health practitioner, type of care: outpatient or inpatient). The aim of the second process is to provide patients’ medical information and to validate it under the responsibility of the MP with the help of patients if needed. During these two processes, the patients’ privacy will be ensured through cryptographic hash functions like the Secure Hash Algorithm, which allows the pseudonymization of patients’ identities. The Medical Record Search Engine we propose will be able to retrieve and to provide upon a request formulated by the MP all the available information concerning a patient who has received care in different health structures without divulging the patient’s true identity. Associated with strong traceability of all access, modifications or deletions, our method can lead to improved efficiency of personal medical record management while reinforcing the empowerment of patients over their medical records. Keywords: medical record, patient identifier, direct access, data security, privacy, E-health
Introduction The concept of empowerment can be defined as a “social process of recognizing, promoting, and enhancing people’s abilities to meet their own needs, solve their own 1
Corresponding Author: Catherine Quantin, MD, PhD; Service de Biostatistique et Informatique Médicale, CHU Dijon, Postal address BP 77908, 21079 Dijon Cedex; Email : [email protected];
C. Quantin et al. / Empowerment of Patients over Their Personal Health Record
69
problems, and mobilize necessary resources to take control of their own lives.” [1]. In the health care context, patient empowerment means promoting autonomous selfregulation so that the individual’s potential for health and wellness is maximized. Patient empowerment begins with information and education and includes seeking out information about one’s own illness or condition, and actively participating in treatment decisions [2]. It points out the passage from an old model of care based on patients’ “compliance” with a health care professional’s “directives” to a new paradigm based on patients’ “adherence” to health care professional’s “recommendations”, through, among other things, active participation of the patient in the management of his personal medical records. That means also to move from the traditional patient’s medical record managed by health professionals and used under their supervision and authority to a Patient Controlled Health Record (PCHR) [3]. As described by L. Rostad [4] a PCHR contains data from multiple care sites, and the patient is in complete control of the information. According to a previous study conducted by SE Ross [5] and to the results of clinical trials, the main benefit of the giving patients direct access to their medical records is the improved communication between the doctor and the patient. Other benefits concerning only modest improvements in adherence, patient education, and patient empowerment were found in certain randomized controlled clinical trials, but not in all. However this lack of efficacy could result from the fact that patients had only access to their medical records, which maintains them in a passive situation; access alone is insufficient as a real active patient control over their personal health records is needed. By empowering him of his personal health information, we may expect that patient will become a real key manager of his own health, beside the Medical Practitioner (MP). But at the same time, responsibility must be shared according to the knowledge of each actor. In this case, management concerns mostly relate to validation (or cancellation) of the information, and any information must be assessed before it can be deemed true or false. Thus, the role of patients is to verify and/or modify their administrative data, while the MP will verify and/or modify purely medical information. However, we aim to underline the strong link between these two kinds of data and demonstrate their interdependence in terms of quality. The main objective of this paper is to propose a new method to reconstitute, update and provide the correct administrative and medical data for patients through the mixed management of their medical records. This will empower patients with regard to the management of their health data by increasing their responsibility [1, 2], and by increasing the accountability of MPs concerning the medical data. We propose to correct and update patients’ health information or data, with their help, so as to reconstitute their medical histories efficiently with no or only a slight increase in MPs’ workload and with the highest level of accuracy.
1. Methods Patients are the best able to validate not only their first and second names as well as their date of birth, but also the health structures they visited and the dates of their visits. MPs are best able to validate medical data as defined below. This means that patients and MPs together should be responsible for the reconstitution and the update of patients’ medical histories (administrative and medical information) of the patient. The division of responsibilities between doctors and patients is very time consuming for decision making [6-12], but has not been extensively investigated regarding the
70
C. Quantin et al. / Empowerment of Patients over Their Personal Health Record
management of patients’ medical records. The methodology we propose for responsibility sharing between MP and patients relies on two processes: The first process aims to provide patients’ administrative data to know where and when patients received care (name of the health structure or health practitioner, type of care: outpatient or inpatient). Patients will be responsible for the validation of this information [3, 4]. The first objective of this validation is to obtain the right information for the right patient at anytime and anywhere and to be sure that the information concerns only the patient, and to avoid providing or considering information from another patient. Secondly, the validation of the information retrieved will constitute the patient’s medical history. During this validation process, the MP may confer with the patient. The validation of this administrative information will be under the exclusive responsibility of the patient, with the help of the MP if needed. The main objective of the second process is to provide patients’ medical information. The MP will be accountable for the validation of these medical data, with the help of the patient, if needed. For this validation process to be successful and efficient, it will be useful to take into account the help patients are able to provide given their thorough knowledge of their administrative and medical information. In fact, patients who have all their mental faculties know the history of the care they have received better than anybody else. Patients could regularly update administrative and medical data, for example, once or twice a year depending on the national public health strategies with regard to periodic health check-ups. Such checkups may be initiated by national health authorities or employers, like, for example, in the offer of a minimum package of services for employees for the prevention and follow up of transmissible and non transmissible diseases. In both processes these, patients’ privacy is a key issue and has to be ensured. Patients’ privacy is one of the main concerns in the storing, sharing or transmission of personal medical data and it is comprehensively covered by legal acts like the HIPAA in the United States or Directive 95/46/EC in Europe. Many companies including employers, insurers or banks are very interested in gaining access to such data. Recently, attempts to gain illegal access to such data have been reported in the network. Thus, systems to share medical data through open networks like the Internet need to prevent access to the identity of the patient. Pseudonymization is one of the solutions that have been suggested for this purpose. It provides a trade-off between patients’ privacy requirements and society’s needs in order to improve health care systems by linking the patient’s identity to a pseudonym from which it is not possible to get back to the patient’s identity [13]. Cryptographic hash functions like the Secure Hash Algorithm can be used to reach this goal of pseudonymization [14]. However, sharing this pseudonym or code without introducing secrecy in its calculation may lead to specific attacks, in particular when the total number of possible messages that could have been hashed (pre-image) is too small. In this case, the authorized receiver as well as a pirate eavesdropping on the communication could retrieve private information. A pirate, who obtains lists of patients’ pseudonyms, by illegal access or simply by listening to the Internet network, may be able to retrieve the real identity of the patients. The solution we have proposed [15] within the framework of a medical record search engine procedure would counter such risks. It relies on two entities: i) the MRSE (Medical Record Search Engine) which sends out one request issued by one MP, and; ii) the aggregator which gathers responses from the health structures that receive requests for information. In the first step of this procedure, our solution encrypts the patient’s pseudonym and the secret key to be used for its decryption before sending them through two different channels. If gaining access to private information
C. Quantin et al. / Empowerment of Patients over Their Personal Health Record
71
becomes more difficult for someone who is spying on the network, such as MRSE intruders, it will be almost as difficult for the MRSE and Aggregator entities involved in information gathering, and which may be tempted to obtain information illegally. Any attackers will face a set of patient pseudonyms without being able to identify those associated with the same patient. In fact, a one-time pseudonym will correspond to each request. This is the main interest of using pseudonyms in our proposal. 1.1 FIRST PROCESS: Reconstitution of a Patient’s History The first step of this process is the pseudonymization of the patient’s identity and the generation of several pseudonymous codes. During a consultation between an MP and a patient, the MP enters all the components of that patient's identification such as his first and last names and his date of birth. Of course, the choice of these components will depend on their availability (exhaustiveness) and their quality. To optimize the request concerning the patient’s data, several identifiers can be generated, based on the different combinations of first names, last names and date of birth of a patient. This information related to the patient’s identity will be rendered anonymous using a robust cryptographic hash function. To improve patient’s privacy, the pad used for hash coding is not the same for each request. A specific pad is generated by the first Medical Record Search Engine (MRSE 1) and sent to the MP, after being encrypted with the MP public key. The MP can then hash the patient’s identifiers with this pad. As a consequence the MP sends not only one pseudonymous identifier per patient but a list of pseudonymous partial identifiers for each patient. The aim of this first step is then to obtain several pseudonymous codes, but, hopefully, always the same ones for a given individual in order to link all of the information concerning any given patient. At a second stage, occurs the search for patient’s administrative data. The Medical Record Search Engine Procedure [15] is used to provide a patient’s administrative data. For one health structure, MRSE 1 encrypts the pad with the public key of this health structure and sends it. Each health structure can then search for administrative information corresponding to these pseudonymous codes (by comparing them with hashed identities of the patients cared for in the structure). All the pseudonymous codes created from the different possible combinations are considered. This administrative information will be sent back to the MP, through the aggregator. The third step is the validation of a patient’s administrative data by the patient. Any patient of sound mind is the only person who can confirm, where, when and for what reasons she/he consulted for health problems. In patients with a mental handicap, MPs will help patients to confirm or not the doubtful administrative information, thanks to their experience with these patients and their diseases. The doubtful administrative information may relate to a long distance between the patient’s residence and the health structure supposedly visited, or to the lack of coherence between the type of health care provided by one hospital or clinic and the patient’s disease. An automatic check for this coherence can be implemented, not to automatically exclude any information, but to help the MP and the patient to highlight potential errors and correct them if necessary. At the end of this validation procedure, the list of administrative information (pseudonymous codes, dates and health structures) to be conserved is transmitted to MRSE 2.
72
C. Quantin et al. / Empowerment of Patients over Their Personal Health Record
1.2 SECOND PROCESS: Access to Relevant Medical Data for Patient Care. Firstly, patient’s medical data are retrieved. The MRSE procedure is used to make available a patient’s medical data. Each health structure that has been contacted or selected during the first process can then search for medical information corresponding to the pseudonymous codes and dates it receives. Medical information retrieved in this search will be sent to the MP, through the aggregator. Secondly, this medical information is validated by the MP, with the close collaboration of the patient, is accountable for the validation of the medical information.
2. Discussion Some MPs may object to our procedure because they believe it will increase their workload as they will have to read all of the information in order to detect possibly false or missing information. However, the burden is not as heavy as they may think for many reasons. First, this enquiry on the patient’s past history will not be necessary for all patients. Secondly, the number of different medical records that MPs will have to summarize for a particular patient will more frequently be one or two (or eventually three) rather than ten or twenty; but that should be verified. Finally, this work has to be done in with the patient collaboration which can speed up this task. An improvement in time efficiency could also be the reduction of gathered data, via a selection tool as implemented in any kind of search engine and based on several criteria such as a time period, a list of hospitals, a kind of ward or pathology. One could also figure that the first process could be done by the patient alone before meeting the MP, subject to his computer skills, but this point has to be assessed, as it regards data security. However, impact of patient behavior on the quality of his health information may happen independently of the management of the medical record, in the daily practice, and the relationship between the patient and the MP is unavoidable based on reciprocal trust. Moreover, more and more health professionals involve patients in discussions and decision-making concerning their health status and future. Considering that this patient empowerment is bound to happen, the question is how we can contribute to this evolution. However, the impact of the collaboration between the patient and the MP to identify potential errors should be evaluated. For some diseases, the understanding of the natural history is fundamental [19]. Obviously, the history taking [20] requires the patient’s help. The study about patients’ access [5] reported that a substantial proportion of patients can identify factual errors in their records, but concluded that it was difficult to assess the rate of finding clinically important errors. In this procedure, we have to find a balance between the potential improvement in the overall quality of the information resulting from patients’ management of data collection from all sources and the risk of altering the accuracy of the data especially since patients will be able to delete information. Naturally, appropriate modifications to our methodology have to be proposed in the case of mental handicap (by calling on trusted third party: next of kin, proxy or legal representative), in case of emergency care (by exceptionally authorization granting even without validation) or even in the case of a patient meeting a new MP (communication between MPs based on the patient’s willingness). One other aspect that our paper is not presenting is related to the safety of EPHR. This aspect has been studied in a previous paper [18] where it was demonstrated that strong traceability is required to build trust in EPHR.
C. Quantin et al. / Empowerment of Patients over Their Personal Health Record
73
3. Conclusion With our proposal for the administration of health records, for the first time, patients will be involved in the management of their medical records, which gives them more responsibility for their health care. Furthermore, the MP would obtain more reliable information, which would provide a better basis for decision making and lead to more appropriate treatment. Moreover, the sharing of these responsibilities would enhance the doctor/patient relationship and hopefully mutual trust. Finally, it appears that sharing responsibility for the management of medical records between the patient and the MP can help to provide better health care by increasing the efficiency of the management process. However, even if have confidence in MPs and patients, strong traceability would be required to complete this organizational scheme.
References [1] [2] [3] [4] [5] [6]
[7] [8] [9] [10]
[11]
[12] [13] [14] [15]
[18]
[19] [20]
P.S. Jones, A.I. Meleis, Health is empowerment, ANS Adv Nurs Sci. 15(3) (1993), 1-14. D.H. Lau, Patient empowerment--a patient-centred approach to improve care, Hong Kong Med J. 8(5) (2002), 372-4. K.D. Mandl, P. Szolovits, I.S. Kohane, Public standards and patients' control: how to keep electronic medical records accessible but private, BMJ 3, 322(7281) (2001), 283-7. L. Rostad (Edr.) An initial model and a discussion of access control in patient controlled health records. The third international conference on availability, reliability and security, IEEE; 2008. S.E. Ross, C.T. Lin, The effects of promoting patient access to medical records: a review, J Am Med Inform Assoc 10(2) (2003), 129-38. J.E. Baars, T. Markus, E.J. Kuipers, C.J. van der Woude, Patients' preferences regarding shared decision-making in the treatment of inflammatory bowel disease: results from a patient-empowerment study. Digestion 81(2) (2010), 113-9. D. Burton, N. Blundell, M. Jones, A. Fraser, G. Elwyn, Shared decision-making in cardiology: do patients want it and do doctors provide it? Patient Educ Couns 80(2) (2010), 173-9. J. Gerris, P. De Sutter, Self-operated endovaginal telemonitoring (SOET): a step towards more patientcentred ART? Hum Reprod 25(3) (2010), 562-8. G.L. Larkin, A.L. Beautrais, A. Spirito, B.M. Kirrane, M.J. Lippmann, D.P. Milzman, Mental health and emergency medicine: a research agenda, Acad Emerg Med 16(11) (2009), 1110-9. J.E. Lopez, M. Orrell, L. Morgan, J. Warner, Empowerment in older psychiatric inpatients: development of the empowerment questionnaire for inpatients (EQuIP), Am J Geriatr Psychiatry 18(1) (2010), 21-32. J.D. Tariman, D.L. Berry, B. Cochrane, A. Doorenbos, K. Schepp, Preferred and actual participation roles during health care decision making in persons with cancer: a systematic review, Ann Oncol 21(6) (2009), 1145-51. M.A. Visse, T. Teunissen, A. Peters, G.A. Widdershoven, T.A. Abma, Dialogue for air, air for dialogue: towards shared responsabilities in COPD practice, Health care anal. 18(4) (2010), 358-73. T. Neubauer, B. Riedl, Improving patients' privacy with pseudonymization, Stud Health Technol Inform, Vol. 136, pp 691-6, IOS Press, Amsterdam 2008. B. Schneier, Applied cryptography, Fourth Edition, 2006. C. Quantin, D.O. Jaquet-Chiffelle, G. Coatrieux, M. Fassa, F.A. Allaert, Medical record search engines, using pseudonymised patient identity: an alternative to centralised medical records. Collaborative meetings on health informatics (CoMHI) 2009, IMIA WG4 (SiHIS); 2009 21-25 November 2009; Hiroshima. F.A. Allaert, C. Quantin, Patients' empowerment of their personal health record requires strong traceability to guarantee patients health care security, Stud Health Technol Inform Vol. 155, pp 43-7, IOS Press, Amsterdam, 2010. A.B. Mehta, Anderson-Fabry disease: developments in diagnosis and treatment, Int J Clin Pharmacol Ther 47 Suppl 1 (2009), S66-74. B.E. Clauser, K. Balog, P. Harik, J. Mee, N. Kahraman, A multivariate generalizability analysis of history-taking and physical examination scores from the USMLE step 2 clinical skills examination, Acad Med 84 (10 Suppl) (2009), S86-9.
Personal Information Protection – Exceptional Challenges of Integrated Systems of eHealth Anka BOLKAa1, Blaž ZADELa and Martina ZORKOa a Health Insurance Institute of Slovenia
Abstract: Informatization has been bringing important, quick and extensive changes into the healthcare environment for years. Individual systems still represent isolated information islands; however, the need for interconnectivity and mutual accessibility has become more pronounced. On the one hand, integration of systems brings numerous, financially measurable advantages, and on the other hand, personal information in such systems becomes more vulnerable. Providing personal information protection is therefore the permanent task of informatisation and, with elimination of national borders and integration of national systems, it is becoming a challenge from the legal, organisational, technical and financial standpoints. Key words: information security, personal information protection, smart cards, on-line system
1. Information support in the healthcare environment in Slovenia The healthcare system in Slovenia is already effectively supported with information solutions. The administrative work of all providers of healthcare services (hereinafter referred to as providers), patient records and procedures for collection and exchange of financial and statistical information, is fully supported with information technology. In addition, numerous providers have already developed individual information solutions for electronic management of medical records and support for medical procedures. However, some providers have remained isolated information islands because the exchange of electronic healthcare information among providers is scarce; only individual and specific solutions are developed; for example, exchange of laboratory test results and radiological scans. Health Insurance Institute of Slovenia (hereinafter referred to as the Institute) is the only holder of compulsory health insurance in the country and has been an important operator of healthcare informatics for more than two decades. The Institute made a particularly important developmental step in 2000 by introducing the national health insurance card system. This system established a system of electronic transfer of information about insurance holders and their health insurance from insurance companies to providers. All providers and all insurance companies - both the Institute
1 Corresponding Author: Anka Bolka, Head of the Health Insurance Card System Sector, Health Insurance Institute of Slovenia, Miklošičeva 24, 1507 Ljubljana, Slovenia; E-mail: [email protected]
A. Bolka et al. / Personal Information Protection
75
and all commercial insurers providing additional health insurance - are incorporated in the system. The system was upgraded in 2010 with the introduction of a card based on public key infrastructure, and with the establishment of an on-line system in which providers can get all necessary information directly from databases instead of from the cards. The introduction of the on-line system brought additional advantages to providers, with the most important advantage being safe access to the Internet from all sites. The infrastructure established in this way represents the information and security foundations on which further safe solutions can be built.
2. Legislative Frameworks of Information Protection Protection of personal information and electronic business in Slovenia is stipulated by the Personal Information Protection Act, the Electronic Commerce and Electronic Signature Act, the accompanying decree and other relevant regulations. At the EU level there are also numerous standards, directives and recommendations on information security and in particular on the protection of information in healthcare, because information on the health condition of individuals is especially sensitive. From this aspect, exchange, storage and use of such information has to be supported with information technology with the highest possible level of security [1]. In establishing the card system and developing the on-line system, the Institute followed the legislative provisions on technological and security requirements applied in Slovenia, recommendations and advice from Slovenian and international experts and the EU legislation, directives and recommendations related to security of healthcare systems. The solutions that have already been established meet the requirements and provisions at the EU level and are an adequate foundation for safe cross-border information integration.
3. Protection of Information from the Aspect of Direct Users The informatised healthcare system is used by patients/citizens and healthcare workers: doctors, nurses, pharmacists, administrative staff in healthcare and others. The expectations of patients directly depend on the level of their education in information technology. They also depend on their experience with the safety and reliability of information solutions in other areas, primarily with electronic commerce and banking. Something common among patients is their expectation that informatisation will simplify procedures while not at all increasing the risk of unauthorised access to sensitive personal information. Every individual should also be, in line with the regulations, aware of information about them entered into databases of individual providers, and to decide whether they will prohibit access to individual sets of information. The health insurance card is used for identification of patients in the healthcare system in Slovenia. The use of the card for this purpose is not password-protected, which enables simple handling with the card in routine procedures and a sufficient level of security for this purpose.
76
A. Bolka et al. / Personal Information Protection
In developing information solutions, it is also very important to take into account the needs and demands of healthcare workers. All solutions have to provide a sufficient level of security, but at the same time be user-friendly enough and acceptable. To access insurance data, healthcare workers use password-protected professional cards. In more than ten years of the use of the card system, the Institute has not detected any abuse or invasion into the system. On this basis and considering the feedback from users on a daily basis, we can say that users trust the system. This trust is a condition for the development of new solutions that will be even more involving for users and for the envisaged cross-border expansion.
4. Experience with Security Components in Slovenia In developing information solutions, the Institute pays particular attention to information security, because it deals with sensitive personal information. In addition to solutions for the protection of information during transfer through telecommunication networks and for access protection, the Institute has installed mechanisms for the protection of central components against invasions from public networks and for comprehensive usage tracking. A series of technical and organisational solutions has been introduced to ensure high availability of the system. Redundancy equipment has been provided, and highly available architectural solutions for communication links, active network equipment, application servers and databases in the central system have been established. Comprehensive supervision over the functioning of all components has been ensured and a supervision service, which activates expert teams in case of problems, has been organised. Particular attention is paid to managing security and reliability of systems used by providers of healthcare services. The Institute has introduced a series of necessary solutions and measures related to upgrading of equipment, information protection and reliability of the functioning of the system, which providers have to ensure. Providers of healthcare services receive earmarked funds for these tasks once a year. 4.1. Security of New Cards (Health Insurance Cards and Professional Cards) The new cards are Java smart cards. They are certified for safe electronic signatures and have Protection Profile Secure Signature Creation Device and EAL 4+ (Evaluation Assurance Level) certificates [2]. When issued, the health insurance card carries two digital certificates issued by the Institute’s certification authority. The first digital certificate on the card is used by providers to access information with the use of professional cards and is used to verify the identity of health insurance card-holders when accessing information through the on-line system. The key size of this digital certificate, which is not password-protected, is 1024 bits. The second digital certificate, whose key size is 2048 bits, is passwordprotected and is used to verify the identity of health insurance card-holders when logging into Internet applications to access their own medical information. Both digital certificates are valid for 10 years. In addition to the mentioned digital certificates, the card provides room for importing two digital certificates with the key size of 1024 bits and two digital
A. Bolka et al. / Personal Information Protection
77
certificates with the key size of 2048 bits, which can be used for electronic banking, remote access to local area network, etc. The professional card has a digital certificate issued by the Institute’s certification authority and is used for verification of the identity of professional card-holders when accessing the on-line system. Such a digital certificate has the key size of 2048 bits and is valid for 5 years. The second digital certificate, which is installed on professional cards of doctors and pharmacists, is a qualified digital certificate issued by certification authority Pošta®CA. Such a digital certificate is used for signing of electronic documents (for example electronic prescriptions). A qualified digital certificate has the key size of 2048 bits and is valid for 5 years. The professional card does not allow import of personal digital certificates, because it is intended exclusively for professional use. The existing digital certificates can also be used for other purposes within the healthcare system, for example for verification in the use of applications. The chip in the new health insurance card complies with international standard ISO/IEC 15408:2005. These are common criteria for security assessment of information technology. The chip complies with the security level EAL 4+ (Evaluation Assurance Level 4+). This is currently the safest (and still economically reasonable) standard level and stipulates that specifications, production and evaluation of the product have to meet the criteria of the standard for that level. Evaluation is made by independent organisations, which is an additional guarantee that a security test has been carried out. The card also meets the criteria of the directive PP-SSCD Type 2 and Type 3 (Protection Profile – Secure Signature-Creation Device) prepared by CEN/ISSS (European Committee for Standardization/Information Society Standardization System). These criteria determine, above all, the functional and security requirements for electronic signatures that cards have to meet. In addition to the mentioned criteria, the new card also meets GlobalPlatform Card Specification 2.1.1, which is recommended by the non-profit and non-governmental organisation, GlobalPlatform. The most important security aspect of GlobalPlatform specifications is the determination of safe architecture of the card and the procedure of card management throughout its entire life cycle. Considering the aforementioned, we can, without a doubt, claim that the new cards are the safest as well as the cheapest option for effective provision of security in the healthcare environment. The used solution with digital certificates in cards allows for the possibility to expand the use of such cards (certificates) for other purposes. Communication among cards in the card systems uses dual slot readers with keyboard, which prevent interception of passwords, because they never leave the reader. The password has to be re-entered with each use, which prevents abuses. 4.2. Safe Transfer of Data in the On-line System The on-line system enables direct and safe exchange of data messages among information systems [3]. It links the application of the provider with back office systems. Providers log on to the on-line system via safe Internet connection (TLS – Transport Layer Security). Information goes through the central entry point that provides for identification, verification and authentication of the users’ authorities. The entry point is an application server using software developed in Java and directs messages towards selected back office systems. The services of the on-line system are
78
A. Bolka et al. / Personal Information Protection
on-line services. Messages between the entry point and users are XML (Extensible Markup Language) messages in SOAP (Simple Object Access Protocol) format. 4.3. Opt-out Solutions In cooperation with the National Medical Ethics Commission, we have introduced two more safety features for information on prescribed medications [4]: x x
HIV medications are not displayed to anyone, because they are so discriminatory in their nature that awareness about such patients would bring more harm than good in their treatment; Insurance holders have the option to prevent specialist physicians and/or pharmacists to access data on medications prescribed to them (opt-out).
The Institute’s solutions have passed security tests carried out by independent expert institutions, such as the Jožef Stefan research institute. Tests have been carried out upon the introduction of the previous card system and the introduction of the new cards and the on-line system. The testers have established that the solutions meet the security criteria.
5. Challenges of Integrated Systems and Solutions The existing infrastructure of the Institute can also be used to exchange medical information, which can be stored in the back office system of the Institute or in another central back office system connected to the entry point of the on-line system. The following possible solutions of national importance have been identified: x x x x x x
eDocuments of compulsory health insurance (ePrescription, eReferrals, etc.), with the use of electronic signature with professional card and electronic archive; eVaccination (immediate collection of information on vaccinations in the central archive, authorised persons have safe access to such information); eAbsenteeism (immediate collection of information on sick leaves in the central archive, authorised persons have safe access to such information); eAllergies (immediate collection of information on a person’s allergies, safe access to such information); eEHRSummary (immediate collection of selected information from the summary of the electronic health record); eRegisteredAthletes (access of the providers that carry out systemic medical examinations to information on the status of registered athletes).
First concrete further steps related to cards are expected to be the following: x
Set up processes and develop application equipment that will enable insurance holders to access information about them kept and managed by the Institute and insurance companies for voluntary health insurance. This will strengthen the role of insurance holders in the care for regularity of insurance.
A. Bolka et al. / Personal Information Protection
x x x
79
The use of cards in solutions of the eHealth projects, such as access to the zVEM portal, electronic registering for healthcare services, insight into the national waiting list, etc. Records of professional cards, health insurance cards and digital certificates can be used for the portal services that request identification, authentications and authorisation of the user. State-of-the-art security mechanisms extend the possibilities to use the cards – both in the healthcare environment (for logging into your personal information system, accessing healthcare applications, digital signatures and encrypting medical documentation) and outside of it (for example, using the card to safely store your personal digital certificates).
6. Conclusion Information solutions that have already been established in Slovenia fully meet the highest requirements on technical security and protection of personal information. An infrastructure has been built enabling safe communication among providers, as well as completely safe exchange of very sensitive personal medical information. All these foundations are sufficient for expansions and upgrades of solutions for international use. What is more, the experience and know-how that Slovenian experts have gained by establishing domestic solutions will be welcome for such upgrades.
