Developments in Healthcare Information Systems and Technologies: Models and Methods Joseph Tan McMaster University, Canada
Medical inforMation science reference Hershey • New York
Director of Editorial Content: Director of Book Publications: Acquisitions Editor: Development Editor: Publishing Assistant: Typesetter: Production Editor: Cover Design:
Kristin Klinger Julia Mosemann Lindsay Johnston Julia Mosemann Casey Conapitski Travis Gundrum Jamie Snavely Lisa Tosheff
Published in the United States of America by Medical Information Science Reference (an imprint of IGI Global) 701 E. Chocolate Avenue Hershey PA 17033 Tel: 717-533-8845 Fax: 717-533-8661 E-mail:
[email protected] Web site: http://www.igi-global.com Copyright © 2011 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher. Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark. Library of Congress Cataloging-in-Publication Data Developments in healthcare information systems and technologies : models and methods / Joseph Tan, editor. p. ; cm. Includes bibliographical references and index. Summary: "This book presents the latest research in healthcare information systems design, development, and deployment,investigating topics such as clinical education, electronic medical records, clinical decision support systems, and IT adoption in healthcare"--Provided by publisher. ISBN 978-1-61692-002-9 (hardcover) -- ISBN 978-1-61692-003-6 (ebook) 1. Medical informatics. 2. Information storage and retrieval systems--Medicine. I. Tan, Joseph K. H. [DNLM: 1. Medical Informatics. W 26.5] R858.D48 2011 610.285--dc22 2010027162 British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library. All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the authors, but not necessarily of the publisher.
Table of Contents
Preface ................................................................................................................................................. xvi Chapter 1 Evaluating Health Information Services: A Patient Perspective Analysis ............................................. 1 Umit Topacan, Bogazici University, Turkey A. Nuri Basoglu, Bogazici University, Turkey Tugrul U. Daim, Portland State University, USA Chapter 2 Gastrointestinal Motility Online Educational Endeavor ...................................................................... 14 Shiu-Chung Au, State University of New York Upstate Medical University, USA Amar Gupta, University of Arizona, USA Chapter 3 Envisioning a National e-Medicine Network Architecture in a Developing Country: A Case Study ........................................................................................................................................ 35 Fikreyohannes Lemma, Addis Ababa University, Ethiopia Mieso K. Denko, University of Guelph, Canada Joseph K. Tan, Wayne State University, USA Samuel Kinde Kassegne, San Diego State University, USA Chapter 4 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR: One Facility’s Approach ...................................................................................................................... 54 Karen A. Wager, Medical University of South Carolina, USA James S. Zoller, Medical University of South Carolina, USA David E. Soper, Medical University of South Carolina, USA James B. Smith, Medical University of South Carolina, USA John L. Waller, Medical University of South Carolina, USA Frank C. Clark, Medical University of South Carolina, USA
Chapter 5 Information Technology (IT) and the Healthcare Industry: A SWOT Analysis ................................. 65 Marilyn Helms, Dalton State College, USA Rita Moore, Dalton State College, USA Mohammad Ahmadi, University of Tennessee at Chattanooga, USA Chapter 6 Using a Neural Network to Predict Participation in a Maternity Care Coordination Program ............ 84 George E. Heilman, Winston-Salem State University, USA Monica Cain, Winston-Salem State University, USA Russell S. Morton, Winston-Salem State University, USA Chapter 7 Can IT Act as a Catalyst for Change in Hospitals? Some New Evidence ............................................ 94 Teemu Paavola, LifeIT Plc, Finland Chapter 8 Informatics Application Challenges for Managed Care Organizations: The Three Faces of Population Segmentation and a Proposed Classification System .................................................. 102 Stephan Kudyba, New Jersey Institute of Technology, USA Theodore L. Perry, Health Research Corporation, USA Jeffrey J. Rice, Independent Scholar, USA Chapter 9 Scrutinizing the Rule: Privacy Realization in HIPAA ....................................................................... 112 S. Al-Fedaghi, Kuwait University, Kuwait Chapter 10 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care? A Comparison among Norway, Denmark, and Sweden .................................................................... 126 Agneta Ranerup, Göteborg University, Sweden Chapter 11 Characteristics of Good Clinical Educators from Medical Students’ Perspectives: A Qualitative Inquiry Using a Web-Based Survey System ................................................................ 145 Gary Sutkin, University of Pittsburgh School of Medicine, USA Hansel Burley, Texas Tech University, USA Ke Zhang, Wayne State University, USA Neetu Arora, Texas Tech University, USA Chapter 12 Open Source Software: A Key Component of E-Health in Developing Nations ............................... 162 David Parry, Auckland University of Technology, New Zealand Emma Parry, National Women’s Health, Auckland District Health Board, New Zealand Phurb Dorji, Jigme Dorji Wanchuck National Referral Hospital, Bhutan Peter Stone, University of Auckland, New Zealand
Chapter 13 An Empirical Investigation into the Adoption of Open Source Software in Hospitals ...................... 175 Gilberto Munoz-Cornejo, University of Maryland Baltimore County, USA Carolyn B. Seaman, University of Maryland Baltimore County, USA A. Güneş Koru, University of Maryland Baltimore County, USA Chapter 14 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching ................. 195 Masoud Mohammadian, University of Canberra, Australia Ric Jentzsch, Compucat Research Pty Limited, Australia Chapter 15 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care ......................................................................................................................................... 214 Jongtae Yu, Mississippi State University, USA Chengqi Guo, James Madison University, USA Mincheol Kim, Jeju National University, South Korea Chapter 16 E-Patients Empower Healthcare: Discovery of Adverse Events in Online Communities .................. 232 Roy Rada, University of Maryland Baltimore County, USA Chapter 17 Towards Process-of-Care Aware Emergency Department Information Systems: A Clustering Approach to Activity Views Elicitation ......................................................................... 241 Andrzej S. Ceglowski, Monash University, Australia Leonid Churilov, The University of Melbourne, Australia Chapter 18 Applying Dynamic Causal Mining in Health Service Management .................................................. 255 Yi Wang, Nottingham Trent University, UK Chapter 19 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems: An Introduction and Literature Survey ............................................................................................... 275 Christos Vasilakis, University College London, UK Dorota Lecnzarowicz, University of Westminster, UK Chooi Lee, Kingston Hospital, UK Chapter 20 TreeWorks: Advances in Scalable Decision Trees .............................................................................. 288 Paul Harper, Cardiff University, UK Evandro Leite Jr., University of Southampton, UK
Chapter 21 Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor ........................................................................................................................ 302 Hussein Atoui, Université de Lyon and INSERM, France David Télisson, Université de Lyon and INSERM, France Jocelyne Fyan, Université de Lyon and INSERM, France Paul Rubel, Université de Lyon and INSERM, France Compilation of References .............................................................................................................. 312 About the Contributors ................................................................................................................... 343 Index ................................................................................................................................................... 348
Detailed Table of Contents
Preface ................................................................................................................................................. xvi Chapter 1 Evaluating Health Information Services: A Patient Perspective Analysis ............................................. 1 Umit Topacan, Bogazici University, Turkey A. Nuri Basoglu, Bogazici University, Turkey Tugrul U. Daim, Portland State University, USA The objective of the chapter is to explore the factors that affect users’ preferences in the health service selection process. In the study, 4 hypothetical health services were designed by randomly selecting levels of 16 attributes and these services was evaluated by the potential users. Analytical Hierarchy Process (AHP), one of the decision making methods, was used to assess and select the best alternative. Chapter 2 Gastrointestinal Motility Online Educational Endeavor ...................................................................... 14 Shiu-Chung Au, State University of New York Upstate Medical University, USA Amar Gupta, University of Arizona, USA Medical information has been traditionally maintained in books, journals, and specialty periodicals. A growing subset of patients and caregivers are now turning to diverse sources on the internet to retrieve healthcare related information. The next area of growth will be sites that serve specialty fields of medicine, characterized by high quality of data culled from scholarly publications and operated by eminent domain specialists. One such site being developed for the field of Gastrointestinal Motility provides authoritative and current information to a diverse user base that includes patients and student doctors. Gastrointestinal Motility Online leverages the strengths of online textbooks, which have a high degree of organization, in conjunction with the strengths of online journal collections, which are more comprehensive and focused. Gastrointestinal Motility Online also utilizes existing Web technologies such as Wiki-editing and Amazon-style commenting, to automatically assemble information from heterogeneous data sources.
Chapter 3 Envisioning a National e-Medicine Network Architecture in a Developing Country: A Case Study ........................................................................................................................................ 35 Fikreyohannes Lemma, Addis Ababa University, Ethiopia Mieso K. Denko, University of Guelph, Canada Joseph K. Tan, Wayne State University, USA Samuel Kinde Kassegne, San Diego State University, USA Poor infrastructures in developing countries such as Ethiopia and much of Sub-Saharan Africa have caused these nations to suffer from lack of efficient and effective delivery of basic and extended medical and healthcare services. Often, such limitation is further accompanied by low patient-doctor ratios, resulting in unwarranted rationing of services. Apparently, e-medicine awareness among both governmental policy makers and private health professionals is motivating the gradual adoption of technological innovations in these countries. It is argued, however, that there still is a gap between current e-medicine efforts in developing countries and the existing connectivity infrastructure leading to faulty, inefficient and expensive designs. The particular case of Ethiopia, one such developing country where e-medicine continues to carry significant promises, is investigated and reported in this article. Chapter 4 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR: One Facility’s Approach ...................................................................................................................... 54 Karen A. Wager, Medical University of South Carolina, USA James S. Zoller, Medical University of South Carolina, USA David E. Soper, Medical University of South Carolina, USA James B. Smith, Medical University of South Carolina, USA John L. Waller, Medical University of South Carolina, USA Frank C. Clark, Medical University of South Carolina, USA Evaluating clinician satisfaction with an electronic medical record (EMR) system is an important dimension to overall acceptance and use, yet project managers often lack the time and resources to formally assess user satisfaction and solicit feedback. This article describes the methods used to assess clinician satisfaction with an EMR and identify opportunities for improving its use at a 300-physician academic practice setting. We administered an online survey to physicians and nurses; 244 (44%) responded. We compared physician and nurse mean ratings across 5 domains, and found physicians’ satisfactions scores were statistically lower than nurses in several areas (p<.001). Participants identify EMR benefits and limitations, and offered specific recommendations for improving EMR use at this facility. Methods used in this study may be particularly useful to other organizations seeking a practical approach to evaluating EMR satisfaction and use. Chapter 5 Information Technology (IT) and the Healthcare Industry: A SWOT Analysis ................................. 65 Marilyn Helms, Dalton State College, USA Rita Moore, Dalton State College, USA Mohammad Ahmadi, University of Tennessee at Chattanooga, USA
The healthcare industry is under pressure to improve patient safety, operate more efficiently, reduce medical errors, and provide secure access to timely information while controlling costs, protecting patient privacy, and complying with legal guidelines. Analysts, practitioners, patients and others have concerns for the industry. Using the popular strategic analysis tool of strengths, weaknesses, opportunities, and threats analysis (SWOT), facing the healthcare industry and its adoption of information technologies (IT) are presented. Internal strengths supporting further industry investment in IT include improved patient safety, greater operational efficiency, and current investments in IT infrastructure. Internal weaknesses, however, include a lack of information system integration, user resistance to new technologies and processes, and slow adoption of IT. External opportunities including increased use of the Internet, a favorable national environment, and a growing call for industry standards are pressured by threats of legal compliance, loss of patient trust, and high cost of IT. Chapter 6 Using a Neural Network to Predict Participation in a Maternity Care Coordination Program ............ 84 George E. Heilman, Winston-Salem State University, USA Monica Cain, Winston-Salem State University, USA Russell S. Morton, Winston-Salem State University, USA Researchers increasingly use Artificial Neural Networks (ANNs) to predict outcomes across a broad range of applications. They frequently find the predictive power of ANNs to be as good as or better than conventional discrete choice models. This paper demonstrates the use of an ANN to model a consumer’s choice to participate in North Carolina’s Maternity Care Coordination (MCC) program, a state sponsored voluntary public health service initiative. Maternal and infant Medicaid claims data and birth certificate data were collected for 59,999 births in North Carolina during the years 2000-2002. Part of this sample was used to train and test an ANN that predicts voluntary enrollment in MCC. When tested against a hold-out production sample, the ANN model correctly predicted 99.69% of those choosing to participant and 100% of those choosing not to participant in the MCC program. Chapter 7 Can IT Act as a Catalyst for Change in Hospitals? Some New Evidence ............................................ 94 Teemu Paavola, LifeIT Plc, Finland This chapter presents a succesful reorganization of a patient care process that was carried out in a middle sized Finnish hospital. The reorganization of the patient care process for joint replacement surgery succeeded in achieving a 50 per cent increase in operations. This study proposes that IT may have an indirect influence on the achievement of goals, such as productivity, as soon as the IT investment has been decided upon; in other words, IT benefits start accruing before the IT component is even in place. This is a new feature to add to the previous definitions, because this particular benefit cannot be logically derived from any of the features of the actual IT system. Paying enough attention to this phenomen at the planning stage can be vital to the success of new IT system investment.
Chapter 8 Informatics Application Challenges for Managed Care Organizations: The Three Faces of Population Segmentation and a Proposed Classification System .................................................. 102 Stephan Kudyba, New Jersey Institute of Technology, USA Theodore L. Perry, Health Research Corporation, USA Jeffrey J. Rice, Independent Scholar, USA Organizations across industry sectors continue to develop data resources and utilize analytic techniques to enhance efficiencies in their operations. One example of this is evident as Managed Care Organizations (MCOs) enhance their care and disease management initiatives through the utilization of population segmentation techniques. This article proposes a classification system for population segmentation techniques for care and disease management and provides an evaluation process for each. The three proposed operational areas for Managed Care Organizations are: 1) Risk Status: early identification of high-risk patients, 2) Treatment Status: compliance with treatment protocols, and 3) Health Status: severity of illness or episodes of care groupings, all of which require particular analytic methodologies to leverage data resources. By applying this classification system an MCO can improve its ability to clarify internal goals for population segmentation, more accurately apply existing analytic methodologies, and produce more appropriate solutions. Chapter 9 Scrutinizing the Rule: Privacy Realization in HIPAA ....................................................................... 112 S. Al-Fedaghi, Kuwait University, Kuwait Privacy policies, laws, and guidelines have been cultivated based on overly verbose specifications. This article claims that privacy regulations lend themselves to a firmer language based on a model of flow of personal identifiable information. The model specifies a limited number of situations and acts on personal identifiable information. As an application of the model, the model is applied to portions of the Privacy Rule of Health Insurance Portability and Accountability Act (HIPAA). Chapter 10 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care? A Comparison among Norway, Denmark, and Sweden .................................................................... 126 Agneta Ranerup, Göteborg University, Sweden The aim of this article is to evaluate the provision of Web support in choice reforms in health care in Norway, Denmark, and Sweden. Two main issues are investigated: (1) What institutional frameworks for choice in health care exist, and how is the exercise of choice supported by Web technology in these countries? (2) As a consequence of this, what roles of the individual are mediated by this technology? The present study provides a critical analysis of current technologies for providing information about health care. It is concluded that in Norway the individual is equipped to be a reasonably informed consumer, customer, and citizen. A similar situation exists in Denmark, but here the consumer role is even more prominent. In Sweden, there has been little technological support for these roles, but recently national actors have initiated a project aimed at creating a national portal for public health care.
Chapter 11 Characteristics of Good Clinical Educators from Medical Students’ Perspectives: A Qualitative Inquiry Using a Web-Based Survey System ................................................................ 145 Gary Sutkin, University of Pittsburgh School of Medicine, USA Hansel Burley, Texas Tech University, USA Ke Zhang, Wayne State University, USA Neetu Arora, Texas Tech University, USA Medical educators have a unique role in teaching students how to save lives and give comfort during illness. This article reports a qualitative inquiry into medical students’ perspectives on the key qualities which differentiate excellent and poor clinical teachers, using a Web-based questionnaire with a purposeful sample of third- and fourth-year medical students. Thirty-seven medical students responded with 465 characteristics and supportive anecdotes. All participants’ responses were analyzed through reviewing, coding, member checking, recoding and content analysis, which yielded 12 codes. Responses from 5 randomly chosen participants were recoded by two authors with an inter-rater reliability coefficient of 0.72, implying agreement. Finally, 3 larger categories emerged from the data: Content Competence, Teaching Mechanics, and Teaching Dynamics. We incorporate these codes into a diagrammatic model of a good clinical teacher, discuss the relationships and interactions between the codes and categories, and suggest further areas of research. Chapter 12 Open Source Software: A Key Component of E-Health in Developing Nations ............................... 162 David Parry, Auckland University of Technology, New Zealand Emma Parry, National Women’s Health, Auckland District Health Board, New Zealand Phurb Dorji, Jigme Dorji Wanchuck National Referral Hospital, Bhutan Peter Stone, University of Auckland, New Zealand The global burden of disease falls most heavily on people in developing countries. Few resources for healthcare, geographical and infrastructure issues, lack of trained staff, language and cultural diversity and political instability all affect the ability of health providers to support effective and efficient healthcare. Health information systems are a key aspect of improving healthcare, but existing systems are often expensive and unsuitable. Open source software appears to be a promising avenue for quickly and cheaply introducing health information systems that are appropriate for developing nations. This paper describes some aspects of open-source e-health software that are particularly relevant to developing nations, issues and problems that may arise and suggests some future areas for research and action. Suggestions for critical success factors are included. Much of the discussion will be related to a case study of a training and E-health project, currently running in the Himalayan kingdom of Bhutan. Chapter 13 An Empirical Investigation into the Adoption of Open Source Software in Hospitals ...................... 175 Gilberto Munoz-Cornejo, University of Maryland Baltimore County, USA Carolyn B. Seaman, University of Maryland Baltimore County, USA A. Güneş Koru, University of Maryland Baltimore County, USA
Open source software (OSS) has gained considerable attention recently in healthcare. Yet, how and why OSS is being adopted within hospitals in particular remains a poorly understood issue. This research attempts to further this understanding. A mixed-method research approach was used to explore the extent of OSS adoption in hospitals as well as the factors facilitating and inhibiting adoption. The findings suggest a very limited adoption of OSS in hospitals. Hospitals tend to adopt general-purpose instead of domain-specific OSS. We found that software vendors are the critical factor facilitating the adoption of OSS in hospitals. Conversely, lack of in-house development as well as a perceived lack of security, quality, and accountability of OSS products were factors inhibiting adoption. An empirical model is presented to illustrate the factors facilitating and inhibiting the adoption of OSS in hospitals Chapter 14 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching ................. 195 Masoud Mohammadian, University of Canberra, Australia Ric Jentzsch, Compucat Research Pty Limited, Australia Radio frequency identification (RFID) is a promising technology for improving services and reduction of cost in health care. Accurate almost real time data acquisition and analysis of patient data and the ability to update such a data is a way to improve patient’s care and reduce cost in health care systems. This article employs wireless radio frequency identification technology to acquire patient data and integrates wireless technology for fast data acquisition and transmission, while maintaining the security and privacy issues. An intelligent agent framework is proposed to assist in managing patients’ health care data in a hospital environment. A data classification method based on fuzzy logic is proposed and developed to improve the data security and privacy of data collected and propagated. Chapter 15 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care ......................................................................................................................................... 214 Jongtae Yu, Mississippi State University, USA Chengqi Guo, James Madison University, USA Mincheol Kim, Jeju National University, South Korea In the advent of pervasive computing technologies, the ubiquitous healthcare information system, or Uhealth system, has emerged as an innovative avenue for many healthcare management issues. Drawing upon practices in healthcare industry and conceptual developments in information systems research, this paper aims to explain the latent relationships amongst user-oriented factors that lead to individual’s adoption of the new technology. Specifically, this study focuses on the introduction of chronic disease U-health system. Using the Ordinary Line Square (OLS) regression analysis, we are able to discover the insights concerning which constructs affect service subscriber’s behavioral intention of use. Based on the data collected from over 440 respondents, empirical evidences are presented to support that factors such as medical conditions, perceived need, consumer behavior, and effort expectancy significantly influence the formation of usage intention.
Chapter 16 E-Patients Empower Healthcare: Discovery of Adverse Events in Online Communities .................. 232 Roy Rada, University of Maryland Baltimore County, USA E-patients can empower themselves and improve healthcare. In online communities, patients may discuss adverse events that are inadequately addressed in the literature. The author as a patient joined various online patient discussion groups and identified several such adverse events. For each such adverse event, the patient findings, the medical literature, and the implications are noted. Extracts from the literature that were provided to the patients were welcomed by the patients. Possible approaches to financially supporting such activities are sketched. Chapter 17 Towards Process-of-Care Aware Emergency Department Information Systems: A Clustering Approach to Activity Views Elicitation ......................................................................... 241 Andrzej S. Ceglowski, Monash University, Australia Leonid Churilov, The University of Melbourne, Australia The critical role of emergency departments (EDs) as the first point of contact for ill and injured patients has presented significant challenges for the elicitation of detailed process models. Patient complexity has limited the ability of ED information systems (EDIS) in prediction of patient treatment and patient movement. This article formulates a novel approach to building EDIS Activity Views that paves the way for EDIS that can predict patient workflow. The resulting Activity View pertains to “what is being done,” rather than “what experts think is being done.” The approach is based on analysis of data that is routinely recorded during patient treatment. The practical significance of the proposed approach is clinically acceptable, verifiable, and statistically valid process-oriented clusters of ED activities that can be used for targeted process elicitation, thus informing the design of EDIS. Its theoretical significance is in providing the new “middle ground” between existing “soft” and “computational” process elicitation methods. Chapter 18 Applying Dynamic Causal Mining in Health Service Management .................................................. 255 Yi Wang, Nottingham Trent University, UK This article describes an application that illustrates the role of data mining technology in identifying hidden causal knoledge from health and medical data repositories. Across the health care and medical enterprises, a wide variety of data is being generated at a rapid rate. Current information technologies tends to focus on a more statical side of causal knowledge and do not address the dynamic causal knowledge. This article shows that the dynamic causal relation data can be captured for treatment, payment, operations purposes and administrative directed insights. Accessing this currently unrealized knowledge potential would enable the delivery of actionable knowledge to medical practitioners, healthcare system managers, policy planners and even patients to make a significant difference in overall healthcare.
Chapter 19 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems: An Introduction and Literature Survey ............................................................................................... 275 Christos Vasilakis, University College London, UK Dorota Lecnzarowicz, University of Westminster, UK Chooi Lee, Kingston Hospital, UK The unified modelling language (UML) comprises a set of tools for documenting the analysis of a system. Although UML is generally used to describe and evaluate the functioning of complex systems, the extent of its application to the health care domain is unknown. The purpose of this article is to survey the literature on the application of UML tools to the analysis and modelling of health care systems. We first introduce four of the most common UML diagrammatic tools, namely use case, activity, state, and class diagrams. We use a simplified surgical care service as an example to illustrate the concepts and notation of each diagrammatic tool. We then present the results of the literature survey on the application of UML tools in health care. The survey revealed that although UML tools have been employed in modelling different aspects of health care systems, there is little systematic evidence of the benefits Chapter 20 TreeWorks: Advances in Scalable Decision Trees .............................................................................. 288 Paul Harper, Cardiff University, UK Evandro Leite Jr., University of Southampton, UK Decision trees are hierarchical, sequential classification structures that recursively partition the set of observations (data) and are used to represent rules underlying the observations. This article describes the development of TreeWorks, a tool that enhances existing decision tree theory and overcomes some of the common limitations such as scalability and the ability to handle large databases. We present a heuristic that allows TreeWorks to cope with observation sets that contain several distinct values of categorical data, as well as the ability to handle very large datasets by overcoming issues with computer main memory. Furthermore, our tool incorporates a number of useful features such as the ability to move data across terminal nodes, allowing for the construction of trees combining statistical accuracy with expert opinion. Finally, we discuss ways that decision trees can be combined with Operational Research health care models, for more effective and efficient planning and management of health care processes Chapter 21 Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor ........................................................................................................................ 302 Hussein Atoui, Université de Lyon and INSERM, France David Télisson, Université de Lyon and INSERM, France Jocelyne Fyan, Université de Lyon and INSERM, France Paul Rubel, Université de Lyon and INSERM, France Recent years have witnessed a growing interest in developing personalized and nonhospital based care systems to improve the management of cardiac care. The EPI-MEDICS project has designed an intel-
ligent, portable Personal ECG Monitor (PEM) embedding an advanced decision making system. We present two of the ambient intelligence models embedded in the PEM: the neural-network based ischemia detection module and the Bayesian-network risk stratification module. Ischemia detection was expanded to take into account the patient ECG, clinical data, and medical history. The neural-network ECG interpretation module and the Bayesian-network risk factors module collaborate through a fuzzylogic-based layer. We also present two telemedicine solutions that we have designed and in which the PEM is integrated. The first telemedical architecture was created to allow the collection of medical data and their transmission between healthcare providers to get an expert opinion. The second one is intended for improving healthcare in old people’s homes. Compilation of References .............................................................................................................. 312 About the Contributors ................................................................................................................... 343 Index ................................................................................................................................................... 348
xvi
Preface
A State-of-the-Art Review of Developments in Health Information System & Technology Models and Methods: The MEDIA Paradigm
IntroductIon For years, misguided health organizational information technology (IT) leadership, and the lack of available expertise and skills in the application of health IT (HIT) models and methods have somehow predisposed administrators of large-scale health maintenance organizations (HMOs) to become generally reluctant to migrate away from legacy health information systems (HISs). For many health provider organizations, the more recent decline in economic activities over the last few years has also further led to new debates on allocating limited public and private resources so gravely needed to build and sustain large-scale interoperable systems infrastructure in order to achieve such a system migration efficiently and effectively. Not surprisingly, progress towards adopting new health IT initiatives such as innovative e-technologies for the health care industry in the United States (US) has been slower than most other industries. Inadvertently, a major gap in the strategic and opportunistic use of emerging health IT models and methods now exists to significantly transform the apparently fragmented nature of US health services delivery system. Aside from the fear of costly systems failure, many key stakeholders in the US health care system have, admittedly, been resistant to invest in new and contemporary enterprisewide systems due to the lack of a well-focused national health IT vision and strategy. On the one hand, attempts to successfully diffuse interoperable, integrative HIT applications throughout the US health care system must now depend on the strength of the health IT leadership shouldered by the current Administration. On the other hand, progress in health IT implementation and diffusion will also depend on how quickly many of these health key stakeholder groups who have been technological laggards for one reason or another can be motivated, challenged, and appropriately enticed to design, develop, and deploy large-scale, complex and interoperable computer-based enterprisewide systems. Such enterprisewide systems include, but are not limited to, those that are designed to incorporate integrative health data management models, new biomedical informatic methods, web-based semantic search capabilities, and emerging clinical decision support methodologies. In the context of today’s complex and largely fragmented US health services
xvii
systems, specific applications of enterprisewide, interoperable systems can range from patient-centric records and information services systems such as electronic medical records (EMRs), electronic health records (EHRs), personal health records (PHRs), payor-based health records (PBHRs) and computer-based physician order entry (CPOE) systems, to health administrative-aided transaction and health information exchange (HIE) systems such as e-prescribing systems (EPS), supply chain management (SCM), customer relationship management (CRM), enterprise resource planning (ERP), and e-payment systems. Briefly, key reasons why the diffusion of interoperable, integrative HIT models and methods is needed in the US health care services sector entail: a. b.
c.
an urgency, as a whole, to contain escalating health care cost - the growing health care cost has become an increasingly unsustainable burden on US taxpayers over the years; the growing complexities and uncertainties in the health information processing and exchange environment of large, medium, and even smaller health organizations populating the US health care system - the expansion of stakeholder groups and increased federal, state, and municipal regulatory oversight mechanisms surrounding health services delivery have and will continue to add to the already intricate health information management (HIM) and services delivery system; and the potential of interoperable, integrative HIT models and methods to aid complex data analysis and semi-structured decision making - such analysis will not only empower care providers, but also enable critical information sharing to occur among referring physicians in consultation with patients, and especially when specialists and a team of caregivers are involved in key administrative and clinical decision making.
