Stead, William W. MD; Searle, John R. PhD; Fessler, Henry E. MD; Smith, Jack W. MD, PhD; Shortliffe, Edward H. MD, PhD
Biomedical informatics is the interdisciplinary scientific field that studies and pursues the effective use of data, information, and knowledge for scientific inquiry, problem solving, decision making, and communication. Although the field dates to the 1950s,1 until relatively recently it lurked largely below the radar screen of academic medicine because deans of medical schools saw it as a disciplinary priority of other schools such as engineering or computer science. Several internal and external factors have altered this landscape. These factors include efforts to increase adoption of electronic health records (EHRs), the growth of outcomes research and genomics (both of which required new informatics methods to manage and interpret massive data sets), and the Clinical and Translational Science Award program.2 At this juncture, the explosive growth of biomedical complexity calls for a shift in the paradigm of medical decision making—from a focus on the power of an individual brain to the collective power of systems of brains. This shift alters professional roles and requires informatics and information technology (IT) infrastructure. This shift changes what individual professionals need to know and how they will learn. We illustrate this shift with a vignette; summarize the evolving understanding of both beneficial and deleterious effects of informatics-rich environments on learning, clinical care, and research; provide a framework of core informatics competencies for health professionals of the future; and conclude with four broad steps for faculty development.
Shifting the Paradigm From Individual Brains to Systems of Brains
Imagine a typical student or resident beginning a shift in a busy emergency department. After a quick sign-out, he grabs the chart of a patient in his perceived order of acuity. He collects information, including a history, exam findings, and the results of laboratory tests. Based on his knowledge and cognitive skill, he assembles a differential diagnosis and makes a plan. He reviews his findings and conclusions with a more senior physician, and they compare their conclusions to arrive together at the best course of action. If time allows, the trainee may consult a set of guidelines or the medical literature, or perhaps he will add a topic to his list to read about later. Thus, the wisdom of a more experienced individual transfers to a less experienced one.
Today's medical education processes and curricula lead to the development of individual experts. Each expert has a base of current medical knowledge, understands scientific methods, is prepared to act on what he or she knows, can interpret new literature, and learns from ongoing medical practice. Learning is individual. Each professional seeks to be the best expert at caring for the cases he or she encounters. As an expert, he or she reasons by recognizing patterns, much the way one recognizes a constellation in the night sky. In addition, this expert works with imprecise data that often only loosely correlate with a biological state (e.g., serum creatinine reflects renal function, but also varies with muscle mass, metabolic state, and other factors). He or she makes life-altering decisions in the face of uncertainty. This practice of depending on individual expertise engenders autonomy, self-confidence, and gracious acceptance of variability in practice.3
However, the cognitive capacity of individual brains, which can correlate only about five sets of facts in a single decision,4 limits expert-based medicine. As discovery increases the amount of biomedical knowledge and information available, the expert trades breadth of expertise for depth. She or he specializes. Each expert manages a subset of the patient's problems with a subset of the data. Figure 1 illustrates the volume of data the expert will need to integrate in a clinical decision as biomedicine fully exploits structural genetics, functional genomics, proteomics, and other discoveries.5 Unlike single genetics tests, which strongly associate one mutation with one disease phenotype, the full genetic sequence will provide many low-power associations that in combination change prior probabilities about both an individual's susceptibility to disease states and the likelihood of that individual carrying a specific diagnosis. Specialization is not a viable approach to managing this complexity. Ironically, the future world of individualized medicine will place many decisions beyond the power of individual physicians.
Technology has brought about this cognitive overload, and biomedical informatics and IT will be part of the solution.6 Imagine the student or resident of 2020 starting his shift in the emergency department. A computer will list and color-code patients in their order of acuity (based on informatics algorithms that survey all available clinical information), and patient information will be updated continuously as test results arrive. The trainee will focus on human interactions, communication, and exam skills. When he proposes a differential diagnosis or course of action, he and the senior physician will immediately see how their choices differ from both the computational model of the patient's biology and the recommended, evidence-based next steps. The trainee and the supervising physician will see the dollar cost of alternative choices (with links to published guidelines or relevant literature) and suggestions based on local disease prevalence or antibiograms. Instead of discussing the difference in expertise between two individuals, they can discuss whether the expertise of the medical community at large seems correct for this particular patient, and why this expertise may or may not concur with the decisions of the senior physician. The trainee of 2020 and his supervisor are empowered by the expertise of a network of brains and computers. Furthermore, the student can easily see how his management of similar cases differs from the average of his peers. The emergency department director can see how his hospital compares to others. The researcher can link her outcome data to process-of-care variations among hospitals in the region and develop hypotheses to improve outcomes. The teacher can see whether either a lecture or a small-group session creates better compliance with care guidelines. The network empowers opportunities for individual, institutional, and medical science learning.