References [1] [2] [3] [4]
B. Zadel, Z novo kartico zdravstvenega zavarovanja do še večje varnosti, Varnostni forum, November 2008, Palsit d.o.o. A: Žlender, Nova kartica zdravstvenega zavarovanja – ključ za dostop do e-zdravja, Moj mikro, nr. 1, January 2010, Delo revije d.d. T. Marcun, A. Bolka, On-line health insurance in Slovenian Healthcare – solutions and pilot implementation, Congress of the Slovenian Medical Informatics Association, Zreče, October 2008 Project documentation, Health Insurance Institute of Slovenia, 2006-2009
XML as a Cross-Platform Representation for Medical Imaging with Fuzzy Algorithms Norbert GAL1 and Vasile STOICU-TIVADAR “Politehnica” University of Timisoara, Department of Automation and Applied Informatics, Faculty of Automation and Computers, Timişoara, Romania
Abstract. Machines that perform linguistic medical image interpretation are based on fuzzy algorithms. There are several frameworks that can edit and simulate fuzzy algorithms, but they are not compatible with most of the implemented applications. This paper suggests a representation for fuzzy algorithms in XML files, and using this XML as a cross-platform between the simulation framework and the software applications. The paper presents a parsing algorithm that can convert files created by simulation framework, and converts them dynamically into an XML file keeping the original logical structure of the files. Keywords: medical imaging, fuzzy system, XML, text parsing.
N. Gal and V. Stoicu-Tivadar / XML as a Cross-Platform Representation for Medical Imaging
1. Medical Image Interpretation Using Fuzzy Inference System A computer assisted linguistic image interpretation process cannot be done without converting the numerical values to semantic values using the fuzzification process. These semantic values are referring to the investigated objects features: shape, size, histogram value (gray-scale or color), and position. Using these features by cognitive reasoning the medical staff can decipher if the patient has an illness or not. Computers on the other hand use complex rules to infer which organ is under investigation and if the organ is healthy or not. Using fuzzy inference and medical knowledge a conclusion, a diagnosis or other indications can be formulated [6]. The schematic of the image interpretation can be viewed in Figure 1. The presented structure uses object features like shape, histogram, size and position.
Figure 1.Fuzzy system for image interpretation.
The editor from MatLab incorporates three major parts of the fuzzy system: the fuzzification process, the rule base in form of an expert system, and the output variables. The editor generates a text file where the structure and the features of the FIS are saved. The text file is divided in four different sections presented in table 1. The first part of the text file is the system section where the basic features of the FIS are described: the name, the type (Mamdani or Sugeno algorithm), the number of the inputs, the number of the outputs, the number of the rules and the AND-, OR-, imp-, agg- and defuzzmethod types. Table 1. *.fis sections and sample code. System [System] Name='Image-Interpretation' Type='mamdani' Version=2.0 NumInputs=5 NumOutputs=2 NumRules=3 AndMethod='min' OrMethod='max' ImpMethod='min' AggMethod='max' DefuzzMethod='centroid'
2. XML as a Cross-Platform Representation In our case the proposed XML representation is used in a particularized way as a crossplatform bridge between a software application and the fuzzy inference system editor from MatLab. Because XML is a set of rules for encoding documents in machinereadable form, it can contain rules, data, logical relationships and mathematical formulas as well [8]. Because the fuzzy inference system is saved in the form of a text file using a text parsing algorithm, an XML file can be constructed. The XML file shares the same logical structure as the “*.fis” file. The main branches are: the system descriptor, the inputs, the outputs and the rules. The leaves of the branches are the values which defines the branches. The logical structure of the parsing algorithm can be viewed in figure 2. The algorithm constructs dynamically the XML file. Because the number of the input, output variables and the rules can wary only the basic structure of the branch can be imposed.
86
N. Gal and V. Stoicu-Tivadar / XML as a Cross-Platform Representation for Medical Imaging
Figure 2. *.fis to xml parser algorithm
First the algorithm analysis the header of the section that is next to be parsed. The syntax for the header is ‘[header name] ’. The headers in the ‘ *.fis ’ file are [System], [Inputnr], [Outputnr] and [Rules]. The algorithm in the first step identifies the important values in the [System] section which refers to the name of the system, the number of the inputs, outputs, rules and other important parameters for the FIS but not important for the XML file construction. The first section of the XML file is constructed this way The second step is to parse the inputs of the FIS. This is done using two repeating loops. The first loop is for the inputs name, range and number of MFs. After the number of the MFs is obtained the algorithm enters in the second loop. This second internal loop takes each MF and according to the presented syntax above, copies it into the appropriate branch of the XML file. After there are no more MFs in the current input the parsing algorithm makes another step in the external loop until all of the inputs were parsed in the XML file. The parsing of the output variables is exactly the same as for the inputs because the definition syntax of the output is the same as for the input. In the “[Rules]” section there are no explicit markers of which MF of which variable is active. The only indicators are the position number on the row and the number itself which indicates the active MF. In this case the algorithm must go character by character and analyze each one separately. It must retain its position and value for the inputs; if it hits the “,” character then the next ones are for the outputs. The number between the “()” is the weight of the rule. The last number stands for the operator, number “1” stands for the AND and number “2” stands for the OR. A sample of the resultant XML file is presented in table 2: System
N. Gal and V. Stoicu-Tivadar / XML as a Cross-Platform Representation for Medical Imaging
87
3. Conclusions and Future Work XML representation offers simple but extensive cross-platform architecture for fuzzy algorithms from MatLab to other programming platforms. XML is a representation standard, only the format of the architecture is imposed, not the content. The user is free to use its own syntaxes, expressions and rules that make this type of architecture the perfect tool for software representation. Today there is gap between simulation and implementation. The proposed cross-platform solution contributes to cross the border between different software environments for a better implementation process. The presented parsing algorithm creates the XML dynamically, using the “*.fis” files structure according to the number of the variables, the number of the MFs in a variable and the number of the rules. This simplifies the conversion process from a simulation framework to a software application, which today is done manually. The main advantage offered by the proposed representation is the possibility to change the system parameters without interfering with the programs source code. Consequently, the information exchange between diverse scientific and technical communities, even in cross-border cooperation, is easier, allowing the rapid dissemination of valuable research results regarding the development of new medical imaging algorithms. The next step is to implement the parsing algorithm and to create a software environment that can interpret and incorporate the algorithms from the XML files.
Acknowledgement This work was partially supported by the strategic grant POSDRU/88/1.5/S/50783 (2009) of the Ministry of Labor, Family and Social Protection, Romania, co-financed by the European Social Fund – Investing in People.
References [1] L.A. Zadeh, Fuzzy sets, Information and Control, vol. 8, pp.338-353, 1965. [2] R. Chbeir, F. Favetta, A Global Description of Medical Imaging With High Precision, IEEE Trans. On Systems, Man, And Cybernetics Part B: Cybernetics (2003), 752-757. [3] The MathWorks, Inc. User’s Guide Version 2: Fuzzy Logic Toolbox User’s Guide For Use with MATLAB®, July 2002 Online only Revised for Version 2.1.2 (Release 13). [4] R.D. Rodrigues, A.J.O. Cruz, R.T. Cavalcante, Aliança: A proposal for a fuzzy database architecture incorporating XML, Fuzzy Sets and Systems 160 (2009), 269–279. [5] Z.M. Ma, Li Yan, Fuzzy XML data modeling with the UML and relational data models, Data & Knowledge Engineering 63 (2007), 972–996. [6] N. Gal, V. Stoicu-Tivadar: Computer Assisted Medical Image Interpretation Using Fuzzy Logic; 4th International Workshop on Soft Computing Applications (2010), 159-163. [7] E.M. Ramirez R.V. Mayorga: A Cascaded Fuzzy Inference System for Dynamic Online Portals Customization, ,QWHUQDWLRQDO -RXUQDO RI (OHFWULFDO DQG &RPSXWHU (QJLQHHULQJ [8] B. Evjen, K. Sharkey, T. Thangarathinam, M. Kay, A. Vernet, S. Ferguson, Professional XML, Wrox, 2007
Fully Connected Emergency Intervention for the Critical Home Care System Draško NAKIK1, Suzana LOŠKOVSKA and Vladimir TRAJKOVIK Ss. Cyril and Methodius University Faculty of Electrical Engineering and Information Technologies, Karpos 2 bb. Skopje, Macedonia
Abstract. The Critical Home Care System – CHCS, we propose, achieves permanent advising, frequent control appointments and quick reaction to critical conditions by constant remote monitoring of patient’s vital signs from the hospital, while staying at his home. Physicians react properly to the developing condition, contacting the patient or a member of the household, or sending an ambulance in an emergency. The CHCS additionally provides constant inspection of the patient’s condition to the ambulance doctor in emergency situations and to the urgent centre staff to prepare better for accepting the patient, enabling a fully connected emergency intervention. In this paper we will concentrate on the data flow during the emergency intervention in this highly collaborative system. Keywords. home-care, remote monitoring, emergency intervention, information retrieval, information presentation, EMR access, remote collaboration
Introduction Patient home monitoring systems [3,4,5,6] generally share some common deficiencies: they either do not provide permanent monitoring, engage multiple devices for sampling vital signs [2,11], or are not firmly integrated with a system for facing emergency situations [12]. In this paper we will focus on a concept for overcoming the lack of integration of such systems in a much larger existing system for emergency intervention. The CHCS system is intended to provide constant monitoring of convalescents with fragile health condition, people with chronicle diseases and old people. It is a model of a permanent monitoring system that will process patient’s data and alert a physician for abnormalities and potential emergency conditions [10]. We are interested to have the patient connected to the hospital not only when staying in his home, but also when he is being transferred from his home to the hospital [1] in emergency situations and have the personnel in the urgent centre prepared for patient’s admission. Our model combines synchronous remote monitoring [9] with transit data transfer and teleconference between the three main actors, monitoring doctor or general practitioner, ambulance doctor and urgent centre doctor [7, 8]. We give a basic description of the system in Section 1, and in Section 2 and in Section 3 we present the mobile applica-
1 Corresponding Author: Draško Nakik Graduate student at the Computer Science and Engineering Department, Phone: +389 70 269 959, E-mail: [email protected].
D. Nakik et al. / Fully Connected Emergency Intervention for the Critical Home Care System
89
tion layer and the data flow of the fully connected emergency intervention, respectively. Finally, we make the conclusion in Section 4. 1. Description of the system The first system we need to elaborate is the already existing emergency intervention system. The whole urban area is divided into sectors, shown in Figure 1, so emergency cases that take place in certain sector must be treated in the hospital that belongs to that sector. The same logic stands for the ambulance service. An ambulance is sent from the emergency intervention centre (EIC) that belongs to a sector where the emergency case takes place. There are some basic rules that dictate the whole scenario: a) when a call is received in the dispatching centre, the closest ambulance vehicle to the location of the call should be sent in action; b) the ambulance vehicle should transport the patient to the closest hospital with urgent centre (UC); c) there is a doctor in the ambulance vehicle.
Figure 1. Sectorization of the urban area by hospitals and EICs
The backbone process of this system is the automatic emergency signal flow depicted in Figure 2. This signal is raised when the Emergency Condition Evaluator (ECE) component in the HS evaluates some abnormal values. The monitoring doctor receives the alert signal, one for potential problem and one for emergency situation, and has patient’s VS displayed immediately on the computer screen. He decides weather to forward the signal to the patient’s doctor: GP or specialist who was responsible for the patient while treated in the hospital. If the alert is of first emergency level (marked red) the monitoring doctor sends alert signal to the Ambulance Dispatching Centre. An ambulance vehicle is activated according to the address of the patient. Another signal is sent to the UC according to the sectorization. If the alert is of second emergency level, the alert signal is sent solely to patient’s doctor. If patient’s doctor is not available, then the monitoring doctor takes his role in the system. Another very important tier in the system is the communication tier. In Figure 2 we recognize two phases of communication. The first when the ambulance is getting from EIC to patient’s home and the other when it is getting from patients home to the UC. In the first teleconference triangle the patient’s doctor and ambulance doctor are included.
90
D. Nakik et al. / Fully Connected Emergency Intervention for the Critical Home Care System
This way the ambulance doctor can get information for the patient from a doctor that knows the patients clinical history well, and from the patient himself for his very current condition. In the second teleconferencing triangle the patient, UC doctor and ambulance doctor are included. This way the UC doctor can gain information for the patient’s condition development in real time. The patient’s doctor can communicate the UC as well. If patient’s doctor is not available, then the monitoring doctor takes his role in the scenario.
Figure 2. Alert signal propagation diagram (left) and CHCS system communication layer (right)
2. The mobile application layer The patient application resides on his mobile phone that collects the data from the body signals sensors, and has four basic modules represented in Figure 3 on the left. Monitor: the patient can monitor his vital signs (single or multiple) and choose the type (curve, bars, line) and representation of the graph (all values or abnormal values) shown in Figure 3.a. Medication Therapy Scheduler: the user creates timetable and is reminded accordingly with a sound signal. The patient can also follow a link to medication user instructions (see Figure 3.b). Critical Condition Manager: the patient can parse a tree by giving answers to multiple choice questions, which will lead him to the closest description of his condition where he can find instructions to manage the situation (see Figure 3.c). Communicator: the patient can communicate with other actors in the system according to the features in Figure 3.d. All possible network interferences are included: the Internet, landline and mobile networks. Patient’s doctor uses the same application as the ambulance doctor, though the later is equipped with a forearm mobile phone holder and hands free gadget. This way the ambulance doctor can perform the on-site intervention undisrupted, while still being connected to the system. This application is designed to strongly convey the bedrock rules for mobile applications that are used in sternly time-limited circumstances: complete information with great relevance and minimum recall, short response time, easy navigation and clear presentation. It is consisted of five modules represented in Figure 3 on the right. Monitor: it has the same presentation functionalities as in the patient application, plus the Medication Therapy Scheduler that appears with high transparency below the graph in focus (Figure 3.a). This way the doctor can see if the deterioration of patient’s condition occurred due to not taking medications on time. Vital Signs Comparator: the doctor can compare the vital signs from two equal time periods (shown in the first graph in Figure 3.b) by investigating the lines of both periods in the same graph. He can also inspect the difference between the vital signs in first and the
D. Nakik et al. / Fully Connected Emergency Intervention for the Critical Home Care System
91
second period of time (second graph in Figure 3.b). The doctor can view the vital signs values in specified period of the day (time slot) during specified period of time (graph three in Figure 3.b). Finally, he is provided to compare unequal time periods – with option for merging the periods together in continuous line or comparing the shorter period with equal time slots from the longer period (graph four in Figure 3.b). Patient Profile: this module gives short and complete presentation of the patient to the ambulance doctor (see Figure 3.c). With this information the doctor is able to prepare himself better for the emergency intervention. EMR Analyzer: this module provides additional information for the patient’s clinical history giving his complete EMR at doctor’s hand shown in Fig. 3.d. There are some predefined queries that appear as the EMR tree is traversed. Aside of this the user is enabled to type his own query and save it for later use (see Figure 3.d). Communicator: the communicator provides connection to all relevant actors in the system for collaboration to enhance the efficiency of the intervention. Various ways of communication are presented in Figure 3.e. The ambulance and the urgent centre applications will not be discussed in this paper.
Figure 3. Alert signal propagation diagram (left) and CHCS system communication layer (right)
92
D. Nakik et al. / Fully Connected Emergency Intervention for the Critical Home Care System
3. The data flow of the fully connected emergency intervention Providing data about the patient’s condition to the actors involved in the emergency intervention, continuously, during the complete cycle of the intervention is very important. By having insight in the patient condition by monitoring his vital signals and his medical history trough EMR, the ambulance doctor can prepare for the intervention and can consult patient’s doctor more precisely while getting to the patient’s home. On the other hand, the medical team from the Urgent Centre can prepare for accepting the incoming patient for further necessary treatment. Figure 4 represents the data providers: Central EMR Manager (CER) or Central EMR Repository (CEM) which are responsible for providing patients’ medical data from hospitals ISs; patient’s home where vital signals are collected from the residing patient and transferred to Central Vital Signal Hub (CVSH); and the ambulance vehicle equipped with application for collecting and forwarding data from its monitoring system while transferring the patient from his home to the Urgent Centre; and the data users: the ambulance vehicle, equipped with application for remote monitoring patient’s vital signals, the monitoring doctor and the patient’s doctor, and the Urgent Centre.
Figure 4. Time and data diagram of the emergency intervention without the communication layer (left) and CHCS vital signals harvesting and distribution during emergency intervention (right)
To have the full functionality and feasibility for this system EMRs should be made available to the users of the system. Two general concepts are proposed form overcoming this imposing difficulty. In Figure 5 we can see a Central EMR Manager, a system module that is responsible to interface the system to the existing hospital ISs for obtaining the EMRs and alternative module which collects the EMRs from different hospitals in a standardized format which is directly available to the users.
4. Conclusion The CHCS system is intended to provide constant monitoring of convalescents with fragile health condition, people with chronicle diseases and old people. The subsystem
D. Nakik et al. / Fully Connected Emergency Intervention for the Critical Home Care System
93
for fully connected emergency interventions provides an efficient way of rescuing people’s lives by supplying the doctors on field with valuable information and establishing remote-collaboration with all relevant actors in the system and vice-versa – providing information from the field to the urgent centre where the patient is being transported. The enhanced level of collaboration and high patient data provisioning will play a major role in saving lives in emergency situations.
Figure 5. Two general concepts for providing patients’ EMRs to the users of the CHCS
References [1]
D. Komnakos, D. Vouyiokas, I. Maglogannis, P. Constantinou, Feasibility study of a joint e-health mobile high-speed and wireless sensor system. 1st international conference on PErvasive Technologies Related to Assistive Environments, 2008. [2] I. Al Khaib, F. Poletti, D. Bertozzi, L. Benini, M. Bechara, H. Khalifeh, A. Jantsch, R. Nabiev, A multiprocessor system-on-chip for real-time biomedical monitoring and analysis: ECG prototype architectural design space exploration. ACM Transactions on Design Automation of Electronic Systems (TODAES) (2005), 294-303. [3] J.E. Bardram, C. Bossen, A. Thomsen, Designing for transformations in collaboration: a study of the deployment of homecare technology. Conference on Supporting Group Work. [4] V. Koufi, F. Malamateniou, G. Vassilacopoulos, A medical diagnostic and treatment advice system for the provision of home care. 1st international conference on PErvasive Technologies Related to Assistive Environments, 2008. [5] S.I. Ahamed, M.M. Haque, A.J. Khan, Wellness assistant: a virtual wellness assistant using pervasive computing. Symposium on Applied Computing (2007), 782 – 787. [6] T. Carl, L. Dajani, The future of homecare systems in the context of the ubiquitous web and its related mobile technologies. 1st international conference on PErvasive Technologies Related to Assistive Environments, 2008. [7] S. Hamek, F. Anceaux, S. Pelayo, M. Beuscat-Zephir, J. Rogalski, Cooperation in healthcare theoretical and methodological issues: a study of two situations: hospital and home care. 2005 annual conference on European association of cognitive ergonomics (2005), 233-240. [8] S.B. Larsen, J.E. Bardam, Competence articulation: alignment of competences and responsibilities in synchronous telemedical collaboration. Conference on Human Factors in Computing Systems (2008), 553-562. [9] S.A. Ballegaard, T.R. Hansen, M. Kyng, Healthcare in everyday life: designing healthcare services for daily life. Conference on Human Factors in Computing Systems (2008), 1807-1816. [10] L. Chittaro, Visualization of patient data at different temporal granularities on mobile devices. Advanced visual interfaces (2006), 484-487. [11] T. Zimmerman, K. Chang, Simplifying home health monitoring by incorporating a cell phone in a weight scale. 1st international conference on PErvasive Technologies Related to Assistive Environments, 2008. [12] Blanchet D. K.: Remote patient monitoring. Medical Conectivity, Telemedicine and e-Health, March, 2008
How to Ensure Sustainable Interoperability in Heterogeneous Distributed Systems through Architectural Approach a
Louise PAPE-HAUGAARD1a and Lars FRANKb Dept. of Health Science and Technology, Medical Informatics, Aalborg University, Denmark b Dept. of Informatics, Copenhagen Business School, Denmark
Abstract. A major obstacle in ensuring ubiquitous information is the utilization of heterogeneous systems in eHealth. The objective in this paper is to illustrate how an architecture for distributed eHealth databases can be designed without lacking the characteristic features of traditional sustainable databases. The approach is firstly to explain traditional architecture in central and homogeneous distributed database computing, followed by a possible approach to use an architectural framework to obtain sustainability across disparate systems i.e. heterogeneous databases, concluded with a discussion. It is seen that through a method of using relaxed ACID properties on a service-oriented architecture it is possible to achieve data consistency which is essential when ensuring sustainable interoperability. Keywords. Medical Informatics, Computer Network, Computer systems, Database, Database Management Systems
Introduction One of the major characteristics in eHealth systems is the requirement of ubiquitous information. How to achieve ubiquitous information in a reliable, sustainable and consistent manner in eHealth is still an unsolved paradox. The objective in this paper is to explore if a theoretical proposed architectural design potentially meets the challenge of ubiquitous information in a practical eHealth setting. The theoretical design includes a refinement of traditional design properties of database management systems and a transformation of the well–known ACID (Atomicity, Consistency, Isolation and Durability) properties [1]. The practical eHealth setting chosen is a Danish national project SMR (‘Shared Medication Record’), between regions, hospitals, GPs (General Physicians) and pharmacies [2]. The requirement of ubiquitous information puts high constraints on the architectural design and database design, because it is needed to apply different replica techniques in order to achieve ubiquitous information. When applying replica techniques to heterogeneous databases the complexity typically escalates due to the fact that coordination of data consistency, updates, synchronization, security, and processing have to happen at multiple sites. However, in the solution we suggest the 1 Corresponding Author: Louise Pape-Haugaard; Fr. Bajers Vej 7D, DK-9220 Aalborg. Email: [email protected]
L. Pape-Haugaard and L. Frank / How to Ensure Sustainable Interoperability
95
theory of the design to handle this complexity by applying relaxed ACID properties as proposed by Frank, 2008 [3] and thereby optimize both ubiquitous availability to EHR information and the consistency in this information. Furthermore to achieve an optimal use of relaxed ACID properties it is also needed to refine the architectural design to access the heterogeneous databases and thereby the systems. We have in order to contextualize whether the theory is suitable in vitro settings included a project as an example of how this can be done. The project had an initial purpose of establishing if a SOA (Service Oriented Architecture) solution can be applied to Danish eHealth, even though there has not been a deliberate focus on applying the described theory of design in the project. To describe the architecture needed to meet the requirement of ubiquitous information there is a focus on high performance for the end-user i.e. the clinicians using the eHealth systems and furthermore on keeping an acceptable level of consistency between heterogeneous systems. A focus on high performance has also been essential for the SMR projects development. The paper is organized as follows: The approach is firstly to explain the theoretical setting being traditional model of transactions in an architecture in distributed database computing, followed by a possible approach to use an architectural framework to obtain sustainability across disparate systems i.e. heterogeneous databases, the theory is then compared to SMR. SMR is furthermore elaborated on in the architectural framework through the Danish example of using services.
1.
Theory: Traditional Design of Transactions
Traditionally, transactions between databases are defined by use of ACID properties [4]. ACID properties were developed to centralized or homogeneous distributed databases, and furthermore developed to business-oriented applications such as banking, because what characterizes ACID properties is the guarantee of reliability in transactions. We will describe the basics in transactions followed by ACID properties. Atomicity is typically a ‘black and white’ world meaning either all transactions are executed or neither is executed, also called the ‘all or nothing’ rule. Consistency is that transactions always operate on a consistent view of the database and that when the transaction has ended the database is left in a consistent state. Isolation is to pretend that each transaction is executed exclusively i.e. it seems that only one transaction is done at a time. Durability guaranties that once a transaction is committed its effects are persistent. The ACID properties are now-a-day broadly used in all kinds of domains and to a certain extend to distributed, heterogeneous and thereby multidatabase environments, however, this is not straight forward, why further methods to support a multidatabase environment is being explored [4, 5]. Recent research indicates methods to handle distributed heterogeneous and multidatabase environments, either by applying a generalization of the transaction model or by relaxing the ACID properties [3, 5]. The consequences of the methods are identical, i.e. to provide flexibility to the performance of transactions. Basically, the objective of transactions is to ensure that all objects managed by a server remain in a consistent state when accessed by multiple transactions.
96
L. Pape-Haugaard and L. Frank / How to Ensure Sustainable Interoperability
1.1 Global Transactions in a Multi-Database Environment by Applying Relaxed ACID Properties A multi-database architecture roughly illustrates today’s organisation of eHealth database architecture, because of the utilization in this domain of broad locally distributed autonomous databases. Global transactions have been applied to access data in multiple local databases [4, 6]. However, in order to access information, and potentially obtain ubiquitous information, in these autonomous and disparate databases one can use a refinement of the methodological approach of global transactions. Recently, traditional transactions models have been designed and evolved to integrate locally but distributed databases without using a traditional organized DBMS. In the study ‘Semantic ACID Properties in Multidatabases Using Remote Procedure Calls and Update Propagations’ [7] a countermeasure transaction model is developed to manage and reduce problems occurring when ACID properties in distributed heterogeneous databases are missing. In this model, a root transaction may manage several sub transactions corresponding to server and client transactions. So normally all communication is handled during the root transaction and information is accessed through the sub transactions. Global transactions are applied to the atomicity property when using relaxed ACID properties and is divided into three parts, where the last part secure that missing updates will be re-executed until all updates have been committed. Handling the consistency property when applying relaxed ACID properties one would assume that this is impossible due to the definition of consistency, because a database with relaxed ACID properties is typically inconsistent. To address this issue it is necessary to apply asymptotic consistency to the database meaning that the execution has relaxed consistency property if it is asymptotic consistent at the beginning of a transaction and when it has been committed [7, 8].
2.
Applying Relaxed ACID Properties to Distributed eHealth Systems
When implementing relaxed ACID properties in eHealth systems the first convention is to avoid normal peer-to-peer communication between distributed systems, because of flexibility issues. Of course, direct communication is acceptable if the peer-to-peer communication is based on streaming or another unlikely database communication technology. However, to implement relaxed ACID properties there is a need for controlling and executing communication between locally distributed eHealth through SOA applications as this gives independency of the internet address of the eHealth service. Furthermore, the second convention is needed when designing SOA applications handling the communication; a) use direct database accesses for local database services and b) use read only, compensable update and retriable update SOA services when the service is managed elsewhere [3, 7, 9]. In Frank and Munck, 2008 [9] it was established that out of six evaluated integration architectures two were chosen as being the best, due to evaluation on technical properties as performance, availability, atomicity, consistency, isolation and durability. Interesting is it, that the two best evaluated architectures (respectively centralized and decentralized structure) both apply a method for storing information on site and furthermore it is possible to be an acknowledged user of this stored
L. Pape-Haugaard and L. Frank / How to Ensure Sustainable Interoperability
97
information from a distance. The main difference is whether or not it is possible to reuse information directly (decentralized) or indirectly (centralized).
3.
In vitro Example of Applying a SOA Solution in eHealth: Shared Medicine Record
To contextualize the proposed solution an example is here given of an in vitro ongoing Danish project: SMR is organized as a centralized architecture meaning that the user will access information from one place and one place only [3]. The user will not access information where the information is recorded, but merely on a centralized platform. In this manner the user will only need to ‘know’ one place to find information regarding medication. This is illustrated in figure 1. The application of read-only services has been implemented on SMR, i.e. if SMR is in use at one site then other sites only have read-only permissions. Looking at this example of sharing information in eHealth it seems possible to develop new healthcare services to be implemented in this manner.
Figure 1. Distribution of communication in a heterogeneous eHealth environment (inspired by [3])
4.