With increasing attention paid to the critical role that HIT models and methods can and will play in reforming the US health care system, we are finally seeing a trend in increased projections on health IT spending, which is currently anticipated to exceed 15 billion dollars annually for the US (Lipowicz, 2009). According to the National Coalition on Health Care, in 2008 alone, the US has spent well over 17% of its Gross Domestic Product (GDP) on health care - a percentage that clearly exceeded those spent by many other OCED countries (National Coalition on Health Care, 2009). Sadly, the fact that the US has outspent almost every other country on health care has not translated into better health or even more convenient, accessible, available or affordable health care services delivery for Americans. In fact, findings abstracted from 2007-2008 data in a study championed by the consumer health advocacy group Families USA has revealed that one out of every three Americans under 65 may still have to live, at one point in time or another, without health insurance coverage (Parisi and Bailey, 2009). Clearly, many of the issues raised here have now taken central stage in the debate raised by the Obama Administration for championing health care reform in the US. As well, various solutions have been considered, chief among them, using and adopting interoperable, integrative HIT applications. Such a strategy is not without merit, as various forms of health technologies have indeed risen over the years to similar challenges; for instance, online claims processing and e-prescribing have been successfully deployed by various HIT vendors to serve as tools to combat rapidly escalating health administrative costs while simultaneously leveraged to reduce wastes, increase efficiencies, eliminate redundancies, and improve the overall quality of clinical care and services. Raghupathi and Tan (2002, 2008) noted that various strategic applications of HIT models and methods can evidently improve the efficiency and effectiveness of US health care services delivery, adding value to existing, legacy-based HISs, and helping to integrate the islands of health services management
xviii
systems. Their arguments not only cited the power of e-technologies to streamline increasingly complex routine HIM processes that may require multi-provider, cross-organizational collaboration, but also the ability of interoperable, integrative HIT capabilities to augment enterprisewide efficiencies and care provider network connectivity. Accordingly, the key underlying argument for migrating from legacy HISs is that both data and processes linked to diverse functioning information systems must now be shared on an increasingly real-time basis between both on-site and off-site caregivers if we want to improve the quality of patient care. These interoperable, integrative systems can also streamline complex administrative and clinical workload, easing the communication needs among health care administrators, HIT personnel, health engineers, health informatic researchers, and HIT consultants, all of whom need to work closely together in today’s health services delivery systems if major systems bottlenecks are to be effectively managed over time. Figure 1 depicts the Model and Evidence Driven Integrated Analysis (MEDIA) paradigm, an integrated model management framework that is applied in this review to provide an integrative conceptualization of the evolving, state-of-the-art developments in HIT models and methods. Against this background, existing HIT models and gathered evidence will not simply aggregate but will be seen as complementing each other alongside the MEDIA integration and analysis process. Take the case of the influenza propagation - a disease model - and imagine how it can intermingle with a patient care process model, such as that of a primary clinic. Given the separate evidence about the flu outbreak and patient arrival pattern, running both models in parallel within the MEDIA framework produces either a more realistic health care supply-demand scenario in the context of a real-world community health setting, and/or offers insights into other related and potential issues needed to be addressed when challenged with health hazard preparation such as when an apparent discrepancy is observed to be operative within the overall system. All such information analysis and its integration can be further quantified and addressed consistently. Such multi-attribute analytic capability is crucial to the success of any future applications and developments of HIT models and methodologies. The rest of this review on health IT models and methods is organized as follows. In Section 2, a high-level systems perspective of HIS information and process flow in the context of applying HIT models and methods is outlined. Fundamentally, the classical information system input-process-output
Figure 1. The model and evidence integrated analysis (MEDIA) paradigm Model Integration Ontology modeling
Probabilistic dependency modeling
Hybrid computational analysis
Evidence Fusion Service Assurance a. Delivery improvement b. Risk management
xix
triad that underlies all systems engineering conceptualization is revisited. In Section 3, the focus will shift to the complexity-uncertainty challenge encountered in the current US health care sector. Following this, in Section 4, the emerging engineering approaches in health services applications at two levels are highlighted. These levels include: (1) the general health engineering level; and (2) the more focused health services modeling level. In Section 5, the MEDIA paradigm, including its three key facets are briefly overviewed: (a) domain ontology modeling and management that lays the foundation for model representation and infrastructural connectivity; (b) hybrid probabilistic modeling that computationally integrates systems and subsystems models; and (c) adaptive knowledge fusion to generate quality assurance that support active information collection and continuing operations management. Essentially, the MEDIA approach is based on a linked series of health engineering concepts, which deviates fundamentally from the traditional health care best practices in which quality is heavily dependent upon the hands-on skills of expert clinical practitioners as well as the deployment of progressively specialized medical instrumentation. The application of the MEDIA paradigm is also illustrated in a case study on risk management. Finally, in Section 6, we conclude this review by speculating on future research directions and practical implications in the field of health IT models and methods.
VIewIng HIt Models & MetHods In tHe classIcal HIs Flow cycle Model Owing to the need to deal with increasingly complex health data processing routines and the growing number of participating stakeholders who may need to share information and provide their particular interpretation of such data taken from the very same electronic health databases, an exciting playground for health researchers and practitioners now exists to incorporate efficient medical information processing models and emerging health decision support systems (HDSS) methodologies into interoperable systems linked to existing health databases, model management bases, and knowledge management bases. Despite recent investments in medical information packages such as the Vista medical information system and the GoogleHealth patient record management system, the lack of system interoperability has made it difficult, if not impossible, to integrate isolated health records stored at varying functional levels. Moreover, a key challenge is the integration of HIT models and methods through a knowledge fusion process so as to enhance the capability of viewing and analyzing any chosen set of discrete and continuous events in a health services delivery system such as relating patient-physician encounters and the progression of different disease stages from an enterprisewide and trajectory perspective, for example, the interactions and optimal interventional strategies for a particular patient cohort within a HMO, built up from its connected elements, including the affiliated hospitals, clinics, practicing physicians, nurses, information and personnel resources, labs, and departments linked to its health services delivery system. Figure 2 depicts a generic health information flow process model in which health administrative, clinical, and service delivery decision and policymaking must necessarily transpire at an enterprisewide level while aided with the use of relevant and applicable HIT models and methods within the traditionally defined input-process-output system triad.In the initial data collection, and information-knowledge gathering stage, a huge amount of raw data, guided protocols, and knowledge elements are often sourced from various input sensors and devices as well as recorded answers to questions asked of the individual patient and/or groups of patients, including underlying reasons for these patients to seek care. All of the
xx
Figure 2. A general health information flow process model with Its input-process-output system triad Data Collection
Intelligent Information Processing
Health IT Models & Methods Knowledge Formation & Feedback
Health Care Services Decisions & Policymaking
collected information is now ready to be pre-processed into some form of meaningful and intelligent datasets. These datasets are then stored appropriately either in a centrally located or in multiple electronic locations, typically via databases and data warehouses, waiting for further processing to serve a variety of clinical, research and administrative goals and purposes. Such purposes could range from clinical testing and follow up, to research, administrative billing, and/or mere reporting for managerial decision making. As a case in point, imagine the gathering of demographical data as well as specific vaccination records for a population of young adult patients to guide the clinical management of an ongoing epidemic, such as the “swine flu”. The data gathered should not only be accurately coded but they should also be securely stored in a more or less structured format to be communicated to the attending physicians for further clinical evaluations, testing, and/or diagnosis. In fact, the same information may be aggregated with other data for research analysis or it may be used to establish insurance co-payments to the various care provider groups. Finally, the information may also be critical to plan health vaccination programming by the different workplace or educational institutional settings tied in with the population of young adult patients. At the data manipulation-information mining-task processing stage, embedded patterns are often sought and hidden knowledge uncovered within the captured datasets to be translated intelligently into meaningful clusters and to provide additional advice and insightful guidance to the end-users, in this illustrative case, the various caregivers. In other words, the same data-information-knowledge collected of the young adult patients may now be further fused with other high-level information, such as expert clinical knowledge about the state of the patient immunization types and dates as well as the state of ongoing epidemics or pandemics, if any. The combined information can then be used to determine the health status of the individual young adults following the immunization, for example, their resistance to prevailing illnesses, their showing signs of some sort of common allergies, and/or their susceptibility to some other atypical reactions. In fact, it is here that the HIT models and methods may be most relevant to aid in the follow-up of crucial clinical and administrative decisions.
xxi
In operationally organizing, aggregating, mixing, dicing, and mining the stored patient data, various HIT models and methods may be combined in a complex fashion to generate a meaningful and “best” fit for clustering and/or grouping the stored data based on some known statistical patterns. The resulting data pattern classification are in turn routed back, on the one hand, to the clinicians to guide them in making prognostic, diagnostic and other therapeutic decisions, and, on the other hand, to the administrative staff to assist them in determining patient diagnostic codes, and thereby, generate accurate and appropriate third-party billings. At the end of the data processing-data mining stage, feedback will also typically occur in terms of second opinions, reviews and/or expert evaluations on model validity to further enhance systems finetuning and performance outcomes. Such enhancements recognize the possibility for more intelligent decision making in the face of missing or incomplete data and/or the fact that inadequate tools were available to appropriately process such data so that adjustments to these decisions could still be made to further improve accuracy or timeliness of key decisions and outcomes. In our case example, the nursing unit may, at this point, requests for feedback from the attending physicians as to who among these young adults would be due for specific and further vaccination visits, be sent for more clinical testing and evaluations, such as generating referrals to other departments or physician specialists, or they should simply be discharged. This commonly encountered health information flow cycle is characteristic of an open health information processing system that follows the input-process-output triad. Hence, despite the complexity of HIT models and methods, the potential for these methodologies to contribute significantly to future health services management is evident and apparent. While their applications represent mostly a part of the wider and far reaching field of HIS, it should not and cannot exist in and of itself. It must go hand in hand with the many different types of decisions on health services management that have to be executed routinely alongside the processing of initially gathered data, stored information, and captured knowledge. With a basic understanding on the flow of information and processes through a generic HIS flow cycle model, it is clear that multiple disciplines – including computer and library sciences, medical and/or health informatics, sub-fields in nursing informatics, teleradiology, telemedicine, telehome care, telematics, biomedical informatics, as well as domains of health IT, health engineering, and health operations research (OR) and management science, HDSS, and even the integration of cognitive sciences, information sciences, and health sciences – must all come together to enrich our ability in using HIT models and methods to meet a diversity and variety of clinical and health administrative decision needs. Ultimately, in order to achieve high quality health services management and related decision making, not only will we need medical informatics tools such as computers, clinical protocols, formal and informal medical terminologies, but also the integration of various informatic resources, tools, models, devices, techniques, methods, decision aids, and methodologies to optimize the collection, storage, retrieval, analysis, design, and use of health information in health services delivery (Tan and Payton, 2010).
ratIonale For IntegratIng HIt Models & MetHods: tHe us HealtH care coMplexIty-uncertaInty landscape While the interplay of multiple disciplines and sub-disciplines in managing a growing body of health services information and processes makes the health care system one of the most complex systems to study, the growing intricacies of the US health services sector may also be seen in many real-world events.
xxii
The 2007 drug-resistant tuberculosis case, for example, shows the high degree of interdependency that exists in the US health care infrastructure, and uncovers the many hidden flaws, previously ignored, in the management of patients, hospitals, the center for disease control and prevention (CDC), as well as the complex sociopolitical forces that link the US government with other national governments (U.S. House of Representatives, 2007)
Key challenges & risk Management in the us Health services sector For the US health care sector to achieve greater accessibility, affordability, and accountability, two key determinants for its continuing growth and future developmental success include: (1) ensuring a 24/7 service availability; and, (2) having a focus on service quality. In the coming years, as the US health care sector experiences sweeping reform, daunting risk and unpredicted challenges will likely arise. Hence, dealing with risk and uncertainty is a critical skill in running any modern-day health services enterprise, and this is precisely where HIT models and methods can and will play a central role. The major threats and challenges facing the US health care sector are depicted in Figure 3. As shown, these challenges include systems complexity, event uncertainty, and the need to share information. Owing to the fact that such threats and challenges will often emerge unpredictably and their diverse manifestations will also typically blur the line of most standard classifications, both health administrators and clinicians will want specialized tools to aid them in their routine policymaking and decision execution, as we moved forward with the US health care reform. First, with regard to systems complexity, there is the overwhelming intricacy of diverse evolving structures alongside the rapidly expanding health services supply networks. With hospital expansions and alliances formed among various stakeholders, increasing complexities have also followed. Again, with stakeholders trying to leverage on increasingly competitive global health supply chains, the resulting system relations are even more complex. Moreover, stakeholder collaboration across diverse industrial sectors may often follow different policy paths and practices. These factors, and the fact that most such collaboration is developed independently, compound the intricacy. Hence, there is the need for integrated HIT models and methods to guide future decisions. Second, in regard to event uncertainty, there is the need for systems configuration flexibility. A health service is subject to risk whenever and wherever the stakeholder cannot predict changes to the typical Figure 3. Major threats & challenges facing the US health care sector industry
• • • •
Complexity Healthcare alliance; Global supply chains; Regional, national, and international coordination; …
Staff
• • • •
Challenges Uncertainty New, rare, and epidemic diseases; Natural disaster and man-made emergency; Terrorist attacks; …
• • • • • •
Information Sharing Incompatibility; Inoperability; Integrity and accuracy; Cyber attack; Privacy regulation; …
Service Availability and Quality Resources Information/Knowledge
xxiii
environmental situation. As event uncertainty can arise, without prior warning, from diverse occurrences, such as (a) variations in staff and patient flow, (b) the onset of new, rare or epidemic diseases, and (c) emergencies, whether natural and/or man-made, all such uncertain events will require appropriate and timely adaptation in health services management. These challenges are further convoluted by evolving structures, novel technologies, the constant threats of natural disasters and/or man-made events, including the unpredictable growing number of participating stakeholder groups. Traditionally, stakeholders including patients, care providers, insurance companies, and the government do not see eye-to-eye. As many health administrators and clinicians tend to focus largely on the details, and likely also to stubbornly stick to their personal heuristic business decision-making biases, challenges of prevailing complexity and uncertainty will become even more evident over time. In light of this, it is not uncommon to have conflicting views shared among the multiple stakeholders. Moreover, health care debates often hinge on highly sensitive issues, and it is unlikely that any lopsided discussions or outcomes will contribute to the overall satisfaction of certain key stakeholder groups. Therefore, given that current management practices rely largely on personal heuristics, qualitative comparisons, and subjective guidelines, without precise accounting and science-based quantitative analysis, the significance of HIT models and methods cannot be overly emphasized. As today’s health services delivery systems often encompass huge numbers of interacting elements, significant interdependency, and enormous uncertainty and risk, it is very difficult to expect quality and outcomes of the decisions to be appropriately calibrated and/or justified quantitatively. Hence, the attempts to incorporate the latest science and methodological advances in HIT models are some of the ways to ensure better health decision outcomes and policymaking. Third, in regard to information sharing, it is in the interest of health provider organizations, at the minimum, to ensure that their patients receive prompt and accurate medical care services by increasing the accessibility and availability of health information services on the one side, and ensuring the secured, confidential, and private storage of personal medical records for sharing among affiliated caregivers on the other side. Technologically speaking, effective information sharing has to do chiefly with systems interoperability and the availability of an interoperable enterprisewide infrastructure. This is why a move towards integrative, interoperable HIT models and methods is key to achieving US health care reform. In summary, the challenges of complexities, uncertainties, and sharing of health data have deterred key stakeholders within the US health care industry sector to work collaboratively, effectively, and productively. But it has also highlighted the importance of having a national health IT strategy and vision as well as building a cumulative effort to apply and diffuse integrative, interoperable HIT models and methods.
Inadequacy of Health services Modeling practice & research Despite the wide range of quantitative models that has emerged to address various performance optimization risks with regards to different parts and/or aspects of health services systems over the years, there is still the need for a system-wide perspective on how these different HIT models and methods may be combined and integrated to aid health care decisions and policymaking. Roberts (2007), for instance, emphasized models to represent the progression of certain disease categories as well as to predict best treatment timing and cost. Denton, Fowler, Schaefer, Batun, Erdogan, and Gul (2009) detailed scheduling systems that model patient arrival and how they may be queued efficiently and effectively to be served by doctors in operating rooms. Finally, Shehad, Bertino, and
xxiv
Ghafoor (2005) discussed the application of computational models to quantify information leakage and availability based on the analysis of user communication via a computer infrastructure. Notwithstanding, each of these isolated attempts represented a series of uncoordinated efforts to address only a small, specific set of health services elements, without adjusting to benefit the greater whole – that is, the limited impact of the individual studies on the total system do not account for the intricate interdependencies among the systems elements. In other words, confined within a limited scope of a system component or sub-system, the performance measures adopted by these studies are typically defined in terms of cost, quality, time, or other specific but narrowly defined criterion, which, without exception, would still impact, one way or another, on other systems elements and components on an unknown basis and scale. Thus, isolated models and methodologies, working in and by themselves, simply cannot address the larger challenges presented in the US health services management system. With health engineering as an emergent field, researchers interested in HIT models and methods are beginning to redefine the boundary of traditional health care management philosophy. The modeling capability to guide flexible reconfiguration of health care structures, such as new investment and/ or rearrangement of provider organizations and facilities, is critical if we are to handle the complexity of systems modeling and analysis appropriately. Arguably, the different levels of local, regional, and national coordination that are actually interconnected (Schnase & Cunnius, 1995) will dictate how the national health infrastructure is to be reconfigured to manage existing resources efficiently, effectively and productively. All in all, for enterprisewide systems modeling, not only should external disturbance be taken into account, but also losses incurred due, simultaneously, to rising health care cost and the need to minimize poor quality medical (patient) care (Hick, Hanfling, Burstein, DeAtley, Barbisch, Bogdan, & Cantril, 2004) In this sense, many industry leaders and practicing engineers have now recognized the significance of integrating the operation, audit and control management of health supply systems (Barbera & Macintyre, 2002). Unfortunately, most, if not all, of these initiatives are still in their infancy. For instance, many health services institutions still lack the required level of modeling capability and organizational preparedness for dealing with external interfering events. Indeed, as we attempt to introduce the latest enterprise paradigms such as the system-of-systems and the reconfigurable enterprises across institutions where a large number of participants interact and evolve on a continuous basis, the lag in health services reengineering practices and health performance management research becomes clearly evident. As early as 2003, Woolhandler, Campbell, and Himmelstein (2003) have noted that, by 1999, 31% of the US health expenditures and 16.7% of the same expenditures in Canada were largely health administrative overhead expenses; clearly, this points to the high potential for HIT models and methods to achieve a significant system performance improvement largely through a reduction in health administrative cost. Moreover, just as we set to data mine and understand the interactive behavior of a complex adaptive system (CAS) (Kiel, and Elliott, 1996; Woolhandler et al, 2003) emerging states can also be generated under various systems configuration to be studied so as to achieve across-the-system improvements.
comparison to service enterprise risk Management From a service supply chain management perspective, the dynamics of the US health care system may be conceived as comprising, with special structures and performance requirements, service supply networks of suppliers on the one hand in terms of caregivers and medical resources, and patient demands on the
xxv
other hand in terms of patient needs and services. As we know, health systems are, without exception, multi-layered and closely interconnected; these systems must necessarily provide high value to their customers in terms of patient benefits and services; as well as the availability of services, security, and privacy of captured and stored patient information. Many of these factors entail the highest priority in risk management. Whereas traditional OR techniques such as game theory (Shubik, 2005) 18 derived from artificial intelligence (AI) and economics, and optimization models 18 taken from quantitative analysis have been extensively used for enterprise modeling, current research focuses largely on core strategic and operational issues such as location configuration, demand response analysis, dynamic pricing and contract management, as well as e-commerce challenges (Graves, Kletter, & Hetzel, 1998; Lambert, Cooper, & Pagh, 1998).19,20 Risk assessment and risk management cover essentially many forms of risks, including, but not limited to, business risk, financial risk, technological risk, and physical risk (Anupindi & Akella, 1993; Argrawal & Nahmias, 1997; Kouvelis & Milner, 2002; Simchi-Levi, Kaminsky, a&nd Simchi-Levi, 2003;) 22, 23, 24, 21. Lately, research into the applications of HIT models and methods has also taken on a broader perspective. Thomas (2002), for example, analyzed the reliability of a supply chain under contingency when it is being impacted by unexpected disasters. Using Bayesian networks, Pai, Kallepalli, Caudill, and Zhou (2003) provided a conceptual business risk assessment framework. Based on industry practitioners’ input, Blackhurst, Craighead, Elkins, and Handfield (2005) summarized the common themes and issues surrounding supply chain disruption. Hale and Moberg (2005) presented a set cover location model that is used in disaster preparation for identifying the minimum number and possible locations of off-site storage facilities for supplies. Lee and Whang (2005) compared the inspection of hazardous goods, such as for transporting explosives, with total quality management (TQM) in manufacturing. A RAND Corporation (2004) report focused on the impact of terrorist attacks on global container supply chain performance and advocated the importance of fault-tolerance or resilience. Despite the stream of well-funded research in the area of service supply chain and enterprise management, the emerging body of knowledge has not adequately addressed the growing complexity and uncertainty-risk management dimensions of the US health care sector. Little, if any, of the past scattered research has provided a comprehensive analysis or given enough attention to systems and systems modeling integration, which has resulted in the following significant challenges: a.
b.
c.
Current research is sparse and isolated, lacking a cumulative agenda. Even so, common terminologies or protocols are not often shared to ease communications among future researchers. Ultimately, there is the lack of a bridging theme to aid the systemic interfacing among heterogeneous models and documented evidence; Past researchers generally hold a narrow view, focusing solely on a small subset of traditional risks. In this sense, most studies deal primarily with individual system components or aspects thereof, without an overall context to account for all interlaced challenges and subsystem interactions; and Enterprisewide modeling and analysis is still in its infancy. Moreover, past studies tend to be mostly conceptual and qualitative, with limited applicability to modern enterprises in addressing current challenges.
Put together, the need for a systemic health services management framework, a representation that would exhibit three fundamental characteristics. First, it would encapsulate a family of heterogeneous
xxvi
models at different levels of detail and for different subsystems, scalable also to the growth of the enterprise and its captured knowledge cannot be overly emphasized. Second, such a representation would bridge high-level qualitative knowledge with quantitative computation by automatic conversion and instantiation, a capability that is crucial to practical deployments of HIT models and methods. Finally, we would also expect such a representation to incorporate quality assurance with confidence and clarity, measured to handle uncertainty in analysis and to guide decision making in all facets of health services management. The MEDIA paradigm, which we will discuss in Section 5, is such an integrative paradigm exhibiting all of the three fundamental characteristics noted above.
state-oF-tHe-art HIt Models & MetHods Before overviewing the MEDIA paradigm, however, we will first review the state-of-the-art HIT models and methods by categorizing and highlighting the various bodies of literature with respect to the emerging engineering approaches in health services applications. Improving a health services delivery system can be achieved through one of two means: (1) relying on medical equipment, technological and/or procedural advances; and, (2) focusing on improved medical, business, clinical and administrative processes. Modern health engineering technologies such as nano-scale materials, genetic informatics, robotics and applications of virtual health have gone a long way to provide effective and reliable prognoses and treatments to a variety of diseases. Such scientific progression foresees an accelerating pattern in light of technological advances. Yet, management theories and practices have taught us that the quality of health services delivery is also dependent on achieving better health information flow and system processes that best fit the type of organizational structure and culture inherent to particular health institutions. In other words, a solid foundation for health care reform may be achieved only through the appropriate balancing of the health resource configuration and process control. In light of this, a large body of models to address medical and health services delivery issues in health system management has evolved through best practices over the years. This approach, which concentrates on the applications of health IT models and methods, is precisely the focus of this review. It has, in the last decade, gained increasing attention among researchers interested in contributing to enhancing the performance of the US health care sector.
Health engineering Today, a new body of health engineering models has emerged. These models are applicable at different system levels, including, for example, at the low level of disease progression of individual patients, that of a clinical session process, or at a higher level involving a health facility of interconnected subsystems and components, or even at the level of a complete life cycle of health services delivery of a particular region or country. In the following section, we survey the extant literature on the new “health engineering” paradigm, following which we shift focus then to the more specific health care modeling and simulation domains. Evidently, health engineering methods and IT models will not only help hospital personnel to reduce medical errors and risk, but aid also in reducing health care costs, improving health services timeliness, and increasing patient satisfaction. In some cases, it is of course possible that the traditional culture and
xxvii
strict legislation imposed on the health care sector, which uniquely distinguishing it from many other industries, may limit the transferability and diffusion of new concepts and methods. Today, opportunities and challenges in the health care sector exist for testing and adopting concepts, best practices, and tools from diverse engineering domains to improve the efficiency of systems processes and the effectiveness of health services delivery in terms of quality, safety, and productivity. In this sense, both the goals and challenges of the health care sector match those of other industries where knowledge engineering has provided a long-term basis for resolving bottleneck issues. Hence, the combination of knowledge engineering and health IT applications promises to provide many more reengineering opportunities for health services delivery, for example, the installation of bedside terminals, the adoption of new informatic methods for infectious disease control, and therapies, and the use of innovative data and DSS technologies in medical and clinical settings, including hospital laboratories and pharmacies. This section reviews emerging operational systems engineering (OSE) research categorized into several different dimensions. Specifically, the focus is on OSE methods abstracted from a mix of engineering disciplines, including human factors, factory and product design, and security engineering. As the managerial processes and services in the health industry are often similar to those of other manufacturing and servicing industries, we argue that the field of health services management can be enhanced through the intelligent applications of traditional health IT OR models, emerging engineering philosophy and methodology, and mature OSE methods.