This vision shows what might be possible in a future informatics-rich environment. However, to move medicine beyond the limits of individual brains to systems of brains which cope with ever increasing complexity, professional roles will also need to evolve. Expert-based practice, with its focus on the individual's performance, will shift to system-supported practice, with a focus on the system's performance. Teams of people, well-defined processes, and IT will work as a system to produce the desired result. Each omission or error will provide data to guide iterative improvement of the system.7 Discipline-specific curricula will shift to align with a learning process that utilizes the systems approach to care.8 If successful, learning will become an unavoidable outcome as interdisciplinary teams go about their work. The teams will assess individual and team competency against upcoming work, and they will be able to use learning modules, remote access to experts, and simulation to close gaps. Outcomes data will be readily available to assess competency and identify areas for improvement. An informatics foundation will support both medical practice and medical education.7,8
Learning, Clinical Care, and Research in Informatics-Rich Environments
Early computer-based learning environments merely mimicked established teaching methods, using technology to perform the same teaching tasks that had been possible without it; for example, “talking heads” gave lectures, and students accessed drills and practice questions via hard drives or the Internet. The resulting improvements did not justify the effort or cost, and few of the tools survived. More recent approaches have attempted to couple modern informatics and learning. Beaumie and Reeves9 propose that a learner, teaching tool, and activity form a joint learning system. The learner performs higher-level cognitive activities (e.g., executing directions, making decisions, and/or changing approach based on assessment results), while the tool performs lower-level tasks (e.g., visually representing data or providing decision options). By varying the control the computer has over the learner's behavior, the cognitive flexibility it affords adjusts to suit the learner's skill level.
Informatics can support curricula and learning more broadly as well. Denny and colleagues describe a Web-based resource to search curricular content.10 Faculty upload lectures and other material in any standard format (HTML, Microsoft Word etc.,). Algorithms process the documents to recognize biomedical concepts using the National Library of Medicine's compendium of controlled vocabularies11 as a knowledge-source. Faculty and students may then search the resource to find concepts across program and school year boundaries, and they can browse the related material as a virtual course.10 The same processing and browsing infrastructure tags students' notes from the EHR and tabulates their experience with core clinical problems.12
Others, too, have suggested uses for the EHR to link medical practice and medical education. Stead proposes a framework of four tiers through which the EHR and related clinical informatics tools could support learning.13 The processes that both measure variation in resource consumption (e.g., length of inpatient stays, tests performed, medications used) and identify intragroup variation in practice provide the framework's foundation. In the next tier, physicians use data from the EHR to free their cognitive energy to focus on synthesis; the informatics allows them to tie their own practices to their own patient outcomes, converting their open-loop practice into a closed loop with feedback.14 The capacity to detect unexpected events and the availability of decision-support systems (with alerts and reminders, patient-specific information about changes in practice, and links to evidence) together make up the third tier. In the fourth tier, extracts from the EHR and bio-banks combine to support correlation and hypothesis generation. That is, first, measure practice variation; second, provide individual feedback on practice and outcomes; third, improve individual practice by placing knowledge and information at physicians' fingertips; and finally, support advancement of medical science by suggesting testable hypotheses to improve care.
Some have already begun to use the EHR extensively. The Veterans Health Administration (VHA) has provided a national laboratory on the use of EHRs to accelerate learning. The VHA has used system-wide, longitudinal data to quantify the impact of care management and to show that site-to-site variation is more significant than physician-to-physician variation within sites.15 Further, EHRs have the potential to support comprehensive postmarket pharmacovigilence and to accelerate detection of unrecognized adverse drug events.16 EHRs also have the potential to generate clinical hypotheses. Hanauer and colleagues demonstrated the feasibility of using gene-mapping software to identify potential associations among free-text clinical problem statements in their EHR.17 They were able not only to reproduce disease associations that had been confirmed in the literature, but also to detect unknown associations that had biological plausibility. The immense scale of information from linked EHRs could build and validate simulation models to answer clinical questions that cannot be answered directly from EHR data, much the way computer modeling has been used in physiology.18
Beneficial and Deleterious Effects of Informatics-Rich Environments
Without question, integration of informatics into medical care and medical education has potential adverse consequences. Many of these are familiar to those who have experienced a transition to EHRs. Both caring for patients and teaching students are human-to-human interactions, which can be impeded by the interposition of a computer.19 Peled and colleagues debate whether EHRs help or hinder medical education.20 Some potential problems include attendings' direct access to source data, which reduces the need for students to think critically beforehand to synthesize and present the data, and the abilities to copy and paste notes and to access results instantaneously, both of which, first, reduce the need for students to decide what and how to document and, second, threaten the narrative in the medical record. The disruption of doctor–patient or attending–student conversation as everyone focuses on the computer, rather than on one another, presents another potential problem.