Discussion and Conclusions
Uniquely for eHealth is the rare update of a specific piece information, instead a new version of information in question is added. This has several reasons, and to mention some; a) the need for historical overview, b) a law stating not to erase information from the EHRs and c) the diversity in healthcare profiles. In Denmark the different hospital regions has had autonomy in selecting EHR systems and therefore these are heterogeneous with incompatible types of health records [10, 11]. However to design, develop and deploy a monopolistic system or even just homogenous system across organizational borders would not only be more costly it would also conflict with the national digitalization strategy of having a multiple vendor strategy and an equal procurement. Therefore it is the hospital regions
98
L. Pape-Haugaard and L. Frank / How to Ensure Sustainable Interoperability
responsibility to select solutions that suits their demands of how to distribute the health records stored in the central database. We have to acknowledge the need for solutions and supporting architectures, which can include already implemented systems. This is shown through the case of SMR (Shared Medication Record) where existing legacy systems are to interoperate through web services in SOA context. Furthermore, it was established that the theory can be applied in practice. The question is whether a technical solution supporting interoperability between existing systems is enough, because the solution of SMR is still not completely embraced by healthcare professionals. This lack of adoption influences the assessment of clinical utility in the solution.
5.
Conclusion
In this paper, the objective was to explore if a theoretical proposed architectural design potentially meets the challenge of ubiquitous information in a practical eHealth setting. To explore this objective we have described a theoretic framework of architecture for integrating the information across distributed heterogeneously designed databases. Furthermore, we have described how the existing theory of relaxed ACID properties may be used to improve the integration of distributed health records in e.g. Denmark. We have seen that the theoretic framework is applicable in a practice eHealth setting, even without a profound familiarity to the theory as through the example of SMR.
References [1] [2] [3] [4] [5] [6] [7] [8] [9]
[10]
[11]
T. Härder and A. Reuter,. Principles of transaction-oriented database recovery. Computing Surveys 15(4) (1983) 287–317. L. Frank, Databases and applications with relaxed ACID properties. Samfundslitteratur. ISBN: 97887-593-8356-8, 2008DigitalHealth, 2008. Projectdescription in Danish of Shared Medication Record – www.sdsd.dk/det_goer_vi/Faelles_Medicinkort.aspx accessed nov.16th. J. Gray and A: Reuter, A., Transaction Processing, Morgan Kaufman, 1993. B. Medjahed, M. Ouzzani, Elmagarmi, 2009. Generalization of ACID Properties in Encyclopedia of Database Systems, p 1221-1222. ISBN 978-0-387-35544-3 S. Mehrotra, R. Rastogi, H. Korth and A. Silberschatz, A transaction model for multi-database systems, Proc International Conference on Distributed Computing Systems (1992), 56-63. L. Frank and T. Zahle, Semantic ACID Properties in Multidatabases Using Remote Procedure Calls and Update Propagations, Software - Practice & Experience, Vol.28, (1998), 77-98. L.C. Rivero, R. Doorn, V. Ferraggine, Encyclopedia of database technologies and applications, Idea Group References, 2006. L. Frank and S. Munck, An Overview of Architectures for Integrating Distributed Electronic Health Records, Proceeding of the 7th International Conference on Applications and Principles of Information Science (APIS2008), (2008) 297-300. L. Frank and S.K. Andersen, Evaluation of Different Database Designs for Integration of Heterogeneous Distributed Electronic Health Records, Proc. of the International Conference on Complex Medical Engineering (CME2010), IEEE Computer Society, 2010. K.H. Rosenbeck, A.R. Rasmussen, P.B. Elberg, S.K. Andersen, ’Balancing centralized and decentralized EHR approaches to manage standardization’ Studies in Health Technology and Informatics. Vol. 160 p151-155, IOS Press, Amsterdam, 2010.
A Preliminary Study on Network Traffic Estimation in Using EPR with Thin-Client Computing over Wide Area Network Kei TERAMOTO a,b, Shigeki KUWATA b,1, Masaki MOCHIDA c and Hiroshi KONDOH b a Graduate School of Medical Sciences, Tottori University, Tottori, Japan b Division of Medical Informatics, Tottori University Hospital, Tottori, Japan c SECOM.SANIN Co., Ltd, Shimane, Japan
Abstract. With the remarkable advantages of thin-client computing (TCC) on security enhancement and cost reduction, the TCC architecture seemed appropriate for EPR utilization in cross-border e-health systems. The advantage in less consumed network bandwidth, however, still remains quantitatively unidentified at present. This study aimed to estimate the network traffic required in using EPR on WAN environments through the comparison of TCC and server-client (SC) models The results indicated that one of representative TCC applications required much less network bandwidth than the conventional SC model. Further studies will be focused on the verification of the adopted scenarios and a combination of applications that would affect the estimation of the network bandwidth. Keywords. electronic patient records, network bandwidth, scenario, server-client model, thin-client computing,
Introduction In recent years, introducing the electronic patient record (EPR) system has spread rapidly in medical facilities. Most of EPR systems were designed based on the conventional server-client (SC) model. Client terminals (PCs) of such EPR systems are generally required to have high-powered hardware, described as fat-clients, since they have to process various types of information in the EPR application. On the other hand, concerns for thin-clients, i.e., inexpensive terminal computers with limited resources in contrast to fat-clients, have been brought to public attention [1]. Applying thin-client computing (TCC) technology to the hospital information system proved effective in actual settings of medical facilities in reducing the total ownership cost (TOC) and enhancing security level of the system, e.g. by diskless PC installation and secure remote access to EPR via the Internet [2] [3].
1 Corresponding Author: Shigeki Kuwata Division of Medical Informatics, Tottori University Hospital, 36-1 Nishi-cho, Yonago, Tottori, 683-8504, Japan; E-mail: [email protected]; +81-859-38-6892; Fax: +81-859-38-6899
100
K. Teramoto et al. / A Preliminary Study on Network Traffic Estimation
Technical Architecture of Thin-Client Computing The methods for introducing the TCC model into existing EPR systems developed based on the conventional SC model can be largely modeled according to the following four groups: (1) Server-based computing (SBC) model, (2) Virtual machine model, (3) Blade server model and (4) Web-based model. SBC enables multiple PC users to share virtualized applications upon a single operating system with virtualization software such as GO-Global [4] and Citrix Presentation Server [5]. In this study, the authors focused on the first model, SBC, among the TCC models and used it in the following experiments, since SBC provided us with the easiest installation by minimizing the additional customization for the setup procedures. For the simplicity, SBC will be referred to as TCC below. Contribution of Thin-Client Computing to Cross-border e-Health A typical example of successful TCC applications to the cross-border e-health was the networked EPR system being open to mutiple medical facilities [6] [7]. The system enabled medical professionals to access the EPR system in other medical facilities from client PCs in their own hospitals with the TCC technology over the Internet. The professionals were able to view all records of the patients referred to the hospital, whereas the hospital previously sent limited information to other hospitals by using portable media. When designing the EPR system utilized for cross-border e-health over the wide area network (WAN), system administrators should be concerned about the system security, required network bandwidth and system response time. 1. Objective With its remarkable advantages, the TCC architecture seemed appropriate for EPR utilization in cross-border e-health systems. The advantage in the required network bandwidth, however, still remains quantitatively unidentified at present in using EPR systems on TCC. Therefore this study aimed to estimate the network traffic required in using EPR on WAN environments through the comparison of the network bandwidth occupied by EPR implemented based on TCC and SC models. 2. Methods The following procedures were carried out to obtain the network bandwidth consumed in practical utilization of EPR over WAN. Firstly, the authors investigated access logs of the “open” EPR system, Oshidori-net [6], in which the introduced TCC architecture allowed users to make cross-reference of patient records between a university-affiliated hospital and a regional medical center in Japan. The log files contained a variety of EPR operations by users, which were grouped into four categories and their frequencies were added up for each “session” defined as the duration from starting opening records of a certain patient to terminating it. Next, the authors organized experimental systems for measuring the network bandwidth on TCC and SC models by simulating several tasks on the EPR system on both models. Finally, the required bandwidth for both
101
K. Teramoto et al. / A Preliminary Study on Network Traffic Estimation
models was estimated and compared for four categories of EPR operations by coupling the frequency of the category with the corresponding consumed bandwidth, on the assumption that every operation corresponding to a certain category could be represented by combinations of the simulated tasks in the experiment. 2.1. Collecting Access Logs from EPR To grasp the operations of medical professionals on EPR systems, access logs were extracted from the EPR system (Clinical Information System [8] by IBM Japan Ltd.) adopted for Oshidori-net. The logs included patient ID, user ID, operation and access time between January 4 and May 28, 2010. During the period 197 sessions in total were observed. The types of operations were categorized as follows: S1: Initial access to patient data, S2: Viewing order data, S3: Viewing progress notes with attached JPEG images and S4: Viewing DICOM images. Table 1 shows the EPR operations by medical professionals and their corresponding categories and frequencies in the logs. The category will also be referred to as a “scenario” below. Table 1. EPR operations by medical professionals and their corresponding categories and frequencies in the access logs for Oshidori-net [11] from January 4 and May 28, 2010 *: Frequency No. S1 S2
Category (Scenario) Initial access to patient data Viewing order data
S3 S4
Viewing progress notes Viewing DICOM images
Typical operations by EPR users Downloading patient data for initial caching Viewing blood test results, vital chart, diagnosis, summary, meal and visit reservation Viewing progress notes and attached JPEG images Viewing radiology images
Freq.* 219 401 23 11
2.2. Experimental Environments An experimental system was developed to measure the network bandwidth that varied with EPR operations on TCC and SC environments. The environments simulated the TCC system used in Oshidori-net. GG and Citrix, well-known TCC applications, were adopted. Experimental devices consisted of EPR servers, centralized servers, client PCs, network switches and packet monitor machines. Switches were placed on the network between EPR servers and centralized servers (referred to as SC zone) and between centralized servers and client PCs (TCC zone), respectively. In order to transfer the copy of all network traffics in both zones to packer monitor machines, the port mirroring function of the network switches was activated. The packet monitor machines were supposed to receive network traffics from all packets through the switches and measure the packet size and its timestamp. The reason of using the packet monitor machines, instead of installing software-based monitoring tools on the servers, was to avoid the disturbance that would affect the performance on the server, which would exhibit a systematic bias in the experimental results. 2.3. Measurement of Network Bandwidth In the experiment, it was targeted to measure the network bandwidth occupied by EPR on SC and TCC zones for each scenario described in Table 1, since it appeared considerably impracticable to simulate all the EPR operations found in the logs. The authors assumed that every operation corresponding to a certain scenario was
102
K. Teramoto et al. / A Preliminary Study on Network Traffic Estimation
represented by combinations of the following tasks: T1: selecting a patient on a bed map so as to open its records (corresponding to S1), T2: selecting a tab of progress notes on the main window (S2 and the first half of S3), T3: selecting a link in progress notes to display pathological images on a web browser (the second half of S3), T4: booting a DICOM viewer application (the first part of S4), T5: selecting DICOM images twice (the second part of S4), T6: magnifying the image twice as large as their original size for each image (the last part of S4). While the network bandwidth was being measured by traffic monitoring machines, all tasks were executed 10 times for three implementations (SC, GG and Citrix) respectively. 2.4. Estimation of occupied network bandwidth Network bandwidth occupied by EPR for each scenario EB(scen) was estimated as total consumption of the measured bandwidth B(task) for the N tasks incorporated in the scenario. To obtain the practical estimation of the bandwidth required for use of EPR over WAN, frequency of each scenario per session F(scen,sess) found in the Oshidori-net logs was multiplied by the estimated bandwidth corresponding to the scenario EB(scen) and summed up for all scenarios to yield the estimated bandwidth required for each session EB(sess) (Equation 1) . Subsequently the required bandwidth for SC and TCC models was compared. N
EB( scen)
¦ B(task) ,
task 1
4
EB( sess)
¦ ^EB( scen) u F ( scen, sess)`
(1)
scen 1
3. Results and Discussion Table 2 shows the representative values of network bandwidth required for executing the tasks B(task) form T1 to T6 for SC, GG, and Citrix. Midranges and quartile deviations were used instead of means and standard deviations when the hypothesis on distributional normality was rejected by the Kolmogorov-Smirnov goodness-of-fit test. For SC, the bandwidth was consumed much more for the events that needed some internal initialization in the EPR application, e.g., downloading data to open a new window (T1, T4 and T5). It appeared true that the downloaded data were partially or fully cached, since the subsequent tasks (T2, T3 and T6) did not transfer much data as compared with those for TCC models (GG and Citrix). For TCC models, the values for all tasks except for T6 were much higher for GG than for Citrix. The differences would be caused by the difference between two applications: The Rapid-X Protocol, adopted for GG, tries to precisely trace the graphic API commands issued in the centralized server in order to transfer to and reproduce them in the client PC. Therefore the bandwidth would be consumed for GG when some drawing commands were frequently transmitted over the network, even if users did not perceive drastic changes in the client display (T1 to T5). On the other hand, for T6, contents in the client display were frequently redrawn in response to the magnification of images. In that case much more bandwidth would be consumed for Citrix, which were developed based on a bitmap transferring protocol. Values of the estimated network bandwidth for all sessions in Oshidori-net, EB(sess), for SC, GG, and Citrix were presented in Figure 1. As the values much differed
103
K. Teramoto et al. / A Preliminary Study on Network Traffic Estimation
depending on whether reference to images in EPR occurred in the session, the sessions were divided into two groups: (a) sessions in which users did not view JPEG or DICOM images (corresponding to sessions including neither S3 nor S4) and (b) sessions in which users viewed images. The analyses were targeted separately for (a) and (b) and together for (c) all sessions. For all cases, the estimated bandwidth for Citrix was significantly lower than that for others. For case (a), the values of EB(sess) in kbps were 285.0 (285.0–876.5), which stands for those ranging from 285.0 to 876.5 with a median of 285.0, 195.2 (195.2–1435) and 16.68 (16.68–72.96) for SC, GG and Citrix, respectively. The values for GG were significantly lower than those for SC. The values were 336.0 (319.9–2669), 971.4 (623.2–2645) and 44.44 (31.24–1630) for case (b); and 290.4 (285.0–2669), 311.2 (195.2–2645) and 21.08 (16.68–1630) for case (c); for SC, GG and Citrix, respectively. Consequently, Citrix, an implementation model based on the TCC architecture, was far superior to the conventional SC model and the other TCC model, GG, in that Citrix required much less bandwidth in using EPR over the network. It was found that GG was superior to SC for (a), the case that users did not view JPEG or DICOM images, though the difference was relatively slight as compared with the difference between Citrix and others. Table 2. Results of measured bandwidth B(task) in kbps. Estimated bandwidth for a scenario EB(scen) equaled the summation of B(task) incorporated in the corresponding scenario (cf. Equation 1). S1 T1 279.6±10.7 79.1±5.0 12.3±1.1*
Model SC GG Citrix
S2
S3
T2 5.4±0.0* 116.0±3.5 4.4±0.8*
T3 34.9±1.2 428.1±9.8 14.6±0.2*
T4 2255.6±25.4 63.0±2.7 4.4±1.3
S4 T5 101.9±1.6* 531.1±88.9 218.2±175.9
T6 0.0±0.0 1275.1±42.7 1369.0±1.3*
*Asterisked values are midrange±quartile deviation and others are mean±standard deviation for 10 iterations. (a) EB(sess) for sessions without S3/S4 1000
300 0
(b) EB(sess) for sessions with S3/S4
700
1000
1000
500
600
700
300
400
500
*
200
*
300
100
*
200
*
200
70
40
70 20
50
N=165 SC
30
GG
Citrix T CC
*
60
100
30
*
100
*
50
(c) EB(sess) for all sessions 200 0
200 0
20
N=32 SC
GG
Citrix T CC
N=197 SC
GG
Citrix T CC
Figure 1. Results of estimated bandwidth EB(sess) in kpbs grouped by three implementations (SC, GG and Citrix) for (a) sessions without S3 and S4, (b) sessions with S3 and S4 and (c) all sessions. A plot represents a session. *: P<0.05 by non-parametric multiple comparisons, Steel-Dwass testing
4. Limitations The authors regarded this study as a preliminary one, as the validity of the essential assumptions made in the study was not examined yet. The study assumed that all the EPR operations by users in Oshidori-net were completely recorded in the logs and that
104
K. Teramoto et al. / A Preliminary Study on Network Traffic Estimation
the actual bandwidth consumed by the operations properly approximated the bandwidth by four scenarios associated with six tasks. Both assumptions will be examined by interventional studies conducted in real or simulated settings such as a screen recording technique, as often adopted in usability testing [9]. The results of the study also depended on the EPR and TCC applications selected in the experiment. Consumed bandwidth varies depending on the mechanism how the EPR application interacts with the server to get patient data: some would get and cache almost all data required for the expected subsequent operations (corresponding to the EPR adopted in the experiment) and others would get them on demand. Generally, the former would be superior in the consumed bandwidth for the users who repeat similar operations on the EPR application. Also TCC applications have various mechanisms for caching and omitting data, which would affect the measured bandwidth. It is noted that a bias given by a combination of scenarios and applications makes it difficult to generalize the results of the study and apply them as-is to other environments at present. 5. Conclusion The results of the study indicated that Citrix as one of representative TCC applications required much less network bandwidth than the conventional SC model in using EPR systems over the network. However, the overall advantage of the TCC models over the SC model was not proved in the study: GG, the other TCC application adopted in the study, roughly consumed as much bandwidth as the SC model. Further studies will be focused on the verification of the adopted scenarios and a combination of applications that would affect the estimation of the network bandwidth.
References [1] [2] [3] [4]
[5] [6] [7] [8] [9]
D. Schlosser, K. Binzenhofer, et al.: Performance comparison of windows-based thin-client architectures, Telecommunication Networks and Applications Conference (2007), 197–202. S. Kuwata, K. Teramoto, et al.: Effective Solutions in Introducing Server-Based Computing into Hospital Information System, Stud Health Technol Inform 143 (2009), 435–440. K. Teramoto, Shigeki K, et al.: Development of a design model of Hospital Information System based on Server-Based Computing, Japan Journal of Medical Informatics 27 (suppl.) (2007), 666–668. GraphOn, Corporation. LEADING JAPANESE HOSPITAL IMPLEMENTS COST-EFFECTIVE SERVER-BASED COMPUTING SOLUTION [internet].2009 [cited 2010 Dec 11]. Available form: http://www.graphon.com/images/file/Tottori_Hosp_Customer_Story.pdf. Citrix Systems, Inc., The virtual desktop revolution is here … for everyone [internet].2010 [cited 2010 Dec 11]. Available form: http://www.citrix.com/English/ps2/products/feature.asp?contentID=2300341. S. Kuwata, K. Teramoto, et al.: Oshidori-Net: Connecting Regional EPR Systems to Achieve Secure Mutual Reference with Thin-Client Technology, E-health 2010 IFIP ACT 335 (2010), 82-89. H. Kimura, K. Nakahara, et al.: The operative experience of five years of the regional medicine cooperation system (Ajisai net). Japan Journal of Medical Informatics 29 (suppl.) (2010), 530-531. IBM Japan, CIS-MR (Medical Record) [internet] [cited 2010 Dec 11]. Available form: http://www06.ibm.com/industries/jp/healthcare/solution/cismr.html. A.W. Kushniruk, E.M. Borycki, et al.: Engineering Approaches to Evaluating the Usability of Health Information Systems. in Human, Social, and Organizational Aspects of Health Information System; eds. Andre W. Kushniruk, Elizabeth M. Borycki; Idea Group Inc (IGI), (2008), 1–22
Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers Ksenija LEKIĆa, Nuša KONEC JURIČIČa, Petra TRATNJEKa and Borut JEREB b 1 a Institute of Public Health Celje, Slovenia b University of Maribor, Faculty of Logistics, Slovenia
Abstract. This is Me is the largest youth counseling web portal in Slovenia, providing teens with a friendly, simple, fast, free, anonymous and efficient public access to expert information and problem-solving assistance. Network of web counselors includes 38 experts (medical specialists, psychologists, social pedagogues, social workers, and teachers), who are volunteers from 12 different institutions. In the course of ten years, experts have answered almost 21,000 questions about dilemmas and problems faced by teens. The program was created by the Institute of Public Health Celje. This is Me supports adolescents in their problem-solving efforts. The program focuses on development of positive mental health, with emphasis on self-image, social and life skills. The program responds to the adolescents needs and makes efficient use of web technology. Youth web portal is a response to the current teen lifestyle. Most users are between ages of 13 and 18. Keywords. Web counseling, youth, health promotion, mental health, preventive program, self-image
Introduction Adolescents often face problems such as lacking of social skills, self-confidence, selfrespect and optimism. Uncertainty in life is increasing. A young person will need many skills for life, amongst which are the ability to communicate, solve problems, adapt, organize and deal with defeats. [1] All these are the reasons because of which the Institute of Public Health Celje developed the health promotion programme called This is Me. [2] The central theme of the project is strengthening of self-esteem as the stronghold in the process of growing up. Aiming to find an effective pathway of communication with adolescents in the field of health promotion, and stemming from research results, the institute has undertaken ten years ago an innovative approach, based on web communication and counselling on-line. The youth web counselling in Slovenia is also supporting the well established fact, that organized care for self-image and social skills has an important preventive and curative role. Web counselling service for teenagers at http://www.tosemjaz.net was opened on the April the 7th, 2001. First asker was a 16-year-old girl who was in love head-over-heels. With the expert’s answer, published the next day, the Institute of Public Health Celje, successfully started web-based youth counseling program. The This is Me portal 1 Corresponding Author: Ph.D., University of Maribor, Faculty of Logistics, Mariborska 7, 3000 Celje, Slovenia; E-mail: [email protected]
106
K. Leki´c et al. / Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers
became a popular focal website for adolescents from the entire Slovenia. The program provides support to adolescents on their way to adulthood and helps them in problem solving. The biggest value of youth counseling website is the possibility of anonymously get counseling from experts. In the case of the program This is Me the web is upgrading classical forms of counseling. [3]
1. Web space as everyday support A lot of dialogues that develop in the safety of e-counseling website would never have happened in the office of a school physician, gynecologist, psychiatrist or psychologist. Many adolescents clearly say that they would never have gathered the courage to open up face-to-face, or would not even have sought help. The majority of questions deal with subjects that preoccupy teens the most: being in love, dating, curiosity about the first sexual experience, contraception, love problems, temporary family disputes. They are followed by more severe issues stemming from poor self-image. The most difficult and pressing questions the adolescents ask us touch the subject of suicidal thoughts, suicide attempts, self-harming behavior, depression, eating disorders, sexual abuse and teen pregnancy. About 5.5% of questions involve the most difficult subjects. [4]
Figure 1. Breakdown of question topics asked through the This is Me web counseling service.
Results and characteristics of the e-counseling service: x Network of web counselors includes 38 experts: 12 medical specialists, 15 psychologists, 11 social pedagogues, social workers, and teachers. All counselors are volunteers from 12 different institutions. x Over 2,500 answered questions and more than 100,000 users were registered in one year. x Every day the editor carefully edits e-content and manage e-contacts between experts and teens. x On average, the answers are published within three days. x Most users are between ages of 13 and 18.
K. Leki´c et al. / Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers
x x x
107
The e-counseling service is further augmented through preventive self-image development workshops in schools. The program is funded by the Institute of Public Health Celje with support from the Ministry of Health. The program has won five national and international awards.
2. Pros and Cons The main limitation that comes with web counseling is the absence of face-to-face contact i.e. lack of social and visual clues. The counselor doesn’t see the asker. The face is hidden, eye contact impossible, body language invisible, the personality and energy can only be guessed at. In fact, the communication is even more restricted than with phone conversations. Contact is reduced to written messages. Language is the sole communication channel. A message example: "Been cutting myself for a year. I want to stop." Such messages are sometimes the only thing the counselors have to work with. They will try to respond to such curt words (perhaps indicative of several problems) in a way that will encourage the user to speak up again. E-counselors are not a virtual shoulder to cry on. Serious problems are never easy to solve and the teens have to work hard. E-counselors carefully and suitably support and motivate the adolescents. They explain possible problem-solving scenarios and available sources of help. [5] Benefits of web counseling: x Anonymity (promotes openness while expressing problems). x Easy access to experts (from the safety of the teen’s room; no referrals, no waiting rooms). x One e-counselor can help several adolescents with a single answer. x Insight into others experience and problem solving approaches. x A quick first aid effect. x Networking of institutions, experts and users. Limitations of web counseling: x Limited communication (insight via written word only). x Limited possibilities for establishing a therapeutic relationship. x On-line relationships are insecure and can be broken off at any moment. x Questions X: insufficient asker info and incomplete problem description. x Adolescents’ abilities to put problems into words. x Unrealistic expectations (hoping for a quick fix through a few mouse clicks).
3. Critical point The most difficult questions, asked by the adolescents, bring to the light limitations of web counseling. Like in the next case of suicidal 17 years teen-age girl, who was writing: "Hey, what I wanna know is, which are the best pills you can use to kill yourself? does it hurt? will a pack of sleeping pills do the job? how much, so it’s all over 100%? been thinking about suicide a lot ... really hope you can give me some kind of answer, ’coz I can’t handle it anymore ..." The counselor’s answer was published the
108
K. Leki´c et al. / Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers
next day, the psychologist was encouraging and arousing her into actively solving the problem; teen-age girl responded, but the third reply of the counselor went unanswered. Unfortunately we don’t know what happened with this girl. This case illustrates the critical point of web counseling. Solving serious problems cannot be achieved by one or two clicks. The web counselor, for example, will not be able to resolve or do away with an adolescent’s cause which leads them to consider suicide; however, by offering understanding and prompt expert advice within the framework of the web intervention, the counselor may be able to prevent an adolescent’s suicide attempt in a given moment and direct them towards considering more constructive strategies of solving their problems. The following dialog illustrates an effectual case, when an e-counseling is being followed by a psychological counseling in person. Is it possible to stop... Author: Saphira Hello! I know you receive many e-mails concerning cutting, but still... I’d like to know if it is possible to stop? I’ve been cutting myself for a very long time and was caught twice doing it. I always said I will quit, I even went to a psychologist once. I can’t stop. I do this regularly and it’s the only thing I can rely on. Maybe it’s worthless for me to seek help cause there must be worse cases than me. Is it possible that I stop by myself? I don’t know what to do. Should I stop at all? I apologize for another e-mail on self harm. I’m everything that I hate. Editorial staff: counselor reply Yes, it is possible, but it is difficult, usually only with professional help over a longer period of time. Only few individuals decide to stop and stick to their decision. Very few can stop without help. You at least need someone who is on your side, who understands and doesn’t judge you, someone who stands by you. This can be a friend, relative or a professional. I advise you to go back to your psychologist or find another therapist. What is necessary is collaboration, a psychotherapeutic or counseling process over a longer period of time, half a year or more. Therefore: Get some help! Should you stop at all? Of course! There’s a strong survival instinct in you, otherwise you wouldn’t have turned to us for help. Unfortunately I do not know you well enough to give you some realistic reasons why it would be good to survive and stand on your own two feet. And if I start philosophizing now, you won’t take me serious anymore. If I’d believe that humans are only left to their destiny then I wouldn’t do what I do. But every day I get to see people who stand up on their own two feet, find happiness and start leading a live in a manner that they don’t regret that they’ve stayed alive. I therefore cheer also for you, don’t give up: get some help, it’s much easier when you’re not alone.