Traditional HIT OR Models Pierskalla and Brailer (1994) discussed the application of OR models and methods in a broad range of health services management tasks. Major applications of health IT OR models include: 1.
2.
3.
4.
5.
Demand Forecasting – A fundamental input to many other analyses in health engineering, demand forecasting, such as predicting daily census for the different types of resource needed, is important to improve the efficiency of health resources allocation. Capacity Planning - The allocation of bed capacity within the hospital is a critical factor in operational efficiency. Thus, health services capacity planning typically focuses on total bed capacity, bed capacity allocation to different services, surgical system capacity, capital equipment capacity, and ancillary service capacity. Discrete event simulations and semi-Markov process models have been used to examine bed allocation and related capacity questions. Patient Screening - Screening of patients for particular disease can improve medical diagnosis on the one hand and disease detection on the other. It may also be applied individually (individual screening) and/or population-wise (mass screening). Specifically, for individuals, the objective is often to prolong a patient’s life, whereas, in mass screening, the objective may be to minimize the cost at the societal level, thereby lowering the prevalence of a specific contagious disease. Clearly, when attempting to achieve any such objective function in OR modeling, there still may be resource constraints and compliance levels to be factored into the solution. Patient Scheduling - Scheduling is critical for matching demand with the supply of available but limited resources. Most scheduling systems attempt to optimize the combined objectives of patient and worker satisfaction, as well as the utilization of facilities. Clinical Decision Making - Clinical decisions can be aided through OR models that incorporate mathematics and structural analysis. Not only can such analytic models assist in the formulation
xxviii
6.
7.
of health care policies, but they can also be applied to the structuring of critical medical decisions, and the fine-tuning of health systems performance. Workforce Planning and Scheduling - Human resources management is one of the most costly and intensely unpredictable activities that is to be managed intelligently across health organizations. For example, Hershey, Pierskalla, and Wandel (1981) conceptualized the nurse staffing process as that of a hierarchy of three decision levels over different time horizons and with different precision – that of corrective allocations, shift scheduling, and workforce planning. Cost Cutting – Kumar, Ozdamar, and Zhang (2008) have developed several reengineering conceptual and simulation models, which were used for cost containment within the Singapore’s health industry in the domain of supply chain management process reengineering.
Emerging Engineering Philosophy and Methodology Over the years, various streams of engineering philosophy and methodology have been applied to improve health systems performance. Process orientation and patient focus are the essential concepts embedded in these methodological philosophies. Key approaches that have been discussed in the extant literature include: (1) Lean Thinking; (2) Six Sigma; and (3) Theory of Constraints. Young (2005) argues how a clinical session for a patient’s treatment would serve as a good analogy for explaining how these different streams of engineering philosophy and methodology can be combined to address various system bottlenecks encountered in the health services delivery industry. 1.
2.
3.
Lean thinking – Kollberg, Dahlgaard, and Brehmer (2007) believed that the idea of lean thinking is applicable specifically to health care systems in a number of ways. For example, just-in-time (JIT), level scheduling, and multi-skilled teams are generic techniques that can create a smooth operation process flow through the matching of the supply-demand level of health care resources. When applying lean thinking to health care, a measurement framework for lean initiatives that reflects both efficiency and effectiveness of health systems performance, such as patient satisfaction, referral management, process mapping, and fulfillment of targets and policies, is needed in order to fully capture lean changes. Six Sigma - This methodology involves standardized data collection and informed reporting protocols based upon a well-controlled quality feedback cycle to minimize variations and quantitatively align production or service quality to a predetermined standard. “Bridges to Excellence” is an example of a national case initiative based on Six Sigma quality feedback methodology launched to improve clinical care quality. This initiative targeted on physicians and their practices to enhance patient care quality (Brantes, Galvin, & Lee, 2003) and had further been incorporated as part of a collaborative product commerce (CPC) approach to health supply chain purchasing (Ford & Hughes, 2007), as highlighted in the next section. Theory of Constraints (ToC) – Similar to Six Sigma, ToC applies the root cause thinking processes for analyzing system bottlenecks. Unlike Six Sigma, however, ToC attempts to deal with managing constraints in CAS not from a technical limitation perspective, but from a “qualitative” and philosophical perspective. The methodology is first applied to identify the most vulnerable constraint, then exploiting and increasing flow through that constraint, working from the weakest link upward to other links between that constraint and the overall system.
xxix
Mature OSE Discipline A maturing OSE discipline focuses on examining, analyzing, and further understanding the operating elements and systems dynamic processes in complex systems to achieve efficient and effective systems performance. This greatly supports the view of health care as a CAS. In this sense, OSE tools and techniques may be applied to achieve a balance for meeting multiple goals, for example, quality patient safety, accessibility, availability, comprehensiveness, and affordability of care. 1.
2.
3.
4.
Supply chain management (SCM) – In recent years, SCM topics have gained significance as health services organizations vie to lower the cost and improve the ease of accessibility and delivery of health services and their associated resource supplies (Brantes et al, 2003; Ford & Hughes, 2007). 38,39 . Ford & Hughes (2007) identified potential barriers that health services organizations must overcome in order to apply SCM principles successfully within the health care sector. In their study, they inferred that physician services are the primary channels in a health care supply chain to provide the relevant expertise and services to group practices, hospitals and pharmacies. These practices, in turn, behave as secondary channels, acting as refineries and production facilities to serve health insurance distributors and purchasing programs in the supply chain. Therefore, the starting point for cost containment of health services in SCM will be determined by how physician services are being managed. A specific form of SCM used by US employers is the collaborative product commerce (CPC), which is discussed next. Collaborative product commerce (CPC) – CPC attempts to extend the limited boundaries of enterprise collaboration for product design by leveraging innovative e-technologies to engage members from internal as well as external constituencies. CPC approach differs from other SCM tools in two significant aspects (Swinehart & Smith, 2005): (1) it permits inter-organizational collaboration feeding on a common supply chain or meeting similar consumer product or services needs; and (2) its processes are transparent to all stakeholders. CPC models, therefore, allow health provider organizations and/or third-party payers, who are located in different health care markets, to share in innovative product life cycle designs. Business process reengineering (BPR) – BPR is a process-driven technique to improve the efficiency and effectiveness of a business through meaningful process redesign, change management, and system reorganization. As an example, Kumar, Swanson, and Tran (2009) conducted BPR using simulation modeling on the complex operating theatre (OT) system in a Singapore Hospital. His case simulation produces two recommendations: (1) a need to redesign the OT process in order to maximize its productivity without altering the current workload of surgeons and anaesthetists; and (2) reviewing the OT utilization data periodically so as to derive a meaningful productivity index and accurately gauge its utilization. Health management information systems (HMIS) – HMIS, which has to do with all aspects of business information systems functions in health care, plays an integral part in any modern health services system.43 As noted earlier, a national health IT strategy to support and ensure systems interoperability to link health services networks throughout America is believed to be a necessary step towards realizing US health care reform (IEEE-USA, 2005). Today, many traditional HMIS functions can be easily and logically augmented to encompass database, model-based and knowledge-based HDSS technologies when trying to build OR models to aid in the management
xxx
of health services systems, or more specifically, to improve administrative productivity, increase clinical decision-making responsiveness, and enhance patient care quality.
Health services Modeling Serving as the foundation for our proposed MEDIA framework to be discussed later, the extant literature in health services models may be further subdivided along the following three dimensions: 1. 2. 3.
the scale (level) of health system problems being studied; the goal being sought or performance measure being evaluated; and the modeling method being applied.
The intelligent applications of health IT models and methods depend largely on understanding how each of these dimensions will impact on a specific health services modeling study. Many variations and types of models exist such as policy models, procedural models, intervention models, graphical models, deterministic as well as stochastic models. As a case in point, Eldabi and Young (2007) indicated that quantitative models, methods or tools can be studied across different organizational levels, from the physical and mechanical design level through the services and policy design level.
The Scale (Level) of Health System Problems This dimension focuses mainly on the extent and/or scale of health services problems to be modeled. Four sub-levels include: (a) Individual patient disease models; (b) Operational process models; (c) Organizational system-level models; and (d) National sociopolitical-level models Individual patient disease models focus mainly on the biological disease processes occurring in individual patients, that is, the infection processes among either healthy or infected person/population. These models can range from microbiological or cellular, to organ, as well as person-to-person transmission level. O’Leary (2004), for instance, simulated person-to-person infection by building mathematical models to predict the epidemic path of disease transmission, to assess possible infectious outcomes of events over time, and to test the effectiveness of different intervention measures such as vaccination strategies and quarantine policy. In general, such disease progression models are applicable to evaluating clinical effectiveness or cost effectiveness of different interventions to particular disease (Brailsford, 2007). The operational process models are devoted to observing and simulating individual patients residing in a ward, a clinic, or a hospital department such as the Emergency Department (ED) or Intensive Care Unit (ICU). Often, such models are used to facilitate business process reengineering, resource allocation, capacity planning, staffing, and scheduling challenges. Tan, Gubaras, and Phojanamongkolkij (2002) for instance, employed a discrete event simulation model for studying capacity planning, staffing, and scheduling of Dreyer Urgent Care Center. Organizational system-level models combine different departments within a large institution or enterprise, and attempt to study the interactions among the different departments. Typically, such models address longer-term and broader issues, similar to the conceptualization of enterprisewide models. System dynamic (SD) is a common example of organizational system-level model that is often applied at a
xxxi
strategic level where the stakeholder is interested to look at the forest more than just the trees (Brailsford, Lattimer, Tarnaras, & Turnbull, 2004). Finally, national sociopolitical-level models encompass and address a wide range of very high-level issues. Having a view at a national sociopolitical level on the state of emergency, for instance, will aid government policymakers and decision makers design a reliable health services delivery policy that will coordinate hospitals, police, and emergency units for a rapid and more readied response. For example, the BioSense Real-Time Clinical Connection Program of the United States (Laudon& Laudon, 2007) builds a national surveillance system model to continuously summarize and analyze the disease and health information by source, day, and syndrome for each ZIP code, state, and metropolitan area. This model is designed to improve the national capabilities for disease detection and monitoring, as well as awareness of real-time health situations.
Goal Sought & Performance Measures In health systems, goals sought are often different from many other for-profit industries, for example, appropriateness of care, safety, and patient satisfaction are critical relative to cost and resource utilization, not just profits. These goals are usually measured by a set of key performance indicators such as patient throughput as measured in system wait time and wait line; length of stay (LOS); system capacity; utilization of staff, equipment, and space; cost; and various other related measures such as service quality in terms of error and patient satisfaction. Major groupings of studies on various health systems performance measures include the following: 1.
2.
3.
4.
Length of Stay (LOS) and Patient Throughput – Ramis, Palma, Estrada, and Coscolla, (2002) created a generic simulator within a network of clinics to reduce patient LOS in the system. The simulator also facilitated the appropriation of resources and reallocated these resources based on the number of patients and how quickly they go through the system. Bosire, Wang, Gandi, and Srihari, (2007) modeled a computer tomography (CT) scan facility in a hospital to study how patient wait time can be minimized and how staffs can be used efficiently to increase patient satisfaction. Resource Allocation and Capacity Planning – Rico, Salari and Centeno (2007) built a model using ARENA simulation software and OptQuest heuristic optimization to increase system capacity, improving nursing staffing allocation, and augmenting the utilization of equipment and space. They suggested multiple ways in which the number of nurses needed for health services delivery during a pandemic influenza outbreak may be combined. Cahill and Render (1999) created a model to assess ICU bed availability. Their model was applicable for excess capacity rebalancing that would otherwise lead increasingly and unjustifiably to wasting limited health resources such as bed space and personnel in the Cincinnati VA Medical Center. Cost Containment - In the Cahill-Render (1999) study cited previously, an additional application of their model was to contain costs by drawing from outside resources in meeting patient needs. Stahl, Roberts, and Gazelle (2003) also investigated the strategy for setting appropriate preceptorto-trainee ratio in the context of a teaching ambulatory care clinic. The key purpose of their study was to achieve an optimal financial feasibility and to cut system operational cost. Policymaking - Policymaking is fueling a resurgence of interest in modeling and simulation to improve health services delivery performance. This is especially applicable in countries such as Canada and the UK, where the system is based largely on a single payer (the government). More
xxxii
specifically, these systems employ a governmental body of agencies to manage the entire system through a set of national performance measures. A specific example would be the star rating used for measuring wait time targets for emergencies such as how long it takes a patient to be served upon admission into the emergency department (ED) of a hospital. Gunal and Pidd (2005) provided an example in their policy-oriented simulation study that aimed to improve the accuracy of the rating system in the UK National Health Service (NHS) through uncovering performance irregularities.
Modeling Method As well, a number of modeling methods and techniques have been identified for health care services implementation. Among these, the most influential, useful and widely applied methods are simulation techniques, which include the mainstream methods as well as simulation algorithms based on artificial intelligence as discussed in Cooper, Brailsford, and Davies (2007) and Kuljis, Paul, and Stergioulas (2007) `s study. Table 1 summarizes the current prevalent modeling technologies and simulation methodologies, as well as their potential application domains in health services analysis. 1.
2.
3.
4.
Discrete-Event Simulation (DES) - Among the most widely used simulation techniques in health care as evidenced by many previously cited studies, DES appears to be tailor-made for hospital systems to study queuing behaviors of patients waiting for appointments, investigations and treatments. In particular, DES allows the modeler to construct more complex, dynamic and interactive systems. Nonetheless, as Cooper et al. (2007) pointed out in their choice of modeling technique for evaluating health care interventions, it may take more time and money to develop DES models. System Dynamics (SD) – To resolve systems bottlenecks and understand emerging systems states, SD modeled patient flow behaviors in complex health services systems by capturing the feedback loops and inventory control rules for patient arrival, discharge and follow-up visits. This is similar to studying how water flows through a heating system. SD has gained popularity in recent years and some researchers who first applied the DES had switched to SD to deal with the dynamic changes in patient flow. For example, after using DES to redesign the phlebotomy and specimen collection centers in Calgary Laboratory Services (Alberta, Canada), Rohleder, Bischak, and Baskin (2007) have later on decided also to implement a SD model to handle the unexpected performance discrepancies found due to the dynamic interactions within these service centers. Markov process models - Markov models can describe changes in the state of patient health over time such as the case of benign vs. malignant tumors. In this sense, Cooper et al. (2007) argued that Markov models are suitable for chronic disease interventions. Accordingly, they employed a Markov model to evaluate the effectiveness of statins, one of the cholesterol lowering drugs, over time for at risk patients of coronary heart disease (CHD), based on a so-called Southampton CHD model (Cooper, 2005). Monte Carlo Simulation – Essentially, Monte Carlo uses repetitive sampling process to make estimates about key performance variables of interests under uncertainty conditions just like throwing a dice repeatedly to predict the chance of picking a winning stock in the Dow Jones Industrial market. For example, Jacobson, Lindberg, Lindberg, Segerstad, Wallgren, Fellstrom, Hulten, and Jensen-Waern (2001) used the Monte Carlo methodology to sample various potential vaccine price distributions determined by the mix of care providers vis-à-vis parent-guardian acceptance so as to assess the economic values of variously combined vaccines for pediatric immunization.
xxxiii
Table 1. Current modeling and simulation techniques and their applications in health Technique
Potential Applications
Discrete-Event Simulation (DES)
Process reengineering, ward layout design, patient pathway design, scheduling, queuing management
System Dynamics (SD)
Strategic and operations management, alteration management, resource and asset management, patient pathway management
Continuous Simulation
Physical/ biological laboratory processes control
Monte Carlo Simulation
Decision making under uncertainty conditions, risk analysis in the long run
Agent-Based Simulation
Demand and supply management, health economics, risk management
Decision tree
Acute interventions but not for disease recurrence modeling
Markov process
Cohorts of patients between health states over time, chronic disease intervention
Operations research
Patient arrival patterns, LOS management, waiting time management, cost management
Human factors and ergonomic models
Workload analysis, safety monitoring, productivity increasing, error reducing
5.
6.
Continuous Simulation - Continuous simulation is primarily employed to accommodate continuous systems variables and therefore has limitation in its applicability for health care studies. Specifically, it is limited predominantly to physical or biological laboratory processes control simulation such as modeling the trajectory of a missile launch. In health care, it can, for example, be used to assess the design of equipment to further enhance the production volumes and manufacturing process efficiencies of pharmaceutical products (Kuljis et al., 2007) 67 Decision tree – Based a hierarchical tree-like structure, decision trees aid decision making through an assessment of various probabilities in terms of possible consequences corresponding to different options and alternatives. Cooper et al (2007) showed that a decision tree can facilitate comparative decisions on mean life expectancy of CHD patients. Here, the current response states for CHD patients are compared to more efficient ambulance delivery services and thrombolysis (“clot busting” drug) intake that may result in the death and/or the survival of the patients.
In summary, whenever a health system problem is encountered, a critical issue then is how to make choose intelligently among various available modeling and simulation techniques. Schriber and Brunner (2007) proposed to explore the nature and logical foundations in every method and software and thus gain a detailed understanding of “how simulation works”. Cooper et al. (2007) stated in their study that the choice of modeling technique depends on several aspects including modeling technique acceptance, model appropriateness, dimensionality, and ease and speed of model development. Generally a decision can be made based on the complexity and dynamics of the system to be modeled in terms of interaction
xxxiv
of systems elements, model size or resource constraints. The introduction of OSE philosophies also helps making a better combination of choices. For instance, Young (2005) studied the philosophies driving changes in health care services delivery such as Lean Thinking and the Theory of Constraints. He proposed that a strategic agenda could be created out of a three-way fusion of health care delivery, industrial process and simulation capacity, on which basis, an appropriate modeling and simulation techniques can then be selected.
tHe Model and eVIdence drIVen Integrated analysIs (MedIa) paradIgM & Its applIcatIon The MEDIA paradigm, which has previously been introduced at the beginning of this review in Figure 1, offers a handle to unraveling today’s highly complex and environmentally uncertain health services enterprises by focusing on the effective integration of operational models and evidence. Three key phases underline the MEDIA paradigm: (a) Referential ontology modeling; (b) Hybrid probabilistic model integration; and (c) Adaptive knowledge fusion for quality assurance. Referential ontology modeling relates to the qualitative representation of model structure that is generally constrained by the health services systems complexity and uncertainty. Nodes, representative of crucial system artifacts, and the interdependencies among nodes, that represent their relations, typically encapsulate the model structure. Figure 4, for example, shows the nodes and the interdependency links among these nodes for a partial risk assessment model of operational and information risk representation in the context of a generic model structure. As the purpose of referential ontology modeling is essentially to support the generation of hybrid probabilistic models, probabilistic modeling requirements should first be examined in terms of the domain of enterprise services and elements, environment risks, and their relations (dependency). The emergent systems complexity is then represented by referential knowledge in formal ontology based on description logic domain and the uncertainty among systems relations is often addressed through probabilistic representation for ontology modeling such as applying the Bayes approach. Accordingly, a critical step is the conversion of ontology to probabilistic models. The next phase, hybrid probabilistic model integration, entails bridging the qualitative ontology representations with computational analysis to quantify the relations among referential nodes and to achieve an integration of heterogeneous models. This quantification process comprises fundamentally an attempt to assign probabilistic distribution to nodes and/or conditional probability to relations among the nodes. It is important to note that boundary nodes are those nodes shared by multiple models and the same quantification process is applied to these nodes to connect among the models. Special procedures such as those algorithms dealing with virtual and soft evidence may have to be instantiated to adjust the associated probabilities to satisfy all of the probabilistic constraints across each of the model to be integrated (Kim, Valtorta, & Vomlel, 2004; Xiang, 2002). In this second phase, the core migration procedures can utilize semantic web techniques such as ByesOWL framework (Ding, Peng, & Pan, 2006) to convert the ontology to probabilistic networks. The final phase, knowledge fusion, aims at improving the accuracy and confidence of hypothetical estimations about key systems variables based on discrete evidence gathered from multiple locations throughout the systems being studied. Furthermore, adaptive knowledge fusion attempts to achieve a high quality assurance of information collection and operations management decisions in the knowledge
xxxv
Figure 4. A model structure represented by nodes and interdependencies among nodes
vaccination
backup
epidemic
HMM Power Model (1)
Markov Cancer Model (2)
power surge computer
staff
patient
OR
Bayesian Parameter Model (3)
Staff B
Staff A Patient arrival
OR
Patient exit
Computer network topology
staff OR Simulation Model
Staff C
Probabilistic Communication Model (5)
(4)
Decision Model (6) node dependency
evidence control
Resource allocation information
utility
resource flow link
fusion process. Appropriate quality assurance strategies are used iteratively to guide the knowledge fusion process and are typically aided by dynamic quality scores or indices calculated from information entropy of evidence and effectiveness of different operations. Essentially, these strategies will attempt to provide insights to key questions regarding the knowledge fusion process, including: (1) how good the resulting diagnosis or prediction may be; (2) when to engage and/or stop the knowledge integration procedures; (3) where and what information is to be collected; and (4) finally, what resource allocation decisions could be best implemented (Li & Chandra, 2006; Li & Ji, 2005).
MedIa procedural Framework Figure 5 overviews a MEDIA procedural framework that can be used to guide the implementation of the different phases involved in the construction of relevant health IT models and methods for a generic risk management system problem. In the context of the MEDIA paradigm, any kind of a management task such as a risk assessment scenario for a specific system like a hospital in a chosen ontology domain, specifically, health care, will therefore pass through a set of generic procedures. •
Phase 1a: Physical constructs and security components are modeled in ontology bases for the chosen domain. These class templates have common attributes and relations sufficient to support the modeling of most management tasks;
xxxvi
Figure 5. Flowchart of the model and evidence driven integrated analysis (MEDIA) Domain Specific Physical Constructs
Security Components
System and Task Specific System Information
Relations
Security Tasks
Relation Instances
Existing Models
Model ontology
Ontology
Instantiate system models
Identify boundary interfaces
Referential Dependency Models
Translate ontology and fit parameters
modify System Specific Evidence
•
•
•
Link models
Hybrid Probabilistic Computation Models Adaptive knowledge integration
Phase 1b: Aided by the domain templates above, two different types of inputs are needed for referential dependency models to be instantiated for the specific system and the risk management task: (i) for an existing model, boundary nodes and their dependency, and (ii) for a new model, complete knowledge about the model structures of nodes and dependencies; Phase 2: Probabilistic computational models are then generated through ontology translation and parameter fitting to capture the interdependency and their associated uncertainty. This is helped by additional descriptions specific to the system and task, for example, hypothesis nodes of interests and available evidence collection positions designated for risk management; Phase 3: When evidence can be found, probability distributions for nodes connected within a network as well as quality scores are updated accordingly through probabilistic inferences. The probabilistic inferences that are carried out here should be specific to the different types of models, for example, Markovian process, Bayesian, or DES models.