We believe the problems lie not in the technology per se, but in its application. Styles of medical documentation will need to be adapted to the electronic format. Faculty and student interactions and processes need to reflect learning goals; for example, requiring a cogent presentation without direct use of the computer except to answer questions that come up after the presentation.
Research from cognitive science can guide improved application of IT. Patel and colleagues review the relationship between cognition and decision support. They show that IT does not merely support or enhance the decision process, but fundamentally transforms it, having an enduring effect on cognitive capacity.21 They draw on work by Eddy22 and Gigerenzer22 to show that the very structure of the presentation of data to a physician affects the physician's judgment. That is, the way data are shown or visualized strongly influences the way physicians interpret the data, the decisions they make based on the data, and even future decisions they will make based on similar data they later encounter. This effect of presentation is, of course, familiar to anyone who has worked in advertising. Data visualization must point the audience to the correct conclusion.
Informatics Competencies for Future Health Professionals
In 2003 the Institute of Medicine's Health Professions Education Summit identified utilizing informatics as one of the five domains of core competency for health care professionals.24 Informatics also underpins the other four competency domains, which are to provide patient-centered care, to work in interdisciplinary teams, to employ evidence-based practice, and to apply quality improvement. Although most health professionals use IT daily in their work, few know how to adapt their roles and work processes to incorporate IT for the greatest benefit.
The optimal use of the EHR at the bedside does require some special skills, but physicians can certainly learn these. Morrow and colleagues reported a pilot study to establish the feasibility and practicality of teaching EHR-specific communication skills to early first-year medical students: students first learned the mechanics of documenting patient histories with the EHR.25 The researchers then randomized students into two groups: one group received training in EHR-specific communication skills, and the other did not. Standardized patients compared the skills of the two groups. The intervention group performed better on 6 of 10 EHR-specific skills (e.g., positioning relative to the patient and engaging the patient in use of the computer) and performed similarly in 10 of 11 general communication skills (e.g., introducing self and establishing rapport).
In 2008, the American Medical Informatics Association and the Association of Academic Health Centers convened representatives of 14 health professions in a DesignShop at the Vanderbilt Center for Better Health to develop a common informatics competency framework.26 The group identified a set of competencies needed for clinical work in any of the health professions. Table 1 regroups these competencies to highlight the informatics aspects of the six core competencies of the Accreditation Council for Graduate Medical Education. All of the competency goals in the field of biomedical informatics have correlates in the familiar competencies of graduate medical education. Informatics subserves, rather than subverts, the goals of medical education.
Broad Steps for Faculty Development
We believe biomedical informatics has become a foundational science essential to the improvement of health and the delivery of high-quality patient care. The future expert will not be able to know everything necessary to make the best decisions. An understanding of biomedical informatics will assist future physicians as they decide not only what they need to know but also how to find what they need to know in the information infrastructure. Today's flawed IT overloads the physician with data. Biomedical informatics research seeks to develop computational models that simplify presentation of complex biological and social interactions. Sufficient understanding of the strength and weaknesses of the computational techniques will help physicians decide when to accept, and when to override, the results of their informatics tools. These coming realities may challenge or even threaten faculty who have trained and practiced in a system that reserves its highest praise for individual expertise; nonetheless, biomedical informatics will help them adapt their practice, so training and preparing faculty will make the evolution smoother.
Investment and support by medical schools can hasten the evolution to informatics-enhanced patient care and learning. We suggest the following four broad steps: (1) create academic units in biomedical informatics; (2) adapt the IT infrastructure of academic health centers (AHCs) into testing laboratories; (3) introduce medical educators to biomedical informatics sufficiently for them to model its use; and (4) retrain AHC faculty to lead the transformation of health care based on a new systems approach enabled by biomedical informatics. We explain these steps further in the next several paragraphs. However, we recognize that the details of implementation will vary widely among institutions and will require local experimentation.