K. Leki´c et al. / Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers
109
Author: Saphira So... it is possible to stop. I don’t know how yet, but I think, I might succeed. I don’t want to go back to that psychologist. Can you recommend me some professional? I come from Maribor. Thank you. Editorial staff: counselor reply Hello again, Saphira! I am glad that you wrote again and that you’d like to know who you can turn to for help. I know Maribor very well and as you probably have noticed, I work here at the counseling centre for children, adolescents and parents. We are five psychologists, two women and three men, you need to call our number 02/... and make an appointment with one of us. If you don’t want to come to the counseling centre you can get a psychologist in a health care center, in a student outpatient clinic, in a private outpatient clinic... I wish you all the best, think about it and give us a call. 7 days later the editorial board has received a note, sent by e-counselor: "Saphira called me and made an appointment." This case lights up the positive side of web counseling. The Web provides adolescents with an encouraging supportive environment of information and communications: "I understand you and I feel your distress. I can advise and help you." However, web counseling cannot replace personal counseling.
4. Conclusion and useful insights Modern technologies open up new possibilities for traditional counseling services. [6] The This is Me portal is one possible response to the current teen lifestyle. After ten years of web communication we can confirm [7]: 1. The Web is indeed a useful counseling channel and problem-solving aid. The e-counseling service provides easy access to expert advice during moments of difficulty or everyday dilemmas. The web counseling site is as a safe virtual space where adolescents can articulate the most difficult of issues anonymously. By posting a single reply the web counselor helps many teenagers with similar problems. 2. Through web communication and counseling we can reach typical adolescents with everyday problems. Before the new communication technologies became widely available, the insight into this particular group was not as good. 3. The Web truly connects people, organizations, and sources of help. The web social network provides us with information, answers and increases the number of problem-solving options. 4. E-counseling cannot replace face-to-face counseling and assistance in cases of more severe issues. 5. The role of e-counseling is above all preventive. E-counselors try to guide the problem solver. They use various information to help adolescents understand their situation, help them create problem-solving strategies and help them take
110
K. Leki´c et al. / Slovenian Practice Story: 10 Years of e-Counselling Service for Teenagers
6.
7.
responsibility for their lives. Whenever the counselors succeed in guiding the teens towards making positive changes in their lives, e-counseling has been a success. The communication context of the website makes adolescents feel well and secure. For young web users there are no referral forms, waiting periods and queues. The youth web counseling site is an additional, auxiliary form of support tailored to the generation of adolescents currently growing up. It is characterized by positive synergic effects of various supportive communications and preventive approaches in the field of health promotion that are available to Slovenian adolescents. The web can, if used creatively in conjunction with strong counseling network of experts, take up a role of preventive co-factor (co-influencer) of health improvement. The web can be creatively used as an interactive and practical counseling tool. In the case of program This is Me, the web communication has shown itself to be an efficient tool in the field of health promotion and strengthening of self-image as well as preventive informing of young people.
References [1] [2]
[3]
[4]
[5]
[6] [7]
M. Ule, M. Kuhar, Sodobna mladina: izziv sprememb. V: V. Miheljak (ur.), Mladina 2000. Slovenska mladina na prehodu v novo tisočletje. Ljubljana: MŠZŠ – Urad RS za mladino, Aristej, (2002), 39-77. D. Podkrajšek, K. Lekić, T. Kopač Vidmar, To sem jaz. In to potrebujem. Rezultati anketiranja mladih celjske regije o problemih odraščanja. V: Mladostnik in zdravje. Tretji kongres šolske in visokošolske medicine Slovenije. Zdravstveno varstvo (2001), 223-228. K. Lekić, N. Konec-Juričič, P. Šafran, B. Jereb, Web counselling as everyday support for teens. V: Kramer, Jeannette (ur.), Man, John-Gavin (ur.), Wammes, Anke (ur.). First International e-Mental Health Summit 2009, Amsterdam. Abstract book. Amsterdam: Trimbos Institute: University of Amsterdam: ISRII: VU University Amsterdam, (2009), 182. K. Lekić, N. Konec-Juričič, A. Tacol, P, Šafran, Primer preventivne prakse: mladinski program To sem jaz. V: S. Gaber (ur.), M. Sardoč (ur.), J. Strel (ur.), A. Lukšič (ur.) Za manj negotovosti: aktivno državljanstvo, zdrav življenjski slog, varovanje okolja. V Ljubljani: Pedagoška fakulteta, (2009), 189200. K. Lekić, N. Konec-Juričič, P. Tratnjek, Najstnik v spletni svetovalnici. V: J. Govc Eržen (ur.), Z. Klemenc-Ketiš, K. Tušek-Bunc, Otroci in mladostniki učno gradivo - monografija, (Poglavja iz družinske medicine). Ljubljana: Zavod za razvoj družinske medicine, (2010), 59-88. M. Cerar, Internet – Virtualno shajališče mladih: primerjave socialnih interakcij v realnem in virtualnem okolju. Diplomsko delo. Ljubljana: Pedagoška Fakulteta; (2006), 60-69. K. Lekić, N. Konec Juričič, P. Šafran, B. Jereb, Web counselling for e-teenagers. V: G. Buijs (ur.). Third European Conference on Health Promoting Schools, held in Vilnius, Lithuania, 15-17 June 2009. Better schools through health: learning from practice: case studies of practice presented during the third European Conference on Health Promoting Schools, held in Vilnius, Lithuania, 15-17 June 2009. [Woerden]: Netherlands Institute for Health Promotion; [Vilnius]: State Environmental Health Centre, (2009), 43-50.
E-Learning for Occupational Physicians’ CME: a Study Case M. Cristina MAZZOLENIa,1, Carla ROGNONIb, Enrico FINOZZIa, Tommaso GRI b, Marco PAGANIc and Marcello IMBRIANIa a Salvatore Maugeri Foundation IRCCS, Pavia, Italy b University of Pavia, Dept. of Computer Science and Systems, Pavia, Italy c CBIM – Consorzio di Bioingegneria e Informatica Medica, Pavia, Italy
Abstract. The present study reports the results of the evaluation of an e-learning CME course in the field of Occupational Medicine. In particular the following aspects have been investigated: If and how the course contents have met the educational users’ needs; The effectiveness of the course in terms of knowledge improvement; Users’ behaviour. Attendance data and results of a sample of 1128 attendees have been analyzed via ad hoc developed tools for direct inspection of Moodle CMS database. The results document the effectiveness of the e-learning course, as regards meeting the educational needs of physicians and also the improvement in terms of knowledge and problem solving skill acquisition. Users’ behaviour has revealed a certain tendency for passing the tests, more than for pursuing the best possible result. Interaction with the tutor is low. Keywords. E-learning, CME, evaluation, occupational medicine
Introduction Computer based learning is as effective as typical lecture based teaching sessions for educating postgraduates [1]. Continuous medical education (CME) via e-learning is taking ground worldwide, and the same is happening in Italy. In particular, after an experimental phase that has lasted some years, the Italian Ministry of Health has recently developed a normative framework for CME that takes into account distance learning and also promotes e-learning because of its recognized potentialities. Salvatore Maugeri Foundation has developed and provided eleven e-learning courses in the field of Occupational Medicine [2] during the just concluded experimental phase. More than 15.000 users have enrolled in the e-learning platform. A pretty large amount of information is hence available now to evaluate the pilot experience. Aim of the present study is to get some hints for the development of new elearning CME courses from the analysis of the attendance data of the just concluded experience. In particular the following aspects have been investigated: if and how the course contents have met the educational users’ needs; the effectiveness of the course in terms of knowledge improvement; users’ behaviour. The present study focuses on one of the provided CME e-learning courses that 1 Corresponding Author: M. Cristina Mazzoleni, Fondazione Maugeri IRCCS, Via Maugeri 4, 27100 Pavia, Italy; Email: [email protected]
112
M.C. Mazzoleni et al. / E-Learning for Occupational Physicians’ CME: A Study Case
deals with the pathologies related to the exposition to mechanical vibrations in occupational environment (MVC). The topic is pretty relevant since mechanical vibrations are a risk factor for 25% of European workers.
1. Background: the e-Learning Course The implemented e-learning system was based on Moodle CMS (www.moodle.org) and has been available free of charge to the users for 34 months. MVC is structured according to the following educational model providing: x An initial test (IT) (mandatory, one attempt allowed) composed by 25 questions with single and multiple choice answers x Three self-learning modules for free navigation (including PDF printable versions) related to anatomy, epidemiology, physiopathology, symptoms, diagnosis, prevention, regulatory and legal aspects x Two hyper-flowcharts of guideline-based decision processes (GL1 and GL2) x Two case-based exercise/tests (CBT1 and CBT2) based on the hyperflowcharts (4 attempts allowed each, minimum score to pass the test: 75%) x One final test (FT) (4 attempts allowed, minimum score to pass the test: 75%) composed by the same 25 questions of IT. A tutor was available for asynchronous communication. To obtain CME credits, the user has to reach a positive result in FT and in each case-based exercise/test.
2. Methods Moodle provides a certain number of functionalities to analyze the learning process of a single user, or to evaluate the global outcome of a course taking into account the entire population of enrolled students. These functionalities don’t allow detailed analysis such as those based on the stratification of students or events according to our defined criteria (for example: student characteristics, test results, number of attempts). Starting from the analysis of the PostgreSQL database connected with Moodle, a set of ad hoc queries has been developed to retrieve data fitting our inspection goals. 2.1. Users’ Needs versus Provided Contents The design of a course starts from the identification of users’ need. Direct inspection of the needs through surveys is possible but quite time consuming, and it points out perceived users’ needs only. Frequently, and this is also the case of the course presented here, the need analysis is performed through group discussion among experts of the specific field. Based on their experience, the experts define the contents of the course and the educational model. To verify to what extent the course contents fit users’ needs, an analysis of the initial knowledge, and consequently of the gaps, of the addressees of the course has been performed through the inspection of IT results. IT scores (range 0-100%) have been analyzed, both as global result, and in relation to the various topics covered by the course itself. At this last purpose, each question of the IT has been attributed to one of the three following areas depending on the subject covered:
M.C. Mazzoleni et al. / E-Learning for Occupational Physicians’ CME: A Study Case
113
Area A): anatomy, epidemiology, physiopathology (9 questions) Area B): symptoms, diagnosis, therapy (7 questions) Area C): surveillance, regulatory and legal aspects (9 questions). The distribution of correct answers for area has been calculated stratifying by initial knowledge level. At this purpose, the sample has been divided according to the score reached in IT. Four groups have been considered: “0-25” (0≤score≤25), “25-50” (25<score≤50), “50-75” (50<score≤75), “75-100” (75<score≤100) group. 2.2. Knowledge Acquisition As indicator of knowledge acquisition, the comparison between IT and FT results has been used, taking into account both the various areas (A, B, C) and the initial knowledge degrees. In addition, the results of the two case-based exercises have been analyzed in terms of mean scores per initial knowledge level. The structure of the case-based exercises doesn’t allow an analysis in terms of different subject areas. Since the FT and the casebased exercises allow multiple attempts, the best results have been considered. 2.3. Users’ Behaviour A not so small but anyway limited number of attempts (four) for FT and case-based exercises has been allowed in order to induce a progressive improvement of the results up to the best possible. The number of attempts for FT, CBT1 and CBT2 has been calculated, stratified by initial knowledge degree. As regards the usage of course resources, the number of events “connection to the module/test” and “accessed PDF version” have been considered. Moodle native functionality has been used to retrieve the total number of events. Messages exchanged between users and the tutor have been inspected.
3. Results and Discussion Out of 1128 physicians who have enrolled in the course, 1116 have passed FT, 6 have failed and 6 haven’t concluded it. The following results refer to those physicians who have passed FT. 3.1. Users’ Needs versus Provided Contents Figure 1 shows the distribution of IT global scores of 1116 physicians (FT score is considered here just for convenience, in relation to the next point). 74% of the sample have an insufficient knowledge, and in particular 44% of the attendees were able to correctly answer less than 50% of the questions. These data document an educational need in the topic addressed by the course. The mean scores of IT for area A, B and C are respectively 50%, 66% and 54%. Figure 2 a) shows the distribution of correct answers of IT for each area, stratifying the sample by the four IT groups. In all but the last group, knowledge gaps for area A and C are particularly evident (maximum mean score 60%).
114
M.C. Mazzoleni et al. / E-Learning for Occupational Physicians’ CME: A Study Case
Figure 1. Distribution of IT and FT results
Figure 2. Correct answers per area and IT group: a) initial test; b) final test
3.2. Knowledge Acquisition Figure 1 shows the comparison between IT and FT in terms of global score. 84% of the users have got a final score in the range 80-100%. Wilcoxon test has been applied to compare the results for each question of the IT with the corresponding question in the FT. All the comparisons were statistically significant (p<0.0001) showing a gain in terms of knowledge. As shown in Figure 2 b), for all the groups and all the areas the percentage of correct answers has, on average, reached and crossed the threshold of 75%. This result supports the effectiveness of the course not only in global terms, but also in relation to each area, and in particular for area A and C where the initial knowledge gap was more evident. The analysis of the percentages of correct answers for each question, whose numeric results are not reported in the present paper, has allowed the identification of a probably not sufficiently clear point in Area C module. First and second case-based test mean scores reported in Table 1 show pretty good results also as regards problem solving skills. The percentage of failures for CBT1 is around 3% (36/1116), higher than the one of CBT2 (2/1116, 0.2%) and of FT (6/1116, 0.5%). These values are anyway very low. 3.3. Users’ Behaviour The number of performed attempts is an indicator of both the effort to achieve a positive result and the tendency for increasing the acquired knowledge level. Table 1 shows, for each test and for each IT group, the score mean values and the
M.C. Mazzoleni et al. / E-Learning for Occupational Physicians’ CME: A Study Case
115
number of performed attempts. From a qualitative point of view, far from giving statistical meaning to the following considerations, the reported mean values support the conclusion “the lower the initial knowledge, the lower the result, the higher the number of attempts and failures”. Anyway the differences are quite small and the number of performed attempts is pretty low. Table 1. Distribution of mean score and mean number of attempts per IT group
IT Group 0-25 25-50 50-75 75-100
#
Mean score
138 346 389 243
88.9% 88.2% 89% 94.3%
FT Mean # of attempts 1.73 1.71 1.48 1.12
Mean score 86.5% 86.1% 88.7% 94.1%
CBT1 Mean # of attempts 1.95 2.3 2.1 1.4
# of failures
Mean score
6.5% 4.6% 2% 1.6%
91.2 % 89.7% 91.2% 94.3%
CBT2 Mean # of attempts 1.6 1.6 1.5 1.2
# of failures 0.6% -
For CBT1, the distribution (Figure 3) of the number of attempts has been calculated versus the different ranges of the score, for those attendees that have passed the case-based test. Most of the users have not fully exploited the still available attempts to improve their result and increase the acquired knowledge. Similar behaviour has been found also in [3]. Reaching a positive result, even if suboptimal but anyway allowing CME credits acquisition, seems to be the main goal.
Figure 3. Distribution of the number of attempts per various result ranges in CBT1
Table 2 shows the results provided by Moodle functionality about the number of connections to the modules of the course, tests included. Statistics refer to all the people enrolled in the course, not only to 1116 physicians that have passed FT. In general, we would have expected a more intense usage of the online modules. Anyway, the data provided by Moodle don’t allow to distinguish users’ behaviour in terms of online module navigation AND/OR usage of the PDF version. The results show that the number of attempts do not coincide with the number of connections to the tests. For example: CBT1 has around 2 mean numbers of attempts, while the mean number of connections is 6.53. This means that the users have suspended the test, maybe to revise the theory module, and reconnected to conclude the task. At present we haven’t developed yet a tool to stratify the users in terms of number of connections, attempts and reached result. A better understanding of users’ behaviour might lead to a revision of the observation about the attitude of the users to stop “studying” when a positive but sub-optimal result is reached. As to the Area C module, the quite low number of accesses could explain the fact that area C is characterized by the lowest mean FT score (see Figure 2a).
116
M.C. Mazzoleni et al. / E-Learning for Occupational Physicians’ CME: A Study Case
Table 2. Mean number of accesses to each module/test. In brackets the mean number of accesses of the printable version of the module Area A module (PDF) 1.12 (0.73)
Area B module (PDF) 1.3 (0.77)
Area C module (PDF) 0.93 (0.63)
GL1
GL2
CBT1
CBT2
FT
1
0.82
6.53
4.35
3.35
As to the communication with the tutor, only 24 messages have been sent. No question has been sent about the subject that was probably unclear in Area C module. Users reveal a very low attitude to exploit the potentialities of the interaction with the tutor provided by the e-learning platform.
4. Conclusions The results of the performed evaluation document the effectiveness of the e-learning course, as regards meeting the educational needs of physicians and also the improvement in terms of knowledge and problem solving skill acquisition. Users’ behaviour has revealed a certain tendency for passing the tests, more than for pursuing the best possible result. On this issue further studies are advisable. As far as our sample could be representative, the interaction with a tutor offered by e-learning systems (and today required also by the national regulation about CME distance learning) is still far from being appreciated and exploited by the Italian physicians. This kind of inspections can be useful to identify eventual weaknesses of a course, in the perspective of continuous quality improvement of the educational offer. From the technical point of view, the ad hoc developed tools have proved to be useful and necessary in order to analyze data with a non-basic granularity level and specific stratification needs. The usage of open source software has allowed, even if not without a certain effort, the inspections presented here, otherwise nearly impossible.
Acknowledgements We remember and thank Prof. Giovanni Catenacci for his precious support in designing the course contents.
References [1] J. Davis, E. Chryssafidou, J. Zamora, D. Davies, K. Khan, A. Coomarasamy, Computer-based teaching is as good as face to face lecture-based teaching of evidence based medicine: a randomised controlled trial, BMC Med Educ (2007), 7:23. [2] M.C. Mazzoleni, C. Rognoni, E. Finozzi, E, I. Giorgi, F. Pugliese, M. Pagani, M. Imbriani, Development of an e-learning system for occupational medicine: usability issues, Stud Health Technol Inform 136 (2008), 579-84. [3] M.C. Mazzoleni, C. Rognoni, E. Finozzi, M. Landro, E. Capodaglio, M. Imbriani, I. Giorgi, Earnings in e-learning: knowledge, CME credits or both? Hints from analysis of attendance dynamics and users' behavior, Stud Health Technol Inform Vol. 160, 576-80, IOS Press, Amsterdam, 2010.
Regional Medical Data Mining System Raul ROBU1 and Vasile STOICU-TIVADAR Politehnica University from Timisoara, Romania
Abstract. This paper suggests a solution to acquire medical data from hospitals located in a region (addressing especially the DKMT Euroregion), and then perform data mining. The medical data from the hospital databases are exported in XML format, according to HL7 CDA standard. Afterwards, they are automatically centralized on a server in a database using web services calls. The data will be analyzed with the data mining tool WEKA. Data of interest are converted into ARFF format and loaded into WEKA. The next stage consists in preprocessing and analyzing the data based on the algorithms provided by WEKA, having as a goal several relevant medical conclusions. WEKA application interface has been improved to facilitate the process of performing predictions. Keywords. medical data mining, HL7 CDA, WEKA
Introduction Data mining techniques are successfully applied in the medical field in order to build a profile for a patient that suffers from a certain disease, in order to improve diagnosis [1], to determine the adverse drug events [2], to ameliorate the general treatment of patients. Data analysis from different regions allows the determination of specific illnesses, finding out their causes and acting in those demographic areas in order to reduce or eliminate the causes. Centralizing data from a vast geographic area in a single medical database offers a substantial source of information, the analysis of which, will lead to relevant medical results. Analyzing data coming from different hospitals is often difficult because hospitals may use different information systems so the structure of the databases can be different from one medical institution to another, even if they reside in the same town. This problem can be solved using the HL7 CDA (Clinical Document Architecture) standard that permits the export of data from the databases of the hospitals in a well established XML format. Data could be obtained in a unitary format from hospitals even located in different countries and they can be centralized and analyzed through some specialized data mining instruments. The solution suggested by the authors in this paper is to use the data mining instrument and machine learning WEKA, in order to extract the knowledge from the medical data obtained from the hospitals inside a geographic region. The authors intend to develop a pilot system in the DKMT (Danube-CrisMures-Tisa) Euroregion which will view the hospitals in Timisoara, Arad, Szeged and Novi Sad.
1
Corresponding Author: Raul Robu, Faculty of Automation and Computers, Bvd. Vasile Parvan, No. 2, 300223, Timisoara, Romania; E-mail: [email protected]
118
R. Robu and V. Stoicu-Tivadar / Regional Medical Data Mining System
WEKA is useful software for education, research and applications. WEKA version 3.6.3 offers 117 classification and regression algorithms, 11 clustering algorithms, 6 algorithms for finding association rules, 3 graphical interfaces: the Explorer, the Experimenter and the KnowledgeFlow [3]. The data files associated with WEKA are ARFF files (Attribute-Relation File Format). Furthermore, data can be loaded directly from databases in WEKA using, for example, the Java Database Connectivity (JDBC) technology. In this case, several conditions must be fulfilled and certain specialized actions must be carried out [4]. The authors consider that the described process is relatively complex so they suggest the use of the ARFF Convertor [4] application in order to load data from the databases in WEKA. The ARFF Convertor connects to the database, displays the structure of the database (tables, fields, type of fields) and its content (see Figure 1, a). The relations between tables are also displayed. The user selects the fields of interest for the data mining study by making a double click on them, even if eventually they are located in different related tables (see Figure 1,b). During the next stage, the ARFF Convertor analyzes the content of the selected fields, chooses the desired type for the fields from ARFF and then generates the ARFF file.
Figure 1a. ARFF Convertor: main user interface
Figure 1b. ARFF Convertor: tables relationships
For an application to access the database on the server, the database has to be accessible from the exterior. If, for security reasons, this condition cannot be accomplished, the appropriate solution is to use the extended ARFF Convertor, which invokes a web service through which data are taken from the server in XML format and are then converted into ARFF format. Based on the data loaded in WEKA, predictive or descriptive models can be built. The suggested system insures access to a substantial volume of data which strengthens the relevance of studies. In WEKA, building and testing a predictive classifier model is very simple, but using it to perform predictions can be realized only indirectly, testing the built model on the dataset on which the predictions would be realized. Creating an ARFF file and adding the instance whose class would be predicted to that file is necessary. The process is not intuitive, that is why the interface of WEKA was modified (WEKA is a Java Open Source) so that predictions could be easily achieved [5].
1. Data Mining System at a Regional Level The suggested system uses at the level of each hospital (see Figure 2) an application (HL7 Connector) able to convert data from the databases of the information systems from the hospitals, in XML format, according to HL7 CDA standard [6]. The conversion in this format is necessary because the databases from hospitals may have very different structures, considering the used information system (Microsoft Amalga
R. Robu and V. Stoicu-Tivadar / Regional Medical Data Mining System
119
Hospital Information System, Hipocrate, Hospital Manager a.s.o. [7]). The data from the XML files are centralized in a database on the server. An XML file in HL7 CDA format includes demographics, active problems, past medical history, surgical history, medications, allergies and adverse reactions, family history, immunizations, social history, labs, diagnostics, etc. HL7 Connector performs the following functions x Downloading the template (i.e. the structure description) of the desired data by invoking a function of the web service x collecting the desired data (according to the template) from the hospital’s database x Generating the XML (HL7 CDA) file x Uploading the XML file to the Server by invoking a function of the web service. All of the previously mentioned functions are executed periodically either automatically or manually by the personnel in charge with the administration of the hospital information system, who launches the application whenever is requested.
Figure 2 - Regional data mining system
The data centralized on the server must be analyzed with the WEKA tool. Because data files associated with WEKA are ARFF files (Attribute-Relation File Format) a conversion of the data from the server’s database in ARFF format is necessary. If the access to the database is allowed through an open port (see A section of Figure 2), the ARFF file can be achieved connecting the ARFF Convertor directly to the database. If, for security reasons, the ports toward the database are closed (see B section of Figure 2), then the extended ARFF Convertor can be used, which downloads through a web service a XML file that contains the data from the database. The extended ARFF Convertor contains a module that performs the conversion of data from XML in ARFF. The programming language used for developing HL7 Connector and Arff Convertor is C#, and the programming environment is Visual Studio.NET 2008. If the access to the database cannot be done directly, the sequential diagram is the one presented in Figure 3 (in order to simplify, the download part of the data template was not introduced). First HL7 Connector automatically accesses, at certain time intervals or at the initiative of a user, the Hospital Information System, in order to read the desired data from the data managed by the system. These data are then converted
120
R. Robu and V. Stoicu-Tivadar / Regional Medical Data Mining System
into HL7 CDA format and sent through the web service. The web service parses the data from the XML files and centralizes them in a database on the server. The ARFF Convertor calls the web service, which sends a SQL command to the database, command through which the data is read and then loaded in a DataSet. Using a TextWriter type object and the GetXml() method of the DataSet the XML file is created and sent to the extended ARFF Convertor. The application loads the data from the XML file in a DataSet type object with the ReadXml method. Next, all the DataTables from the DataSet are explored and their names are added in the TreeView control. Inside each DataTable all the DataColumns are scanned and their names are added in the TreeView control. When the AfterSelect event and the TreeView control are triggered, the content of the selected table is displayed in the DataGridView control, from the bottom part.
Figure 3 - Sequence diagram for indirect access to database of extended ARFF Convertor
After converting the content of the XML file into a table form (by loading it in a DataSet) (see Figure 4) the fields of interest for the ARFF file may be selected by simply double clicking on the name of these fields. The content of the fields will be analyzed and if it repeats itself, the nominal type will be suggested for these fields.
Figure 4 - Extended ARFF Convertor
For the instance of direct access, the sequence diagram is similar, only that the ARFF Convertor connects directly to the database, through the already existing facilities in the application [4]. The list of functions in the web service defined for the suggested system is presented in the followings: x Registration of new partner x Notification call for the acceptance of the partner
R. Robu and V. Stoicu-Tivadar / Regional Medical Data Mining System
121
x x
Download of the necessary data template Loading data extracted from the database of a hospital (with a temporal filter), with the conversion in HL7 CDA and saving them in the database on the server x Download of data from the central database, in order to be used by WEKA application. The web service is currently being developed, using Visual Studio.NET 2008 and C# language. For the exploitation of the system, a desktop application for the administration, and definition of the data templates is conceived to run on the server, using an interface similar to the one utilized by ARFF Convertor. This application can be extended with specific modules for filtering and preprocessing other than the ones contained by WEKA application. Building a classifier model can be performed very easily in WEKA. The data source is chosen (ARFF file), the data are preprocessed (through techniques such as the analysis of the main components, factorial analysis, etc.), the algorithm is chosen (Naïve Bayes, C4.5, k-Nearest Neighbor, etc), the set of data is divided into the training set and test set. If the built and tested model has satisfying results, it can be used for predictions. The process of achieving predictions has been improved by modifying the Classify interface of the Explorer (see [5]), transforming it into a dynamic interface that depends on the analyzed data set.