If a model needs to be modified in terms of both its structure and parameters due to changes in the underlying systems dynamics, or a new model needs to be introduced, the process goes back to Phase 1b to repeat or iterate on the instantiation process (Cooper & Herskovits, 1992; Heckerman, Geiger, & Chickering, 1995). The rest of this section will turn to discussing the application of MEDIA paradigm in a multi-tier capacity planning system.
xxxvii
MedIa application for a Multi-tier capacity planning system In the past, mathematical models for the optimal timing prediction of individual disease treatments such as cancers and for the best health services scheduling such as emergency and operation rooms scheduling have been developed respectively. Yet it is important for these models to work together as may be necessary for setting up new global optimization strategies. In the real world, for example, the service appointment and operation room scheduling do not always guarantee that the patient gets the screening test or the surgery at the best chosen timing as dictated by the individual disease progression model. Suboptimal timing for treatment may therefore have to be chosen and analyzed again for potentially undesired consequences to patients. As this might further delay the treatments for patients and jeopardize their life quality while increasing medical cost, these two types of models may have to be reconnected and integrated through their interdependency to optimize the two processes of the overall system. In addition, the challenging issue of uncertainty arises when we consider the dynamic changes such as an epidemic or more disturbing natural and man-made emergencies. In a normal operating environment, the discussed risk in the abovementioned case may not be very serious. However, if epidemic cases such as SARS or events such as the 9/11 terrorist attack occurring simultaneously, for example, excessive demand on medical aids in responding to these events may easily deprive regular patients of “optimized” treatments. The strategy here then is to factor in the uncertainty by integrating and incorporating additional appropriate models. Indeed, this integration can even be expanded either horizontally or vertically into as many channels of the health care system as needed such as the sub-units within the same health care enterprise, or the various business processes and institutional structures of a regional health alliance. Naturally, instead of building a new, but highly sophisticated model from scratch by applying the same algorithmic approach or method, applicable existing models and/or the most suitable type of model for each of these processes can now be reused or variously combined to satisfy the overall systems model optimization requirements. Such an application shows the dominant complexity in terms of a large number of participating systems, subsystems, and basic units, and more significantly, their interdependency. Figure 4, which was also discussed earlier, illustrates a partial risk assessment model of operational and information risk representation in the patient care process. This hybrid probabilistic modeling solution employs a variety of models. A hidden Markov process model (1) predicts the state of power supply for potential power surge. A heterogeneous Markov process model (2) captures the progression of a given type of cancer and estimates the best treatment timings. The predictions from these models enter a Bayesian parameter model (3) that estimates the states of corresponding variables of staff, patient, operation room, and computer network, with the help of other available evidence. Then a simulation model (4) can run to collect statistics on current operating room capacity, using these parameters. Another model in the form of a probabilistic network graph (5) accounts for computer network topology and user configuration. It calculates the risk of information leaking and availability for the different collaboration scenarios based on staffing and computer network situations. Actually this availability prediction can be feedback to the operating room simulation model, if applicable. Moreover, a computational decision model (6) will enable the computation on the utility based on performance information for different allocation actions for given resource constraints. This decision model can also be distributed to more nodes across the system, if and as needed.
xxxviii
Imagine how a comprehensive model of a medical center can be generated through the application of the MEDIA framework. Various sources of evidence can be introduced into this model on a continuing basis, in terms of errors, delays, and clinic starvation alarms from physical, software or human sensory channels, signaling emerging or potential problems. The administrators or doctors, facing an uncertain and complex operational environment, have to ask first, given the evidence, what the big picture really is and where the weakest links may be situated. Then the best decision (where and how) to invest resources (capital and manpower) to avoid further deterioration of the system state and to mitigate the problems can be made intelligently. Moreover, this process will be carried out, not haphazardly and without knowledge, but in an active and timely manner to handle existing incomplete and uncertain information. Imagine also that this model can further be updated with real-time system status by monitoring facilities/sensors in an “online” fashion. The key decision makers and policymakers are not asked to just (always) rely on the “average” profile about user/disease from collected historical data. For example, patient arrival may fluctuate from day to day; different treatment results of a special disease may require different actions to be taken; and different caregivers as well as health administrators may also be motivated to stick stubbornly to their personal information processing styles and biases in making certain decisions. Therefore the latest evidence, collected on a continuing basis to update information relating to individual “special” patient, at a prescheduled pace and/or for significant events, will all be made available as feedback to the existing model structure. Over time, the model structure or parameters may, of course, be no longer appropriate, relevant or accurate enough for current or future predictions (so called “concept drift”). Then, model learning and knowledge fusion will again be executed at these times as well. Put together, the MEDIA paradigm is clearly a useful and practical framework that can be applied intelligently across any type of a real-world scenario to enable the mixture and integration of various health IT models and methods to facilitate the making of short-term key health services delivery and management decisions as well as to aid longer-term health systems planning and policymaking.
conclusIon Essentially, our review on existing bodies of knowledge about the state-of-the-art health IT models and methods has led us to become increasingly aware of the need for and significance of having an integrative platform for leveraging existing models and methodologies intelligently such as the MEDIA paradigm. This paradigm aims to offer an appropriate framework for conjugating heterogeneous probabilistic models including the Bayesian network (Pearl, 1998), Markov process models (Doob, 1953), and Monte-Carlo simulation (Fishman, 2001), through virtual or soft evidence update of distributions over boundary nodes. The rationale for achieving such an integration is that: (1) inherent probabilistic representation and inference is able to deal with uncertainty; (2) many such probabilistic models have built-in mechanisms, such as in dynamic Bayesian networks and Markov process models to deal with the temporal dependency; and (3) many of these models support localization of information processing and can be easily implemented into agent-based architectures, aided by the interdependency present in enterprise constructs and model components. More specifically, just as languages spoken by people, we have seen that ontology representation is necessary for computer models to communicate and interact with each other. Ontology defines the concepts and the relations of generic elements in a domain and provides a common “language” shared
xxxix
by components and subsystems of MEDIA. It is more than just a classification or a dictionary because dynamic behaviors of the systems elements are also captured using description logic and language such as RDF and OWL (Baader, Calvanese, McGuinness, Nardi, & Patel-Schneider, 2003; Barwise, 1997). We have also seen how the integration of health IT models and methods can benefit through the application of such a shared “language” for accurate and practical representation of various constructs in information theoretic terms to represent risk and uncertainty in complex systems environments, thereby permitting the accurate and rapid assessments of emergency situations and allowing efficient and effective health services to be delivered productively. Currently, there is a dire shortage of major studies in the applications of interoperable, integrative health IT models and methods. As well, inadequate knowledge has yet to be derived from other industrial and system engineering sectors such as manufacturing enterprises into aiding comparative processes in health services delivery and management. For instance, the surgery department is always at the core of many hospitals. Managing operation rooms requires a variety of expensive resources including surgical space and facility, equipment and material supply, nurse, technician and doctor, to provide timely appointments to patients. It is a challenging task in and by itself. Yet, the people manning the surgery department has still to interact with all the inpatient-outpatient departments and clinics, as well as to be linked to a huge number of medically interconnected components that will affect their own behaviors and dynamics at any moment. Reports and requests are largely the only primary information artifacts interfacing between the operation rooms and other care providers in the same hospital, and mostly for discrete interactions updated at a fixed time interval. The surgeons and surgical administrator will often not be familiar or have knowledge of the status of other systems and organizational components, and without the aids provided to them from an integrated platform of health IT models and methods, they cannot possibly make the best decisions, including the operation room assignments. Ironically, most hospital departments and clinics nowadays have management models, policies, and supporting software packages in place, but all running individually, to “efficiently” schedule patient appointments and make medical decisions in an isolated manner. With the MEDIA framework application, decisions on such localized “best” operations can now be optimized for the entire system. Feasible assignments satisfying all of the required health policies for the various connected health provider institutions can then be assured. Our case example may now be further generalized to many other equally complex scenarios that commonly occur in health services delivery. For example, even for a specialty clinic such as the urology or cardiology unit located within a hospital, it has to deal with many other elements including primary care, inpatient, outpatient, emergency, surgery, lab testing, logistics, and many other areas in terms of the different nodes and types of information flow, including patient, provider and material flows. Similarly, an insurance company has to navigate through patients, medical providers, pharmaceutical companies, and other vendors or suppliers in order to keep routine financial claim transactions in order. For all of these health provider organizations, if we do not attempt to empower the different health services managers with the necessary, interoperable health IT capabilities, there will be no assurance that a well-coordinated effort in patient care delivery will be sustained. The need for such a well-coordinated effort and enterprisewide collaboration across the spectrum of health services delivery is even greater when we consider health alliances, which may be formed at varying scales, levels and sizes. Frequently emerging out of the cooperation among multiple health provider institutions of primary clinics, hospitals, research institutes, insurance management organiza-
xl
tions, and even government health agencies, health alliances are dependent on the separate units to work together in order to achieve high quality patient care. Piece-by-piece examination of related management models and systems does not and will not offer much help when local operation decisions have to be made based upon disparate evidence. In such networked systems and sub-systems, a medical provider or manager needs to know far more than his or her own world in order to make better, if not the “best,” decisions for the patients and the health alliance organization as a whole. Often, the critical need is not more models, or even more sophisticated yet isolated models – where abundant, if still not adequate, models have already been developed, implemented, and studied – instead, the primary need here may just be how to integrate existing models productively and intelligently so as to allow interoperable health IT applications to be shared. Then, and if necessary, new models and methods can be added to enhance those aspects of decisions that may still be lacking or have yet to be addressed. A wonderful developing story for applying the MEDIA paradigm is the 2009 case of the “swine flu” (H1N1) pandemic that is ongoing at this point in time (Esterl, 2009). It is just impossible to have one model or system to capture everything relating or reported about this pandemic, from the private clinics to the US governmental health services sector plan for controlling H1N1 to the different magnitude of H1N1 developments spreading throughout the globe. In other words, it is just not enough to harbor the narrow perspective of a tiny component within the entire global health service system, but having an interconnected worldview of even just how the H1N1 is affecting the different nations will significantly improve the patient survival chance and H1N1 prevention for a particular or entire country. Even so, the demand to integrate models and evidence for seamless analysis and management of a pandemic like H1N1 may be attributed largely to the following facts: 1.
2.
3.
The complexity and uncertainty are greatly increasing in the system of systems context of health services supply chains and global health information exchange networks to which any analysis and management systems have to appropriately respond. The rapidly changing medical and health technology landscape and innovative practices in medicine, health management, industry and government policy, as well as the evolving field of health IT models and methods require the constant incorporation of new systems perspective and integrative model approaches. Unfortunately, as noted previously, vast legacy HISs have typically been built piece by piece, resulting in the preservation only of disparate health IT models without the capability to leverage intelligently from evidence aggregated over time from diverse sources embedded in the overall global health care system.
Development of novel technology and unconventional business model has inspired complex enterprise structures to evolve as in the case of the US health services delivery system. These enterprises usually form close partnerships in order to survive and grow under increasing competition, for which decision making and policymaking relying on existing and past discrete models for isolated system segments is no longer sufficient. Another obstacle many of these health enterprises encountered in across-the-organization management is the complicated systems dynamics. This needs to be dealt with by systematically increasing the connectivity of existing models, resulting in the constant understanding and monitoring of the evolving systems states comprising rapid changes in participants and configurations along both temporal and spatial dimensions. As characterizing these different systems states efficiently can simply become overwhelming, health services quality assurance is undoubtedly a very demanding task. Interoperable,
xli
integrative health IT models and methods via the MEDIA paradigm aim at reducing such complexity to meet the demands of high quality health services. The MEDIA paradigm also addresses these systems challenges through referential ontology modeling, hybrid probabilistic model integration, and adaptive knowledge fusion that adopt a real system-wide and process oriented perspective. Thus, the introduction of the MEDIA framework to emerging health IT models and methods research represents a key step to unifying our knowledge for modern health services enterprise management. The future of health IT models and methods is dependent on the development of even more powerful integrative frameworks beyond the MEDIA paradigm. Such a paradigm will seek to provide a consistent approach to reusing and integrating available knowledge, dealing with complexity and uncertainty dimensions in multiple ways, and encouraging the sharing and exchanging of health information privately and securely across organizational boundaries that do not currently exist. This systematic methodological conjugation represents a departure in traditional integrated modeling and analysis from fitting all systems into just “one type of model” to an emphasis on knowledge interpretation, interaction and fusion among models. This makes it much easier to achieve balance between the local and global representation of different models. In other words, the hybrid probabilistic model integration can actually make use of any available models and/or choose the most suitable ones for each subsystem/task, rather than having to start the modeling process from scratch each and every time one or more new systems problem(s) is encountered. In closure, the MEDIA methodology or an expansion of such a methodology can further be applied to create novel solutions to new health systems challenges faced in a wide range of practical contexts that vie to achieve the convergence of enterprise systems and processes. The trend toward preventive health, alternative and integrative medicine, e-health and healthy lifestyle promotion, for example, are all virgin grounds for the applications of integrative, interoperable health IT models and methods. Imagine how a self-monitoring community health care system that can be equipped with a regional center that uses interoperable and integrated health IT models and methods to aid in guiding individual residents throughout the community to cope with undue stresses arising from risky driving behaviors on a daily basis, become quickly alerted and fully prepared for emergencies by deploying the community health resources efficiently and effectively on an as needed basis, and channeling any and all unused health resources to evolve the community in multiple ways to adopt a healthy, low risk and more active lifestyle environment within the community. Joseph Tan McMaster University, Canada Xiangyang Li University of Michigan - Dearborn, USA Yung-wen Liu University of Michigan - Dearborn, USA
xlii
reFerences Agrawal, N., & Nahmias, S. (1997). Rationalization of the supplier base in the presence of yield uncertainty. Production and Operations Management, 6, 291–308. Anupindi, R., & Akella, R. (1993). Diversification under supply uncertainty. Management Science, 39, 944–963. Baader, F., Calvanese, D., McGuinness, D. L., Nardi, D., & Patel-Schneider, P. F. (2003). The Description Logic Handbook: Theory, Implementation, Applications. Cambridge, UK: Cambridge University Press. Barbera, J., & Macintyre, A. (2002). Medical and health incident management system: A comprehensive functional system description for mass casualty medical and health incident management, Retrieved from http://www.gwu.edu/~icdrm. Barwise, J. (1997). An Introduction to First-Order Logic. Handbook of Mathematical Logic (pp. 5-46). Amsterdam. Blackhurst, J., Craighead, C.W., Elkins, D., & Handfield, R. B. (2005). An empirically derived agenda of critical research issues for managing supply-chain disruptions. International Journal of Production Research, 43(19), 4067-4081. Bosire J., Wang, S., Gandhi, T., & Srihari, K. (2007). Comparing simulation alternatives based on quality expectations. In Proceedings of the Winter Simulation Conference (pp. 1579-1585). Brailsford, S.C. (2007). Tutorial: advances and challenges in healthcare simulation modeling. In Proceedings of the 39th Conference on Winter Simulation (pp. 1436-1448). Brailsford, S., Lattimer, V., Tarnaras, P., & Turnbull, J. (2004). Emergency and on-demand health care: modeling a large complex system. The Journal of the Operational Research Society, 55(1), 34-42. Brantes, F.D., Galvin, R.S., & Lee, T.H. (2003). Bridges to Excellence: building a business case for quality care, Journal of Clinical Outcomes Management, 10, 439. Cahill, W., & Render, M. (1999). Dynamic simulation modeling of ICU bed availability. Winter Simulation Conference, 2, 1573-1576. Cooper, G.F., & Herskovits, E. (1992). A Bayesian method for the induction of probabilistic networks from data. Machine Learning, 9, 309-347. Cooper, K (2005). A comparative study of modeling approaches for the evaluation of healthcare intervention with reference to coronary heart disease. PhD Dissertation (unpublished), University of Southampton. Cooper, K., Brailsford, S., & Davies, R. (2007). Choice of modeling technique for evaluating health care interventions. Journal of the Operational Research Society, 58(2), 168-176. Denton, B., Fowler, J., Schaefer, A., Batun, S., Erdogan, E., & Gul, S. (2009). Optimization of the design and operation of surgery delivery systems. In Proceedings of the 2009 NSF engineering research and innovation conference.
xliii
Ding, Z., Peng, Y., & Pan, R. (2006). BayesOWL: Uncertainty modeling in semantic web ontologies. Soft Computing in Ontologies and Semantic Web. Springer-Verlag. Doob, J. L. (1953). Stochastic Processes. New York: John Wiley and Sons. Eldabi, T., & Young, T. (2007). Towards a framework for healthcare simulation. In Proceedings of the 39th Conference on Winter Simulation. 1454-1460. Esterl, M. (2009, June 11) Swine Flu pandemic is closer, says WHO. Wall Street Journal. Fishman, S. (2001) Discrete-Event Simulation: Modeling, Programming, and Analysis. New York: Springer-Verlag. Ford, E.W., & Hughes, J.A. (2007). A collaborative product commerce approach to value-based health plan purchasing. Supply Chain Management: An International Journal, 12(1), 32-41. Graves, S.C., Kletter, D.B., & Hetzel, W.B. (1998). A dynamic model for requirements planning with application to supply chain optimization. Operations Research, 46(3), 35-49. Gunal, M., & Pidd, M. (2005). Simulation modeling for performance measurement in healthcare. In Proceedings of the 37th Conference on Winter Simulation (pp. 2663-2668). Hale, T., & Moberg, C.R. (2005). Improving supply chain disaster preparedness: A decision process for secure site location. International Journal of Physical Distribution & Logistics Management, 35(3), 195-207. Heckerman, D., Geiger, D., & Chickering, D. (1995). Learning Bayesian networks: The combination of knowledge and statistical data. Machine Learning, 20(3), 197-243. Hershey, J., Pierskalla, W., & Wandel, S. (1981). Nurse staffing management. In D. Boldy (Eds.), Operational Research Applied to Health Services, (pp. 189-220). London: Croon Helm. Hick, J. L., Hanfling, D., Burstein, J. L., DeAtley, C., Barbisch, D., Bogdan, G. M., & Cantrill, S. (2004). Health care facility and community strategies for patient care surge capacity. Annals of Emergency Medicine, 44(3), 253-261. IEEE-USA, (2005) Interoperability for the National Health Information Network, Whitepaper by the Medical Technology Policy Committee (MTPC) of The Institute of Electrical and Electronics EngineersUnited States of America (IEEE-USA), November. IT Governance Institute (ITGI). (2007). COBIT4.1-Control Objectives for Information and Related Technology. Retrieved January 20, 2010, from http://www.itgi.org/ Jacobson, M., Lindberg, J. E., Lindberg, R., Segerstad, C.H., Wallgren, P., Fellstrom, C., Hulten, C., & Jensen-Waern, M. (2001). Intestinal cannulation: model for study of the midgut of the pig. Comp Med., 51(2), 163–170. Kiel, L. D., & Elliott, E. (Eds.). (1996) Chaos theory in the social sciences: Foundations and applications. Ann Arbor, MI: University of Michigan Press.
xliv
Kim, Y.G., Valtorta, M., & Vomlel, J. (2004). A prototype system for soft evidential update. Applied Intelligence, 21(1), 81-97. Kouvelis, P., & Milner, J.M. (2002). Supply chain capacity and outsourcing decisions: The dynamic interplay of demand and supply uncertainty. IIE Transactions, 34, 717–728. Kollberg, B., Dahlgaard J. J., & Brehmer, P. (2007). Measuring lean initiatives in health care services: issues and findings. International Journal of Productivity and Performance Management, 56(1), 7-24. Kuljis, J., Paul, J., & Stergioulas, L. (2007). Can health care benefit from modeling and simulation methods in the same way as business and manufacturing has? In Proceedings of the 2007 Winter Simulation Conference (pp. 1449-1453). Kumar, A., & Ozdamar L. (2004). Business process reengineering at the hospitals: A case study at Singapore hospital. In Proceedings of the 18th European Simulation Multiconference (CD version). Kumar, A., Ozdamar L., & Zhang C. (2008). Supply chain redesign in the healthcare industry of Singapore. Supply Chain Management: An international Journal, 13(2), 95-103. Kumar, S., Swanson E., & Tran, T. (2009). RFID in the healthcare supply chain: usage and application. International Journal of Health Care Quality Assurance, 22(1), 67-81. Lambert, D.M., Cooper, M.C., & Pagh, J.D. (1998). Supply chain management: Implementation issues and research opportunities. The International Journal of Logistics Management, 9(2), 1–19. Laudon, K. and Laudon, J. (2007). Can information systems help prevent a public health crisis? In Management Information Systems (10th ed., pp. 77-78). Lee, H. L., & Whang, S. (2005). Higher supply chain security with lower cost: Lessons from total quality management. International Journal Production Economics, 96, 289-300. Li, X., & Chandra, C. (2006). Efficient knowledge integration to support complex supply network management. International Journal of Manufacturing Technology and Management, 10(1), 1-18. Li, X., & Ji, Q. (2005). Active affective state detection and user assistance with dynamic Bayesian networks. IEEE Transactions on Systems, Man, and Cybernetics-Part A, 35(1), 93- 105. Lipowicz, A. (2009, October 23). Government agencies to spend $15 billion on health IT by 2014. Federal Computer Week, Retrieved January 20, 2010, from http://fcw.com/Articles/2009/10/23/Governmentagencies-to-spend-15-billion-on-health-IT-by-2014.aspx?p=1 National Coalition on Health Care. (2009). Facts on Health Insurance Costs, Retrieved from http://www. nchc.org/facts/cost.shtml O’Leary, D. (2004). Models of infection: person to person, Computing in Science & Engineering, 6(1), 68-70. Pai, R.R., Kallepalli, V.R., Caudill, R.J., & Zhou, M. (2003). Methods toward supply chain risk analysis. In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, 5, 4560-4565.
xlv
Parisi, L., & Bailey, K. (2009, April). Too Great a Burden: Americans Face Rising Health Care Costs. Retrieved from http://hdl.handle.net/10207/18507. Parlar, M., & Perry, D. (1996). Inventory models of future supply uncertainty with single and multiple suppliers. Naval Research Logistics, 43, 191-210. Pearl, J. (1998). Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. San Mateo, CA: Morgan Kaufmann Publishers. Pierskalla, W.P., & Brailer, D.J. (1994). Applications of operations research in health care. Handbooks in OR & MS, 6, 469-505. Raghupathi, W., & Tan, J. (2002). Strategic IT Applications in Health Care. Communications of the ACM, 45(12), 56-61. Raguphati, W., & Tan, J. (2008). Information Systems and Healthcare XXX: Charting a Strategic Path for Health Information Technology. The Communications of the Association for Information Systems, 23(1), 500-525. Ramis, F., Palma, J., Estrada, V., & Coscolla, G. (2002). A simulator to improve patient’s service in a network of clinic laboratories. In Proceedings of the Winter Simulation Conference (Vol. 2, pp. 1444-1447). RAND Corporation (2004). Evaluating the Security of the Global Containerized Supply Chain, Technical Report, Santa Monica, CA Willis, H.H., & Ortiz, D.S. Rico, F., Salari, E., & Centeno, G. (2007). Emergency departments nurse allocation to face a pandemic influenza outbreak. In Proceedings of the Winter Simulation Conference (pp. 1292-1298). Roberts, S. D. (2007). Simulation of medical decisions, IERC 2007, Nashville, TN. Retrieved January 20, 2010, from http://74.125.95.132/search?q=cache:GYvFAgiR6rIJ:ccehub.org/resources/121/download/ steve-roberts-simulationofmedicaldecisions-crc.ppt+Simulation+of+medical+decisions,+IERC+2007& cd=1&hl=en&ct=clnk&client=safari Rohleder, R., Bischak, P. & Baskin, B. (2007). Modeling patient service centers with simulation and system dynamics, Health Care Management Science, 10(1), 1-12. Schnase, J. L., & Cunnius, E. L. (Eds.). (1995). Proceedings from CSCL ‘95: The First International Conference on Computer Support for Collaborative Learning. Mahwah, NJ: Erlbaum. Schriber T. J. & Brunner, D. T. (2007). Inside discrete-event simulation software: how it works and why it matters. In Proceedings of the Winter Simulation Conference (pp. 113-123). Shehad, M., Bertino, E., & Ghafoor, A. (2005). Secure collaboration in mediator-free environment. In Proceedings from CCS’05: ACM Conference on Computer and Communications Security, Alexandria, VA. Shubik, M. (2002). Game theory and operations research: Some musings 50 years later. Operations Research, 50, 192-196. Simchi-Levi, D., Kaminsky, P., & Simchi-Levi, E. (2003). Designing and Managing the Supply Chain: Concepts, Strategies, and Case Studies (2nd ed.). McGraw-Hill Irwin.
xlvi
Stahl, J., Roberts, M., & Gazelle, S. (2003). Optimizing management and financial performance of the teaching ambulatory care clinic. Journal Gen Intern Med, 18(4), 266-274. Swinehart, K. D., & Smith, A. E. (2005). Internal supply chain performance measurement: a health care continuous improvement implementation International Journal of Health Care, 18(7), 533-542. Tan, B., Gubaras, A., & Phojanamongkolkij, N. (2002). Simulation study of Dreyer urgent care facility. In Proceedings of the Winter Simulation Conference (Vol. 2, pp. 1922-1927. Tan, J., & Payton, F. C. (2010). Adaptive health management information systems: concepts, cases, & practical applications (3rd ed.). Sudbury, MA: Jones & Bartlett Publishers, Inc. Tan, J., Wen, J.H., & Awad, N. (2005, May). Health care and services delivery systems as complex adaptive systems. Communications of the ACM, 48(5), 36-44 Thomas, M.U. (2002). Supply chain reliability for contingency operations In Annual Proceedings of Annual Reliability and Maintainability Symposium (pp. 61-67). U.S. Bureau of Economic Analysis. (2009). Manufacturing Industries Led Slowdown in 2008. Retrieved from http://www.bea.gov/newsreleases/industry/gdpindustry/2009/gdpind08.htm. U.S. House of Representatives, (2007). The XDR tuberculosis incident: a poorly coordinated federal response to an incident with homeland security implications. Hearing of U.S. House of Representatives Homeland Security Committee, June 6, 2007. Woolhandler, S., Campbell, T., & Himmelstein, D. U. (2003). Costs of health care administration in the United States and Canada. The New England Journal of Medicine, 349(8), 768-775. Xiang, Y. (2002). Probabilistic Reasoning in Multiagent Systems: A Graphical Models Approach. Cambridge University Press. Young, T. (2005). An agenda for healthcare and information simulation. Health Care Management Science, 8, 189-196.
1
Chapter 1
Evaluating Health Information Services:
A Patient Perspective Analysis1 Umit Topacan Bogazici University, Turkey Nuri Basoglu Bogazici University, Turkey Tugrul U. Daim Portland State University, USA
aBstract The objective of the chapter is to explore the factors that affect users’ preferences in the health service selection process. In the study, 4 hypothetical health services were designed by randomly selecting levels of 16 attributes and these services was evaluated by the potential users. Analytical Hierarchy Process (AHP), one of the decision making methods, was used to assess and select the best alternative.
IntroductIon Healthcare service providers benefit from different technologies so as to reduce cost and improve quality of the medical procedures (Gagnona, Godinb, Gagnéb, Fortina, Lamothec, Reinharza, & Cloutierd, 2003). In particular, telemedicine resides on the center of these technologies. The American Telemedicine Association defines telemedicine as “the use of medical information exchanged from one site to another via electronic communications to improve patients’ health status” (ATA, 2009). Telemedicine applications were used in a broad
range including consultation (Berghout, Eminovic, de Keizer, & Birnie, 2007), education and training (Chen, Yang, & Tang, 2008), and home care (Biermann, Dietrich, Rihl, & Standl, 2002). Selecting the best telemedicine service among given ones is a complex task. The process needs considerations of trade-offs between cost and benefits of the service. Analytic Hierarchy Process (AHP) (Saaty, 1977; Saaty, 1996) is an outstanding method that can be used in multifactor decisionmaking environments. It presents a structured approach to determine individual weights of multiple attributes of a product or service so that
DOI: 10.4018/978-1-61692-002-9.ch001
Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Evaluating Health Information Services
they can be compared in a simple way. Then, it simplify decision-making in the selection process. Recent technological developments enable advancements in delivery of medical services, appropriate healthcare at a reasonable cost, and access to quality healthcare in underserved areas in the medical sector. Telemedicine is one of these developments that “enable remote medical procedures and examinations between patients and medical providers via telecommunication technologies like the Internet, or telephone” (AlQirim, 2007). Moreover, many previous researches show that compared to traditional medical care, telemedicine services present many benefits to the patients and physicians (van den Brink, Moorman, de Boer, Pruyn, Verwoerd, & van Bemmel, 2005; Chae, Lee, Ho, Kim, Jun, & Won, 2001). Diffusion of intelligent monitoring systems in the medical industry has gather speed with the help of recent developments in the information and communication technologies. “Smart homes’ for telecare by means of movement detector, oxymeter, tansiometer and various other devices (Rialle, Lamy, Noury, & Bajolle, 2003), a ringsensor that monitors patient’s blood oxygen saturation (Yang & Rhee, 2000) and a web based electrocardiogram monitoring application facilitating collect, analyze and storage of patient data (Magrabi, Lovell, & Celle, 1999) were designed by researchers to follow-up of patients’ health status at home. Evaluating health technologies is a complicated procedure because people face with some difficulties while evaluating trade-offs between alternatives. Analytic Hierarchy Process (AHP) is a potential decision making method to deal with complex decisions. The aim of AHP is to qualify relative priorities for a given set of alternatives on a ratio scale. In the literature, many applications of AHP have published in different fields including planning, resource allocations, and selecting a best alternative, etc. (Magrabi, Lovell, & Celle, 1999). AHP method also widely used in vendor selection problems (Nydick & Hill, 1992; Tam & Tummala, 2001).