2. Scenarios WEKA can be used to build a regression model that enables the determination of the values of medical parameters based on others (for example expressing the resistance index of the respiratory muscle PEmax according to the predictive variables represented by height, weight, age, sex, body mass percent, forced respiratory volume per second, residual volume, residual functional capacity and total capacity of the lung for a part of the patients ill with cystic fibrosis [8]). With WEKA’s aid a classifier model, capable to predict the risk of premature birth can be built [9]. The classification algorithms can be used for determining if a solitary pulmonary nodule (SPN) is benign or malignant, using information from noninvasive tests. The diagnosis is perceived to depend on many features, such as the SPN diameter, border character, presence of calcification, patient’s age, smoking history, results of CT densitometry, and overall prevalence of malignancy within the population [10]. The suggested system allows a lot of statistic and data mining studies to be made, based on a substantial volume of data. The number of people who are sick with a certain disease for each county of the region (DKMT, for example) can be determined. If a certain anomaly is determined (a lot more sick people with a disease in a county than in others, related to the number of inhabitants) supplementary investigation can be made and actions to prevent and reduce the number of diseased people can be taken. The authors suggest performing a study regarding the gastro-duodenal ulcer, for the DKMT region, by using the suggested infrastructure. This study will entail: x The analysis of the number of people affected by all types of ulcer at county level and related to the number of inhabitants x The analysis of the incidence of the disease based on age and social categories x The analysis of the manner in which factors such as nutrition, stress, medication influence the apparition and evolution of the two types of ulcer (gastric and duodenal)
122
R. Robu and V. Stoicu-Tivadar / Regional Medical Data Mining System
x
Building the profile of the patient that presents an increased risk to become sick with ulcer The study will contain the following steps: x Defining the structure of the necessary data, selecting the sample of interest x Gathering data from the hospitals in Timisoara, Arad, Szeged, Novi Sad x Performing with WEKA’s help the statistic analysis on each of the attributes of interest x Determining the most appropriate data mining algorithms for achieving the studies x Performing the studies and drawing the relevant medical conclusions x Disseminating the results by publishing them at the highest possible level This action plan will constitute the basis for a regional project.
3. Conclusions This paper suggests a solution to build a database of considerable dimensions with the data acquired from the hospital information systems from a region. The data from the information systems of the hospitals are first converted into XML format, according to HL7 CDA standard and they are then saved in a database through web service calls. The centralized data are then transformed into ARFF format, and analyzed with the data mining program WEKA, in order to extract new knowledge. The interface of WEKA was modified to allow an easy prediction. The ARFF Convertor application was extended with a module that allows the transformation from XML into ARFF. Hereinafter, the authors pursue the realization of a study regarding gastro-duodenal ulcer in the DKMT region, using the suggested system.
References [1]
Y. Atilgan and F. Dogan, Data Mining on Distributed Medical Databases: Recent Trends and Future Directions. IT Revolutions (2009), Volume 11, 216-224. [2] E. Chazard, C. Preda, B. Merlin, G. Ficheur, The Psip Consortium and R. Beuscart, Data-Mining-Based Detection of Adverse Drug Events, Medical Informatics in a United and Healthy Europe (2009), 552556. [3] K. Hornik, C. Buchta and A. Zeileis, Open-Source Machine Learning: R Meets Weka, Computational Statistics (2009), 225-232. [4] R. Robu and V. Stoicu-Tivadar, Arff convertor tool for WEKA data mining software, Proceedings of the IEEE International Joint Conferences on Computational Cybernetics and Technical Informatics (2010), 247 – 251. [5] R. Robu, C. Hora and V. Stoicu-Tivadar, Improving the Classify User Interface in Weka Explorer, Proceedings of the 21st International DAAAM Symposium, Intelligent Manufacturing & Automation: Focus on Interdisciplinary Solutions (2010), 171-172. [6] P. Schloeffel, T. Beale, G. Hayworth, S. Heard, and H. Leslie,. The relationship between CEN 13606, HL7 and openEHR, HIC 2006 and HINZ 2006:Proceedings (2006), 24-28 [7] Hospital Manager Suite, Info World Company, 2010, Available at: http://www.infoworld.ro/dnn/Default.aspx?tabid=56 [8] F. Gorunescu, Data Mining, Concepte, Modele si Tehnici, Editura Albastra, Cluj-Napoca, 2006. [9] D. Zaharie, S. Holban, D. Lungeanu and Dan Navolan, A Computational Intelligence Approach for Ranking Risk Factors in Preterm Birth, Proceedings of the 4th International Symposium on Applied Computational Intelligence and Informatics (2007), 135-140. [10] A. Kusiak, J. Kern, K. Kernstine and B.Tseng, Autonomous Decision-Making: A Data Mining Approach, IEEE Transactions on Information Technology in Biomedicine (2000), 274-284.
Assessment of Hospital Information Systems Implementation: a Case Study a
Dimitrios ZIKOS 1, Athanasios MITSIOS and John MANTAS Laboratory of Health Informatics, faculty of Nursing, National and Kapodistrian University of Athens, Athens, Greece
Abstract. The use of integrated Hospital Information Systems is related with many benefits for the healthcare system, increasing the effectiveness of the provided services and assuring quality of care. Aim of this study is to investigate the types of Information Systems, the level of integration and the problems identified during the implementation phase, in three public hospitals. The above are expected to contribute to the understanding of the organizational, human resource and technical factors related with the successful implementation of a hospital IS. In order to investigate those elements, an assessment questionnaire was developed and completed by nine hospitals IT employees of the three hospitals. In addition, open interviews were organized with the same employees to further formulate an overall aspect, while in one hospital case, observation and discussion with four different categories of involved staff was undertaken. It was found that the implementation problems are mainly related with the underfunding, inadequate use of standards, lack of skilled IT experts, insufficiently trained personnel and users’ reserve. The problems may be tackled with a supportive hospital administration committed to the successful implementation. The external contracting company working on its own, without any participation of the hospital IT department seems to be a failure recipe. It is evident that an active management support and skillful hospital IT employees, are expected to result to success stories during the implementation of integrated hospital information systems.
Keywords. Hospital Information System, Implementation, Integration, Barriers
Introduction The introduction of Information Systems (IS) in large hospitals drastically changed the management of patient care. Increased computer speed and processing power, local area networks and use of standards have led to many possibilities in terms of access and availability of information for health professionals. However, although integration of hospital information systems lead towards a more efficient use of resources, therefore increasing the effectiveness of the provided services, the improvement of quality of care and the rise in productivity has not been drastically increased in many countries [1]. Lack of a strategic plan, organizational issues, incompetent IT departments and bureaucracy appear to be bottlenecks towards the successful introduction of IT systems far more than any, technical and financial barriers [2]. The 1
Corresponding Author: Dimitrios Zikos, Health Informatics Laboratory, Faculty of Nursing, University of Athens, 123 Papadiamantopoulou str. 11527 Athens, Greece; E-mail: [email protected].
124
D. Zikos et al. / Assessment of Hospital Information Systems Implementation: A Case Study
workflow process within the hospital should be thoroughly examined, in order to identify factors related with the impending the successful introduction of IS [3].
1. Scope Aim of study is to make a qualitative assessment of the implementation process of IS and of the integration level in the case of three hospitals. The assessment focuses on problematic IS implementation in the selected hospitals, pinpointing to the nature of the problems. Design, implementation and level of integration have been investigated using of a questionnaire, based on evidence that factors leading to the success or failure of a hospital IS are not only technical, but also related to the existing organizational models, education, managerial and evaluation issues.
2. Material and Methods The case study was performed in two mid-sized general hospitals and one oncology hospital. Hospital A has an integrated administrative, financial and clinical IS. Hospital B is in process of getting its own Integrated IS, while Hospital C is supported by nonhomogeneous, subsystems, lacking a clinical section. The qualitative assessment included employee perceptions analysis, in two phases. Firstly, a literature research has been performed in order identify critical items with regard to the design and implementation of IS. An Assessment Questionnaire including close-ended, selfexploratory items has been developed and distributed to nine IT department employees of the three hospitals. The second phase included open interviews, and a further overview of the IS implementation process, via daily observation and discussion with four types of employees, in one of the three hospitals. The assessment questionnaire includes questions on human resources and IT services within the hospitals and captures the current status regarding the existence of integrated administrative, clinical IS [4], [5], [6], [7], LIS and Pharmacy IS subsections [6]. Next section is about the system specifications, and the use of coding systems to achieve integration [8]. The role of the hospital management in terms of planning and financing and the existence of contracts with developer [9] are also investigated. The questionnaire also asks about education and training of the IS users [8], user support during the implementation, and also investigates the existence of user motivators and external consultants [10], also aiming to outline whether there have been contracted concrete task agreements between the supplier and the hospital management to guarantee the efficient and constant system performance [8], [11]. Next section captures the hospital workflow, possible changes caused by the newly installed system and whether the system would be altered and parameterized according to the workflow needs [12]. It is evident that issues emerging from workflow changes should be overcome relatively easily [13]. Data migration from previous to the new systems was also investigated [14]. There are also questions about coding standards and users’ reluctance to use the system. Also included are questions about the central planning and specialized IT staff [11], [15], about the fulfillment of a needs analysis, exploratory, feasibility and risk analysis study prior to implementation [8]. Finally the questionnaire explores whether the IS have been evaluated in terms of patient satisfaction, employee satisfaction and cost-productivity [16].
125
D. Zikos et al. / Assessment of Hospital Information Systems Implementation: A Case Study
3. Results 3.1. Phase 1: The Assessment Questionnaire Regarding the technical specifications, it has been found that hospital A satisfies all requirements of the assessment questionnaire, while hospital C IS lags behind. Hospital A also has a coding system for the materials, using a unique combination which includes the GMDN nomenclature. On the contrary, other than the Patient ID, hospital C makes no use of any other standard. All three hospitals have a project implementation team constituted by members of the administration, the developer, the IT department, medical-nursing staff. Hospital B also recruits an external consultant. Regarding the role of hospital management, its active participation in the case of hospital A, during the IS implementation, the utilization of key employees to support the system and the adoption of a long term plan have been found to be success factors. This was not the case for hospitals B and C. The fundamental need for user training, education and support provided by the IT staff, is considered as a high priority only in the case of hospital A and partially in hospital B. Of the three cases, hospital A has been found to provide adequate user support and training during the implementation phase. With an exception of case A, the other two hospitals did not contract a full task agreement with the developer to commit to the provision of IS support. It is positive though that hospital B regularly proceeds into support and maintenance contracts. Regarding the workflow adjustments and data migration, in the case of hospital A, the typical workflow has been altered due to the introduction of the IS. On the other in the case C, there was followed a reverse procedure, where the IS was actually adjusted in order to adapt to the existing workflow. In the case of hospital B this issue has been tackled through system parameterizations. In the case of Hospital A, changes due to the introduction of the new IS have been managed by the department of Informatics and the working group, whereas in hospital B the IT department or the working group were inactive. In hospital C data could be successfully transferred to the new IS while maintenance of the old one was still possible. In the case of hospital B the old system was reported to be interoperable with the new one in terms of data exchange. The IT employees believe that the main obstacles found during the IS implementation include lack of central planning, difficulties in the acceptance and incorporation of IT and non use of standards. In two hospital cases there is a common belief that barriers include the existence of specific human interests within hospitals, inadequate central planning and insufficient number of capable IT specialists and health informaticians. Regarding the pre-implementation studies and evaluation of the provided services, hospital cases B and C made no serious studies prior to the IS introduction, while only one out of the three cases (hospital A) evaluated the IS services using three different evaluation approaches. In the case of hospital B costbenefit has been the only measurement indicator (Table 1). Table 1. Assessment Questionnaire responses and level of agreement Hospital Role of the hospital management Active Participation of hospital management Hospital Management contributed to the IS planning Hospital Management efficiently utilised skills of IS Users Education and Training IT department providing active support Training provided during the IS introduction
A
B C Rate of positive IT staff responses
3/3 3/3 2/3
-/3 1/3 2/3
-/3 -/3 3/3
3/3 3/3
-/3 3/3
-/3 3/3
126
D. Zikos et al. / Assessment of Hospital Information Systems Implementation: A Case Study
Table 1. Continued User Support during the implementation Management provided support to users Training-education was provided to users External consultant (for IT Department) Motivators were offered to the employees Contracted Task Agreement Schedule-deliverables agreed Revision-modification was agreed Criteria for system specs specified Formal documentation System followed Contract-maintenance support provided Specialists beliefs regarding implementation hurdles Difficulty in acceptance of IT Difficulties in incorporating standards Users reluctant to handle sensitive data Existing Individual Interests Lack of central planning Insufficient specialized IT staff IT Services Internet-email is available for all staff Hospital IS Infrastructure Existing administrative/financial IS Existing clinical IS Existing LIS Integrated Complete Information System Technical Specifications The menus are localized IS is fast and flexible Data security is adequate Data back-up mechanisms are supported Specific Use instructions are provided Maintenance -System scalability is ensured Standards and Codings HL7 ICD10 ATC GMDN DICOM Pre-implementation studies Exploratory Study was carried out Feasibility Study was carried out Risk analysis was carried out Evaluation of IS Patient Satisfaction evaluation was carried out Employee Satisfaction evaluation was carried out Measurement of cost-productivity carried out
3/3 3/3 -/3 -/3
1/3 -/3 -/3 -/3
-/3 3/3 -/3 -/3
3/3 3/3 3/3 3/3 3/3
2/3 -/3 3/3 -/3 3/3
1/3 -/3 -/3 -/3 -/3
-/3 -/3 -/3 2/3 3/3 2/3
3/3 3/3 2/3 -/3 3/3 3/3
3/3 3/3 1/3 2/3 2/3 3/3
3/3
-/3
3/3
3/3 3/3 3/3 2/3
3/3 3/3 3/3 -/3
3/3 -/3 3/3 -/3
3/3 3/3 3/3 3/3 3/3 3/3
3/3 -/3 2/3 3/3 3/3 -/3
3/3 -/3 -/3 -/3 -/3 -/3
3/3 3/3 3/3 3/3 -/3
3/3 3/3 3/3 3/3 -/3
-/3 -/3 -/3 -/3 -/3
3/3 3/3 3/3
-/3 -/3 -/3
-/3 -/3 -/3
3/3 2/3 3/3
-/3 -/3 -/3
-/3 -/3 -/3
3.2. Phase 2: Open Interviews-Observation of Case C During the interviews with the IT employees, there have been identified three broad categories of problems, namely resource related problems, human element and organizational issues. Regarding the human factors, there has been reported the existence of indifference of users and lack of motivators to support the process and use the system functions; therefore there is an imperative need for training, support and encouragement of the employees. As far as the resource related issues are concerned, the interviews indicated that there exist inadequate number of staff in the IT department, inadequate working space and financial resources to effectively support the IS utilization. Finally, many employees referred to the lack of evaluation and concrete specification of the IS scope, the inadequate hospital management support and the lack of a consistent plan for the introduction and acceptance of new technologies. The belief that the IT department was not utilized by the hospital management was also expressed.
D. Zikos et al. / Assessment of Hospital Information Systems Implementation: A Case Study
127
During the implementation of the IS in hospital C, observation and discussion sessions have been undergone with four different groups of employees, namely representatives of the supplier company, members of the hospital management, employees working in hospital departments (nurses, MDs) and hospital IT department staff. According to the supplier company who developed the IS, it contacted the hospital IT department to be informed about the hospital structure and technical facilities also collaborating with head employees regarding the coding of the hospital material. All the potential IS users were invited to a presentation of system functions, a special computer training room for user training with real scenarios, was prepared. The second group, namely the administrative personnel, expressed the concern that the decision to introduce an integrated IS has been solely taken by the hospital manager. The hospital management activities and the financial department strategy were also reported to be a cumbersome process with unnecessary bureaucracy. Employees working in departments where the IS would be installed (clinical nurses and MDs), reported that it was not until late when they were informed about the new system. Despite having been invited by the management to presentation sessions of the IS benefits in the past, most of them were found to be distrustful about the success of the system. Skepticism may be related with the expected workflow changes by the new IS, while most reactions were found within nursing departments. Finally, employees working at the hospital IT department reported to be positive towards the introduction of an integrated IS in their hospital and supported the overall process. More specifically, they supported the supplying company that developed the system, during the training sessions. The IT department employees also assisted with the development of a training program, alongside with the contribution of the hospital’s education office and the nursing management.
4. Discussion According to the study results, the successful implementation of an integrated IS, shows that problems related with the use of standards, insufficient funding, lack of skilled personnel, denial and indifference of users and inadequate training [13], can be overcome, if the hospital administration is committed to the success of the project and place specialized and trained employees, to lift the burden of implementation and future support [9]. Inadequate IT department personnel cannot support the development process efficiently, while information systems co-developed by many different companies may be related with implementation issues. Such a “Babel tower” development approach, coupled with an incompetent and powerless to act IT department may have contributed to the lack of support services within hospitals. In some cases, users’ reactions are not taken into account by the administration, bearing a limitation to the success of the IS. Related to the above is the fact that the introduction of IS is only supported by employees with little or no expertise. IT departments should play a particularly important role in the context of increasing needs for informational organization of hospitals. Hopefully, in one case IT employees reported a strong desire and commitment to the introduction of integrated IS support combined with the longterm administration support, and this already shown positive potential. Observation and discussions different hospital employee categories also highlighted useful opinions. Strong points towards successful implementation have been found to be the management’s will to introduce integrated IS, and a cooperative
128
D. Zikos et al. / Assessment of Hospital Information Systems Implementation: A Case Study
environment between the management, the developer and the hospital IT department. It is evident that support from the IT department during the process, planning and organization of training sessions are key factors towards successful implementation, while cumbersome bureaucracy, lack of organization within the departments, the communication and interoperability issues between departments are the some of the existing barriers. Other issues include underfunding and officials’ reluctance to assist with encoding issues or alter their work motive. The case of hospital C shows that different roles appreciate the implementation of an IS based on their own needs and perspective. Workflow changes, participation in the decision making, user support and professionalism are factors affected by the introduction of a new IS, though not in the same manner for administrative, clinical, technical staff and external providers. It is therefore very important that a flexible multidisciplinary working group should be established, while a health informatics specialization in the IT department should be mandatory. Seminars should regularly be organised regarding use of the new technology, standards, interoperability issues etc. The provided services should be constantly evaluated. A Quality Control Board consisting of officials from all specialties with responsibilities including training and support, identify and confront problems during the transition period, is expected to ensure that the IS functions properly, prepare periodic and annual performance reports and proposals for further improvement and finally monitor medical-nursing and administrative-financial indicators. In conclusion, administration support, expert personnel and mobilization of the available human resources, continuing education, and a long-term strategic plan against bureaucracy, seem to be key success factors.
E. Mourtou, The EMR in Greek public hospitals, Health Review:Sciences-Technology-Policy 17 (101)(2006), 29-35. A. Baggelatos, I. Sarivourgioukas, Success factors for the Introduction of Information Systems in Health Inspection (2006), issue 101. R. Heeks, Health information systems: Failure, success and improvisation, International Journal of Medical Informatics 75 (2006), 125-137. N. Tsaloukidi, E. Papageorgiou, The role of Health Information Systems in the organization of Nursing Practice, Hellenic Journal of Nursing 47 (3) (2008), 313-319. M. Berg, Implementing information systems in health care organizations: myths and challenges, International Journal of Medical Informatics 64 (2006), 143-156. A. Lazakidou, Hospital Information Systems and e-Health Services, Kleidarithmos, Athens, 2005. M. Stuart, Twenty-Five Years of National Information Systems in the NHS, Public Money & Management 27 (2) (2007), 135-140. R. Haux, Health information systems-past, present, future, Intl Journal of Medical Informatics 75 (2006), 268-81. C. Papanikolaou, F. Malamateniou, I. Vidalis, A. Vagelatos, A new role for the IT department in Greek Hospitals, International Conference on Information Communication Technologies in Health (2003). A. Al-Mudimigh, Successful Implementation of Integrated Health Information Systems: The Role and Impact of Business Process Management, International Journal of Computer Theory and Engineering 1 (3) (2009), 251-57. M. Houser, J. Barlow, R. Tedeschi, M. Spicer, D. Shields, L. Diamond, The implementation of hospital information systems-change, challenge and commitment, Proc Annu Symp Comput Appl Med Care 7 (1984), 221-224. M. Berg, Implementing information systems in health care organizations: myths and challenges, International Journal of Medical Informatics 64 (2001), 143–156. M. Bush, A. Lederer, X. Li, J. Palmisano, S. Rao, The alignment of information systems with organizational objectives and strategies in health care, International journal of medical informatics 78 (2009), 446-456. G. Southon, C. Sauer, K. Dampney, Lessons from a failed information systems initiative: issues for complex organizations, International Journal of Medical Informatics 55 (1999), 33-46. J. Wyatt, S. Wyatt, When and how to evaluate health information systems?, International Journal of Medical Informatics 69 (2003), 251-259. B. Gavish, O. Sheng, Dynamic file migration in distributed computer systems, Communications of the ACM 33 (2) (1990), 177-189.
A Methodology to Assess Experiences in Implementing e-Health Solutions in Croatian Family Medicine Damir KRALJa,1, Miroslav KONČAR b and Stanko TONKOVIĆ c a Ministry of Interior / PD Karlovac, Karlovac, Croatia b ORACLE Croatia, Zagreb, Croatia c Faculty of Electrical Engineering and Computing, Zagreb, Croatia
Abstract. The central information system of primary health care of the Republic of Croatia is in an early stage of implementation which for now covers integration of all family doctors' offices into a single comprehensive eHealth network connecting their software solutions with the national payer institute and public health authority. Measuring the quality and efficiency of information systems at an early stage of development is a very difficult task. The main goal of this work is establishing the foundation for a formal methodology to measure and quantify the experience of family doctors in the current use of this system. A questionnaire has been created to support the work which, on one side carefully follows our assumptions for quality criteria, and on the other collects valuable input from the users of the technology and solutions implemented. Our work is closely aligned with worldwide accepted standards and recommendations carefully analyzed and localized to reflect the current environment and health policy. This paper presents some preliminary results based on the survey conducted with family doctors on the field. Keywords. Experiences based assessment methodology, e-Health, family doctor offices, electronic health records, reliability, certification criteria
Introduction The process of implementation of the national e-Health infrastructure in the Croatian public health care system started in 2006 by introduction of the Croatian primary health care information system (PHCIS or CEZIH in Croatian language) [1]. The first areas of the system implementation includes the integration of family doctor's offices (FDO) into a comprehensive system, that includes the integration of various types of FDO specialized solutions with national infrastructure, Croatian Institute for Health Insurance and Public Health Authority. The system is generally tested "in vivo" i.e. real production conditions and real patient data, and without comprehensive feasibility study performed on the field. The central part of the information system was delivered by a renowned company specialized in this kind of projects, as a very stable and quality system based on well defined business processes, legal and semantic rules, and communication and messaging standards such as ENV 13606 and HL7v3 [2]. Design 1
D. Kralj et al. / A Methodology to Assess Experiences in Implementing e-Health Solutions
of the applications for managing the electronic healthcare records (EHR) in FDO was left to a number of small IT companies competing on the Croatian market, which had to undergo certification process defined by the Ministry of Health and Social Welfare (MHSW). Certification currently includes communication and basic data exchange with the central part of the system only, which are hardly single quality criteria for such solutions, especially from the perspective of final end users – doctors and nurses. The concept and functionality of these applications has been left to the manufacturers of these applications [3]. With such situation in place, it seems very difficult to measure the quality and effectiveness of an information system which is in the early stage of development. For those reasons, our motivation was to establish a methodology to quantify and qualify overall quality criteria. The performed actions can be summarized in the following steps: analysis of previous surveys and field researches on e-Health implementation among general practitioners' (GPs') and family doctors' (FDs') offices; analysis of known e-Health assessment models; analysis of European pilot studies among GPs/FDs; analysis of known standards, industrial initiatives and certification criteria in the world; construction of the formal assessment tool; survey with qualified sample of GPs/FDs; assessment of reliability, validity and accuracy of the tool; and, finally, drawing conclusions on acquired results i.e. point out the problems and propose specific changes.
1. Relevant Work in Assessing e-Health Implementation Experience One of the earliest frameworks for assessing community readiness for eHealth solutions refers to "Framework for rural and remote readiness in telehealth" that was conducted in 2002 by Canada's advanced research and innovation network CANARIE [4]. This paper describes the basic assumptions that derive from the theory of change and stages of change. The next interesting example can be found in association of the Aga Khan University in Pakistan and the University of Calgary in Canada [5]. The subject of the project was development of tools for e-Health readiness assessment in developing countries. Project has proposed methods for validation and reliability testing of the tool for e-Health readiness assessment. A third interesting example is found in the study "e-Health Readiness framework from Electronic Health Records Perspective" conducted on the University of New South Wales in Australia [6]. However, all the models referenced above are based on an analysis of isolated cases ("in vitro") by gathering the elements to assess the readiness of a small part of health system for the introduction of e-Health concept, while our framework requires experiences assessment in large scale deployment with very limited prior analysis ("in vivo"), which leaves us with highly challenging environment that requires careful assessment and offers less change maneuver space. For this reason this work takes more recommendations from some more recent works, including European EuroRec EHR-Q TN project, EuroRec EHR Quality Seal Level 1 and Level 2 [7], and American ONC Meaningful Use of EHR Stage 1 certification project [8]. The contents of European studies that were conducted by Empirica [9] and Health Consumer Powerhouse [10] were also of great help in the making process of the framework. Based on these foundations we made a rather comprehensive questionnaire. The questionnaire consists of seven main units. While the first unit contains general questions about the FDO, the remaining six modules measure major dimensions which
D. Kralj et al. / A Methodology to Assess Experiences in Implementing e-Health Solutions
131
our work has identified as needed (see Table 1). In total there are 118 questions of which 103 issues have been defined for assessment. These 103 questions for the assessment consist of 32 multiple choice questions in a Likert scale of 1-5, 54 questions with dichotomous answers (YES-NO i.e. 1 or 0), and 17 questions with offered answers that should be marked. The range of quantitative ratings for each question is from 0-1, so the total predicted mean quantitative score for the entire questionnaire is less than or equal to 1.The time allocated for completing the questionnaire was estimated at 20 minutes. Table 1. Structure of the questionnaire with general description of its categories. A) BASIC x At ti tud e ab out us e of c ompu t ers in FDO; x Organization of work on computer support; x Impact assessment of use of computers to work process; x Attitude about basic elements of e-Health. D) DOMAIN x Domain usability and functionality of EHR applications; x Structuring and encoding of information in EHR application; x Implementation of advanced decision support systems; x Monitoring and quality of work assessment according to working guidelines; x Overall satisfaction with the EHR applications from the domain view.
B) TECHNOLOGICAL x Problems with hardware and network support in FDO; x Quality and reliability of EHR applications in FDO; of diagnostic x Readiness equipment for use in e-Health; x Data protection and patient safety. E) ORGANIZATIONAL x Use of e-mail communication with other health institutions; x Possibilities of migration to a paperless business; x Elements of e-business integrated into EHR application; x Forms of electronic reporting built in existing application support; x Interoperability and compatibility of EHR applications with current diagnostic systems.
C) ENGAGEMENT x Self-assessment of the IT and medical domain knowledge; x FD engagement in process of new system implementation; x Use of EHR for evaluation of doctor's work and research; x Care about the safety and security of the EHR. F) SOCIETAL x The impact of use of computers and EHR applications on patients' satisfaction; x The impact of health contents available on the Internet on behavior of patients in FDO; of electronic x Forms communication between doctors and patients.
2. Analysis of Survey Process and Data Collected The survey was conducted during the period from mid-December 2009 until the end of January 2010. Questionnaire was made in electronic PDF/FDF form with the ability to automatically return to the sender via e-mail, and in the classical paper form. The questionnaires in electronic form were offered via dedicated mailing list, which has approximately 1100 formal users (assuming that the number of active users is much smaller), while about 70 questionnaires were distributed in paper form at the professional meetings and collected on spot or received by post. Random sample selection depended on FDs' free will to fill the questionnaire. A total of 115 complete and correctly filled questionnaire forms were collected (87 or 75.7% of 115 in electronic and 28 or 24.3% of 115 in paper form). Therefore, we included approximately 4.7% of total 2450 Croatian FDs. By analysis of general data about the respondents and their offices, we got the structure of the analyzed sample, which is showed in Table 2. By comparison of data from well-known official Croatian health statistical publications [11], and data known from some previous analyzed works [3] with data showed in Table 2, it can be concluded that analyzed sample is representative enough to draw conclusions from the study.
132
D. Kralj et al. / A Methodology to Assess Experiences in Implementing e-Health Solutions
Table 2. Characteristics of tested sample of the Croatian family doctors and their offices. Category Age Years of working Gender Specialization FD office autonomy FD office type
3. Results Analysis and Discussion The results of qualitative and quantitative analysis of the categories and total experience are shown in Table 3. Due to limited space, the qualitative ratings are summarized for the most important elements, while the quantitative rates provide fairly realistic overall scores on a scale from 0 to 1. Table 3. Major qualitative results, average quantitative results and reliability coefficients of experiences assessment presented by categories. Category
A
Results -40% of FDs believe that the new system and EHR applications slow down their work; -In 4.3% of FDOs nurses write medical information in the EHR, while in 10.4% of FDOs the doctor updates the administrative and demographic data of patients; -34% of FDs do not support e-prescribing, while 35% do not support e-referral; -50% of FDs are mainly against the secondary use of medical data; -66% of FDs do not believe in the security and confidentiality of data in a central EHR; Mean rate: 0.696
B
Mean rate: 0.555
C
Cronbach α: 0.667
Items: 24
Cronbach α: 0.694
-13% of FDs believed to be overloaded with unnecessary knowledge of IT technologies; -26% FDs attended IT schools or courses in the past 5 years; -17% of FDs assess their IT knowledge as very high; -57% of FDs use the data from their EHR applications for quality evaluation of their work; -75% FDs give to their replacement doctors to work on their user account (security risk); -Only 60% of FDs make daily backup their data (EHR); Mean rate: 0.427
D
Items: 12
-All contracting FDOs are equipped with the necessary ICT equipment; -Automatic remote software update is provided for all EHR applications; -All EHR applications use the same formal structured and coded lists of health registers and nomenclatures that are automatically updated on a regular basis; -All EHR applications have authorized access and role specific access rights; -41% of FDs have some diagnostic devices that provides results in an electronic format suitable for inserting in EHR; -Transfer to another EHR application is rather difficult due to portability and “data lock” issues => EHR is not longitudinal in its most important part;
Items: 18
Cronbach α: 0.722
-In 61% of EHR applications is possible atomized (structured) input of the physical status; -All applications contain regularly maintained classifications and nomenclatures; -In 72.6% and 75.4% cases EHR applications offer support for chronic disease and allergies monitoring, respectively -Decision support systems are in their beginnings as a simpler forms of work assistance; -In 41.5% cases EHR applications have built-in clinical and pharmacological guidelines;
D. Kralj et al. / A Methodology to Assess Experiences in Implementing e-Health Solutions
133
Table 3. Continued Category
Results
D
-In 51.9% cases EHR applications have built-in visual indicators for the financial indexes for diagnostic-therapeutic procedures, drug prescribing and the rate of sick leave; -35.7% of FDs are very satisfied with the overall domain properties of their EHR application; Mean rate: 0.430
E
Mean rate: 0.324
F
Overall
Items: 23
Cronbach α: 0.822
-All EHR applications are capable for e-prescribing and e-referral -All EHR applications have the ability to add scanned paper-based diagnostic test results into the EHR, but only 22.6% of FDs use this feature; -EHR applications support some forms of electronic reporting; -All EHR applications are capable to remotely check the patient's health insurance status; Items: 17
Cronbach α: 0.689
-In 23.5% cases patients are satisfied with the implementation of the new information system; -EHR applications currently do not provide patients with reports on their health status; -Only 27% of FDs communicate with patients via e-mail and other electronic media; -Only 9.6% FDs collect information about chronic diseases of their patients via e-mail. Mean rate: 0.446
Items: 9
Cronbach α: 0.541
Mean rate: 0.48
Items: 103
Cronbach α: 0.886
To determine and prove the reliability of our measurement tool, we used a calculation of the Cronbach α coefficient of correlation for each of categories [12]. The recommended amount of this coefficient for a high degree of reliability, i.e. internal consistency of questionnaire, is ≥0.7. Before calculating of the coefficient, we conducted verification of the required sample size with Bonett's formula [13] using null hypothesis value of 0.7 and estimated value of 0.8 (two-sided, α=0.05, β=0.2, N=103). We calculated a minimum sample size of 41, which is significantly less than our 115. Our population sample was not previously prepared for the testing. For this type of testing are common slightly smaller amounts of the Cronbach α than in controlled or clinical conditions. As we see from Table 3, the lowest Cronbach α has a category of social experience (5.41), however, it is a common occurrence in the questionnaires that have fewer than ten questions. So called face validity and content validity [5] of our measurement tool were confirmed through interviews and commentaries of the doctors. Comments were positive, and confirm the relevance of all categories in over 75% of cases. To determine accuracy, it would be necessary to carry out additional field researches. We see this as the subject of further research. As we see from the results presented in Table 3, categories A, C and partly F reflect the doctors' views about essential objectives of the e-Health concepts and doctors' engagement in achieving these goals. From the results of all other categories we can see how EHR applications meet the current worldwide certification criteria. Based on identified system performance, and current Croatian certification criteria [1], we can conclude that Croatian EHR applications would be able to almost entirely meet the criteria of EuroRec EHR-Q Seal 1 and in some parts even the Seal 2 criteria, which is subject to more detailed analysis. However, in domain functionality, which is better covered by ONC Meaningful Use of EHR, is still necessary to significantly improve the functionality. Furthermore, we see some encouraging first results in applying of the working guidelines, guideline-based decision support systems and monitoring of chronic diseases and allergies [14]. A similar situation is with monitoring and indication of the quality of doctor's work. These are definitely significant areas of further improvement.
134
D. Kralj et al. / A Methodology to Assess Experiences in Implementing e-Health Solutions
4. Conclusion In this paper we have presented some preliminary results of what is envisioned to be a comprehensive methodology and criteria to measure quality of EHR system implementations at FD’s offices. The focus of this paper is a measuring tool which is the basis for data analysis that serves to identify some key areas of quality to measure. Croatian certification criteria are still mainly based on the local requirements and needs of current developments and do not draw direct reference to some of the internationally recognized quality indicators and frameworks or take into account clinical protocols, experts practice, and expectations on readiness and experience by users. Since the certification of EHR applications is performed in successive stages of development, we expect to be relatively easy to fully comply with worldwide technical criteria, however it remains to be seen what additional requirements we will identify as important or how would international certification processes apply to localized environments and large scale deployment. The preliminary results give us confidence that our assessment methodology could be used as the potential tool for monitoring of further improvements of Croatian certification criteria, also in respect to forthcoming development phases of the Croatian healthcare information system.
References [1] [2] [3]
[4]
[5] [6]
[7] [8] [9] [10] [11] [12] [13] [14]
Croatian Institute for Health Insurance: CEZIH PZZ. Available at http://www.cezih.hr/. Last accessed: November 2, 2010. M. Končar, D. Gvozdanović, Primary healthcare information system — The Cornerstone for the next generation healthcare sector in Republic of Croatia, Int J Med Inform 75 (2006), 306-314. D. Kralj, S. Tonković, Implementation of e-Health Concept in Primary Health Care - Croatian Experiences, ITI2009 Posters Abstracts, 31st Int. Conf. on Information Technology Interfaces, Cavtat, Croatia (2009), 5-6. CANARIE, Final report: Framework for rural and remote readiness in telehealth. Written by the alliance of building capacity, June 2002. Available at: http://www.fp.ucalgary.ca/.../Projects-CanarieFinal%20Report,%20June%202002.htm. Last accessed: May 10, 2010. S. Khoja, R. Scott, A. Casbeer, et al., e-Health readiness assessment tools for healthcare institutions in developing countries, Telemedicine and e-Health 13[4] (2007), 425-431. J. Li, E-Health Readiness Framework from Electronic Health Records Perspective – Master Thesis. University of New South Wales, Sydney, Australia, November 2008. Available at: http://handle.unsw.edu.au/1959.4/42930. Last accessed: June 25, 2010. EuroRec: European Institute for Health Records. Available at: http://www.eurorec.org/. Last accessed: November 2, 2010. Centers for Medicare and Medicaid Services: ONC Meaningful Use of EHR. Available at: https://www.cms.gov/EHRIncentivePrograms/. Last accessed: November 2, 2010. A. Dobrev, M. Haesner, T. Hüsing et al, Benchmarking ICT use among General Practitioners in Europe – Final Report, Empirica, Bonn, 2008. A. Björnberg, B.C. Garrofé, S. Lindblad, Euro Health Cosumer Index 2009 – Report, Health Consumer Powerhouse, Brussels, 2009. Ž. Baklaić, V. Dečković-Vukres, M. Kuzman, Croatian Health Service Yearbook 2007, Croatian National Institute of Public Health, Zagreb, 2008. L.J. Cronbach, Coefficient alpha and the internal structure of tests, Psychometrika, 16[3] (1951), 297334. D.G. Bonett, Sample size requirement for testing and estimating coefficient alpha, Journal of Educational and Behavioral Statistics, 27[4] (2002), 335-340. D. Kralj, S. Tonković, M. Končar, Use of Guidelines and Decision Support Systems within EHR Applications in Family Practice - Croatian Experience, 12th Mediterranean Conference on Medical and Biological Engineering and Computing (MEDICON 2010), Chalkidiki, Greece: Springer, Heidelberg, Germany, (2010), 928-931.
What are the Barriers to Conducting International Research Using Routinely Collected Primary Care Data? Simon de LUSIGNANa1, Christopher PEARCEb, Nicola T. SHAWc, Siaw-Teng LIAWd, Georgios MICHALAKIDISa,e, Marília T. VICENTEa,f, Michael BAINBRIDGEg, International (IMIA) and European (EFMI) Medical Informatics Association and Federation Primary Care Informatics Working Groups (PCI-WG) a Primary Care Informatics, St. George’s University of London, London, UK b Melbourne East GP Network, Melbourne, Australia c Health Informatics Institute, Algoma University Ontario, Canada d University of New South Wales, Australia e Department of Computing, University of Surrey, Guildford, UK f Hospital das Clínicas Fac. Medicina Univ. de São Paulo, São Paulo, Brazil g University of Victoria, British Columbia, Canada
Abstract: Background: Primary care is computerized with routine data recorded at the point or care. Secondary use of these data includes: genetic study, epidemiology and clinical trials. However, there are relatively few international studies. Objective: To identify the concepts that might predict readiness to collaborate in international research using routinely collected primary care data Method: Literature review and data gathering exercise, from international Primary Care Informatics working group workshops, and email modified Delphi exercise. Results: To establish whether primary care data are fit for use in a collaborative study information is needed at the micro-, meso-, and macro-level. At the microor data level we need to use documented standards for interoperability, computerized records, to facilitate linkage of data. At the meso-level we need to understand the nature of the electronic patient record (EPR) and specific study requirements. At the macro-level: health system, social and cultural context constrain what data are available. The framework defines the information needed at the point of expression of interest, and joining a study. The initial assessment of readiness should be by self-assessment followed by an in depth appraisal more immediately prior to the start of the study. Finally, a sensitivity analysis should be conducted to test the robustness of the data model. Conclusions: The literature focuses on technical issues: interoperability, EPR and modeling; the workshops on socio-cultural and organizational. This framework will form the basis for developing a survey instrument of the initial assessment of readiness for collaboration in international research. Keywords. Primary care data, electronic health record, research, clinical system, computerized, computing methodologies, classification, vocabulary controlled
1Corresponding Author: Simon de Lusignan, Professor of Primary Care and Clinical Informatics; Faculty of Management and Law - University of Surrey, Guildford, Surrey GU2 7XH, UK; Email: [email protected]; Phone: +44(0)1483683089; http://www.clininf.eu
136
S. de Lusignan et al. / What Are the Barriers to Conducting International Research?
Introduction Primary care is responsible for first contact, continuing and generalist care of the population. In many countries primary care is computerized with data entry at the point of care and thus provides a potential goldmine for research [1]. However, there are many barriers to using these data [2]. There are differences between primary healthcare systems: organizational arrangements (e.g. List based system vs. Fee for service); the availability and supply of IT to primary care, the willingness of family doctors to use technology at the point of care in a way that maximizes the data potential, and the nature of the clinical task. Additionally, many family doctors feel that they don’t have time or the human resources available [3]. Computerization of primary care data facilitates research, data are readily available for observational and for case ascertainment for trials [4]. Practices, usually grouped into primary care research networks (PCRN). These PCRN have other functions: facilitate networking between primary care researchers [5] and quality improvement activities [6]. The standard approach to conducting observational studies has been to transfer data into a central secure data vault; however agents and cloud computing techniques [7, 8] may be more effective methods of identifying subjects for research [9]. There are three principal ways routine data are used in research: 1. Genetic studies: Studies linking clinical and genetic data aim to understand how illnesses and genes might be related. SNPs (single nucleotide polymorphism), a DNA sequence variation, differs between individuals, and these differences must be compared with phenotypes. As spurious correlations are common, sample sizes of >10,000 are required. 2. Epidemiological studies: Cross-sectional studies are either observational or a descriptive exploration of health, disease, and risk factors. Cohort studies look at groups of individuals who have been exposed to the same risk. A casecontrolled study compares cases with a condition with controls, very often matched for age and gender. 3. Clinical trials: Studies designed to test a medical intervention. The most robust randomize their subject to receive the intervention or not, and may use placebos. Appropriate recruitment is a significant issue in the community The objective of this study was to identify the barriers to conducting international primary care research and propose a framework for the development of an instrument to assess the readiness of data to be used in international studies.
1. Method We developed a framework to identify the factors which predicted readiness for of a primary care database for research, and explored the how the process of approval might work. The framework for assessing readiness was based on one developed by Greenhalgh to describe the impact of technology [10]. This framework divided impact into the micro- which we interpreted as the data and interoperability level; meso- the individual study, electronic patient record, and data extraction level; and macro- the societal level where the nature of the health system, social, and cultural values impact on what is included within the health record. We updated the literature review of a previous study [2] searching PubMed, ISI web of Science, and hand searched the proceedings of international informatics
S. de Lusignan et al. / What Are the Barriers to Conducting International Research?
137
conferences. We used the International and European Federation for Medical Informatics (IMIA and EFMI respectively) primary care informatics working groups (PCI WG) workshops during 2010 to share and develop an understanding of the issues and potential barriers to sharing routine data for research; testing these using a modified Delphi approach by email to develop a consensus. We also utilized meetings within TRANSFoRm (Translational Research and Patients safety In Europe) a Seventh Framework Programme (FP7) project, [www.clininf.eu/projects/transform] which aims to develop systems to increase the volume of research across Europe. TRANSFoRm has developed use-cases for two simulated research projects: one a genetic study of diabetes linking genetic and primary care data; the second linking oesophageal cancer, reflux disease and use of anti-indigestion medicines. The instrument to be developed needs to be able to identify European locations able to participate in these use-cases. We excluded from this study: constraints arising from risks to confidentiality and privacy; and what is considered “legitimate” use of data.
2. Results 2.1. Micro-Level: Data Sharing and Interoperability Standards We identified technologies used to achieve interoperability some customized for research (Table 1). These included the Interoperability Toolkit (ITK) - an English NHS interoperability framework [11]; Health Level-7 (HL-7), an international interoperability organization who’s Reference Information Model (RIM) underpins much interoperability in healthcare; Clinical Data Interchange Standards Consortium (CDISC) including the Biomedical Research Integrated Domain Group (BRIDG) [12] an internationally recognized standards body [13]. CDISC has also developed a relationship with HL7 who has adopted the BRIDG data analysis model (DAM). Table 1. Relevant standards for interoperability and communication Standard Acronym / meaning
Description
Purpose
Relevance
HL7
CDISC
BRIDG
Health Level Seven
Clinical Data Interchange Standards Consortium
Biomedical Research Integrated Domain Group
Platform-independent standards supporting acquisition, exchange, submission, and archiving of clinical trials data
Coordinate HL7 and CDISC, by harmonizing existing, & developing specifications for new standards Produces a shared view of the dynamic and static semantics for protocoldriven research and its associated regulatory artifacts Move to more pragmatic data sharing. However, the context of use is much more structured and less flexible than needed using routine data.
System-independent data interchange specifications and messaging standards to support clinical applications The Reference Information Model (RIM) representation of the HL7 clinical data (domains),used to create clinical messages Large operational experience of providing pragmatic data exchange and linkage. Likely generate pragmatic solutions to data problems
The CDISC Operation Data Model (ODM) defines a base XML representation for data collected from clinical trials Wide experience of management of fixed dataset defined within a trial protocol. Able to collect defined datasets may be less flexible with routine data
138
S. de Lusignan et al. / What Are the Barriers to Conducting International Research?
2.2. Meso-Level: Study Level Information, Computerized Record, and Data Extraction We have identified a number of conceptually useful models: generic, specific to primary care research, and of the electronic patient record (EPR, Table 2). Unified modeling language (UML) is a well established modeling method which we recommend is used as standard. The Primary Care Research Object Model (PCROM) identifies use-cases and uses UML activity diagrams to model the core study components. The model used by OpenEHR [14] separating interface, coding system, archetypes, and database can be applied generically to model EPR system components. We have specifically looked at the OpenEHR archetypes as a way of exploring whether there is complete and compatible collection of study relevant data [15, 16]. Table 2. Models with relevant to combining research data Standard Acronym Description Details
Relevance
PCROM Primary Care Research Object Model Standardized language and software development for research in primary care
openEHR Open Electronic Health Record To guarantee interoperability between electronic health records (EHR)
“Use cases” and activity diagrams describing: functions, activities, and interactions of actors and other entities First published attempt to draw together use-cases of idealized trial datasets and primary care data. Not operationalized or used to link primary care & genetic or registry (e.g. cancer) registry data
Open specifications open-source software & knowledge resources; & international standards development Value as a model that conceptually separates interface, clinical archetype, coding schema, and database. However, non-hierarchical archetypes & proprietary use limit scope for use
The data extraction technology further constrains what data are available for research, and we used an error reporting taxonomy to codify the possible sources of data loss [17]. The technical barriers are summarized with macro-level barriers in Table 3. Table 3. Key domains: barriers to sharing routinely collected primary care data.
Technical Organisational Sociocultural
Macro
Meso
Level Domain
Sub-headings
Key issues Method and process for generating queries Data extraction Query translation layer and proxy specification Shape and complexity of EPR system Top-level system design Data collection and translation and schema Data analysis repository Software & Includes links with other systems (e.g. Central registration communication system, pathology laboratory, imaging, hospital system) Hosted or local system Hardware and Access control systems infrastructure Structures to support interoperability Human errors Factors which can effect data quality Nature of health system and impact on data Health service Primary care What cases are seen in primary care Type of clinical records, coding system Medical record system Availability of outcome and longitudinal data Data quality Legal & societal Human rights, Legal barriers to practice, Language Attitude to illness and participation in research. Health seeking behavior and fragmentation of health provision. Funding system: Cultural state, insurance, privacy and acceptability of central registries.
S. de Lusignan et al. / What Are the Barriers to Conducting International Research?
139
2.3. Macro-Level: Health System Organization, Social and Cultural More descriptive models bridge the meso and macro levels. For example: (1) Clinical task; (2) Organization, (3) Individual; and (4) Technological factors [18]; describes the potential barriers to using routinely collected data [2]. The workshop participants highlighted that in their experience macro-level problems were the principle barriers to conducting studies. The social and cultural elements included legal and ethical constraints within a given jurisdiction, as well as the social influences on diagnosis. The macro-level components are: Organizational, at the health system level, social and cultural. The organizational components on the health system from which the data originated and system level standards. 2.4. Process of Assessing Readiness to Participate in International Research Finally we defined three steps in the process of assessing readiness to participate in International research: (1) A registration process whereby researchers and database owners set out their preparedness to undertake research and the scope of the available data; (2) Provide more detailed information about key variables needed for a specific study; (3) Sensitivity analysis to quality assure the data model data for the specific research proposal; including an assessment of risk (Table 4). Track record was considered extremely important and should be weighted heavily. Table 4. Stages of readiness to participate in international research PCRN/ Data repository - Size 1. Registration / Generic Type of: - Expression of interest - Study - UID - Self-assessment - Datasets - Routine data coded - Fully automated/ On line - Duration & Funding - Linkages 2. Study specific Data required: Data provision: - Considering a specific study - Population - Population - Protocol / near to completion - Inclusion & exclusion - Eligible cases, - Coverage & - Key variables known criteria, - Partially automated process - 10 & 20 outcome mapability of 10 & - Final stage requires mapping measures 20 outcome measures - Population 3. Sensitivity analysis / Risk - Track record - Assumptions in linkage model - Seniority - Time using EPR - Institutional support - Data quality - Broad risk assessment - Publications - Research track record - Distorting factors Steps in assessment
Researchers
Genetic database - UIDs held - Scope for linkage - Phenotype - Range of SNPs Phenotype links - Size of sample relating to specific study >10K people - Links & phenotype info for study cohort - Population’s representativeness - Test linkage to health data
3. Discussion Our literature review primarily identified technical and interoperability issues; while the workshops highlighted and the importance of considering organizational, social and cultural information. Our final consensus was that data should be collected in three steps: Firstly, generic data; secondly study specific data; and thirdly a sensitivity analysis including a risk assessment of the likely success of the project. This three level model has been used to create a TRANSFoRm International Research Readiness Instrument (TIRRE – http://www.clininf.eu/projects/tirre.html) which is currently being used to identify potential study sites across Europe.
140
S. de Lusignan et al. / What Are the Barriers to Conducting International Research?
4. Conclusions Discussions among informatics experts across IMIA and EFMI provided insights into factors which may restrict the ability to conduct international research not apparent within a literature review. An instrument designed to assess readiness for international research should take into account broad organisational, social and cultural influences on data quality as well as technical issues.
Acknowledgements TRANSFoRm is supported by the European Commission - DG INFSO (FP7 247787)
References [1]
[2] [3]
[4] [5]
[6] [7]
[8]
[9]
[10] [11] [12] [13] [14] [15] [16]
[17]
[18]
S. de Lusignan, J.F. Metsemakers, P. Houwink, V. Gunnarsdottir, J. van der Lei, Routinely collected general practice data: goldmines for research? A report of the European Federation for Medical Informatics Primary Care Informatics Working Group (EFMI PCIWG) from MIE2006, Maastricht, The Netherlands. Inform Prim Care 14(3) (2006), 203-9. S. de Lusignan, C. van Weel, The use of routinely collected computer data for research in primary care: opportunities and challenges. Family Practice 23 (2006), 253 263. E. Hummers-Pradier, C. Scheidt-Nave, H. Martin, S. Heinemann, M.W. Kochen, W. Himmel, Simply no time? Barriers to GPs' participation in primary health care research. Fam Pract 25(2) (2008), 10512. K. Peterson, Practice-based primary care research—translating research into practice through advanced technology. Family Practice 23 (2006), 149–150. Z. Nagykaldi, C. Fox, S. Gallo, J. Stone, P. Fontaine, K. Peterson, T. Arvanitis, Improving collaboration between Primary Care Research Networks using Access Grid technology. Inform Prim Care 16(1) (2008), 51-8; discussion 59-60. C. Pearce, M. Shearer, K. Besleaga, J. Kelly, A Divisions Worth of Data. Australian Family Physician. 2010; in press. M.T. Whitstock, C.M. Pearce, S.C. Ridout, E.J. Eckermann, A retrospective analysis of VIOXX in Australia: using clinical trial data and linked administrative health data to predict patient groups at risk of an adverse drug event. Aust N Z J Public Health 34(4) (2010), 431-2. S. Treweek, E. Pearson, N. Smith, R. Neville, P. Sargeant, B. Boswell, F. Sullivan, Desktop software to identify patients eligible for recruitment into a clinical trial: using SARMA to recruit to the ROAD feasibility trial. Inform Prim Care 18(1) (2010), 51-8. S. de Lusignan, F. Sullivan, P. Krause, Vault, cloud and agent: choosing strategies for quality improvement and research based on routinely collected health data. Inform Prim Care 18(1) (2010), 14. T. Greenhalgh, K. Stramer, T. Bratan, E. Byrne, Y. Mohammad, J. Russell, Introduction of shared electronic records: multi-site case study using diffusion of innovation theory.BMJ 337 (2008), a1786 NHS Interoperability toolkit. URL: www.connectingforhealth.nhs.uk/systemsandservices/interop Biomedical Research Integrated Domain Group (BRIDG). URL: http://www.cdisc.org/bridg Clinical Data Interchange Standards Consortium (CDISC) URL: http://www.cdisc.org OpenEHR. URL: http://www.openehr.org/home.html C.D. Kohl, S. Garde, P. Knaup, Facilitating Secondary Use of Medical Data by Using openEHR Archetypes. Stud Health Technol Inform 160, 1117-21, IOS Press, Amsterdam, 2010. D. Wollersheim, A. Sari, W. Rahayu, Archetype-based electronic health records: a literature review and evaluation of their applicability to health data interoperability and access. HIM J 38(2) (2009), 717. G. Michalakidis, P. Kumarapeli, A. Ring, J. van Vlymen, P. Krause, S. de Lusignan, A system for solution-orientated reporting of errors associated with the extraction of routinely collected clinical data for research. Stud Health Technol Inform 160, 724-8, IOS Press, Amsterdam, 2010. S. de Lusignan, The barriers to clinical coding in general practice: a literature review. Med Inform Internet Med 30(2) (2005), 89-97.
Evaluation of Possibilities in Demographic Data Exchange Support in Czech Healthcare Miroslav NAGY1, Libor SEIDL and Jana ZVAROVA Center of Biomedical Informatics, Department of Medical Informatics, Institute of Computer Science AS CR, v.v.i., Prague, Czech Republic
Abstract. This paper summarizes the evaluation of two standardized approaches to implementation of messages for demographic data exchange between the preventive cardiology outpatient department located at the Institute of Computer Science AS CR, v.v.i. in Prague and the Outpatients Department of Cardiology of Municipal Hospital in Caslav. Our setting consists of four independent systems maintaining different clinical data (scheduling system, hospital information system, EHR system and a digital ECG). The aim is to avoid repetitive patient demographic data entry. We evaluate the suitability of IHE Patient Administration Management Profile (including HL7 v.2.5) and Czech national standard DASTA using Standard Evaluation Framework proposed and published in 2008 by J. Mykkänen et al. Besides the evaluation of standards, we also discuss some aspects of the framework. Keywords. IHE PAM, DASTA, evaluation, framework, demographics
Introduction Teams of researchers and engineers all over the World have been making a considerable effort in the process of standards’ usage in healthcare for a long time. In the Center of Biomedical Informatics (CBI) we support such efforts and are trying to make the best of application of standards in our research. The main aim of our research is to develop a protocol for the preparation of an oligonucleolytic chip (ONC) with the optimal set of genes for diagnostics and prognosis of cardiovascular diseases [1]. Within the frame of our research our institute cooperates with the Outpatients Dept. of Cardiology of the Municipal Hospital in Caslav, where patient data and blood samples for ONC analysis are gathered. Since patient data are analyzed and processed at our institute in Prague various transport and interoperability issues had to be solved. In this paper we analyze and propose a solution to the given issue: repetitive patient demographic data entry. The amount of demographic information about patients involved in the project entered into various systems (hospital information system –HIS, experimental EHR, etc.) started to be unbearable. Before we could propose an implementation of a system solving the patient demographics interoperability, we had to start with rigorous comparison and evaluation of possible candidate standards. 1
Corresponding Author: Center of Biomedical Informatics, OMI ICS AS CR v.v.i., Pod Vodarenskou vezi 2, 182 07 Prague 8, Czech Republic; E-mail: [email protected]
144
M. Nagy et al. / Evaluation of Possibilities in Demographic Data Exchange Support
1. Materials and Methods 1.1. Preventive cardiology outpatient department The preventive cardiology outpatient department located in the building of the Institute of Computer Science, AS CR, v.v.i. employs four systems capable of maintaining patient demographic data: digital electrocardiograph (ECG), ADAMEKj [2] – an scientific EHR system, WinMedicalc2000 [3] – a commercial HIS system used in many Czech hospitals and finally a proprietary web-based patient scheduling system. Our setting in a wider scope is depicted in Figure 1. The left side shows typical medical examinations done in outpatient department. In the middle are placed all systems that should be interconnected. The right side of the picture describes the context of our work in the scope of the CBI [1] project.