2
There are many applications of AHP in the medical field. In one of the researches, AHP is used to develop an human resource planning model for hospital laboratory personnel (Kwak, McCarthy, & Parker, 1997). Turri applied AHP approach to select a magnetic resonance imaging vendor by using the criteria like price, service, and technology (Turri, 1988). Another application of AHP was designed by Kahen and Sayers for selection of medical expert systems (Kahen & Sayers, 1997). In the study, assessment of a health service prototype have done in the design phase of the service development so as to make clear the factors affects users attitude toward health services. Usability tests of the health services (Kaufman, Patel, Hilliman, Morin, Pevzner, Weinstock, Goland, Shea, & Starrena, 2003), affects of ergonomics on the medical device development (Martin, J.L., Norris, Murphy, & Crowe, 2008) and self-measurement satisfaction of hypertension patients (Bobrie, Postel-Vinay, Delonca, & Corvol, 2007) are just some of them. These studies have applied different techniques in the analyses phase of the research. However, it is difficult to find a study that uses AHP method. Topacan et al (2008) identified 37 different criteria for health information service (HIS) adoption including cost, time factor, content, language, security, customizability, output quality, menu items, input type, sound quality and availability of face-to-face communication. The studies of Karahanna et al (1999), Simon et al (2007), Chang et al (2007) identified a number of criteria with respect to triability, medical provider and vendor support that are applicable for selecting health information service. Tung et al (2008) define financial cost as “the extent to which a person believes that using the electronic logistics information system will cost money”. It was found that financial cost has an influence on adoption of electronic logistic information system (Tung, Chang, & Chou, 2008). Chang et al also examined security protection and
Evaluating Health Information Services
vendor support while studying electronic signature adoption. They found that vendor support affects adoption, and there is no relation between security and e-signature adoption (Chang, Hwang, Hung, Lin, & Yen, 2007).
eValuatIon process
•
•
•
evaluation Model
•
Based on the previous studies, 21 different health service selection criteria were selected for the research. We conducted a mini survey involving randomly selected 3 male and 3 female potential users. The purpose of this mini survey is to assess and identify critical factors and reduce the number of criteria. In this phase, potential users ranked these criteria based on importance level of them. These participants also made suggestions about the levels of the attributes. Finally, 16 attributes were selected and used in the AHP model. Attributes used in the study is explained below;
•
•
•
•
•
•
Implementation Cost of the service is the purchase price of it. This price is paid only once while buying the service. Operating Cost is the price that was paid by the user in every month to be able to use the service. Usage Time is the amount of time required to use the service. 10 minutes usage time means that the user spends 10 minutes in a day while entering meal information and other data. Response Time is the doctor’s response time to the patients’ requests. As an example, in the alternative 1, the service provider guaranteed that doctor could respond the patients’ questions in 45 minutes. Language is the set of visual and auditory signs of communication.
•
•
• • •
•
Security means whether the service provides secure communication between patients and doctors, or not. Triability is the degree to which patient may be experimented with the service on a limited time or functionality. Sound Type is the voice style of the service. It can be one of computerized, male or female voices. Input Type refers the data entry procedure of the patient. Availability of Face-to-Face Communication means whether the service enable the patient to visit the doctor in the hospital or not. Technical Support is how to assist the user in order to solve the specific problems with the service. User Training means the demonstration the features and functionality of the service to the potential users. Medical Provider is the hospital that supports the medical related operations. Customizability is the ability of the service to be changed by the user preferences. Clarity of Menu Items is the degree to which the user may understand the function of the menu item. Output Quality refers how well the system performs the jobs (Venkatesh & David, 2000).
These attributes can be grouped into three major categories of cost, service characteristics and service provider characteristics. The cost factors include implementation and operating cost. The service characteristics consist of mobile device, triability, content, availability of face-toface communication and time. Similarly, service provider characteristics include technical provider and medical provider. Moreover, some of these sub criteria include one more level called sub criteria 2. Mobile device consists of input type, customizability, clarity of menu items and sound
3
Evaluating Health Information Services
Figure 1. AHP Model
type. Content attribute, on the other hand, includes output quality, language and security. Time factor contains usage time and response time. Technical provider consists of technical support and user training. The hierarchy can be visualized as a diagram shown in Figure 1. The diagram contains an overall goal and criteria broken down into two levels of sub-criteria. AHP model was applied on 4 different hypothetical home based telemonitoring services. Each of the services has 16 different attributes. Levels of these attributes were proposed by the potential users in the mini research phase of the study. Based on these levels, 4 different alternative services were designed by randomly selecting among levels. All of the alternatives and their levels are shown in Appendix A.
4
participants profile 6 male and 8 female potential users from different age groups were selected to conduct the research. Age of the participants was in a range between 23 and 60. Average age of the male and female group was 37. Table 1 contains details about the profile of participants.
FIndIngs Each of the participants have compared the attributes and their levels two by two in order to incorporate judgments about them. They responded 69 different pair wise comparison questions. Sample questions shown in Table 2 and Table 3. In these tables, scale number “9” refers that the attribute has extreme importance with respect to counter attribute. Intensity of importance decreases while
Evaluating Health Information Services
Table 1. Participants Profile Age Groups
Average Age
20-30
30-40
40-50
50+
Total
Male
2
2
1
1
6
36
Female
2
2
2
2
8
38
Total
4
4
3
3
14
37
Table 2. Sample Questionnaire Item for Attributes Implementation Cost
9
8
7
6
5
4
3
2
1
2
3
4
5
6
7
8
9
Operating Cost
Table 3. Sample Questionnaire Item for Attribute Levels Implementation Cost 200 YTL
9
8
7
6
5
4
3
2
1
2
3
4
5
6
7
8
9
500 YTL
200 YTL
9
8
7
6
5
4
3
2
1
2
3
4
5
6
7
8
9
800 YTL
500 YTL
9
8
7
6
5
4
3
2
1
2
3
4
5
6
7
8
9
800 YTL
Table 4. Sample Pair-wise Comparison Matrix for Cost Attribute Implementation Cost
Operating Cost
Weight
Implementation Cost
1
1/4
0.20
Operating Cost
4
1
0.80
moving down to number “2” in the scale. Moreover, scale number “1” refers that both of the attributes has equal importance for the participant. The following steps illustrate the AHP method applied in this research. 1.
2.
Participants responded questionnaire form. In the first part of the questionnaire (a sample pair-wise comparison question shown in Table 2), they compared attributes against each other to calculate weight of them. In the second part of the questionnaire (sample pair-wise comparison questions shown in Table 3), the participants compared level of attributes.
3.
4.
5.
Weight of attributes (Table 4) and priority vectors of alternatives (Table 5) was calculated for each of the participants separately. A table that shows the preferences of a single participant was produced for each of participants separately (Table 6). As an example, Table 6 shows that operating cost is the most important factor for the selected participant. In order to calculate total affects of 14 participants, average of the weight and priority vector values calculated in a separate table. Tables 7, 8 and 9 shows the overall results that is the average values of 14 participants.
5
Evaluating Health Information Services
Table 5. Sample Pair-wise Comparison Matrix for Implementation Cost Levels A1 (200 YTL)
A1 (200 YTL)
A2 (500 YTL)
A3 (800 YTL)
A4 (500 YTL)
Priority Vector
1
4
8
4
0.57
A2 (500 YTL)
1/4
1
6
1
0.19
A3 (800 YTL)
1/8
1/6
1
1/6
0.04
A4 (500 YTL)
1/4
1
6
1
0.19
Table 6. Selection Matrix of a Participant Attribute
Weight
A1
A2
A3
A4
Implementation Cost
0.040
0.573
0.191
0.045
0.191
Operating Cost
0.161
0.050
0.138
0.138
0.673
Usage Time
0.020
0.654
0.062
0.142
0.142
Response Time
0.156
0.046
0.121
0.713
0.121
Language
0.014
0.658
0.106
0.118
0.118
Security
0.099
0.250
0.250
0.250
0.250
Triability
0.033
0.188
0.154
0.188
0.470
Sound Type
0.007
0.250
0.250
0.250
0.250
Input Type
0.087
0.538
0.174
0.115
0.174
Availability of Face-to-Face Comm
0.130
0.375
0.125
0.125
0.375
Technical Support
0.013
0.713
0.121
0.121
0.046
User Training
0.003
0.228
0.291
0.384
0.097
Hospital
0.076
0.429
0.071
0.071
0.429
Customizability
0.043
0.313
0.313
0.063
0.313
Clarity of Menu Items
0.014
0.232
0.117
0.418
0.234
Output Quality
0.104
0.083
0.417
0.083
0.417
Sum
1.000
0.246
0.185
0.226
0.343
In Table 7, the local weights represent the relative weight of the nodes within a group of siblings regarding their parent. The global weights are calculated by multiplying the local weights of the siblings by their parents’ local weights. It can be seen from the table that local weights of each group add up to 1 and the global weights of all the 16 attributes add up 1. According to final calculations, Table 7 shows that potential users pay more attention on the characteristics of the service like face-to-face communication, content, and time in the health service selection process. Moreover, operating cost
6
of the service is more important than purchasing cost. Also, content and availability of face-toface communication are two of the significant service characteristics. Medical provider affects decision of potential user more than technical provider. Clarity of menu items, security and response time are other important factors. And, as a characteristic of the technical provider, users attach slightly more importance to user training than technical support. Compared to usage time, short response time has more positive effects on patients selection decision of HIS.
Evaluating Health Information Services
Table 7. Overall Local and Global Weights of Attributes Criteria
Local Weight
Sub Criteria 1
Local Weight
Service Chars
0.39
Face to Face Comm.
0.27
Content
0.26
Ser. Prov. Chars
Cost
0.37
0.24
Time
0.22
Triability
0.15
Mobile Device
0.10
Medical Provider
0.78
Technical Provider
0.22
Local Weight
Global Weight 0.105
Security
0.48
0.049
Output Quality
0.35
0.035
Language
0.17
0.017
Usage Time
0.23
0.020
Response Time
0.77
0.066 0.058
Clarity of Menu Item
0.40
0.016
Customizability
0.28
0.011
Input Type
0.25
0.010
Sound Type
0.07
0.003 0.289
Technical Support
0.57
0.046
User Training
0.43
0.035
Operating Costs
0.65
0.156
Implementation Costs
0.35
0.084
Table 8. Overall Global Weights of Attributes Attribute
Sub Criteria 2
Weight
Medical Provider
0.289
Operating Cost
0.156
Face-to-Face Communication
0.105
Implementation Cost
0.084
Response Time
0.066
Trialability
0.058
Security
0.049
Technical Support
0.046
Output Quality
0.035
User Training
0.035
Usage Time
0.020
Language
0.017
Clarity of Menu Items
0.016
Customizability
0.011
Input Type
0.010
Sound Type
0.003
Sum
1.000
Table 8 presents weight of the 16 attributes in the decreasing order. According to Table 8, medical provider is the most important factor for health service selection. Secondly, people give importance on the operating cost of the medical service. Thirdly, they choose services that make available face-to-face communication with the physician. On the other hand, input type, customizability and sound type was found three of least important attributes according to overall values. Table 9 presents overall priority vector values of alternatives. According to the table, people significantly choose to pay least amount of money and not spare much usage time for the health information services. Moreover, alternatives that provide secure data communication and storage environment, offer one week unlimited trial period, and include seven days twenty-four hours technical support were highly preferred by the participants. People prefer consulting about their health status to private hospitals instead of public ones and
7
Evaluating Health Information Services
Table 9. Overall Priority Vectors of Alternatives Attribute
A1
A2
A3
A4
Implementation Cost
0,60
0,17
0,06
0,17
Operating Cost
0,05
0,17
0,17
0,61
Usage Time
0,62
0,06
0,16
0,16
Response Time
0,09
0,16
0,58
0,16
Language
0,31
0,10
0,30
0,30
Security
0,41
0,09
0,41
0,09
Triability
0,07
0,25
0,07
0,61
Sound Type
0,19
0,25
0,37
0,19
Input Type
0,29
0,12
0,47
0,12
Availability of Face-to-Face Comm
0,41
0,09
0,09
0,41
Technical Support
0,66
0,15
0,15
0,05
User Training
0,21
0,46
0,28
0,05
Hospital
0,36
0,14
0,14
0,36
Customizability
0,36
0,22
0,06
0,36
Clarity of Menu Items
0,23
0,16
0,23
0,38
Output Quality
0,12
0,38
0,12
0,38
receiving guidance conducted by professionals instead of online education and manual. Table 10 shows the total decision weights of the alternatives. Total decision weight of a specific alternative was calculated by multiplying weight of the attribute with overall priority vector value of the attribute for that alternative and summed up these values calculated for 16 attributes. According to the table, people have a tendency on selecting alternative 4 which is followed by alternative 1. Alternatives 2 and 3 are the least preferred services by the potential users. Total Decision Weight = Σ Global Weight of the Attribute* Priority Vector of the Attribute Alternative 4 has the finest attribute values for the most important three attributes listed in Table 8. Medical provider of the alternative is a private hospital. Although its implementation cost is higher than alternative 1 and equals to alternative 2, it has cheapest operating cost that has valuable
8
impact on potential users’ health service selection decisions. Moreover, it provides patients to face-to-face communication with the physician.
conclusIon As explained in section 1, electronic health information service selection is an important and difficult problem to a medical company and patients. We first identified 16 criteria and then formulated an AHP-based model to select health information service as shown in Appendix B. As for HIS selection, all potential users agreed that a service characteristic was the most important factor. Service providers and cost followed as the second and third most important consideration. The availability of face-to-face communication of the HIS was ranked first among service factors, followed by content, time, triability and mobile device. Moreover, designers should pay more attention on medical provider and cost of the service while developing a telemedicine service. Instead of public hospital, private hospital should be selected as a medical provider, because people take into confidence in private hospitals. Also, patients prefer the services that enable doctor visits. The most important attributes, found in the research, enable the service designers and providers to know the characteristics of the most preferred e-health service. Furthermore, governments, especially in developing countries, face with some difficulties while providing quality healthcare in underserved and rural areas. Even if they achieve to serve, they suffer from high cost of the health service. With the rapid development of information and communication technologies, they benefit from internet to serve the healthcare service. But, at this time user adoption problems becomes one of the obstacles in spreading of the electronic health services. The proposed AHP framework will help the authorities to overcome adoption problems so as to reduce health care
Evaluating Health Information Services
Table 10. Final Decision Matrix Total Decision Weights
costs, to improve health care and access to health care, to improve patience obedience. For a future study, a comprehensive qualitative research with large sample size can be conducted to test validity of the results. Moreover, attribute values of proposed alternatives can be changed and examined the affect of them on the health service selection process.
reFerences Al-Qirim, N. (2007). Championing telemedicine adoption and utilization healthcare organizations in New Zealand. International Journal of Medical Informatics, 76, 42–54. doi:10.1016/j. ijmedinf.2006.02.001 ATA. (2009). Americal Telemedicine Association. Retrieved from http://www.atmeda.org Berghout, B. M., Eminovic, N., de Keizer, N. F., & Birnie, E. (2007). Evaluation of general practitioner’s time investment during a store-and-forward teledermatology consultation. International Journal of Medical Informatics, 76, 384–391. doi:10.1016/j.ijmedinf.2007.04.004 Biermann, E., Dietrich, W., Rihl, J., & Standl, E. (2002). Are there time and cost savings by using telemanagement for patients on intensified insulin therapy: A randomised, controlled trial. Computer Methods and Programs in Biomedicine, 69, 137–146. doi:10.1016/S0169-2607(02)00037-8 Bobrie, G., Postel-Vinay, N., Delonca, J., & Corvol, P. (2007). Self-Measurement and SelfTitration in Hypertension. American Journal of Hypertension, 20, 1314–1320. doi:10.1016/j. amjhyper.2007.08.011
A1
A2
A3
A4
0,30
0,17
0,18
0,35
Chae, Y. M., Lee, J. H., Ho, S. H., Kim, H. J., Jun, K. H., & Won, J. U. (2001). Patient satisfaction with telemedicine in home health services for the elderly. International Journal of Medical Informatics, 61, 167–173. doi:10.1016/S13865056(01)00139-3 Chang, I. C., Hwang, H. G., Hung, M. C., Lin, M. H., & Yen, D. C. (2007). Factors affecting the adoption of electronic signature: Executives’ perspective of hospital information department. Decision Support Systems, 44(1), 350–359. doi:10.1016/j.dss.2007.04.006 Chen, I. J., Yang, K. F., & Tang, F. I. (2008). Applying the technology acceptance model to explore public health nurses’intentions towards web-based learning: A cross-sectional questionnaire survey. International Journal of Nursing Studies, 45, 869–878. doi:10.1016/j.ijnurstu.2006.11.011 Gagnona, M. P., Godinb, G., Gagnéb, C., Fortina, J. P., Lamothec, L., Reinharza, D., & Cloutierd, A. (2003). An adaptation of the theory of interpersonal behaviour to the study of telemedicine adoption by physicians. International Journal of Medical Informatics, 71, 103–115. doi:10.1016/ S1386-5056(03)00094-7 Kahen, G., & Sayers, B. M. (1997). Health-care technology transfer: Expert and information systems for developing countries. Methods of Information in Medicine, 36, 69–78. Karahanna, E., Straub, D. W., & Chervany, N. L. (1999). Information technology adoption across time: a cross-sectional comparison of preadoption and post-adoption beliefs. Management Information Systems Quarterly, 23, 183–213. doi:10.2307/249751 9
Evaluating Health Information Services
Kaufman, D. R., Patel, V. L., Hilliman, C., Morin, P. C., Pevzner, J., & Weinstock, R. S. (2003). Usability in the real world: assessing medical information technologies in patients’ homes. Journal of Biomedical Informatics, 36, 45–60. doi:10.1016/S1532-0464(03)00056-X
Simon, J. S., Rundall, T. G., & Shortell, S. M. (2007). Adoption of Order Entry with Decision Support for Chronic Care by Physician Organizations. Journal of the American Medical Informatics Association, 14, 432–439. doi:10.1197/jamia. M2271
Kwak, N. K., McCarthy, K. J., & Parker, G. E. (1997). A human resource planning model for hospital/medical technologists: An analytic hierarchy process approach. Journal of Medical Systems, 21(3), 173–187. doi:10.1023/A:1022812322966
Tam, M. C. Y., & Tummala, V. M. R. (2001). An Application of the AHP in vendor selection of a telecommunications system. Omega, 29(2), 171–182. doi:10.1016/S0305-0483(00)00039-6
Magrabi, F., Lovell, N. H., & Celle, B. G. (1999). A web-based approach for electrocardiogram monitoring in the home. International Journal of Medical Informatics, 54, 1145–1153. doi:10.1016/ S1386-5056(98)00177-4 Martin, J. L., Norris, B. J., Murphy, E., & Crowe, J. A. (2008). Medical device development: The challenge for ergonomics. Applied Ergonomics, 39, 271–283. doi:10.1016/j.apergo.2007.10.002 Nydick, R. L., & Hill, R. P. (1992). Using the analytic hierarchy process to structure the supplier selection procedure. Journal of Purchasing and Materials Management, 25(2), 31–36. Rialle, V., Lamy, J. B., Noury, N., & Bajolle, L. (2003). Telemonitoring of patients at home: a software agent approach. Computer Methods and Programs in Biomedicine, 72, 257–268. doi:10.1016/S0169-2607(02)00161-X Saaty, T. L. (1977). A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology, 15, 234–281. doi:10.1016/00222496(77)90033-5 Saaty, T. L. (1996). The analytic hierarchy process. Pittsburgh: RWS Publications.
10
Topacan, U., Basoglu, A. N., & Daim, T. U. “Exploring the Success Factors of Health Information Service Adoption. In Proceedings of Portland International Conference for Management of Engineering and Technology’08 (pp. 2453-2461). Tung, F. C., Chang, S. C., & Chou, C. M. (2008). An extension of trust and TAM model with IDT in the adoption of the electronic logistics information system in HIS in the medical industry. International Journal of Medical Informatics, 77, 324–335. doi:10.1016/j.ijmedinf.2007.06.006 Turri, J. J. (1988). Program eases decision making. Health Progress (Saint Louis, Mo.), 69, 40–44. Vaidya, O. S., & Kumar, S. (2006). Analytic hierarchy process: An overview of applications. European Journal of Operational Research, 169, 1–29. doi:10.1016/j.ejor.2004.04.028 van den Brink, J. L., Moorman, P. W., de Boer, M. F., Pruyn, J. F. A., Verwoerd, C. D. A., & van Bemmel, J. H. (2005). Involving the patient: A prospective study on use, appreciation and effectiveness of an information system in head and neck cancer care. International Journal of Medical Informatics, 74, 839–849. doi:10.1016/j. ijmedinf.2005.03.021
Evaluating Health Information Services
Venkatesh, V., & David, F. D. (2000). A theoretical extension of the technology acceptance model- Four longitudinal field studies. Management Science, 46, 186–204. doi:10.1287/ mnsc.46.2.186.11926 Yang, B. H., & Rhee, S. (2000). Development of the ring sensor for healthcare automation. Robotics and Autonomous Systems, 30, 273–281. doi:10.1016/S0921-8890(99)00092-5
endnote 1
This chapter is based on a paper presented at Portland International Conference on Management of Engineering and Technology 2009 in Portland Oregon USA by the same authors
11
Evaluating Health Information Services
appendIx a Table 11. Proposed Health Service Alternatives Attribute
Alternative 1
Alternative 2
Alternative 3
Alternative 4
Implementation Cost
200 YTL
500 YTL
800 YTL
500 YTL
Operating Cost
100 YTL / month
50 YTL / month
50 YTL / month
25 YTL / month
Usage Time
10 min / day
50 min / day
30 min / day
30 min / day
Response Time
45 min
15 min
5 min
15 min
Language
Turkish
English
Turkish / English
Turkish / English
Security
Available
Not Available
Available
Not Available
Triability
Not Available
1 week limited functionality
Not Available
1 week unlimited functionality
Sound Type
Computerized
Male
Female
Computerized
Input Type
Sound
Text
Selection among alternatives
Text
Availability of FacetoFace Communcation
Available
Not Available
Not Available
Available
Technical Support
7 / 24
09:00 – 17:00
09:00 – 17:00
Not Available
User Training
Operating manual is given to the user
Training is conducted by a professional
Online Education
Not Available
Medical Provider
Private Hospital
Public Hospital
Public Hospital
Private Hospital
Customizability
Frequently used menu items are automatically shown on top of the menu
Users are selecting frequently used menu items manually
Menu items are shown in the same order
Frequently used menu items are automatically shown on top of the menu
Clarity of Menu Items
Service does not contain menu
Textual menu items
Graphical menu items
Both of the graphics and text are used in the menu items
Output Quality
Meal list is given as a output
User preference are taken into consideration in the meal list
Meal list is given as a output
User preference are taken into consideration in the meal list
12
Evaluating Health Information Services
appendIx B Figure 2. AHP Model
13
14
Chapter 2
Gastrointestinal Motility Online Educational Endeavor Shiu-chung Au State University of New York Upstate Medical University, USA Amar Gupta University of Arizona, USA
aBstract Medical information has been traditionally maintained in books, journals, and specialty periodicals. A growing subset of patients and caregivers are now turning to diverse sources on the internet to retrieve healthcare related information. The next area of growth will be sites that serve specialty fields of medicine, characterized by high quality of data culled from scholarly publications and operated by eminent domain specialists. One such site being developed for the field of Gastrointestinal Motility provides authoritative and current information to a diverse user base that includes patients and student doctors. Gastrointestinal Motility Online leverages the strengths of online textbooks, which have a high degree of organization, in conjunction with the strengths of online journal collections, which are more comprehensive and focused. Gastrointestinal Motility Online also utilizes existing Web technologies such as Wiki-editing and Amazon-style commenting, to automatically assemble information from heterogeneous data sources.
IntroductIon For the last several decades, Harrison’s Principles of Internal Medicine, published by McGraw Hill, has served as a major source of information in the field of Gastrointestinal Motility. This book and its online presentation have been, and continue to be, used by many medical colleges to train the
next generation of medical doctors; practitioners in this field also frequently refer to them. Traditionally, papers and articles in specialty medical journals supplemented the material in textbooks like Harrison. The latter book would itself be updated periodically to reflect the state of the art in medicine and the various specialties, providing a consensus opinion of the standard of care.
Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Gastrointestinal Motility Online Educational Endeavor
The advent of computers and Internet has given rise to online sources of information such as UpToDate (http://www.uptodate.com/) and WebMD (http://www.webmd.com/). While gaining tremendous following and being updated frequently, these sources of online information relate to the medical field as a whole and not to particular specialties. Furthermore, the information on these sites is generally maintained by personnel of the respective organizations, not by specialists in specific disciplines of medical science. These organizations are usually set up as commercial entities, rather than not-for-profit ones. The progressive transformation of information has seen many journals that were previously in paper format opting to use new electronic technologies; most of them now come out both in paper and electronic formats. Searchable electronic archives, such as PubMed (http://www. pubmedcentral.nih.gov/), now place a plethora of information into the hands of researchers and physicians. However, such searches are very time consuming and often produce irrelevant or poorly supported articles. Sites like Harrison’s Online (http://www.accessmedicine.com/) serve as information directories that can be searched, hoping to place the most suitable information on a medical topic in a user’s hand. Students have gradually come to expect information in quick and readily available forms without having to bother about inter-library loans or even hardcopy versions at all. The goal of the endeavor described in this article was to adapt emerging technologies to improve methods of teaching gastrointestinal material to students and to serve as a more effective source of relevant and accurate information for medical practitioners and specialists.
evidence-Based Medicine A study from the School of Information Management and Systems at UC Berkeley estimates that, in 2003, the World Wide Web contained
about 170 terabytes of information on its surface alone, equivalent to seventeen times the size of the information in the Library of Congress (Lyman & Varian, 2003). With this increasingly information-rich society, the most precious ability for students and learners is no longer to find the information, but to discern the most relevant pieces of information and to integrate them into practice. The American Library Association describes “information literacy” as the ability of individuals to “recognize when information is needed and have the ability to locate, evaluate, and use effectively the needed information” (American Library Association, 1989). The medical domain version of information literacy is evidence-based medicine. Evidence-based medicine (EBM) is the integration of best research evidence with clinical expertise and patient values (Guyatt et al., 1992). The Centre For Evidence-Based Medicine in Toronto, Canada, states that the origins of evidencebased medicine date back to post-revolution Paris (CEBM, 2007), but that the current growth is most closely attributed to the work of a group lead by Gordon Guyatt at McMaster University in Canada in 1992. EBM publications, reflecting interest in this field, have grown from a lone publication in 1992 to thousands in 2007. Studies have become increasingly critical of the value of textbook sources (Antman et al., 1992). Didactic continuing medical information may be ineffective at changing physician performance (Davis et al., 1997), and clinical journals may lack practical application (Haynes, 1993). In addition, physicians are faced with an increasing burden on their time, forced to diagnose patient findings within a matter of minutes (Sackett & Straus, 1998), and can only afford to set aside half an hour or less per week for general medical reading (Sackett, 1997). The staggering mass of information being discovered is also daunting: 500,000 articles are added to the commonly used Medline medical journal database every year, and “if a physician read 2 articles each day, every day
15
Gastrointestinal Motility Online Educational Endeavor
for a year, (s)he would still find herself or himself 648 years behind” (Lindberg, 2003). As research increases the quantity of information available, medical practitioners are compelled to find efficient methods to educate themselves. The Centre for Evidence Based Medicine has cited several examples of strategic, educational, and technical improvements in medicine that have enabled the current explosion in interest in this field. These include the emergence of new strategies for evaluating information; the creation of systematic reviews; the growing emphasis on continuing medical education and lifetime learning; and the advent of online journals, meta analysis of multiple studies, and ready access to such resources through electronic archives. Rapid dissemination of accurate and comprehensive compilations of research results enables medical practitioners to make informed decisions that are supported by the latest research results, and not by outdated trials. In light of the increasing number of medical journals, especially journals that focus on specialties, the sheer quantity of information threatens to overwhelm medical practitioners. The concept of information mastery has been coined to describe the set of skills that physicians must nurture in approaching, analyzing and incorporating or rejecting new information. The issues mentioned above are not limited to the medical arena alone: Former Vice-President Al Gore described the state of information management as “resembling the worst aspects of our agricultural policy, which left grain rotting in thousands of storage files while people were starving” (Gore, 1992). The Center for Information Mastery at the University of Virginia asserts that the usefulness of medical information is dependent upon
its relevance, validity, and the work required to obtain it, as specified in Equation 1 (Slawson et al., 2007). Further, the increasing quantity of research being performed by commercial enterprises, as well as other organizations with potential conflicts of interest, requires information be filtered for validity before incorporation into medical the canon. Finally, increased effort involved in accessing the relevant pieces of information reduces the accessibility of such information. In addition, healthcare organizations are siloed, gaining the advantage of sub-optimizing local departments, possibly at the cost of the whole (Senge, 1980); this complicates the problem further.
Information retrieval and decision-Making As Stephen Hawking observes in The Universe in a Nutshell, the rate of growth of new knowledge is exponential. While 9,000 articles were published annually in 1900 and around 90,000 in 1950, there were 900,000 scientific articles published per annum in 2000 (Hawking, 2001). The explosive growth of information is challenging both the information repositories designed to hold it, and the ability of users to access relevant information. At the time of this writing, Wikipedia serves as the de-facto standard for online general encyclopedias, and is among the top ten most-visited Web sites (Alexa Internet, 2007a). Its open-source, volunteer-without-accountability approach led to initial concerns about information validity, but these concerns have been largely addressed. Nature magazine studied Wikipedia and Encyclopedia Brittanica and found that the two of them
Equation 1. Usefulness (Slawson et al., 2007)
Usefulness of Medical Information =
16
(Relevance)(Validity) Work
Gastrointestinal Motility Online Educational Endeavor
were largely similar in accuracy (Giles, 2005). The growth of Wikipedia’s information base further enhances the quality and breadth of coverage, and supports the possible future use of Wikipedia or Wiki-style architecture as an academically respectable source of reference. In contrast to the indexed and contributed semi-structured format of Wikipedia, Google relies on search-keyword phrases. The usefulness of Internet-crawling indexers, like Google, is based upon the ability to retrieve and capture information from many sites, and to retrieve relevant pages on query. Google’s initial strength and rise to stardom was achieved through its superior PageRank algorithm, which still remains a carefully guarded trade secret; this algorithm provides an uncannily relevant list of matches to any user query, ranging from commonplace query phrases to obscure esoteric trivia and even misspellings. In a small supermarket today, shoppers are bombarded with a selection of 285 varieties of cookies and 95 varieties of chips, leading the consumer to a state of decision overload (Schwartz, 2004). There is a growing need to restructure data to meet the informational and management requirements of an organization or group of people (Carlson, 2003).
Medical Information repositories For the medical arena, PubMed is the most widely used information database in the world, accounting for 1.3 million daily queries by 220,000 unique users (Lindberg, 2003). It is a free access search engine, provided by the U.S. National Library of Medicine as the main access point to the Medline database, a cataloged repository of medical
literature classified using the descriptors known as Medical Subject Headings (MeSH). A broad range of search features are offered, including combined searches, exclusions, classification by type of article (original research versus review) and related articles. Another feature of the Medline database, known as MedlinePLUS, provides generalized information on health topics, and is aimed at the public or at practitioners outside their specialty domain. At the current time, PubMed serves as the gold standard in comprehensive medical information, despite its dated interface. HighWire Press, a Stanford-originated endeavor, distributes thousands of journals, and provides its own search engine. In a recent study, the relevance of articles retrieved from HighWire was found to be greater than that of PubMed, but with the disadvantage of a slower retrieval Speed (Vanhecke et al., 2006). In order to make a comparative evaluation between different approaches, it is appropriate to characterize the information recall ability using three parameters: precision, recall, and fall-out. Precision (Equation 2) can be defined as the proportion of all retrieved documents that are relevant. Recall (Equation 3) captures the concept of complete retrieval of all relevant documents. Fall-out (Equation 4) is a measure of the number of documents that are retrieved but are unrelated to the issue being searched. Medical research, while generally emphasizing maximal precision and minimal fall-out, occasionally requires increased recall, in the case of obscure diseases, or unusual side effects of medications. Medline serves as a canonical list for such purposes, but at the cost of significantly lower precision.
Equation 2. Precision (Wikipedia, 2007) precision =
{relevant documents}∩ {retrieved documents} {retrieved documents}
17
Gastrointestinal Motility Online Educational Endeavor
Equation 3. Recall (Wikipedia, 2007)
recall =
{relevant documents}∩ {retrieved documents} {relevant documents}
Equation 4. Fall-out (Wikipedia, 2007) fall-out =
{non-relevant documents}∩ {retrieved documents} {non-relevant documents}
Medical practitioners using Google for their searches will often find themselves frustrated at the large quantity of articles on obscure and irrelevant topics. A researcher searching for a pharmacologic treatment of a syndrome will turn up with thousands of articles dealing with various sub-types, biochemical-signaling processes involved, and even support groups, before finding a therapeutic treatment. Due to the nature of the search engine and the storage methods, there are concerns about Google’s or any search engine’s ability to maintain a collection of such information. Carlson (2003) showed that due to the relatively small collection of documents indexed by an average search engine, a significant amount of relevant information would not be returned even in the presence of a perfectly formulated search phrase. In order to accommodate domain-specific areas, Google has introduced the concept of “Refine Your Search” (http://www.google.com/coop). Without altering its main core search methodology, Google allows users to more quickly locate the type of information desired (i.e., treatment or symptoms). These refinement tools, provided by vendors and other private individuals or agencies that are deemed authoritative, subsequently label Web sites with appropriate descriptor tags. The potential conflict of interest created by these corporate associations is a matter of concern, due to omissions or maliciousness of the labeling.
18
The range of challenges and issues that characterize the medical domain include: •
•
•
The presence of an extensive array of synonyms for various drugs and diseases that require semantic knowledge to be encoded into the search engine in order to link concepts that are not lexically related; The naming of disease subtypes (often after a major contributor or discoverer), requires that hierarchies be constructed to allow users looking for the subclasses of disease to find information on the main umbrella disease, and vice versa; The growth in the understanding of generalized syndromes results in a corresponding need for reclassification based on new etiologies of disease, thereby suggesting a dynamic organizational structure for the online medical information systems.
In view of the growing difficulty in locating desired pieces of information, individuals performing research are in increasing danger of information overload. As such, the next generation of medical information access tools must aim to improve the ability to retrieve the right chunks of information quickly, with zero or minimal extraneous information; this concept is termed as increasing the signal to noise ratio in the field of electrical engineering.
Gastrointestinal Motility Online Educational Endeavor
VIsIon and goals For gastroIntestInal MotIlIty onlIne Gastrointestinal (GI) Motility Online is an example of a medical information system that seeks to provide access to high quality medical information online related to a particular medical specialty by centralizing the information and presenting relevant information that is customized to the user’s information requirements. The field of Gastrointestinal Motility is complex and interdisciplinary, involving a variety of experts. The possible user base includes laypersons, patients, medical students, biomedical scientists, physiologists, pathologists, pharmacologists, biomedical students, researchers, pharmaceutical staff, house staff, specialty fellows, internists, surgeons, and gastroenterologists. Each role requires a different approach to depth, scholastic relevance, and clinical direction in terms of information presentation. For example, students are interested in innovative research or review papers; researchers would like to know the most recent developments, and practitioners might be more interested in using the information for differential diagnosis purposes. GI Motility aims to serve as a collaboration of medical professionals, approaching diseases and patients from different angles. In a library, a user interacts with the data in books very differently from the way that she or he interacts with data in an online presentation. The user expects the book to be focused and to address the topic in a linear fashion. Online, the same user navigates quickly, using hyperlinks, to explore secondary topics. In fact, the user expects a different presentation and a different style of information; as a result, the nature of the interactions will differ even with the same content. One of the aims of Gastrointestinal Motility Online is to address these different styles and to present the desired subset of information in the manner that the user might expect. For example, a user
might be interested in viewing articles from the perspective of case-based, symptom-based, or test result-based diagnosis in order to apply the information to a particular problem at hand. In essence, the vision of Gastrointestinal Motility Online is to present information as framed by the interaction with the particular user at the particular point in time. The first step in the vision of Gastrointestinal Motility Online was to collect the information in a manner consistent with the goal to acquire the reputation for the highest quality of knowledge. The information base is assembled entirely from material provided by internationally acknowledged experts. All chapters including synopses, articles, and reviews are written by reputed authorities. The pool of information is envisaged to be shared between different types of users and for different purposes. The design of the system emphasizes a one-stop information approach that enables the users to derive information at various depths. This applies to onsite information, as well as to information at offsite locations.
details of effort A two-phase approach is being utilized for the creation of the information system: the first involves full leveraging of commercial technology as it exists today, and the second involves further research on aspects that can be incorporated in future versions of our system. In the absence of a better term, the term gastrointestinal knowledge repository is used for the final system, as well as for the initial concept-demonstration prototype system. For the first phase, the acquisition of knowledge proceeded with the establishment of titles and themes for chapters, as determined through discussions involving the concerned authors and the editors for this project (Dr. Raj Goyal of Harvard Medical School and Dr. Reza Shaker of the Medical College of Wisconsin). The creation of the gastrointestinal motility knowledge repository
19
Gastrointestinal Motility Online Educational Endeavor
began with calls to key gastrointestinal experts inviting them to submit a chapter, in electronic form, for inclusion in this knowledge repository. The inputs from the contributing authors were reviewed by these two editors and by others, on an anonymous peer review basis. Under the aegis of an unrestricted grant from Novartis Corporation, the two editors worked closely with the staff of Nature Publishing Group on a number of tasks that ultimately led to the creation of the following Web site: http://gimotilityonline.com. The creation of the site involved an automated conversion process to adapt MS-Word and rich text documents into a Web presentable format, with special emphasis placed on images, tables, and video. The majority of the investment for development lay in formatting and typesetting; the design itself was of less concern as it followed existing Web branding and style guidelines of Nature Publishing Group. Rights for images needed to be obtained; further, images, tables and video needed to be edited to fit a standard look and feel. After receiving author contributions, the project required approximately 18-24 months to complete, with 9-12 months required for the editorial process itself. The current site consists of 1,000 HTML pages, 1,000 images, 500 Powerpoint presentations and 40 videos. Articles are cross-linked by topics, and the current volume is equivalent of about 700 pages of text. This volume is increasing as this endeavor continues to progress.
site purpose The ostensible purpose of the site is to disseminate information on the specialty of gastrointestinal medicine, primarily to gastrointestinal medical practitioners and to other interested parties. The ambition for the site is for it to become the central hub of information on this specialty, aggregating information from many sources, authors, and journals into one central location with the objective of becoming a de facto standard for online gastrointestinal information.
20
A secondary role is the development of a community of gastrointestinal specialists and other interested parties, who can form an online collaborative, expand upon the information repository, and facilitate peer communications and discussions. In addition, the site aims to explore and to expand the concept of online information repositories, especially the optimal integration of the current generation of electronic journals and online textbooks.
site audience The primary site audience is the set of gastrointestinal specialists, associated medical staff, and staff under training. Members of this primary user base would initially visit the site because of a recommendation from a colleague or because another site (search or advertisement) directed the user to it. Occasional patients are expected, but are unlikely to become the primary users of this site. Initially, the core attractions for users to this site are the quality and depth of information coupled with the ease of access and the low cost; these same factors will also help to retain the user base. Most physicians and healthcare specialists spend relatively little time online (six hours per week on average); in order to impress the user base, the signal to noise ratio of the site must be high, along with the ease of locating a desired piece of information (through the information architecture) (Friedman, 2000). The primary user of the site is unlikely to demand high interactivity; instead, the interest will be on locating and extracting the information as quickly as possible. As such, the site must be efficiently organized, streamlined, and equipped with powerful search and index functions. After drawing a user initially, repeat visitors would use the site to browse new topics of interest, as well as to entrust the editorial staff to select articles that represent innovative research in the relevant field. An interacting community base would evolve over time and eventually cause a
Gastrointestinal Motility Online Educational Endeavor
change in the workflow. Specialists would then visit the site to explore the comments from their colleagues on the topics expressed and to leave their own authoritative inputs on different articles, eventually contributing entire articles as much as a good electronic journal aspires to do today. While relatively low in terms of being technically savvy, a typical user of Gastrointestinal Motility is likely to be very comfortable with using the Internet for retrieval of scholarly information through PubMed, online textbooks, and ready references such as UpToDate. A sleek, uncluttered design is likely to be the most attractive, even though it may not support applications involving high load times, such as Flash or embedded videos. Users are comfortable reading large articles online, but they also expect printable versions to be available, on an as-needed basis. The user base, while highly intelligent, is relatively small in numeric terms, and is characterized by a small presence on the Web. The process of attracting a critical mass to build and to maintain a user community is a high priority task; this involves contacting a high percentage of all members of this specialist community.
site content Gastrointestinal Motility Online is a site that provides information on both medical practice and the fundamentals of the gastrointestinal tract. The information must provide both breadth and depth on the topic, and should ultimately serve as an encyclopedia of the domain-specific knowledge. Information Architecture: As in a library, information must be properly accessible in order to have value. The architecture of the information is partly determined by the methods that the users will use to query the evolving knowledge base. In a library, the name of the author and the title of the book are important. In a journal, the age of the article or the issue in which it was published may be essential. In Gastrointestinal Motility Online, the most likely user scenarios involve searching for
information by anatomical section or syndrome. The timing is also important; recent material is favored over older material. Since relevance of the article is also important, the name of the journal and the name of the author are also used prominently in searches. Furthermore, classification schemes or hierarchies to group-related topics are essential to narrow the branching process used by the search technique. The determination of related information is a complex topic. In order to address the latter issue, medical organizations are creating medical ontologies, such as unified medical language system (UMLS), to quantify and to impose structure on conceptual relationships. With this classification, the selection criteria can be hierarchically built up or finely specified to obtain desired results. A lengthier discussion of this topic appears later in this article. Look and Feel: The look and feel of the site needs to convey the sense of scholarly authority, but with a sleek technological approach. Medical personnel have high standards concerning the accuracy of articles, and a professional presentation aids in supporting this perception. A site that is gaudy or shows too many bells and whistles, or involves a long load time, reflects poorly on the content, as do garish colors or lack of color. Images and video must be consistent and should be available to download as needed for reference and for closer examination. The level of interactivity available should be low, as most gastrointestinal specialists have small online presence. Natural language queries, such as those used with AskJeeves (http://www.ask.com/), are unnecessary, as long as the search and browse features are precise and efficient. Pages may be presented as either textbook or journal articles, based on the preference of the user. Extracting Content, Metadata, and Cross Referencing: One of the most powerful functions of the Internet is the ability for Web users to span several related articles quickly because of cross-referenced links. A user interested in
21
Gastrointestinal Motility Online Educational Endeavor
the preparation of a particular chicken recipe can quickly reference how to sauté and with what form of pan, moving quickly to sources to buy the appropriate cast-iron skillet or wok to benefits and comparisons of different brands and retailers. A Wiki of only gastrointestinal specialists, with limited control from an editor, would be appropriate for the collection and dissemination of information: Gastrointestinal Motility Online is striving towards that goal. A particularly useful feature in Gastrointestinal Motility Online, not available in online journals, is its cross-referencing tool. Articles that are closely related or broach a topic in greater depth can be quickly accessed by a cross reference within the Gastrointestinal Motility Online domain. These links are established by content extraction tools that create metadata and relate that data between different documents. Footnotes are available at the bottom and allow for a broader topical search; however, the inline linking of articles is particularly appealing for tracking particular items or syndromes of interest.
auxiliary technologies Search features are incorporated in Gastrointestinal Motility Online, but are considered secondary to the organization of the information. A dynamic keyword search for anatomical sections of the gastrointestinal tract is less likely to reveal useful information on general function than a manual perusal of the literature through the prepared subject browse option. The efficient organization of the information, based on the anticipated needs and access patterns of users, is an essential feature for building a knowledge repository. By acknowledging that specialists would be more likely to query by anatomy or syndrome, one needs mechanisms for structured order, as compared to mechanisms that order by article size, author name, or recent usage. An additional feature, incorporated within the community-building module, is a user rating system that allows users
22
to rank the importance of articles so that users browsing information can be directed to the most useful and informative articles. One difficulty with searches is the likelihood that a particular phrase will appear in nearly all documents unless the search is very specific. In cases where the gastrointestinal tract is analyzed as an interactive system, the phrases for anatomical locations may occur multiple times as reference points, but the central theme of the article may not be easy to determine by word frequency. As a result, a number of semantic tools, described later in this article, are used to analyze articles and to classify them appropriately. Experts of the Nature Publishing Group (NPG), who possess prior experience in online information presentation, based on the online version of Nature magazine, helped to develop the initial vision of the online knowledge repository. The base site is hosted by Nature Publishing Group. While the site was being developed, commercial tools were available to handle both the production of electronic journals and static textbook efforts like AccessMedicine/Harrison’s Online (http://www. accessmedicine.com/) and WebMD. However, the gastrointestinal knowledge repository falls somewhere in between these two cases; accordingly, few over-the-shelf tools and algorithms were available for immediate use. As such, a significant fraction of the interface and architecture had to be innovated and refined, through experimentation. Many of the existing tools for creating electronic journals are geared towards collation of articles, graphics, and layout work. These tools reduce the time needed by the authors and editors for the processes of uploading, formatting, and editing. In the development of Gastrointestinal Motility Online, the use of a software suite facilitated handling of images and consistency of look and feel. One area where tools are lacking is the ability to organize information into a coherent topical fashion, as in a textbook. Searching by keyword is especially difficult on a physician specialty site, where the dialect is limited and
Gastrointestinal Motility Online Educational Endeavor
the concepts are reused multiple times. As such, editorial staff must impose additional control to prevent the site from becoming a write-only knowledge repository. Collections of electronic journals, such as Ovid (http://www.ovid.com/site/index.jsp), have been primarily targeted for libraries and research centers. The primary purpose of Ovid is to serve as a warehouse of information, albeit uncategorized. Gastrointestinal Motility Online’s current state differs from that of Ovid in terms of the presentation of the material: the former system is specifically formatted to provide an online view as well as a hardcopy output. The long-term goal of Gastrointestinal Motility Online is to collect a comprehensive knowledge on subjects (as Ovid does), as well as to add more intelligent search tools or information utilities. While Ovid does not organize information except into broad categories, Gastrointestinal Motility Online refines classifications, provides responses to search queries that are more accurate, and supports tools that use the knowledge in compelling ways, such as for differential diagnosis. The advantages of sites built in a textbook style, such as AccessMedicine, is the hierarchical organization and ready access to information. The evolution of online textbooks has generated significant activity on the sites, as teaching tools. Gastrointestinal Motility Online uses over 40 illustrative videos, which are not typical of a journal, but fit the online textbook paradigm. Online textbooks are excellent repositories of information, except that updating the sites to incorporate new information is generally cumbersome because of the level of interactivity involved. “Gastrointestinal Motility Online” is a hybrid, adopting the best qualities of online textbooks, such as Harrison’s Online, and journal collections, such as PubMed. Organized, Well Edited, Frequently Updated and Comprehensive Information are self-explanatory. Search features specifically refer to the ability to search for a keyword or phrase. Differential diagnosis refers to the abil-
ity to integrate clinical patient presentations and create a list of possible problem diagnoses. At a broader technological level, eBooks represent a technology that has been adapted to deliver information electronically. Medical eBooks can be argued to be a natural outgrowth of the eBook movement to electronic media: volume and space requirements are reduced, key phrase searching can be performed, and portability is enhanced. Nevertheless, few medical texts are adapted as eBooks. Although eBooks have grown in popularity, they have not grown as rapidly as projected by consultants; this could be because of the following reasons: 1. 2.
3.
4.
Most readers see no need to replace print books. Due to the limited screen size, limited battery life, and navigation interface issues with eBooks, many people still find paper books easier to handle. Digital rights management causes compatibility and portability problems when attempting to move the eBook from desktop to PDA or laptop. Current pricing of eBooks does not account for the reduced value relative to paper books. When readers finish reading a paper book, they can give it to a friend or sell it to a used bookstore; neither is possible with most eBooks. (Crawford, 2006)
Additionally, delivery of information may not be simply online, but online and to a mobile user using a PDA or other portable device. The constraints involved in transmitting and displaying information on a limited display panel create a new set of challenges. In most markets other than healthcare, the primary applications for PDAs are for scheduling and contact management (as a busy executive might use in lieu of a pen-and-paper daily planner), or as a portable browser or e-mail client (as in the case of technologists and engi-
23
Gastrointestinal Motility Online Educational Endeavor
neers). In such cases, the application of the PDA works within the bounds of the limited display and the modern constraint of minimal bandwidth, often serving as a surrogate cell phone of sorts. However, in medicine, the PDA is often stretched beyond its limits. The current trend is the delivery of detailed information pages into a portable format, downloadable to PDA. Since medical practitioners can no longer maintain a complete mental catalog of all drugs and particularly obscure clinical symptoms or diseases, PDAs assist physicians in their duties without requiring a quick trip to a computer terminal or a large paper binder archive. Harrisons and UpToDate have both moved rapidly in this area, and the list of available drug databases is already large. PDAs fill the role of drug lookup very well, as well as serve as a primer for obscure diseases. The difficulty lies with more graphically intense data that may not display properly, or may need to be downloaded on the fly. In such cases, medical information systems are pushing the technological limits of PDAs. PDA sales as a whole, however, are in decline, except as a niche application. Analysts at organizations such as IDC and Gartner have predicted downtrend trend in sales of PDA. Dell has withdrawn its PDA line from production (Mechaca, 2007). In the long term, the PDA may carefully constrain its niche to feature more of the portable planner features and less multimedia and display power, rendering it less useful to medical practitioners. Until the advent of revolutionary new
display technologies, such as holographic displays or direct-to-eye projection technologies, the ability for PDAs to contribute to medical reference appears to be technologically limited. As shown in Table 1, Gastrointestinal Motility Online is a hybrid that lies between the two models of electronic journals and online textbooks, exclusively focused on providing authoritative information in an organized fashion within a specific domain. Using this system, a gastrointestinal researcher can find the most recent journal articles because of the frequent updates, and a gastrointestinal clinician can easily locate a detailed diagram of the lower esophageal sphincter. In addition, topical information can be cross-linked between papers—a typical feature of textbooks and very pertinent for teaching and presentations.
concepts and InnoVatIons The exploration of the technological space between electronic journals and online textbooks is a relatively new idea. All new ideas face challenges in terms of deployment and adoption. Consider the fax number. As the number of fax machines increases, the value of the fax increases—this illustrates the fact that networks attain greater value with larger number of users. The difficulty faced by Gastrointestinal Motility Online and other specialty interest sites is in terms of initial growth and development of specific communities of interest. These sites must be aesthetically
Table 1. The online endeavor Online Textbook
Hybrid – GI Motility Online
X
X
Well Edited
X
X
Frequently Updated
X
X
Comprehensive Information
X
X
X
X
Search Features Differential Diagnosis
24
Journal Collections
Organized
X
X
Gastrointestinal Motility Online Educational Endeavor
attractive, informative, efficient, and up-to-date. The case of Gastrointestinal Motility Online illustrates one form of evolution of online journals and textbooks into an active online scientific community. Site loyalty is achieved and maintained by the reputation of the authors and contributors. Gastrointestinal Motility Online needs only to achieve a critical user mass before gaining the benefits of Web sites like e-Bay (http://www.ebay. com/) or Amazon (http://www.amazon.com/) in terms of de facto authority and brand recognition. For sites which are less commercially oriented, the loyalty of the user base is perhaps even more heavily emphasized. Two notable sites, which have grown rapidly without such a strong commercial bias, are Wikipedia and Imdb. Wikipedia has been both maligned and praised for its loyal community and efforts to create a free encyclopedia that is accurate and up to date without any commercial affiliations. Wikipedia began in 2000 as a complementary project for Nupedia, in which articles were written by experts and reviewed by a formal process (Wikipedia, 2007). In 2001, Larry Sanger proposed on the Nupedia mailing list to create a wiki as a “feeder” project for Nupedia (Sanger, 2001); this spawned rapid growth. By 2001, Wikipedia contained approximately 20,000 articles and 18 language editions. As of 2007, English Wikipedia contains over 1.7 million articles, making it the largest en¬cyclopedia ever assembled. The site relies on the goodwill of its members to write, update and contest articles, and has thus far proven that the Internet community as a whole is willing to contribute towards the database, albeit haphazardly (recent events are more likely to be covered in detail, while significant historical figures languish). Given the size and relatively stable growth of the project, the prospect of a peer-reviewed and written information repository might not be so cynically doomed. The International Movie Database (IMDB), a site in the top 20 (Alexa Internet, 2007b) U.S. Web sites visited, is the largest Internet compilation of
movie information (approximately 900,000 titles and 2.3 million names) (IMDB, 2007). IMDB draws a significant portion of its information from the participation of its user base. Beyond subjective reviews, users are also asked to supply cast and crew lists, production details, and actor biographies. IMDB grew from two lists that started as independent projects in early 1989 by participants in the Usenet newsgroup rec.arts.movies. Each list was maintained by a single person, recording items e-mailed by newsgroup readers, and posting updated versions of his list from time to time. The lists were eventually combined, and by late 1990, the lists included almost 10,000 movies and television series. As the contributions continued to grow rapidly, the IMDB formed as an independent company in 1995 and was later purchased by Amazon Inc. (Wikipedia, 2007b). The approach of the IMDB system closely resembles the envisioned system for GI Motility Online, with the user population submitting proposed changes, followed by an editorial process. Database content is generally provided and updated by a vast collection of volunteer contributors. There are only 17 members of the IMDB who are dedicated to monitoring received data, although 70% of IMDB’s staff serve as editors (IMDB, 2007), reviewing changes, verifying the information before posting the changes, and policing the forums. Peer review is considered to be one of the most even-handed and least biased methods of scrutinizing articles for publication. The development of a community, as well as the ability for members of that community to voice their opinions about professional papers is pivotal to the dissemination of accurate information. The model of Amazon or eBay is to allow users to comment, and thereby to signify reliability and approval. Gastrointestinal Motility Online allows users to provide feedback, both critical and supportive, in order to enhance the relevance of articles. According to Harris Interactive poll, only one-fourth of physicians use the Internet to
25
Gastrointestinal Motility Online Educational Endeavor
communicate with their patients (Computing in the Physician’s Practice, 2000). In the same poll, although 89% of physicians use the Internet in their practice in search of information, they spend only six hours per week to browse medical developments. Accordingly, Gastrointestinal Motility Online has been configured to serve as an encyclopedic source, as well as a high-value news feed. AccessMedicine uses a Podcast update model, with 10-minute broadcasts generated daily for use in family practice. Gastrointestinal Motility Online caters to a smaller specialty group with correspondingly less frequent pace of developments, and is therefore less time sensitive.
automation Automated content generation or extraction from other publications is not feasible, due to the stringent need to maintain relevance and quality. Thus, deployment of a pure peer-reviewed wiki-style community is precluded by the need to maintain quality. If membership of the wiki is restricted to those users whose credentials are accepted, or if changes must be approved at an editorial level, then a wiki would facilitate the rapid exchange of information as well. In the second part of the endeavor, the goal is to enable machine-assisted updating of the material in the gastrointestinal knowledge repository. Currently, all updates must be initiated manually. The pressure to update, but to update accurately applies to many situations: addition of new material, editing of existing material, and deletion of parts of existing material as new study results become available. Initial thoughts and test results are documented in Sharma (2005). As the site evolves, the need to keep the site relevant to a particular group of specialist physicians may conflict with the preferences of another significant category of users: the researchers. Gastrointestinal Motility Online intends to use user-based customized layout, as presented by Sarnikar et al. (2005). The filtering of details from
26
the main knowledge repository is not intended to block or hide information, but to provide a more relevant source. Classification and weighting are accomplished by a rules-based system that can assess whether an article is clinical- or basic science-related. Login will also provide both the ability to contribute and comment on articles, and to obtain a specialized view depending on the type of user. The application of user-based site layouts is not innovative, but is important to the domain of medical informatics because physicians attach such a high priority on relevance. Given the large number of articles published weekly and the difficulty in ascertaining relevance and quality, a number of automated tools will be used to optimize the updating process. Sarnikar et al. (2005) present one technique that will assist in filtering journal results and maintaining and updating the site. Their method selects articles ranked by relevance using a combination of both rule-based and content-based methods, using the following principles: 1. 2.