Figure 1. The outpatient department at the Institute of Computer Science, AS CR in the scope of CBI research.
1.2. Demographic data synchronization scenarios The synchronization of patients’ demographic data among various systems could be in general done in two ways: x The system, which accepted a change in patient demographic, will push it to all participating systems. Thus each system has a complete patient database – IHE PAM profile and DASTA are capable of such “PUSH scenario”. x “PULL scenario” – Any system can ask a master system for patient demographic data using query or a previously obtained Patient ID. Changes in data are immediately forwarded to the master system. Such scenario is described in IHE PDQ and IHE PIX profiles with underlying HL7 v2 or v3. Unfortunately, DASTA does not contain specification for queries. As we are interested in “PUSH scenario” only, we introduce DASTA and IHE PAM profile briefly in the following two sections.
M. Nagy et al. / Evaluation of Possibilities in Demographic Data Exchange Support
145
1.3. DASTA DASTA [4] is an acronym for (DAta STAndard of Czech Ministry of Health). It is a Czech national standard with a wide support in the healthcare information systems available on the Czech market. Semantics of DASTA is based on the national nomenclature called the National Code-list of Laboratory Items (NCLP) [4]. The development of DASTA was supported mainly by the Ministry of Health CR in the past. DASTA was specialized mainly in the transfer of requests and the results of laboratory analysis. Nowadays, the latest version of DASTA 04.06 is developed by HIS vendors only and contains a code-lists used, among others, in statistical reports to the Institute of Health Information and Statistics of the Czech Republic (IHIS CR) [5]. DASTA 04.06 is an XML-based standard which supports the exchange of clinical information in free text form (Clinical Events). Unfortunately, DASTA has no relation to the international communication standards such as HL7 [6] or European standards like EN13606 [7]. In principle, DASTA only specifies data structure divided into so-called blocks that have constrained optionality. DASTA also defines a datatype of every field and constrains values, in most cases as pointers to NCLP/IHIS lists or ICD-10 nomenclature. DASTA doesn't provide any proposal in on-site specific setup (component model) and doesn't provide any messaging specification either. Exchange of information is done only on per installation agreement of involved vendors. Usual but non-written scenarios involve pushing XML files into shared SQL table as blobs or creating standalone XML files in shared folders. In our case of four systems involved, there will be 4 shared folders serving as inbox for each system. When a physician changes patient's demographic data in the system A, the system will create three identical XML files in folders devoted to the systems B, C and D. 1.4. IHE PAM profile IHE stands for Integrating the Healthcare Enterprise®. It is a world-wide initiative by healthcare professionals and industry representatives, which deals with constraining of existing communication standards implementable in typical (hospital) settings. IHE Profiles describe clinical information management use cases and specify how to use existing standards (HL7, DICOM, NTP, ...) to address the use cases [8]. Patient Administration Management (PAM) profile is part of the IHE IT Infrastructure domain. This profile enables applications to share accurate patient demographic data. We focus on the Patient Identity Feed part of PAM. The underlying messaging standard involved is HL7 v. 2.5, namely ADT messages. The detailed description of the implementation is left out due to limited space of this text, although it can be found in the specification of this profile. 1.5. Comparing methodology When some project is aimed on implementation of interoperability of two or more systems, usually the utilized standards are enforced by some political/legal/religious aspects of the project donor. Besides, two attempts of interoperability implementation are never the same. They differ in budgets, in applications involved, vendor's knowledge, and standards available. The published successful stories typically don't
146
M. Nagy et al. / Evaluation of Possibilities in Demographic Data Exchange Support
show results for standards that were left out from the solution. Thus there is no relevant debate about suitability of standards for a particular purpose. We try to establish interoperability within patient demographics between two (or more) standalone applications. Each application contains fully featured database of patients. This situation is common in Czech hospitals and outpatient departments. IHE and HL7 standards are widely used abroad, and DASTA is typically used in Czech environment only. How do these standards differ? Which standard is better? We did an evaluation of applicable standards using an evaluation framework created for such purpose and published in [9]. This evaluation framework consists of 9 forms containing a set of considerations, which should be examined in the context of ongoing interoperability implementation effort. Form 1 provides basic information on evaluated standard and general overview. Form 2 focuses on information and semantics. Form 3 focuses on functionality and interactions. Forms 4 and 5 are focused on in-standard proposals of application infrastructure and technical aspects. Forms 6 and 7 assess flexibility, maturity and dissemination phase of the standard. Form 8 covers life cycle of applications. Finally, form 9 covers domain-specific considerations. Our walk-through of the evaluation process is derived from general workflow proposed in the framework and contains following steps. At first, we identified possible standards, and obtained their specifications. Then we chose implementation level of evaluation. If the given standard has appeared as suitable in Form 1 (an overview) we continued with Form 9 (domain-specific) and then stepwise filled in Forms 2, 3, 4, 5, 6, 7 and 8. After filling in each form we made a rough evaluation and could drop any standard that clearly contradicted our requirements. In the end, we updated the overview of the standard (Form 1) and reiterated through all forms again. As proposed by authors of [9], we assign weights wik for all fields (i=1..nk, where nk is the count of fields in form k) in the form (k) from the evaluation framework with scale of 0-3 (0 – not key factor, 1 – desirable, 2 – highly desirable, 3 – mandatory). Each standard (j) will receive a score sijk for each form (k) field (i) on scale from -3 to 3 (-3: standard contradicts the requirement, 0: not specified in the standard, +1: feature can be supported with an extension mechanism, +2: partially supported, +3: fully supported). After evaluating each form (k), cumulative scores scjk for each standard (j) can be calculated: nk
sc jk
¦ (w
ik
* sijk )
(1)
i 1
The higher total score a standard receives, the more suitable it is. Scoring proposed by equation (1) is heavily dependent on the size of each form. The more fields (questions) one form has the higher score it assigns to the given standard. Thus the normalization wik' of each wik by the number of fields (nk) in each form (k) was proposed: wik' =wik / nk. Also normalized scores scj' using wi' can be computed giving more accurate overview. It can be shown how important a particular form was for the evaluator. This can be done by comparing sums of normalized weights wik' for each form (k).
M. Nagy et al. / Evaluation of Possibilities in Demographic Data Exchange Support
147
2. Results According to the needs of our setup we assigned weights to each filed of all forms. Then we evaluated DASTA and IHE PAM standards using all 9 forms in proposed order. The scores obtained from the evaluation are shown in Table 1. This table also shows corrections of weights and computed corrected scores as proposed in the evaluation methodology [9]. Table 1. Scores of candidate standards evaluated via the evaluation framework published in [9]. Maximum values in each column are written in bold font. DASTA sc1k sc1k'
IHE PAM sc2k sc2k'
2.9
58
1.9
126
4.2
29
1.8
116
4.0
135
4.7
3 Functionality and Interactions
27
1.5
41
1.5
96
3.6
4 Application Infrastructure
14
1.4
16
1.1
53
3.8
5 Technical Aspects
12
0.8
17
1.4
12
1.0
6 Flexibility
4
1.5
5
1.3
17
4.3
7 Maturity, Usage, Official Status
9
0.9
17
1.9
20
2.2
8 System Lifecycle
9
1.2
16
1.8
15
1.7
9 Domain-specific Features
17
0.9
22
1.3
42
2.5
308
16.2
516
27.8
k Form name
#of fields nk
∑wik'
1 Basic Info and Scope
30
2 Information and Semantics
Total:
151
When we compare the succession in respect of the size of forms (nk) and the sum of normalized weights wik', we can see no significant difference in the first three biggest Forms (1, 2, 3). On the other hand, when we compare forms 4 to 9 in the same way the values don’t correspond. This is natural, as we require some flexibility from involved standard. Domain specific features are not so important, as we try to solve patient demographics interoperability, which is really a narrow domain and tested standards are pre-chosen to fit. We can see how flexible IHE PAM (and HL7 v2.5 of course) is, when the unweighted score sc2 6=17 can be found at the 7th place, but after the corrections it jumped to the 2nd place (sc2 6'=4.3). The ratio of total scores of standards expresses how many times DASTA is better than IHE PAM profile: sc1 / sc2 = 0.60,
sc1' / sc2' = 0.58,
where for standard j the total cumulative score is
(2)
sc j
¦
9 k 1
sc jk and scj' is defined
analogously. As we can see in (2), the ratios are very similar. This is due to the nature of our project and the fact, that the most important forms for us are the biggest ones. Proposed correction in this case doesn't bring significant change in overall success of particular
148
M. Nagy et al. / Evaluation of Possibilities in Demographic Data Exchange Support
standard. If somebody had done an evaluation depending mainly on Forms 6-8, the effect of correction would have been much more significant.
3. Conclusion According to the scores obtained we can clearly conclude that the IHE PAM Profile with HL7 v.2.5 can be a recommended part of our solution. Contrary to IHE PAM Profile, DASTA will fail in Form 2. During the evaluation process we had to use implicit (undocumented) knowledge of how things work in DASTA, which leads to conclusion, that any implementation based on DASTA for an abroad newcomer is practically impossible. We can state, that the evaluation methodology proposed in [9] is feasible including additional weighting of particular forms. Due to the structure of forms, such weighting is probably not necessary when evaluating standards in really narrow domain. Providing average knowledge of given standard, one can fill in forms in no more than 3-5 hours for each standard. Therefore we hope the proposed methodology will enable more frequent rigorous standards’ comparisons in Czech healthcare and thus the healthcare management decision-makers might be familiarized with various approaches in an easier manner. Having found a winning standard the implementation process can be started, which is, however, beyond the scope of this text.
Acknowledgement The paper was supported by the project no. 1M06014 of the Ministry of Education, Youth and Sports CR.
References >@ >@
>@ >@ >@ >@ >@ >@ > @
Centre of Biomedical Informatics (CBI) (homepage on the Internet). Available from: http://www.euromise.org/cbi/cbi.html (cited Nov 12, 2010). Nagy M, Hanzlicek P, Preckova P, Kolesa P, Misur J, Dioszegi M, Zvarova J. Building Semantically Interoperable EHR Systems Using International Nomenclatures and Enterprise Programming Techniques. In: B.Blobel, P.Pharow, J. Zvárová, D. Lopez (eds) eHealth:Combining Health Telematics, Telemedicine, Biomedical Engineering and Bioinformatics to the Edge, 105-110, IOS Press; Amsterdam, 2008. Medicalc Software Ltd. (homepage on the Internet), WinMedicalc 2000. Available from: http://medicalc.cz/winmedicalc Ministry of Health of the Czech Republic (homepage on the Internet). Data Standard of MH CR – DASTA and NCLP. Available from: http://ciselniky.dasta.mzcr.cz (cited Nov 12, 2010). Institute of Health Information and Statistics of the Czech Republic (homepage on the Internet). Available from: http://www.uzis.cz (cited Nov 12, 2010). Health Level Seven, Inc. (homepage on the Internet) Available from: http://www.hl7.org (cited Nov 12, 2010). European Committee for Standardization (CEN), Technical Committee CEN/TC 251: European Standard EN 13606, “Health informatics – Electronic health record communication”. IHE Wiki page. (homepage on the Internet) Available from: http://wiki.ihe.net (cited Nov 12, 2010). J.A.Mykkänen, M.P. Tuomainen, An evaluation and selection framework for interoperability standards, Information and Software Technology 50 (2008), 176–197s.
Character Sets: an Invisible Pre-requisite towards Cross-Border Interoperability? Frank OEMIGa b 1 and Bernd BLOBELb a Agfa Healthcare, Bonn, Germany b eHealth Competence Center, University Hospital Regensburg, Regensburg, Germany
Abstract. Working interoperability not only requires harmonized system’s architectures, but also the same interpretation of technical specifications in order to guide the development process. This paper analyzes the role individual characters play as the bearer of data in general and information in specific. This aspect of interoperability is especially important in the context of cross-border communication and collaboration. Keywords. Interoperability, Languages, Character Sets
Introduction The fact, that information is represented in form of character strings is self-evident and unanimously accepted today and does not require any additional explanation. The applications starting with the operating systems provide comfortable functions, so that the requirement having a character set for representing the characters in form of bit sequences is not present any more. The user simply selects the characters he needs, but the underlying code is not taken into account. This may cause problems when it comes to cross-border information exchange.
1. Methods 1.1. Character Sets Some countries need some additional guidance on this topic like Russia [1] with its Cyrillic character set. For this and other purposes, HL7 [2] has crafted an implementation support guide [3]. Chapter 1.5.1 “HL7 Encoding Figure 1. Unicode-snippet for Cyrillic Rules” presents further details on how to handle different character sets. Unfortunately, bullet point 3.a in section 1.5.1 mentions Unicode, but has UTF-8 in mind. It also restricts (constrains) the delimiters to be in the range of 0x20 to 0x7F, i.e. 7 bits. 1
Table 1. German specific characters in different character sets
1.1.1. Unicode Unicode is a huge list of characters assigned to certain codes (code points). Figure 1 gives an idea of how it works: To achieve a unique byte sequence, an algorithmic mapping from every Unicode code point must be provided, which is called Unicode transformation format (UTF) (Table 1).
German Umlaut
Windows 1250
ISO 8859-1
ASCII
UTF-8
Ä
E4
E4
84
C3 A4
Ä
C4
C4
8E
C3 84
ö
F6
F6
94
C3 B6
Ö
D6
D6
99
C3 96
ü
FC
FC
81
C3 BC
Ü
DC
DC
9A
C3 9C
Table 2. Differences between UTFs Name
UTF-8
Smallest code point Largest code point Code unit size Byte order Fewest bytes per character Most bytes per character
The possible UTFs differ in the algorithm these code points are serialized. UTF-8 is a sequence of 8 bit characters. UTF-16 uses pairs of 2 bytes only, whereas UTF-32 makes use of 4 bytes. The appropriate code point (scalar value) is of course represented differently then (Table 2). Table 3. Example of Unicode Encoding Forms
Table 3 is a list of a few code points and their representation in different UTFs. As can be seen quite easily, they are not compatible.
Table 4. UTF-8 bit Distribution Scalar Value 00000000 0xxxxxxx 00000yyy yyxxxxxx zzzzyyyy yyxxxxxx 000uuuuu zzzzyyyy yyxxxxxx
First Byte 0xxxxxxx 110yyyyy 1110zzzz 11110uuu
Second Byte
Third Byte
Fourth Byte
10xxxxxx 10yyyyyy 10uuzzzz
10xxxxxx 10yyyyyy
10xxxxxx
Table 4 exemplifies that UTF-8 is equivalent to ASCII, at least in the range of x00 to 0x7F. The higher values (>= x80) are represented as a sequence of two bytes then. 1.1.2. Character Sets in Messages The easiest way to create a message instance is to place the information from the record set out of the database directly into a message instance – without any further processing. Instead of this wrong procedure, a correct encoding is required. Hence, the escaping of delimiters occurring in the data, the declaration of the character set in use (HL7 v2.x MSH-18 Character Set) and the correct encoding of individual characters of the data is essential (Figure 2).
F. Oemig and B. Blobel / Character Sets
151
Most of the implementers would data (in character set data (in character set of database) of database) suspect, that 8 bits are allowed per default in HL7 v2.x messages. Chapter 2.14.9.18 explains why this encoding decoding assumption is wrong: If the field is left blank, the data (encoded for data (encoded for transmission in transmission in character set in use is understood to message) message) be the 7-bit ASCII set, decimal 0 transmission through decimal 127 (hex 00 through hex 7F). This default value may also Figure 2. Handling of Character Sets be explicitly specified as ASCII. Common practice (e.g. in Germany) allows for using standard 8-bit ISO 8859-x characters which must be declared appropriately. However, for the Russian environment this behavior is even not sufficient, because Unicode as a representation for all Cyrillic characters becomes necessary. HL7 Table 0211 “Alternate character Sets” lists all character sets which are allowed in HL7 v2.x messages. The original code “UNICODE“ has been replaced by “UNICODE UTF-8” in order to remove the ambiguity between code points and their transformation format. 1.2. Transliteration If a specific application is unable to understand and process e.g. Cyrillic information, the characters must be converted into another representation – the so-called transliteration. The information conveyed with MSH-18 allows for transliteration into ASCII/Latin characters according to the following algorithm [6, 7]. This should be combined with the declaration of a standard character set like Latin 8859/1 or ASCII given the correct character conversion. 1.2.1. Transliteration Algorithm The transliteration algorithm for Cyrillic is based on the national standard GOST 7.792000 [8]. This standard is accepted by the following countries: Azerbaijan, Armenia, Belorussia, Kazakhstan, Kyrgyzstan, Russia, Tajikistan, Turkmenistan, and Uzbekistan. This standard defines 2 possible ways for transliteration: a) This system defines one Cyrillic character to one Latin character, some with diacritics; b) This system defines one Cyrillic character to one or many Latin characters without diacritics. For transliteration the second option should be used with the following transliteration rules: 1.
Following Latin characters are used for transliteration: а, b, с, d, е, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y, z, ’, `
2.
Cyrillic characters should be replaced by Latin as it is shown in a Table 5.
3.
If the word contains a combination of uppercase and lowercase letters, transliterating an uppercase letter as a combination of letters should use a capital letter only for the first letter of the resulting combination. If a word contains only uppercase letters, all letters in a transliterated word should be uppercase. (Examples: Жуков => Zhukov; МЧС => MCHS).
152
F. Oemig and B. Blobel / Character Sets
4.
Arabian numerals (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0) and Roman numerals (I, V, X, С, М, L, D) should be kept without changes.
5.
Non-literal Cyrillic character № should be replaced by character #.
1.2.2. Transliteration Table The following table provides the transliteration from Unicode to ASCII/Latin, but not vice versa. Upper and lower case letters are defined by the algorithm above (Table 5): Table 5. Latin equivalents of Cyrillic Characters used in Russian Language (extract)
Upper Unicode Й К Л М Ч Ш ..
Case Character Й К Л М Ч Ш ..
1.2.3. Un-transliteration Algorithm
Lower Unicode й к л м ч ш ..
Case Character й к л м ч ш ..
Transliterated Character j k 1 m ch sh ..
Table 6. Replacement of characters in un-transliteration (extract)
Text being converted using the second option above allows for an un-transliteration back to Cyrillic characters. This untransliteration should be done in 3 steps: replacement of 3letters combination → replacement of 2-letters combination → replacement of 1-letter combination (Table 6):
Step 1 2
3
Step order Searching and replacement of 3-letter combinations Searching and replacement of 2-letter combinations
Searching and replacement of 1-letter combinations
Latin shh
Cyrillic щ
Ya yo yu .. a b ..
я ё ю .. a б ..
2. Results 2.1. Encoding Options Generally, different options for an encoding exist. If more than a single character set is used, an escape sequence to switch among them is needed (Table 7): Table 7. HL7 ER7 Escape Sequences (extract) Escape Sequenz \C2842\ \C2D4C\ \C2D4D\
Character Set ISO-IR 6 G0 (ISO 646 : ASCII) ISO-IR 144 (ISO 8859: Cyrillic) ISO-IR 148 (ISO 8859 : Latin Alphabet 5) UTF-8
Comment
no switch allowed
As can be seen, no sequence is defined for Unicode. As a prerequisite, the message header (MSH-18) has to state implicitly which option is implemented. In the following,
F. Oemig and B. Blobel / Character Sets
153
the different options when used in a message instance and their implications are analysed in more detail: 2.1.1. ASCII as the only Character Set This option does not fulfill the requirements because ASCII does not provide Cyrillic characters. It can be used only in combination with transliteration. 2.1.2. ASCII with Latin 8859/5 as an Alternative This would work, but then both character sets should be mentioned in MSH-18. 2.1.3. Latin 8859/5 as the only Character Set This works as well, but then only this character set must be mentioned in MSH-18. 2.1.4. UTF-8 as the only Character Set This option would fulfill the requirements, but a) the delimiters must be constrained to come from the singe byte range, i.e. between 0x20 and 0x7F, and b) a few parsers may be unable to handle the used characters properly, but it should not harm the parsing process in general. 2.1.5. UTF-8 as the Alternative Character Set to ASCII This option is not allowed, so that no escape sequence exists to switch in between. 2.1.6. Built-in Escape Sequences To cite from the Implementation Support Guide [3]: JIS X 0202 - ISO 2022 provides an escape sequence for switching among different character sets and among single-byte and multi-byte character representations. Japan has adopted ISO 2022 and its escape sequences as JIS X 0202 in order to mix Kanji and ASCII characters in the same message. Both the single- and multiple-byte characters use only the low order 7 bits in JIS Kanji code with JIS X 0202 in order to ensure transparency over all standard communication systems. When HL7 messages are sent as JIS X 0202, all HL7 delimiters must be sent as single-byte ASCII characters, and the escape sequence from ASCII to Kanji and back again must occur within delimiters. In most cases the use of Kanji will be restricted to text fields It is recommended to use a single character set only. Therefore, all characters should be encoded either with Latin 8859/5 or Unicode in UTF-8. The selection must be declared in MSH-18. 2.2. Use of Transliteration The algorithm explained above should be provided in case if one of the receiving systems cannot handle it. Depending on the system’s capabilities either transliteration or un-transliteration must be done. This can be provided either internally by a system or externally by another third party product/service (Figure 3).
154
F. Oemig and B. Blobel / Character Sets
3. Discussion DB
An open issue with (file-based) transliteration messaging is the question whether according to the Byte-Order message creation Management (BOM) [9] the message should be pre-fixed with a special message processing character sequence? Unicode transliteration regulations highly recommend the use of this information to avoid conflicts DB in determining the correct UTF. On the other hand, this sequence enforces Figure 3. Handling of Transliteration that the HL7 v2.x messages start with an unknown (and probably unexpected) set of characters. Standard XML encoding demonstrates that it also works without. 4. Conclusions The analysis about handling character sets properly identifies some additional constraints onto possible delimiters. To ensure a compatibility with UTF-8 they must be in the range of printable 7-bit characters. Furthermore, each vendor has to state clearly in his documentation, which character sets are supported and whether his application supports (un)transliteration or not. In addition the character set in use must clearly be identified in MSH-18. Otherwise (un)transliteration becomes impossible. Therefore, MSH-18 must be required (“R”).
References [1] [2] [3]
[4] [5] [6] [7] [8] [9]
HL7 Russia: www.hl7.ru, www.hl7-russia.org, last visited 15.Nov 2010 HL7: Health Level Seven, Inc., www.hl7.org , last visited 15.Nov 2010 HL7 v2.3.1 Implementation Support Guide: http://www.hl7.org/DocumentCenter/private/standards/v2xig/Imp.%20Support%20Guide%20%20V2.3.1.zip, last visited 15.Nov 2010 Unicode for Cyrillic characters: http://www.unicode.org/charts/PDF/U0400.pdf, last visited 15.Nov 2010 Unicode FAQ: http://unicode.org/faq/utf_bom.html, last visited 15.Nov 2010 Lingua Systems: How to transliterate Russian Text, http://www.linguasystems.com/blog/2010/07/30/How-to-Transliterate-Russian-Text.html, last visited 15.Nov.2010 Online Russian Transliterator, http://translit.us, last visited 15.Nov.2010 GOST 16876-71 (7.79 2002) , last visited 15.Nov 2010 -BOM: Byte-Order Mark, http://unicode.org/faq/utf_bom.html, last visited 11.February 2011
The Role of Basic Data Registers in CrossBorder Interconnection of eHealth Solutions Mirjana KREGAR, Tomaž MARČUN1, Irma DOVŽAN and Lojzka ČEHOVIN Health Insurance Institute of Slovenia
Abstract. The increasingly closer international business cooperation in the areas of production, trade, transport and activities such as tourism and education is promoting the mobility of people. This increases the need for the provision of health care services across borders. In order to provide increasingly safer and effective treatment that is of ever higher quality in these cases as well, it is necessary to ensure that data accompanies patients even when they travel to other regions, countries or continents. eHealth solutions are one of the key tools for achieving such objectives. When building these solutions, it is necessary to take into account the different aspects and limitations brought about by the differences in the environments where such a treatment of a patient takes place. In the debates on the various types of cross-border interoperability of eHealth solutions, it is necessary to bring to attention the necessity of suitable management and interconnection of data registers that form the basis of every information system: data on patients, health care service providers and basic code tables. It is necessary to promote well-arranged and quality data in the patient’s domestic environment and the best possible options for transferring and using those data in the foreign environment where the patient is receiving medical care at a particular moment. Many of the discussions dealing with conditions for the interoperability of health care information systems actually start with questions of how to ensure the interconnectivity of basic data registers. Keywords. data registers, data quality, interoperability, online access.
1. Basic Data Registers The term »basic data register« we define as a register of data that represents the basis for other registers. Such basic data include code tables and data on main entities that operate in a business system. The health care information system uses many code tables defining medical data (e.g. diagnoses, medicinal products) and code tables that designate business or administrative data (e.g. services, materials). Some code tables are defined at the international level by authorised organisations or standardisation bodies. Other code tables are defined by entities at the state or regional levels. The third type of code tables are defined entirely at the local level, for example within a hospital. 1
Corresponding Author: Tomaž MARČUN, Head of Application Development Department, Health Insurance Institute of Slovenia, Miklosiceva 24, 1507 Ljubljana, Slovenia, E-mail: [email protected].
156
M. Kregar et al. / The Role of Basic Data Registers
Among entities that operate in the health care system we classify patients, health care service providers and health care workers. Patients are natural persons using health care services. Providers are legal entities or natural persons that have obtained licenses from the competent body to perform a certain health care service. Health care workers are those workers who have obtained suitable medical education and passed the certification examination. When setting up basic data registers it is very important to define unique identifiers for every data entry. This eliminates the need to store large set of data in other registers, but rather to keep only the identifiers which point to relevant entries in basic data registers – e.g. when storing data of patient visits it is wise to store data of patients separate and the register of visits to contain only the identifiers of patients. When recording medical or administrative data in health care, we should insist on the basic data register identifiers being used as consistently as possible, meaning that they are used for the majority of entries. It is also important that the identifiers are chosen very precisely – for example, that an entry on medical treatment is tied to an identifier of the right patient. The use of basic data registers is an indispensable basis for the development of medical statistics, managing health care insurance claims and the management of health care systems because only data furnished with basic register identifiers enables quality reporting, data analyses and control procedures [11].
2. Challenges in Using Basic Data Registers in Health Care The use of basic data registers in health care is essential as it importantly affects data quality and thereby the safety and quality of health care services. For example, the result of an error in the record of the history of prescribed medications of a particular person can lead to an incorrect professional medical decision. The awareness of the importance of well-arranged basic data registers in health care is insufficient. Different areas still mostly use different data registers. There are many opportunities for improving the quality and use of basic registers. There are frequent cases when, for example, the linking of information from different environments makes it impossible to ascertain whether this information belongs to one and the same patient. A great challenge is also the provision of these data in environments where the patient is currently located and is using health care service. Code tables are prescribed at the national level, but are not accessible in electronic format, which would make it possible to automatically transfer the data and its changes to health care information systems in all locations where health care services are provided. A national data register on the citizens or insured persons has been set up in several countries, however in the health care system data of patient’s name, surname and address are copied from personal identification documents or other written certificates or plastic cards, even though the telecommunications network and the World Wide Web technologies make it possible to transfer data between information systems rapidly, automatically and securely.
M. Kregar et al. / The Role of Basic Data Registers
157
3. Experience in the Compilation of Basic Registers in Slovenia We have long-standing experience in the compilation of basic data registers at the national level in Slovenia [9]. Emphasis is placed on the care for the quality and security of data. Information solutions for electronic access to these registers have been provided. The use of basic data registers is not intended exclusively for compiling statistical and accounting data but also provides support to the professional work of doctors and other medical staff. The Institute of Public Health of the Republic of Slovenia and the Health Insurance Institute of Slovenia compile and manage national code tables that are available to the providers of health care services in electronic format. They also keep a central register of medicinal products that are used on the market of the Republic of Slovenia. A regularly updated abovementioned register is available to the health care service providers in electronic format, which they can download and transfer to their information systems. The two mentioned institutions also compile a register of health care service providers and health care workers [10], which is included in the nationally connected information solutions. The Health Insurance Institute of Slovenia compiles registers on persons who are included in compulsory health insurance. The database is linked to the Central Population Register that is compiled by the Ministry of the Interior and to the Business Register managed by the Agency of the Republic of Slovenia for Public Legal Records and Related Services. Before patients are provided health care treatment, they identify themselves with the health insurance smart card and the health care service providers obtain the basic (administrative) data on that patient as well as some other data that assist them in taking medical decisions (e.g. data on prescribed medications) and assist them in the communication between health care workers (e.g. information on the patient’s personal general practitioner, dentist or gynecologist) from the central health insurance registers[4,5]. Such direct data exchange is made possible by the online system introduced in the 2009–2010 period, into which all the health care service providers from the public health care network have been connected [2]. The system involves 21,000 health care workers who exchange over 1,100,000 data messages daily [1]. The described basic registers contain national identifiers that are used in the recording of other data in health care. Every medicinal product, for which an approval for use was obtained in accordance with the procedure of the national Agency of Medicinal products or centralised procedure of the European agency, has a national code allocated in the central register of medicinal products. The health insurance number is provided for every insured person in addition to their citizen identifier. National identifiers are also defined for health care service providers and health care workers. The method for the provision and use of basic data registers in Slovenia is shown in greater detail in Table 1. 4. Basic Data Registers as the Foundation for Cross-Border Interconnection When linking information solutions across borders, numerous questions spring up, namely how to ensure uniformity and accessibility to basic data registers in the expanded environment and thereby make use of the indispensable advantages these registers provide for the quality and effectiveness of health care processes.
158
M. Kregar et al. / The Role of Basic Data Registers
Table 1. The method for the provision and use of national basic data registers in Slovenia Type of basic data Identifiers
Source
Administrator
Method for transferring data into health care information systems
Data on patients Personal registration number, health insurance number
Record on compulsory health insurance, voluntary health insurance. Records are linked to national population registers and business entity registers.
The Health Insurance Institute of Slovenia, insurance companies offering voluntary health insurance. National registers are compiled and kept by the Ministry of the Interior and the Agency of the Republic of Slovenia for Public Legal Records and Related Services.
The identifier is stored on the health insurance card. Other data is obtained by the providers directly in the online system.
Data on health care service providers Provider’s number, health care worker’s number
Register of providers
The Institute of Public Health of the Republic of Slovenia and the Health Insurance Institute of Slovenia.
Data are included in the national interconnected solutions such as the online system.
Data on the medicinal products on the market National medicinal product code
Central register of medicinal products
The Institute of Public Health of the Republic of Slovenia and the Health Insurance Institute of Slovenia. There is a project underway that will incorporate the Agency for Medicinal Products and Medical Devices of the Republic of Slovenia into the data provision scheme.
Data are used in the national interconnected solutions and are available in the form of a Web application or in XML format for downloading and transferring into the information systems of health care service providers.
Various code tables Code table identifiers
Code tables
The Institute of Public Health of the Republic of Slovenia and the Health Insurance Institute of Slovenia.
Data are used in the national interconnected solutions and are available in electronic format on the Web for downloading and transferring into the information systems of health care service providers.
An ideal solution would be a joint agreement at the international level on the use of uniform basic data registers, meaning that a uniform patient register would for example be established at the European level as well as a uniform register of health care service providers and health care workers in addition to as many uniform health care code tables as possible. Because of the high requirements regarding the protection of
M. Kregar et al. / The Role of Basic Data Registers
159
personal data, we cannot expect that a joint European patient register will be established. But with the gradual opening up of the health care service market, we can expect calls to emerge for the establishment of joint code tables and the compilation of primarily regional and national records on the health care service providers and health care workers as these databases are the precondition for building interconnected information solutions (e.g. for the compilation of records on persons authorised to access personal medical data). Many code tables in the area of health care have already been uniformly defined in the international environment such as for example the International Classification of Diseases. Efforts are already underway that are aimed at drafting definitions of new code tables within the scope of international institutions and standardisation bodies. In the EU, we have gradually been establishing databases that are linked to centralised procedures such as for example the compilation of a database of medicinal products kept at the European Medicines Agency. It will eventually probably be necessary to set up uniform code tables for procedures and services. The majority of the code tables are not available in arranged electronic format (e.g. in XML format) which would make it possible for the data to be transferred into health care information systems. It is usually necessary to transfer the data manually and reformat them prior to importing them into the information system. An information service that would enable the highest possible level of automation in the acquisition of health care code tables would be highly welcome. It would be necessary to ensure accessibility of data in such a format that would make it possible for the users to acquire changed data in a timely fashion. Data on patients, health care service providers and health care workers are recorded at various levels in different countries. In some places, the registers are set up at the national level, while in others they are kept at the regional level, whereas in countries with health insurance systems they are usually kept at the level of an individual health care fund. Different identifiers and different agreements regarding the data type and format (e.g. permitted length of data fields for storing last and first name of a person, different code formats) are used in each country. In case services are made use of across borders, the health care service provider usually gathers a patient’s basic data from the personal identification document such as a passport or identity card. Most of these documents contain a record that can be read automatically using devices for optical text recognition; however health care workplaces are usually not equipped with devices for machine reading of such texts. Modern passports contain a chip, but the data stored on it are mostly intended for internal affairs purposes. The European Unions has prescribed the use of the European Health Insurance Card for patient identification and for facilitated procedures for refunding the costs of using urgent health care services abroad. This is a plastic card, on which data is not stored so that it would be possible to read them electronically. The set of data on the card is quite limited. Data on the patient comprise the patient first and last name as well as an identifier. Other electronic data can be obtained using a portable medium (e.g. smart card) or via direct cross-border access to data. Before the patient has been provided health care treatment, health care workers would obtain the patient’s detailed personal information by way of direct remote access to the register of patients in another region or country. One of the preconditions for the use of such a solution is the acquisition of the right identifier or other data that uniquely identify the patient and are used for enquires in the remote register.
160
M. Kregar et al. / The Role of Basic Data Registers
Direct, cross-border access to data has an advantage in that it enables the acquisition of completely up-to-date data. There were concerns years ago that such access was not economical due to the slow speed of telecommunications, security risks and high costs of data transfer. With the increase in the reliability of communications, decrease in prices and the increasingly more effective data security mechanisms, such access is becoming an increasingly more viable solution. Similarly as in the case of exchanging patient it is possible to ensure transfers of data on health care providers and health care workers. Needs are arising primarily in the provision of data for communication between the health care service providers and health care workers – e.g. for the establishment of communication between the patient’s general practitioner in the home environment and the specialist doctor abroad where the patient has recently made use of health care services.
5. Conclusion Basic data registers are a very important basis of every health care information system. Suitable quality and accessibility of basic data registers at the local level is a precondition for the effective use of such databases in cross-border interconnected eHealth solutions. Closer cross-border business interconnection of health care processes requires the definition of uniform international code tables and the provision of efficient crossborder access to basic data registers. The development of telecommunications and the expansion of the use of telecommunications networks in health care will make it possible to directly and immediately acquire remote access to the basic data registers on patients, health care service providers and health care workers. Slovenian experience shows that the development of technologies is already making it possible to set up such solutions.
References [1] T. Marčun, After the Implementation of the On-line System, Knowledge for Successful eHealth, Proceedings of the Congress of the Slovenian Medical Informatics Association, Zreče, October 2010 [2] T. Marčun, On-Line Data Exchange in Slovenian Health Care and Health Insurance. K.-P. Adlassnig, B. Blobel, J. Mantas, I. Masic (Edrs.) Medical Informatics in a United and Healthy Europe, Proceedings of MIE 2009, 48-52, SHTI Vol. 150, IOS Press, Amsterdam, 2009 [3] T. Marčun, After the Implementation of the On-line System, Knowledge for Successful eHealth, Proceedings of the Congress of the Slovenian Medical Informatics Association, Zreče, October 2010 [4] M. Sušelj, T. Marčun, New Generation of the Slovene Health Insurance Card, Smartcard workshop 2008, Darmstadt, February 2008 [5] T. Marčun, A. Bolka, On-line Health Insurance in Slovenian Health Care – Solutions and Pilot Implementation, Congress of the Slovenian Medical Informatics Association, Zreče, October 2008 [6] eHealth 2010, National Strategy, Republic of Slovenia, Ministry of Health, 2006 [7] Strategic Development Plan of the Health Insurance Institute of Slovenia, 2008 [8] Project documentation, Health Insurance Institute of Slovenia, 2006-2009 [9] T. Marčun, F. Košir, Integrated Databases – The Foundation for the Information Linking of the Actors in the National Health Care and Health Insurance Systems, Proceedings of MIE 1999 [10] T. Albreht, M. Paulin, National Health Care Providers’ Database (NHCPD) of Slovenia – Information Technology Solution for Health Care Planning and Management, Proceedings of MIE 1999 [11] A. Bolka, Knowledge and Data for Sustainable eHealth Insurance, Congress of the Slovenian Medical Informatics Association, Zreče, October 2008
Kristiina HÄYRINENa,1 and Juha MYKKÄNEN b University of Eastern Finland, Department of Health and Social Management, Kuopio, Finland b University of Eastern Finland, School of Computing, HIS R&D, Kuopio, Finland
Abstract. The unified content of EHRs promote shared understanding of patient data, information exchange between information systems and health care organizations, data retrieval from EHR and reuse of data for administrative purposes, statistical analysis or clinical research. The purpose of this study was to analyze to what extent Finnish national headings can be coded with the current version LOINC. Ten (37%) of national headings can be mapped to LOINC terms in clinical class. There were LOINC terms in other classes which correspond to headings. Furthermore, inconsistency exists in the names of headings. The need for mapping national headings to all terms in LOINC is needed. Keywords. Electronic health records, classifications
Introduction The content and structure of electronic health record (EHR) have been developed during a long period of time in various initiatives [1-3]. An EHR should have a clear structure to promote shared understanding of patient data, information exchange between information systems and health care organizations, data retrieval from EHR and reuse of data for administrative purposes, statistical analysis or clinical research [2,3]. The free text form introduces barriers to search, summarization, decision support, or statistical analysis. Information extraction from narrative documents of an EHR is still rarely used outside laboratories where information extraction systems have been developed [4]. The balance between structured and coded data in relation to unstructured (narrative, free text) data is one of the challenges in EHR development work. The unified content of the Finnish national EHR has been defined based on proposals for paper-based patient records and information content of widely used EHR systems [5]. Agreement on the national minimal unified structure of the EHR was reached by means of nationwide consultation and expert groups, which represented different domain experts: physicians, nurses, computing specialists, statisticians, health care administration experts and researchers. The need to model content of the Finnish national EHR within the context of a standard vocabulary has been identified [6].The
1
Corresponding Author: Kristiina Häyrinen, University of Eastern Finland, Department of Health and Social Management, Kuopio, Finland; Email: [email protected].
162
K. Häyrinen and J. Mykkänen / Mapping the Finnish National EHR to the LOINC
purpose of this study was to analyze to what extent Finnish national headings can be coded with the current version LOINC.
1. Background In Finland, the unified content of EHR consists of documents composed from harmonised data elements. The internal organisation of the documents follows the idea of grouping meaningful data under section headings which provide the context for narrative text, e.g. [7], [8], [9]. The core data elements (structured data entries) which require the use of vocabularies, nomenclatures and classifications, are located under their corresponding headings. One of the aims was to achieve semantic interoperability of health information systems. The national recommendations and guidelines for EHRs have been agreed in 2007 [10]. A list of multi-professional national headings (specifies names, codes and descriptions) is available through the national code server. Table 1. National headings of EHR in Finland Heading
Example of content
Aims for care
Treatment goals
Anamnesis
Medical, family and social history.
Assessment, end
A description and an analysis of the patient's received treatment at end of the episode of care.
Assessment, intermediate Consultation
A description and an analysis of the patients' received treatment during the episode of care. A request to consultation or consultation response.
Diagnosis
Medical diagnosis.
Functional status
Patients' ability to cope with physical psychological social and cognitive demands related to activities of daily living. Health check-up. For example special inspections of growing children and young people in monitoring the development or examinations the capability of work or education. Manners of living, e.g. smoking, the use of alcohol.
Health examination
Health patterns (life style) Health status Intensity of care Medical statements (certifications)
Medication Nursing diagnosis
Systematic and thorough inspection of the patient for physical signs of disease or abnormality. Degree of patient needs for nursing care. A medical statement is based on medical experts’ assessment of the patient, e.g. for a court of law. A medical certificate is a structured written document on the patient’s disease written by a physician, e.g. for the patient’s employers. Information related to medication.
Outcomes of care
A clinical judgment about individual, family, or community responses to actual or potential health problems/life processes. Nursing action (treatment, procedure or activity) designed to achieve an outcome to a diagnosis, nursing or medical, for which the nurse is accountable. Results of care.
Physiological measurements
Results of physiological measurements e.g. vital signs, oxygen saturation.
Nursing interventions
K. Häyrinen and J. Mykkänen / Mapping the Finnish National EHR to the LOINC
163
Table 1. Continued
Problems
Specific practices for the prevention of disease e.g. health promotion, vaccination. Patients’ problems.
Reason for care
Reason for care or visit.
Rehabilitation
Restoration of human functions in a person or persons suffering from disease or injury. An aspect of personal behavior or lifestyle, environmental exposure, or inborn or inherited characteristic, which, on the basis of epidemiologic evidence, is known to be associated with a health -related condition considered important to prevent.
Preventive measures
Risk factors
Surgical procedures Technical aids
Operations carried out for the correction of deformities and defects, repair of injuries, and diagnosis and cure of certain diseases. Technical aids for disabled persons.
Test and assessment results
Results of different tests such as psychological tests or nutrition assessment.
Tests and examination
Results of laboratory tests and radiology examinations.
The information about followup treatment
Follow-up treatment plan designed by medical nursing or other health care professional.
The HL7 CDA release 2 has been adopted as the standard for XML document data exchange, encoding and structure. In HL7 Clinical Document Architecture (CDA), CDA release 1 (CDA R1) focused on the structured header, and CDA release 2 (CDA R2) introduced the concept of structured elements within the document body. The body of a CDA document contains one or more sections. Section title represents the human readable label of a section (e.g. heading). Label of a section describes section text which is unstructured data and contains the human readable content of the section [11]. Logical Observation Identifier Names and Codes (LOINC) vocabulary is a organised set of terms or words with associated codes developed for use as clinical observation identifiers in standardised messages exchanged between information systems. LOINC version 2.32 contains 58,967 terms. The clinical portion of the LOINC database covers the major headings of history and physical, discharge summary, and operative note reports [12]. Therefore, we decided to map Finnish national headings to LOINC.
2. Methods The national headings or synonyms were entered term by term to Regenstrief LOINC Mapping Assistant (RELMA). The mapping was made by one author ensuring that the definitions of LOINC components were conceptually consistent with national headings. Furthermore, we selected that type of scale of LOINC was narrative and class type was clinical class.
3. Results Ten (37%) of national headings can be mapped to LOINC (see Table 1).
164
K. Häyrinen and J. Mykkänen / Mapping the Finnish National EHR to the LOINC
Table 2. National headings mapped to LOINC codes. Heading
LOINC
Component
Scale
Type
Diagnosis
11535-2
Hospital discharge Dx
Nar
2
Anamnesis
11329-0
History general
Nar
2
Intensity of care
56839-4
Acuity assessment
Nar
2
Health status
29545-1
Physical findings
Nar
2
Problems
57852-6
Problem list
Nar
2
Reason for care
29299-5
Reason for visit
Nar
2
Aims for care
61146-7
Goals
Nar
2
The information about followup treatment Physiological measurements
11544-4
Hospital discharge follow-up
Nar
2
29545-1
Physical findings
Nar
2
Surgical procedures
10223-6
Surgical operation note surgical procedure
Nar
2
Part of headings such as “Reason for care”, “Anamnesis”, “Health status”, “Diagnosis” and “Physiological measurements” can be mapped to several terms of LOINC. Headings are less specific than the corresponding LOINC term. (see examples in Table 2) For example the heading "Reason for care" can be mapped to three different terms, "Diagnosis" can be mapped to six different terms or "Anamnesis" can be mapped to 57 different terms. Table 3. Example of headings mapped to LOINC Heading
LOINC
Component
Scale
Reason for care
46239-0
Chief complaint+Reason for visit
Nar
2
57827-8
Reason for co-payment exemption
Nar
2
29299-5
Reason for visit
Nar
2
10219-4
Surgical operation note preoperative Dx
Nar
2
10218-6
Surgical operation note postoperative Dx
Nar
2
59769-0
Postprocedure diagnosis
Nar
2
11535-2
Hospital discharge Dx
Nar
2
46241-6
Hospital admission Dx
Nar
2
29548-5
Diagnosis
Nar
2
Diagnosis
Type
4. Discussion The purpose of this study was to analyze to what extent Finnish national headings can be coded with the current version LOINC terms in "clinical" class. One limitation of this study is the translation of Finnish national headings to English and selection of synonyms for headings by one author. This poses risk of subjective translation and mapping. Based on this study less than half of national headings can be mapped to one or more LOINC terms in "clinical" class. These headings include reason for care, anamnesis, status, follow-up treatment plan and diagnosis which have been noted to be meaningful headings in physician documentation in Finland [5]. Several issues were found when other headings were mapped to LOINC. First, the level of specification between the headings and LOINC terms is not equivalent.
K. Häyrinen and J. Mykkänen / Mapping the Finnish National EHR to the LOINC
165
Headings "Tests and examinations" and "Nursing interventions" are broader concepts than the closest matching LOINC terms. For example, the LOINC classification includes term "relevant diagnostic tests/ laboratory data". Although, the "relevant diagnostic tests/ laboratory data" does not cover radiology examinations. Correspondence to "Life style" was not found but terms concerning e.g. alcohol use or smoking are included in LOINC although the type of scale is not narrative. Furthermore, history of smoking or history of alcohol use which type of scale are narrative can be found in LOINC. Second, LOINC include also other classes than “clinical” class namely “laboratory” "claims attachments" and "surveys". Some of the national headings can mapped to terms in "Claims attachments" and "Surveys" class instead of "clinical" class. In "Claims attachments" class there are terms which correspond to headings "Technical aids", "Nursing diagnosis", "Medical statements", "Consultation" (consultation request) and "Functional status". Each nursing diagnoses and nursing intervention has a corresponding term in "Survey” class. Third, LOINC also includes a classification of whether LOINC code can be used for a full document, a section of a document, or both. Some headings namely “Medication”, "Risk factors", "Assessment end", "Functional status" and "Radiology examinations" match the names of documents in LOINC instead of sections. Finally, some headings are missing in LOINC, namely the “Test and assessment results”, “Rehabilitation”, “Health examination”, "Assessment, intermediate", “Preventive measures” and “Outcomes of care”. The purpose of the national headings in Finland is to structure narrative text in a coarse level. Based on this study, the LOINC terms do not cover all national headings of health professionals such as nurses or physiotherapists [8]. In Finland these healthcare professionals document their discharge summaries independently or in some cases related to physicians’ documentation. The further validation of national headings to LOINC is necessary. The national headings should map to terms of LOINC without presumptions of type of scale or class of LOINC. In future, it is also necessary to find out which information is necessary to structure with headings at all. Some parts of EHR such as medication and risk factors are entirely structured and in these cases the use of headings in narrative text might be unnecessary.
References [1] [2] [3] [4]
[5]
[6] [7]
H.J. Tange, A. Hasman, P.F. de Vries Robbe, H.C. Schouten, Medical narratives in electronic medical records, Int.J.Med.Inform 46 (1) (1997), 7-29. A.M. van Ginneken, The computerized patient record: balancing effort and benefit, Int.J.Med.Inf. 65 (2) (2002), 97-119. J Grimson, Delivering the electronic healthcare record for the 21st century, Int.J.Med.Inf. 64 (2-3) (2001), 111-127. S.M. Meystre, G.K. Savova, K.C. Kipper-Schuler, J.F. Hurdle, Extracting information from textual documents in the electronic health record: a review of recent research, Yearb.Med.Inform. (2008), 128144. The standardized data content of national EHR. Guide to implementation of the core data elements, headings, and documents as well as structured data elements of specialties and activities in EHR. Version 3. Finnish. Available from: https://www.kanta.fi/c/document_library/get_file?uuid=46b8b38a3488-4c6e-81d2-ae8dcfeaf848&groupId=10206 K. Häyrinen, K. Harno, P. Nykänen, Use of Headings and Classifications by Physicians in Medical Narratives of EHRs -An Evaluation in a Finnish hospital (Accepted 2011). H. Åhlfeldt, M. Ehnfors, L. Ridderstolpe, Towards a multi-professional patient record--a study of the use of headings, Stud.Health Technol.Inform. 68 (1999), 813-817.
166
[8]
K. Häyrinen and J. Mykkänen / Mapping the Finnish National EHR to the LOINC
S. Hyun, S. Bakken, Toward the creation of an ontology for nursing document sections: mapping section names to the LOINC semantic model, AMIA.Annu.Symp.Proc. (2006), 364-368. [9] S. Kay, Ontological and epistemological views of 'headings' in clinical records, Stud.Health Technol.Inform. 84(Pt 1) (2001), 104-108. [10] A. Iivari, P. Ruotsalainen, eHealth Roadmap – Finland, Ministry of Social Affairs and Health Reports 2007:15, Ministry of Social Affairs and Health, Helsinki. [11] R.H. Dolin, L. Alschuler, S. Boyer, C. Beebe, F.M. Behlen, P.V. Biron, et al., HL7 Clinical Document Architecture, Release 2, J.Am.Med.Inform.Assoc. 13 (1) (2006), 30-39. [12] C. McDonald, S. Huff, K. Mercer, J.A. Hernandez, D.J. Vreeman, Logical Observation Identifiers Names and Codes (LOINC®), Users' Guide - December 2010.
Subject Index Adaptive Neuro-Fuzzy Inference System (ANFIS) 37 archetypes 43 architecture 57 barriers 123 certification criteria 129 character sets 149 classifications 135, 161 clinical system 135 CME 111 computer network 94 computer systems 94 computerized 135 computing methodologies 135 cross-border collaboration 21 DASTA 143 data quality 155 data registers 155 data security 68 database 94 database management systems 94 demographics 143 direct access 68 education 21 eHealth 57, 68, 129 e-learning 3, 111 electronic health record (EHR) 43, 63, 129, 135, 161 electronic patient record 99 emergency intervention 88 EMR access 88 evaluation 111, 143 experiences based assessment methodology 129 family doctor offices 129 framework 143 fuzzy logic 37 fuzzy system 83 Generic Component Model 11 Genetic Algorithm (GA) 37 graduate- and post-graduate studies 21 guidelines 49
health promotion 105 Health-Related Quality of Life 31 HL7 CDA 117 home-care 88 hospital Information System 123 HPLC (High-Performance Liquid Chromatography) 37 Human-Computer Interaction 31 IHE PAM 143 impact evaluation 63 implementation 123 information presentation 88 information retrieval 88 information security 74 integration 123 interoperability 149, 155 ISO/EN 13606 43 Knowledge Management Systems 31 languages 149 medical data mining 117 medical decision 49 medical imaging 83 medical informatics 21, 94 medical record 68 mental health 105 model 3 national infrastructure 57 national strategy 57 network bandwidth 99 nicotine detection 37 occupational medicine 111 oncologic patients 31 online access 155 on-line system 74 ontology 11 patient identifier 68 patient portal 63 personal information protection 74 pHealth 11 preventive program 105 primary care data 135 privacy 68 protocols 49
168
regions reliability remote collaboration remote monitoring research review scenario self-image semantic interoperability server-client model smart cards
3 129 88 88 135 63 99 105 11, 43 99 74
standards sustainable telemedicine syntax text parsing thin-client computing vocabulary controlled Web counseling WEKA XML youth
Author Index Allaert, F.-A. 68 Ammenwerth, E. 43, 63 Auverlot, B. 68 Avdagic, Z. 37 Bainbridge, M. 135 Bari, F. 21 Begic Fazlic, L. 37 Benzenine, E. 68 Blobel, B. v, 11, 149 Bolka, A. 74 Čehovin, L. 155 Coatrieux, G. 68 de Lusignan, S. 135 Dovžan, I. 155 Duftschmid, G. 43 European Federation for Medical Informatics (EFMI), Primary Care Informatics Working Group 135 Finozzi, E. 111 Forczek, E. 21 Frank, L. 94 Gal, N. 83 Gomoi, V. 49 Gonçalves, J. 31 Gri, T. 111 Hantos, Z. 21 Häyrinen, K. 161 Hoerbst, A. 63 Hübner-Bloder, G. 43 Imbriani, M. 111 International Medical Informatics Association (IMIA), Primary Care Informatics Working Group 135 Jaquet-Chiffelle, D.-O. 68 Jereb, B. 105 Kohler, M. 43 Končar, M. 129 Kondoh, H. 99 Konec Juričič, N. 105 Kralj, D. 129
Kregar, M. Kuwata, S. Latifi, R. Lekić, K. Liaw, S.-T. Looser, H. Loškovska, S. Lovis, C. Mantas, J. Marčun, T. Mazzoleni, M.C. Michalakidis, G. Mitsios, A. Mochida, M. Mykkänen, J. Nagy, M. Nakik, D. Oemig, F. Orel, A. Pagani, M. Pape-Haugaard, L. Pearce, C. Quantin, C. Rinner, C. Robu, R. Rocha, Á. Rognoni, C. Saboor, S. Schmid, A. Schnell-Inderst, P. Seidl, L. Shaw, N.T. Silveira, A. Stoicu-Tivadar, L. Stoicu-Tivadar, V. Teramoto, K. Tonković, S. Trajkovik, V. Tratnjek, P. Vicente, M.T. Vida, M. Wagner, J.