3. 4. 5. 6.
Profiles are modeled in the form of rules. The purpose of the rule-based profile is to identify a sub-set of documents of interest to a user. Each role has a set of predefined rules associated with it. Rules specify knowledge sources to access (e.g., nursing journals for Nurses). Rules can specify knowledge depth and knowledge breadth. Rules can specify semantic types of primary importance to roles.
Profiles are used in the gastrointestinal motility context to separate information into categories: for example, new clinical findings versus basic science. Articles may be assigned a category and a relevance weight, given categorization rules based on Unified Medical Language System (UMLS) synonym lists and the categories sign or symptom, diagnostic procedure therapeutic or preventive
Gastrointestinal Motility Online Educational Endeavor
procedure and disease or syndrome semantic types. In addition to a text search in the abstract, Sarnikar and Gupta (2007) also assign weights to the type of journal. These tools can select and filter relevant articles for presentation as an RSS XML news feed to editors or automatically assemble relevant articles for use by the editors or Web site administrators. While these tools will aid the editor, there is no replacement for the role of humans in decisively selecting and classifying information. Ontologies and semantic networks are prerequisites to the development and classification of information repositories. Ontologies serve many purposes (Kumar, 2005) including: • • •
Reuse and sharing domain knowledge Establishing classification schemes and structure Making assumptions explicit
They further enable analysis of information and complement the stricter terminology that is used in straightforward text searches. Examples of ontologies in use today include the National Library of Medicine’s Medical Subject Heading (MeSH), disease specific terminologies such as the National Cancer Institute’s PDQ vocabulary, drug terminologies such as the national drug data file, and medical sociality vocabularies such as the classification of nursing diagnoses and the current dental terminology. In Gastrointestinal Motility Online, the ontological hierarchy will be used to distinguish between sections of the gastrointestinal tract, from the stomach and esophagus to the large intestine and colon (Kumar, 2005). A key enabler for development of automatic information processing is the set of ontologies presented in the UMLS semantic network that rely upon the concepts built in the UMLS concept hierarchy. Hierarchical and clustered ontologies allow software to construct knowledge trees, conglomerating relevant knowledge. An overview of the UMLS is available at the unified
medical language system: What is it and how to use it? (http://www.nlm.nih.gov/research/umls/ presentations/2004-medinfo_tut.pdf). UMLS is an aggregate of over 134 source vocabularies, including the classifications from such lists as ICD-10 and DSM: IIIR-IV. It represents a hierarchy of medical phrases that can be used to classify most medical articles and textbook entries. For example, using UMLS, the following phrases are grouped similarly: Deglutition Disorders, Difficulty in swallowing, Difficulty in swallowing (context-dependent category), Dysphagia NOS, Dysphagia NOS (context-dependent category), Can’t get food down, Cannot get food down, Difficulty swallowing, Difficulty swallowing (finding), Dysphagia, Dysphagia (Disorder), Swallowing difficult, Swallowing Disorders (Aronson, 2001). The system described in Sharma (2005) uses techniques of Natural Language Processing (NLP) to construct a semantic understanding that surpasses text searching. Using the automated integration of text documents in the medical domain (ATIMED) system, the content and order of phrases are related lexically using a concept called Word-Net. Word-Net operates on the verbs, subjects and objects of the sentences, comparing sets and subsets of subject-verb-objects collections in order to determine topic relatedness. Word-Net further uses a lexical dictionary to determine similarity in all verb pairs and then subject-verb-action pairs. Sharma (2005) uses the following two sentences as examples: Dysphagia is a disease and defined as a sensation of sticking or obstruction of the passage of food. Dysphagia is related to obstruction of passage of food. Since both sentences contain similar objects and subjects, and use the verb “is”, the sentences are deemed similar. However, within the current mechanism, the phrase, “Dysphagia relates to obstruction of passage of food” would result in a poorly scored correlation or low match because the action verb is not similar (Sharma, 2005). Finally, the same technique allows the creation of new documents by collating sentences and
27
Gastrointestinal Motility Online Educational Endeavor
Figure 1. Integration of semantic understanding to generate new information (Adapted from Sharma, 2005) Sentence Syntax Definition Tables
Feature Selection
Feature 1
Feature 2
Feature 3
Feature 4
WordNet
Concept Formulation and Creation
paragraphs from various documents. An initial method of grouping sentences uses the quantity of concepts expressed. This method is further refined by evaluating the sentences based on the following criteria: similar-subjects, similar-objects, similarsubjects-objects, and similar-objects-subjects. Based on the structure of English grammar, these techniques have been reliably shown to collate relevant data into a readable format. A diagram of the method is shown in Figure 1. A sample output, based on the use of this technique, is provided below: REGURGITATION is defined as the spontaneous appearance of gastric or esophageal contents in the back of the throat or in the mouth. In distal esophageal obstruction and stasis, as in achalasia or the presence of a large diverticulum, the regurgitated material consists of tasteless mucoid fluid or undigested food (Sharma, 2005). The context interchange of heterogeneous sources of information being collated for different classes of users leverages tools from an additional
28
branch of information technology. One level of development is the creation of maps between inputs and outputs, in much the same way that a dictionary might map between languages. The conversion of n sources to m outputs can grow to be an exponentially difficult problem; this can be addressed with the use of intelligent heuristics and protocols to selectively prune the data. Difficulties become particularly apparent with changes in any client schema that cause a cascade of changes in the mappings (Sarnikar, 2005). Using an independently developed predefined mediating schema would restrict the amount of information that could be exchanged. Apart from relational schema, a client schema could also be specified as a hierarchical schema or as an XMLbased message (Sarnikar & Gupta, 2007). In the context of Gastrointestinal Motility Online, the specific source contexts are innovative journal papers, reviews, and textbook articles. The output contexts are specialist clinicians, researchers, students, and other health professionals. Developing a schema to accurately represent journal abstracts and determine the relevance of those abstracts is
Gastrointestinal Motility Online Educational Endeavor
another method of exchanging contexts. Innovation in this domain will allow Gastrointestinal Motility Online to maintain updated, consistentquality references without requiring an editor to read every journal article published immediately.
lessons learned The GI Motility Online site benefited from a mature development environment for Web-based information retrieval. Interface: In the case of GI Motility Online, the user base is a readily identifiable group, trained in a similar fashion, with specific needs and expectations of organization and formatting. The following two aspects influenced the design process: •
•
Pagination is a critical issue for online didactic materials. As people do not like the interface to online books (Crawford, 2006), the development of the site must reflect the reading style and needs of its users. In the case of GI Motility Online, there are currently no page delimiters; this is encouraging users to print the material to be used as a reference. A solution of delivering a formatted print-ready PDF could be applied, but the cost begins to climb with the number of different formats and delivery options supported. Tables must be handled intelligently. Tables are used in many medical publications to rapidly and clearly present information. The trend in the mid-1990s for netiquette was to inline the tables with the text, but this can create awkward gaps or poor formatting choices. GI Motility Online chooses the path of Harrison’s Online, in linking to the tables outside the main document to preserve readability.
Based on their respective training, many physicians expect a particular language and lay-
out of medical information. The highest-value information for each physician may vary based on specialty, and user-customization of graphical widgets may enhance information value. Since GI Motility Online aims to support a broad base of users with many different roles and information needs, reusable widgets, of the type available in customizable Web-portals, may potentially solve these varying needs. Workflow: As stated previously, over nine months was required for simply editing the site information, moving documents back and forth between editors, experts and developers. The method involved several inefficient technologies, such as mailing CDs and sending large files through mail servers. The advantage of such hands on interaction and communication is the result: the authors of the site are particularly pleased by the polished appearance of the delivered product. Electronic collaboration between different sites was difficult due to the nature of document formats, which does not produce a consistent print layout on different computers, and layout formats, which do not allow easy markup or revision to the document. The use of a standardized input format and efficient conversion of the RTF-formatted documents into Web viewable formats was crucial to the development of the Web site. Though conversion of documents is a minimally complex task, the production of a site that allows conversation on the fly to support concurrent editing and proofing between multiple users is a challenge. Online publishing companies, such as Atypon, offer suites of software to facilitate publishing workflow, allowing multiple authors to upload articles, multiple editors to revise articles, and art editors to manage graphics, all in an organized fashion. Meta-Information: The World Wide Web provides a tremendous asset in connecting related articles seamlessly. In determining related pages, an active agent—a human, machine or some combination thereof—who must isolate the crucial elements of a page and capture that as
29
Gastrointestinal Motility Online Educational Endeavor
meta-information, preferably categorized. Organized meta-information allows automated tools to develop connections and links to potentially relevant information. Extracting crucial elements of a page is best handled by encouraging authors to define keywords, as with scientific articles. Although many automated agents have historically been unsuccessful in information extraction, the medical domain provides some assistance with standardized ontologies, which allow agents to categorize information in a framework, reducing some identification errors. Maintenance: Many articles in mainstream sites, such as ESPN.com, offer the opportunity for users to comment and leave feedback. This increases user participation and value to the site, but raises additional issues that need to be addressed, such as: • • •
Profanity or other inappropriate comments must be censored Sites are more open to attack and security leaks due to their increased functionality Bandwidth usage increases, and may debilitate the site
In addition, a version control framework must be established for sites that allow updates, in order to rollback unwanted changes. Furthermore, hardware resources must be allocated to store site changes.
current usage and Future dIrectIons Anecdotal evidence provided by the lead creator of GI Motility Online, Dr. Raj Goyal, highlights that the representatives of Nature Publishing Group describe the site as being popular, and that site visits to GI Motility Online are increasing. Due to the delicate nature of conversation with his peers in gastroenterology, Dr. Goyal cannot be
30
certain, but he states that initial impressions are unanimously positive and enthusiastic. Dr. Goyal notes that the majority of the traffic to GI Motility Online currently arrives from Google, and indeed, Google’s first returned site in response to “GI Motility” is GI Motility Online (Google, 2007). GI Motility Online does not currently permit any advertising; if this policy is revised, the usage patterns may change. Further, the numbers for site visits are expected to increase if the site were to be more supported by Nature Publishing Group. The stakeholders are satisfied with the quality of the product, and are ambitiously pursuing an extension of the product to broaden coverage to other parts of the GI tract. Dr. Goyal has received inquiries to publish the material in a hardcopy format. Custom-built books or full-page colored slides offer additional utility to medical educators, specialists, nurses, and students, and may also offer a revenue stream. These custom-built printable books are a logical extension of the current “print what you want” photo fulfillment services that are popular on photo sharing sites or album hosting services. Given that Harrison’s Principles of Internal Medicine contains over 2,000 colored pages in book form, there may be a future for customized books in medicine. Retrieval: New algorithms for search and ranking are not merely lexically-based, but also combine closely related concepts and relationships between ideas. This facilitates the creation of a richer search language that can account for relationships, such as causation, consequence, association, treatment, or opposite. For example, hypothermia is directly opposite hyperthermia, and may be caused by thalamic alterations. Such basic chains of relationships could be easily captured if the language and storage of meta-information about articles could contain the necessary underlying details. Presentation: Using Tufte’s principles (Tufte, 2001), the real estate of a screen needs to be more efficiently used. Currently, most search-engine
Gastrointestinal Motility Online Educational Endeavor
results do not often give the sense of the relevance of the article, of the correct sections, or the tone of the article. The information density of the return pages is low, causing users to scroll through potentially hundreds of articles to find relevant information. Graphical interfaces are likely to be a solution, as the ability of the Internet to handle higher bandwidth applications grows. In medical research in particular, an interface, that allows users to quickly see search results of the primary concept and even related concepts, may dramatically affect usability. Online Sharing: Bulletin board systems, which flourished during pre-Internet days, have re-emerged on the Internet as forums, and are a popular source of information. Corporations with significant Internet presence, and especially gaming companies like Nintendo (http://forums. nintendo.com/nintendo/), have begun adding and using these forums as a method of improving public relations and offering support. Intranet and Internet file sharing systems have also flourished, as bandwidth rates increase. The rise of Youtube is one phenomenon, but the comfort of using the Internet to disseminate multimedia (such as GI endoscope video) appears to be more solidified. These high bandwidth applications have grown in acceptance, and despite the threat of viruses in downloaded files, file-sharing traffic is increasing daily. Megaupload and Rapidshare, two prominent file sharing services, are now ranked #18 and #27 in the reliable Alexa ranking (Alexa Internet, 2007b) of most visited Internet sites in the U.S. However, these services are predominantly used by noncommercial entities. Commercial enterprises may be hesitant due to slow adoption, security issues, bandwidth maximums, or unprofessional presentation. Such file sharing sites will certainly cater in the future to commercial entities, perhaps by providing specially developed sites, or providing branded services. The future of the Internet, and especially the ability to deliver high bandwidth, will increasingly rely upon specialized sites with
high-capacity and high-bandwidth connections close to Internet backbones. Collaboration: Collaboration in the domain of GI motility will drive improvements in language tools, with many GI specialists throughout the world who may need to collaborate via the Internet or phone services. Currently, Altavista and Google offer Web-based text translation, but more intelligence or domain-specific knowledge may be required. One common example in GI translation is translating the world “oral” to mean “verbal”, when in fact, the correct reference is to the mouth cavity. Overall, however, text translation between languages is generally adequate for initial communication. Oral translation is a high value direction to pursue, but at the same time, is a very difficult task. Psycholinguistics research is still unraveling the complexities of language parsing, and the ability of current artificial intelligence to understand language is severely limited. Nevertheless, translators in this area will prove essential and highly desirable for the next generation of online collaborators. Similarly, one of the holy grails of artificial intelligence development is the creation of an artificial system capable of interpreting human language. Within specialized domains with limited vocabularies, artificial readers become more feasible, but the medical domain is particularly difficult, due to its large specialized vocabulary. Development in this area would provide rewards for medical researchers, allowing the creation of agents, which would allow researchers to process more information by selecting and even summarizing articles. Organizationally, scientific research in domains such as medicine would benefit tremendously from the creation of a centralized authority to monitor, synthesize and rate research. The current system of research funding in America places research at the whim of special-interest private funding and sometimes misdirected public funding in overly popular or extremely esoteric
31
Gastrointestinal Motility Online Educational Endeavor
areas. Regulating and directing research might also help avoid repeating inconclusive research, which does not get published (and thus may be repeated).
conclusIon Gastrointestinal Motility Online is an evolving knowledge base related to Gastrointestinal Motility disorders. The current phase of the endeavor focuses on the collection and organization of knowledge from many different sources. Knowledge-mining tools are being developed to utilize this information as it becomes available to add fast, relevant access and other utilities to the information repository. The continuous change in standards of care and knowledge due to rapid discoveries in the basic and clinical sciences prompts for a system that is more flexible than a textbook, while demanding thoroughness and accuracy. Knowledge mining tools and other advanced technologies to aid in the conversion and integration of articles and research into the mainstream science are being integrated into Gastrointestinal Motility Online, and look to impact the breadth and speed of knowledge-base upgrades. Gastrointestinal Motility Online serves to balance the needs of its user base while embracing the academic rigor in a novel application of technology to the science of medicine.
acKnowledgMent The authors would like to thank Dr. Raj Goyal for his invaluable input and contribution to GI Motility Online and this article. The authors also thank Richard Martin for his helpful comments and proofreading of the text.
reFerences Alexa Internet. (2007a). Three-month traffic statistics for wikipedia.org. Retrieved from http://www. alexa.com/data/details/main?q=&url=wikipedia. org. Alexa Internet. (2007b). Top Sites United States. Retrieved from http://www.alexa.com/site/ds/top_ sites?cc=US&ts_mode=country&lang=none. American Library Association. (1989). Presidential Committee on Information Literacy - Final Report. Retrieved from http://www.ala.org/ala/ acrl/acrlpubs/whitepapers/presidential.htm. Antman, E., Lau, J., Kupelnick, B., Mosteller, F., & Chalmers, T. (1992). A comparison of results of meta-analyses of randomized control trials and recommendations of clinical experts. Journal of the American Medical Association, 268, 240-248. Aronson A. (2001). Effective mapping of biomedical text to the UMLS Metathesaurus: The MetaMap program. Proceedings of the 2001 AMIA Symposium, pp. 17-21. Carlson, C. (2003). Information overload, retrieval strategies and Internet user empowerment. In: L., Haddon (Ed.), The Good, the Bad and the Irrelevant (COST 269) 1(1) (pp. 169-173). Helsinki, Finland. CEBM. (2007). Why the sudden interest in EBM?. Retrieved from http://www.cebm.utoronto.ca/ intro/interest.htm. Computing in the Physician’s Practice. (2000). Retrieved from http://www.harrisinteractive.com/ harris_poll/index.asp?PID=58. Crawford, W. (2006). Why aren’t ebooks more successful?. Retrieved from http:// www.econtentmag.com/Articles/ArticleReader. aspx?ArticleID=18144. Davis, D., Thomson, M., Oxman, A., & Haynes, R. (1997). Changing physician performance: A
32
Gastrointestinal Motility Online Educational Endeavor
systematic review of the effect of continuing medical education strategies. Journal of the American Medical Association, 274, 700-705. Fridman, S. (2000). Doctors lag when it comes to computer use - Industry trend or event. Retrieved from http://www.findarticles.com/p/articles/ mi_m0HDN/is_2000_March_28/ai_60907019. Giles, J. (2005). Internet encyclopedias go head to head. Retrieved from http://www.nature.com/ news/2005/051212/full/438900a.html. Google. (2007). Search returned http://www. google.com/search?hl=en&q=GI+Motility. Gore, A. (1992). Infrastructure for the global village. Scientific American, 265,150-153. Guyatt, G. Cairns, J., Churchill, D., Cook, D. Haynes, B., Hirsch, J., & et al. (1992). EvidenceBased Medicine Working Group: Evidence-based medicine. A new approach to teaching the practice of medicine. Journal of the American Medical Association, 268, 2420-2425. Hawking, S. (2001). The universe in a nutshell. New York, NY: Bantam Books. Haynes, R. (1993). Where’s the meat in clinical journals [editorial]? ACP Journal Club, 119, A22-A23. IMDB. (2007). How/where do you get your information? How accurate/reliable is it? Retrieved from http://imdb.com/help/show_leaf?infosource. Kasper, D., Fauci, A., Longo, D., Braunwald, E., Hauser, S., & Jameson, J. (2005). Harrison’s principles of internal medicine, 16th ed. New York, NY: McGraw-Hill. Kumar, A. (2005). Ontology-driven access to biomedical information (ODABI). Undergraduate Thesis, University of Arizona. Lindberg, D. (2003). NIH: Moving research from the bench to the bedside. Presentation to the Subcommittee on Health.
Lyman, P. & Varian, H. (2003). How much information?. Retrieved from http://www2.sims.berkeley.edu/research/projects/how-much-info-2003/ index.htm. Mechaca, L. (2007). Goodbye, Axim. Message posted to http://direct2dell.com/one2one/ archive/2007/04/11/11397.aspx. Sackett, D. & Straus, S. (1998). Finding and applying evidence during clinical rounds: The evidence cart. Journal of the American Medical Association, 280, 1336-1338. Sackett, D. (1997). Using evidence-based medicine to help physicians keep up-to-date. The Journal for the Serials Community, 9, 178-181. Sanger, L. (2001). Let’s make a wiki. Message posted to http://web.archive.org/web/20030414014355/ http://www.nupedia.com/pipermail/nupedial/2001-January/000676.html. Sarnikar, S., Zhao, J., & Gupta, A. (2005). Medical information filtering using content-based and rulebased profiles. Proceedings of the AIS Americas Conference on Information Systems (AMCIS 2005), Omaha, NE. Sarnikar, S. & Gupta, A. (2007). A context-specific mediating schema approach for information exchange between heterogeneous hospital systems. Forthcoming in International Journal of Healthcare Technology and Management. Schwartz, B. (2004). The paradox of choice. New York, NY: HarperCollins Publishers. Senge, P. (1994). The fifth discipline: The art and practice of the learning organization. New York, NY: Doubleday/Currency. Sharma, R. (2005). Automatic integration of text documents in the medical domain. Undergraduate Thesis, University of Arizona. Slawson, D., Hauck, F., Strayer, S., & Rollins, L. (2007). What is information master?. Retrieved from http://www.healthsystem.vir-
33
Gastrointestinal Motility Online Educational Endeavor
ginia.edu/internet/familymed/docs/info_mastery. cfm#Information. Tufte, E. (2001). The visual display of quantitative information. Cheshire, CT: Graphics Press. Vanhecke, T., Barnes, M., Zimmerman, J., & Shoichet, S. (2006). PubMed vs. HighWire Press: A head-to-head comparison of two medical literature search engines. [Epub ahead of print] Computers in Biology and Medicine.
Wikipedia. (2007a). Information retrieval. Retrieved from http://en.wikipedia.org/wiki/Information_retrieval. Wikipedia. (2007b). Wikipedia. Retrieved from http://en.wikipedia.org/wiki/Wikipedia.
This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Issue 1, edited by J. Tan, pp. 24-43, copyright 2008 by IGI Publishing (an imprint of IGI Global).
34
35
Chapter 3
Envisioning a National e-Medicine Network Architecture in a Developing Country: A Case Study
Fikreyohannes Lemma Addis Ababa University, Ethiopia Mieso K. Denko University of Guelph, Canada Joseph K. Tan Wayne State University, USA Samuel Kinde Kassegne San Diego State University, USA
aBstract Poor infrastructures in developing countries such as Ethiopia and much of Sub-Saharan Africa have caused these nations to suffer from lack of efficient and effective delivery of basic and extended medical and healthcare services. Often, such limitation is further accompanied by low patient-doctor ratios, resulting in unwarranted rationing of services. Apparently, e-medicine awareness among both governmental policy makers and private health professionals is motivating the gradual adoption of technological innovations in these countries. It is argued, however, that there still is a gap between current e-medicine efforts in developing countries and the existing connectivity infrastructure leading to faulty, inefficient and expensive designs. The particular case of Ethiopia, one such developing country where e-medicine continues to carry significant promises, is investigated and reported in this article.
Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Envisioning a National e-Medicine Network Architecture in a Developing Country
IntroductIon Healthcare consumers in general tend to seek access to affordable health services that will meet their needs. From an ethical standpoint, healthcare has to be available when and where consumers need it; physical separation between consumers and healthcare facilities must not pose severe limitations on the delivery of efficient healthcare, even if patients are located in remote areas. In this sense, information and communications technology (ICT) has been demonstrated to offer a competitive choice for accessing affordable and effective health services, especially when access is difficult and limited (Horsch & Balbach, 1999; Kirigia, Seddoh, Gatwiri, Muthuri, & Seddoh, 2005; Tan, Kifle, Mbarika, & Okoli, 2005). More recently, with the continued maturity of network such as Integrated Services Digital Network (ISDN) and Asynchronous Transfer Mode (ATM) networks and related technologies (Perednia & Allen, 1995; Tan, 2001), e-medicine implementation has entered a stage where both the health providers and consumers can now benefit significantly. IT-based horizontal and vertical communications among the healthcare facilities within the organizational structure of the healthcare system are essential. Such communications facilitate efficient information exchange and help the delivery of essential health services to underserved rural areas. These communications can be supported through a nationwide e-medicine network that is based on affordable telecommunications infrastructure. The network should connect all regional clinics to urban area hospitals. The benefits of such a network include: (a) establishing reliable horizontal-vertical communications and information sharing among facilities, thereby driving up quality, improving efficiency, and enhancing cost-effectiveness of services; (b) achieving e-health commitments and bringing healthcare closer to underserved and un-served rural areas; (c) strengthening collaboration among hospitals
36
within a multi-provider care management context; (d) minimizing long distance travels among rural people in need of proper medical care to urban areas or the capital city; and (e) providing medical information to clinical practitioners that will help them keep abreast of clinical breakthroughs as well as new technological advances. For Ethiopia, a lesser developing country with significant challenges in meeting basic healthcare needs, it is argued that e-medicine development is emerging and can be fruitfully cultivated over the coming years if a vision and long-term strategy for this technology can be used to help increase the number of citizens receiving care and decrease the subsequent healthcare costs. This article lays out such a vision and strategy for a nationwide e-medicine infrastructure to be designed. It is organized as follows. First, the background and various design considerations for a nationwide e-medicine network is presented. Next is an overview of the requirements for the network design followed by a more in-depth description of the local and wide area network (LAN/WAN) architecture envisioned. The focus of the discussion will then shift to the existing Broadband Multimedia Network (BMN) and Very Small Aperture Terminal (VSAT) infrastructures and how these networks may be integrated into the nationwide e-medicine infrastructure. Finally, the article concludes with insights into potential future work in e-medicine for developing countries.
e-MedIcIne networK desIgn consIderatIons E-medicine refers to the electronic delivery of healthcare and sharing of medical knowledge over a distance employing ICT. A national e-medicine network allows sharing and exchanging of clinical data among physicians, administrators, even patients or other participating health professionals regardless of physical distance separation or geographical terrain of the whereabouts of these
Envisioning a National e-Medicine Network Architecture in a Developing Country
network participants within the national boundaries. The network also facilitates communications among physicians and academics across diverse cultures, affiliated healthcare organizations, and publicly or privately funded research institutions. Since there is lack of transportation and communication infrastructure in developing countries, medical and clinical data exchanges can be further secured and facilitated through an existing e-medicine network. In developed nations, e-medicine services (Wright, 1998) can benefit remote locations that may not be easily accessed due to unpredictable or harsh weather conditions found during certain times of the year, for example, parts of North America and Scandinavian countries are often heavily affected by snow and other natural hazards such as avalanches, falling boulders and closure of highways due to multi-vehicle accidents or other calamities. Mountainous terrain in certain parts of North American regions such as Alaska, British Columbia, Alberta and New Territories implies the need for viable distance healthcare delivery solutions. E-medicine allows health professionals around the world to establish faster communication and exchange information with clients and regional authorities irrespective of geographical locations. It may also support rural dwellers to get healthcare services delivery similar to their urban counterparts. A mobile e-medicine system, for instance, provides a convenient platform for acquisition, transmission and delivery of health-related data to healthcare providers through 2G/3G-based wireless networks (Wootton, 2001). Recognizing these benefits, the International Telecommunications Union (ITU) has set a global agenda to promote e-medicine applications in developing countries. Ethiopia, one of the beneficiaries of such an initiative, has commissioned some ICT projects such as SchoolNet, WoredaNet and BMN to enable fully-fledged connectivity to make better use of the ICT in the health and education sectors.
network architecture The design of suitable national e-medicine network architecture in a lesser developing country such as Ethiopia requires several key components to be integrated into a dynamic and enterprising communication infrastructure. Major architectural components include: (1) LAN architecture for local networking and sharing of health-related information; here, communications may be established using wireless cellular or ordinary fixed telephone lines; (2) WAN architecture for national networking and sharing of health-related data; this will allow communications among local and national physicians, healthcare workers and clients covering urban and rural communities; and (3) designing a suitable back-end database and front-end user interface applications that integrate seamlessly with the (prototype) implementation of the proposed architecture. The overall goal of the nationwide e-medicine network architecture is then to provide an affordable and a low-cost system that facilitates uninterrupted communications among physicians and health professionals across the country. The system bolsters connectivity among rural clinics and urban area hospitals to support primarily clinical e-consultation and maintenance of stored patient records. This network should also be cost-effective, expandable, and secure. It must support a stateof-the-art ICT access schema and connectivity to rural area clinics. Existing ICT infrastructure will be given priority to minimize the cost of implementing the nationwide network. In the Ethiopia design, expandability is a concern. First, few hospitals are built in the country while more clinics are being added every year. Moreover, there is a chance to incorporate private hospitals in the nationwide e-medicine network as and when necessary, which will further increases the number of future connected sites. As well, the area of e-medicine applications will not be limited to just some specific diseases, but will be expected to increase in type and number over
37
Envisioning a National e-Medicine Network Architecture in a Developing Country
the long haul. In fact, the network should also support advanced applications, which require real-time connectivity such as videoconferencing capabilities for future use. During e-consultation or patient referral, most of the data exchanged over the network are sensitive patient information. Confidentiality of patient information must therefore be respected. For secure communications, protocols such as Secure Socket Layer (SSL) could be used. SSL ensures secured communications over web-based applications and provides the ability to safely exchange patient information across the network (Elmasri, 2000). When doctors exchange patient information, they could adhere to medical protocol that defines the rules to be followed during this process. In addition, the network and accompanying servers could be protected by firewall against hacking from external parties. Firewalls are software or hardware for the sole purpose of keeping digital pests such as viruses, worms, and hackers out of the network (http://www.cisco.com, Tanenbaum, 2004).
networK desIgn reQuIreMents As cost must be one of the driving factors for choosing among existing or emerging ICT infrastructures in the country, implementing nationwide e-medicine network infrastructure may seem at first to be more expensive than building clinics or supplying existing regional clinics with medical personnel and equipment. Yet, a cost-benefit analysis comparing various IT investment approaches will provide best directions to achieving a lower cost solution to the problem of delivering adequate and proper healthcare and disseminating confidential health information to and from various connecting points throughout the country. With today’s oil prices at a premium, network connectivity among the healthcare facilities, both in the urban and rural areas over an existing ICT
38
infrastructure is now considered a cost-effective solution. Of course, set-up costs depend on the type of WAN to be used—to ensure low installation cost, it is proposed that the network design will incorporate an existing WAN provided by the Ethiopia Telecommunication Corporation (ETC). In Ethiopia, most of the inter-hospital communications are traditionally dependent on telephone and hand-delivered referral messages. During referrals patients have to travel afar to one of the urban referral hospitals, carrying the written messages of the referring physician. Clinics located in the telephone coverage areas communicate using telephone to exchange information about availability of specialist(s) or bed in another hospital. Yet, the communication needs of hospitals have grown over the years ahead of its technological capabilities. Geographically dispersed clinics lack modern telecommunication technology access. Among them are instantaneous access to patient information, access to electronic medical records, and access to the Internet. These and other communication needs of health providers also require the development of e-medicine application software backed by electronic patient record systems. Design of such communication networks will also require the understanding of organizational structure of the clinics involved in the network. Since the government/public clinics are owned and organized under their respective regions, the WAN design should follow the organizational structure of the administrative regions in the country. A detailed study about the inclusion of various clinics, their locations relative to the nearest access point to existing ICT infrastructure, traffic load and its characteristics, security, LAN/WAN protocol, topology and bandwidth requirements and utilization, and allocation of bandwidth, among other issues, have to be considered while trying to design a nationwide e-medicine network architecture. For example, issues of communicating patient information electronically may further raise question on medical
Envisioning a National e-Medicine Network Architecture in a Developing Country
ethics, the need for developing standard medical protocols, and detailing policies for use in routine activity via the e-medicine network.
lan architecture To design the LAN for each hospital, we consider the central site, Tikur Anbassa Specialized Hospital located in the capital Addis Ababa, as a model. The hospital is organized into 16 departments with each department further divided into smaller units as necessary. For instance, the Internal Medicine department has several units such as the Renal Unit, the Cardiology Unit, the Neurology Unit, and other units. Physicians in these departments and units need to communicate whenever a patient visits more than one of the units. It is proposed that the LAN follows the hierarchical structure of the hospital. The decision to make the selection between various LAN technologies was based on: (a) expected application to run on the network and their traffic patterns; (b) physical locations of the offices and users to be connected in campus; (c) the rate of network growth; (d) the abundance of the network technology in the market; and (e) simplicity of installation and maintenance. Each of these criteria will now be explored in more depth.
Expected Application to Run on the Network and their Traffic Patterns Currently we expect a Web-based e-medicine application to run on the network. The application will use a central database server where all the user and patient information will be stored. The type of data to be transmitted on the network should accommodate both text and image formats. Since all communications are to be channeled through the server, the traffic pattern around the center is expected to be heavy. Higher speed devices should be installed at the center of the LAN where servers will be located.
Physical Locations of Offices and Users to be Connected on Campus The sample hospital (Tikur Anbassa Specialized Hospital) is housed in a series of five buildings (Blocks A-E). These blocks are not physically separated. Even though precise figures were unavailable, these five buildings are built on roughly 8,000 to 10,000 square meters. While the main offices and departments in the hospital are located in one of the respective blocks, most of these offices are in either of the first two stories of the block they belong. Having routers switches in each of the departments is ideal to design a high-speed and expandable LAN, but it will also make the design expensive to install, support and maintain. A more cost-effective approach is to put switches per building and then get the departments to be connected into various groups by using Virtual LAN technology.
The Rate of Network Growth The rate of the hospital LAN growth depends on the level of computerization in the hospital. Currently in this central hospital site, there is a LAN that connects a few offices and a computer room. The network employs star topology, using a centrally located hub and Unshielded Twisted Pair (UTP) cables forming a peer-to-peer LAN. The purpose of this LAN was to enable offices to share printer and students to get access to research documents. In this design, it is anticipated that as the use of Web-based applications becomes commonplace, there will be opportunities to add more applications and connect more computers and offices to the LAN. The switches-routers selected in this design should therefore have many free ports to help cascade the growing number of anticipated future connections.
39
Envisioning a National e-Medicine Network Architecture in a Developing Country
The Abundance of Network Technology in the Market
layer (core layer); (b) second layer (distribution layer); and (c) third layer (access layer).
Capitalizing on the abundance of emerging network technology in the Ethiopia market, we gathered data from existing network technology vendors and organizations that implement computer networks in the capital city, Addis Ababa. Ethernet technology is common in organizations that implement computer networks, such as Addis Ababa University (AAU Net). AAU Net is a network backed by triangular shape fiber optic cable connecting the three main campuses. The topology is an extended star topology that fastens together fiber optic cables for vertical cabling (backbone cabling) between buildings that house various faculties and departments. These backbone cabling provide interconnection between wiring closets and Point of Present (POP). The zones that fall within a departmental area are served by internetworking devices such as hubs and UTP cables. Interestingly, what is described so far appears to be the dominant design of the small number of networks existing in the capital Addis Ababa. As such, it is not surprising to find that suppliers of network technology devices and support in Ethiopia are restricted to only a limited number of vendors.
Core Layer
Simplicity of Installation and Maintenance To design the LAN architecture we have therefore selected the hierarchical model. This enables us to design and arrange the inter-network devices in layers. Figure 1 depicts the hospital-based LAN architecture. It is a model preferred by most of network design experts for its ease of understanding, expandability and improved fault isolation characteristics (http://www.cisco.com). The model encompasses the following three layers: (a) first
40
Core layer high performance switches, capable of switching packets as quickly as possible, are to be deployed. Essentially, this layer connects the LAN backbone media as well as connects to the outside world via a firewall through WAN. In this design, the devices in the core layer will be placed at a central location in the hospital. The devices will then be connected with high-speed cables such as fiber optics, or fast Ethernet cables. The servers will also be connected to switches, shielded by a firewall.
Distribution Layer Distribution layer will contain switches and routers capable of Virtual LAN (VLAN) switching and allow defining departmental workgroups and multicast domains. The devices should also support connectivity of different LAN technologies since they also serve as the demarcation point between the backbone connections in the core layer and the access layer. In this hospital-based LAN design, the distribution layer represents switches/routers at each building connected to the core layer on the one end and to the access layer on the other. Redundant links will be used for maximum availability and the departments could be grouped forming their own Virtual LAN.
Access Layer Access Layer is where the end-users are allowed into the network. This layer contains switches/ hubs from which PCs in each department gain access to the hospital-based LAN. Each department will have at least one switch/hub, which will in turn have redundant links to more than two of the switches in the distribution layer.
Envisioning a National e-Medicine Network Architecture in a Developing Country
Figure 1. Hospital LAN design
WAN architecture Designing the WAN architecture for a nationwide e-medicine network raises the issue of WAN service provider. Unlike LAN, WAN connectivity depends on the availability of WAN infrastructure in the country. The sole WAN service provider is the Ethiopian Telecommunications Corporation (ETC). ETC provides a number of services (http:// www.telecom.net.et) from which the WAN infrastructure suitable for the e-medicine network may be derived. Existing WAN services include: (a) Internet Services, or, providing basic Internet services over dial-up or leased lines; (b) Digital Data Network (DDN), supporting dedicated Internet, ISDN and frame relay services; (c) SchoolNet VSAT, covering services for secondary schools and institutes of higher learning; (d) WoredaNet VSAT, covering services for districts (Woreda) administrations; and (e) Broadband Multimedia Network (BMN), offering high-speed optical communications to major cities.
To choose among these possible infrastructures for nationwide e-medicine network, the parameters to be considered include the geographical coverage, bandwidth, mode of communication, rental cost of WAN connection and capacity to add more LANs. Table 1 summarizes the data comparing among the available ICT infrastructures in Ethiopia. Based on the data, it appears that WoredaNet is best suited to the national e-medicine network, as long as the existing infrastructures are functioning efficiently and effectively. However, as noted in Table 1, there may be a tradeoff between coverage and capacity, that is, when the coverage is acceptable the capacity may be somewhat limited. For example, BMN coverage is ideal as it represents state-of-the-art service and higher bandwidth. However, it is centered primarily in the urban areas. It is also under development and we have thus considered it as a potential option to be used when integrated with the VSAT-based networks to enhance nationwide e-medicine network. Finally, the SchoolNet needs to be upgraded to support two-way interactivity.
41
Envisioning a National e-Medicine Network Architecture in a Developing Country
Table 1. Summarized comparison of existing ICT infrastructure Internet
Coverage
Bandwidth
Interactivity
Cost
Capacity to scale
DDN
SchoolNet
Telephone coverage areas only
The capital and regional Urban areas only
About 500 schools covered. There are Woredas that do not have schools
571 Woredas out of 594 are covered
The Capital city and 13 regional towns.
Maximum of 56k dialup and 1Mbps in Leased line
Maximum of 1Mbps
Can be upgraded to 384k upstream
Downstream/ upstream 45Mbps/ 256k downlink
ADSL Services: Variable bandwidth Downstream/ upstream 512k/128k and 1024k/256k
Two-way
Two-way
One-way broadcasting
Two-way
Two-way
Free for schools
Free For Woredas
Not yet determined, under development
Will have more than 10 ports free at each Woreda
Can be expanded
0.11 birr/min dialup 1000 birr/ month leased line
Not scalable
Not scalable enough
Thus, one alternative approach is using a combination of VSAT networks and terrestrial BMN. VSAT-based connectivity is believed to be cost-effective and in the case of WoredaNet and SchoolNet, it enables connectivity to the public, even in the rural areas. In addition to serving the rural areas, it also covers urban areas. Together, this will provide modern, convenient as well as economical connectivity to hospitals. For improving state-of-the-art applications such as videoconferencing, connectivity via the emerging BMN is proposed to connect urban area hospitals in the capital city and in the regions where the network can be easily accessible. Figure 2a shows the e-medicine network as a first alternative. Note that this approach requires that urban hospitals maintain two WAN connections. Having more than one WAN connection, however, may become expensive in the long run. If the two WAN infrastructures could be integrated,
42
WoredaNet
BMN
an improved WAN design will result with only one WAN connection to the urban hospitals through which the hospitals will be connected to BMN and the rural area clinics through the WoredaNet. Figure 2b depicts the second alternative solution of the WAN design.
current etc InItIatIVes The recent development in the ETC in providing multimedia network infrastructure is the integration of the VSAT-based networks (SchoolNet and WoredaNet) with the BMN (Tiruneh, 2006). In other words, these VSAT-based networks can now be used as a point of access to the BMN. As part of a longer term vision and mission of broadband initiatives for socioeconomic development in Ethiopia, ETC has also planned an e-health setting that tries to cover rural areas,
Envisioning a National e-Medicine Network Architecture in a Developing Country
Figure 2a. Logical WAN design based on BMN and VSAT networks (1st Alternative)
schools, clinics, hospitals, prisons, and nursing homes, including assisted living with several requirements: (a) high quality patient data, video and images to be exchanged between different medical institutions; (b) ICT infrastructure to connect geographically dispersed institutions,
nationwide or worldwide; (c) infrastructure that supports data, video and voice/audio (multimedia) services; (d) high quality, secured and fast delivery of medical information; and (e) high speed (BW) connectivity or the deployment of broadband infrastructure.
Figure 2b. Logical WAN design based on BMN and VSAT networks (2nd Alternative)
43
Envisioning a National e-Medicine Network Architecture in a Developing Country
Figure 3. ETC’s broadband-integrated infrastructure for e-health applications
Key challenges faced are the need to encompass multiple locations, to use multiple access technologies, to deliver multiple services and to address multiple user markets.
Current BMN Development ETC has already completed building the Core Terrestrial Broadband Infrastructure, which is capable of providing data, video, and voice services with 24 Points of Presence. This infrastructure services key business sites (urban areas) and supports Multiple Broadband Access via ADSL, FWA, WIFI, and fiber networks.
Current VSAT Development In Ethiopia, a Broadband VSAT Network platform, which supports integrated services such as video, data, Internet, and voice on a single infrastructure, is currently in place. It has countrywide coverage (450+ schools and 550+ woredas) as part of SchoolNet and WoredaNet deployments. It is integrated with the core multimedia network,
44
also serving as broadband access means. Figure 3 shows the two recent developments. Apparently, the new developments in WAN infrastructure support the 2nd alternative emedicine WAN architecture discussed here as depicted in Figure 3.
tHe prototype Based on the specified network requirements and architecture, a working prototype for the national e-medicine network is now presented and its operations highlighted. The prototype is a basic e-Medicine service (BEMS), which provides a Web-based graphical user interface (GUI) for health providers. BEMS facilitates the information exchange between remotely located health providers for the purpose of e-consultation, as well as for maintaining electronic patient information. The traditional paper-based forms and patient cards used in the hospitals will be digitized and reproduced electronically. Web-based technology is chosen for its
Envisioning a National e-Medicine Network Architecture in a Developing Country
ubiquity. Using Web-based technology constitutes not only a network that can be used universally, but the technology also supports system-independent platforms, thus providing access to many different computer systems at client sites (http://java.sun. com/products/jsp). Key requirements in these client sites are simply the availability of Web browser software and network connectivity. For a secured network, password protected system ensures that user login is needed to access capabilities of the system. In addition, user types are defined so that there will be a role-based access to database and system functions in BEMS. To ensure compatibility with most legacy systems, a relational database is advocated for storing user and patient information. Beyond e-mails, this approach allows the users to mobilize structured information exchange among the communicating health providers.
Major Features of BeMs Basically the BEMS prototype may be conceptualized as a database-driven Web site with the following main features and functions: (a) providing user management services where administrator can register users, assigning username and password, and defining user type, as well as searching and editing user information; (b) providing patient management services where health providers can register patients, search patients and view patient information on a Rolodex-like interface, as and when necessary; (c) providing, on the one hand, referral systems where physicians can write referral messages to a particular department and hospital, and, on the other, a system whereby a physician can retrieve and study the list of referrals forwarded to the department s/he is working in and allowing the physician to write feedback instantaneously after examining the referral message and patient information; (d) providing a system by which physicians can request and schedule lab test at any hospital laboratories so that patients can get tested in the clinic/hospital
they are being treated; and (e) providing a list of lab test requests to laboratory technicians and allowing them to input lab test results.
BeMs architecture The BEMS architecture is built on three-tiered, client-server architecture. The first layer is where the client machines run Web-browser software. This layer is used to display the user interface (Web pages) of the system and send secure HTTP request to the Web server in the second layer. Along with the Web server, application server resides in the second layer. This application server manages the clinical business logic. The bottom layer contains the persistent data of the system. All data of patients, physicians and other communicated messages will be stored and maintained in this third layer. This layer runs the database management system (DBMS) software. Put simply, BEMS functions as a Web-based application connected to a Web server to provide all the interfaces of the system and that of a database server to manage all the knowledge and information elements stored in the system. Figure 4 charts the BEMS system architecture. The BEMS prototype is constructed with a combination of open source products and freely available software components. The Web server suggested is the Apache Jakarta’s Tomcat Web server (http://jakarta.apache.org) with the func-
Figure 4. BEMS prototype architecture Web Browser HTTP
Web Server
Application Server
Database
Mysql:JDBC
45
Envisioning a National e-Medicine Network Architecture in a Developing Country
tionalities as well as the mandated business rules programmed in Java (Haile-Mariam, 2002). Java Server Pages (JSP) is used to capture the user interface and the text of Web pages (http://java. sun.com/products/jsp; http://www.coreservlets. com). Some scripting is included on the Web pages in JavaScript. JSP has a capability to import java classes and run them from the Web pages when the pages are downloaded to the client machine (http:// www.coreservlets.com). Unlike other server side languages such as Active Server Pages (ASP), JSP makes the system platform independent. It also allows users to take advantage of the full power of java programming language which overcome some of the limitation of other scripting languages such as PHP (http://www.coreservlets.com). The database conceptualized is the open source MySql to back up the database driven application. MySql works on many different operating system platforms and is known for its speed of data retrieval (http://www.mysql.com). It provides Application Program Interfaces (API) for many programming languages including Java. Passwords are secure because all password traffic is encrypted when connecting to the MySql server. For database connectivity, we use mm.mysql driver, which is a Java Database Connectivity (JDBC) driver, from MySQL AB, implemented in 100% native Java (http://www.mysql.com/ products/connector-j).
database design Issues BEMS needs to keep track of information about patient and related medical records, user’s information, and messages for both medical referral requests and feedbacks. A well-designed minimal database is needed to manage this information. A relational database model is selected to store the persistent data of the system, as it could be easier to manage, and provides better management for complex query of such data (Amenssisa & Dabi, 2003). This database is expected to maintain and
46
manipulate basic entities such as users, patients, and medical records. Each component of the medical record of a patient is an aggregation of different types of data, which are stored in the database. In the traditional paper-based system, the medical record of a patient is identified by an Out Patient Card (OPCard) Number, which is usually called patient record number. OPCard is a four-page hard-paper card, which contains patient’s generic information, such as name, sex, age, address on the first page and a table of two columns for date and clinical note so as to record chronologically the compilation of health providers’ notes. All other components such as laboratory test results and x-ray reports, among other pieces of information, are stored inside the hard-paper card referenced by the card number or name of the patient. The lab test results may contain zero or more test request forms along with the results for Urine, Parasitology, Blood Chemistry, Hematology, Serology, Bacteriology, Fine Needle Aspiration Cytology and Biopsy. When a patient is admitted to the hospital, admission and social services information is stored. The admission data include identification information and name and address of next-of-kin, marital status, and number of siblings (children) information, besides occupational information and other demographics, as and when provided by the patient. Subsequently, follow-up data such as vital sign measurements, fluid balance information and other measures will be collected and recorded. Order sheet, which contains a list of treatments to be ordered following admission, is also part of the inpatient medical record. In addition to these, information about the hospitals, departments and laboratories are also stored and captured in respective entities. To minimize connectivity cost and increase system performance, a distributed database is recommended. Horizontal partitioning that splits tables along rows, based on the location of patient and healthcare facility, is seen to be an ideal choice in the e-medicine application that tries to create
Envisioning a National e-Medicine Network Architecture in a Developing Country
nationwide connectivity. Finally, to use the database, transparent data access schemes must be defined for applications that run over the network.
Figure 5a. Administrator’s main page within BEMS
BeMs Interfaces In this part of the discussion, the design of various BEMS interfaces is presented. BEMS is accessed when opening the initial Web page where user authentication is first performed. The initial page contains a typical login screen for specifying username with authenticated password. There is no need of menu or different buttons to be submitted based on the user types. Since the user types are defined in the database when the user registered, the page corresponding to the specific user type will automatically be opened upon successful login. Currently, administrators, physicians, and lab technician user types are defined and all of these user types will have their own main pages as described below.
Figure 5b. User registration page
Administrator’s Main Page The administrator’s main page is used for managing users. The functionalities accessible from this page include: register new user and search user by a combination of name, father’s name, and user name. Figure 5a shows the administrator’s main page whereas Figure 5b depicts the user registration page. The other function provided to administrators is the search user function. It is possible to search users by keying in any combinations of name, including father’s name and/or user name. Note that username is treated as “unique” in the user table with the search result being quickly displayed. Although not shown here, the full name of the search result is captured as a link. This link then leads to a page containing the relevant user information from which the administrator can edit a particular user.
Physician’s Main Page The physician’s main page contains a button to open the patient manager page and has the capability of retrieving a list of referrals forwarded to the department where the physician is located. To open a new page, the physician is free to click on the manage patients button or link to one of the referrals. Figure 6a illustrates that the patient, Ato Andualem Lemma, is one of Dr. Aman’s referrals. If Dr. Aman chooses to treat this patient in the hospital where he is privileged, he can simply
47
Envisioning a National e-Medicine Network Architecture in a Developing Country
Figure 6a. Physician’s main page within BEMS
Figure 6b. Patient manager page within BEMS
open the patient manager page by clicking on the manage patients button. Figure 6b shows that the patient manager page has two options, that is, physicians can “register new patient” or “search patient” for those who were previously registered. When selecting “register new patient,” a patient registration form, similar to the user registration page, will be opened. In contrast, if the physician wants to look for a patient, s/he can input one of the search criteria such as name or record number of the patient, resulting in the display of a matching record number or name and the hospital where the patient was first
4