1. Create academic units in biomedical informatics that have a seat at both the academic and operational tables.
A critical mass of faculty who understand biomedical informatics can provide the nucleus to teach noninformatics faculty and students how to use informatics techniques and tools in their work. If included at the academic decision making table, biomedical informatics faculty can help school and departmental leaders balance resources between faculty who advance knowledge within a discipline and faculty who advance approaches to working with knowledge across disciplines. If included at the operational decision making table, biomedical informatics faculty can help the school and its health system avoid the problems that often plague a new administrative process and database.27 Biomedical informatics faculty can ensure that clinical information systems are fully integrated with and support the education and research missions of the medical school. Equally important, they will be positioned to structure affiliation agreements that give the school access to the data to support quality improvement and learning across a geographic area as envisioned in health reform efforts such as the Association of American Medical Colleges' Health Innovation Zones. Finally, their peers at the table can hold biomedical informatics faculty accountable for aligning their plans with the institution's missions, and vice versa.
2. Adapt the IT infrastructure of AHCs into testing laboratories to evaluate and utilize emerging biomedical informatics techniques for data aggregation, systems analysis, and visualization support.
Today's clinical information systems automate work within health care facilities. Although these systems currently get data to physicians, they rarely help the physician with cognitive tasks as information systems in the future will28 (e.g., the student in 2020 uses the computer to help him visualize what is most important and evaluate alternatives systematically). AHCs with informatics units can supplement simple automation by using computational techniques such as connectivity, social networks that connect people to one another and to systems; statistical decision support, in which multiple weak signals contribute to robust answers; and data mining, through which relationships are discovered among data from diverse highly dimensional data sets. These approaches allow the clinical information system infrastructure to serve as a laboratory where physicians, physician educators, and physicians-in-training may evaluate and apply informatics “interventions.”29
3. Introduce medical educators to biomedical informatics sufficiently for them to model its clinical and research uses, to modernize curricula appropriately, and to evaluate trainees and teaching methods.
Today's approaches to teaching and clinical work evolved in an era when the physician's head or pocket had to carry all the information he or she needed to make a medical or practice decision. With ubiquitous access to information, physicians still need to know the concepts that allow them to make decisions, but they can now retrieve facts as needed. This generational shift in the approach to information is familiar to faculty who have rounded with housestaff bearing smartphones or laptops. This shift changes curricular priorities as well as learning and evaluation strategies. One strategy, for example, is using clinical outcomes to measure the effectiveness of a learning intervention. Faculty, bringing their experience with biomedicine, and students, bringing their familiarity with the digital age, can partner to help each other through the change.
4. Retrain faculty in AHCs to lead the transformation to health care that incorporates systems approaches enabled by biomedical informatics.
We encourage faculty to embrace new roles so that they use biomedical informatics to its fullest extent. New roles allow collaboration among physicians and within the information infrastructure. Academic medicine faces a sea change equal to the one that occurred a century ago when science became the basis for medicine. The shift in focus, from individual brains to systems of brains, challenges professional norms and roles. Prepared faculty can lead this transformation of practice by learning and teaching within the new systems of care and education. The first step is to build awareness within the leadership and faculty of academic medicine that the change is unavoidable. Like many changes, this one offers opportunities—in this case, to develop new roles that are productive and satisfying. Growing dissatisfaction with current roles, under the dual pressures of cognitive overload and payment reform, may ignite the burning platform and motivate the leap needed to reach a sustainable next generation model for the profession.
The authors would like to thank Pauline T. Alexander, MLIS, and Tracy C. Shields, MSIS, who performed a systematic search to identify the most significant publications related to four questions: (1) How does learning in an informatics-rich world change what people learn? (2) What are the differences between the human brain and the computer? (3) What are the best examples of the electronic medical record as a learning resource? and (4) How is informatics being used to discover connections?
Participants in the American Medical Informatics Association/Association of Academic Health Centers 2008 informatics curriculum DesignShop were Frederick Albright, PhD, Brownell Anderson, Elizabeth Bishop, Connie Delaney, PhD, Don E. Detmer, MD, Elise Eisenberg, DDS, Rick Forsman, MSLS, Cynthia Gadd, PhD, Karen Greenwood, Kevin Johnson, MD, Robert K. Lepac, PhD, Chellappa Kumar, PhD, Michael Lardiere, MSW, Vimla Patel, PhD, Lydia Reed, MBA, Rebecca Reynolds, PhD, Steve Wartman, MD, PhD, and Jeffrey J. Williamson.
Jim Owens, MD, Stephan Spann, MD and Stephen Weinberger, MD, worked with the authors to refine the broad steps for faculty development during the 2010 Baylor Faculty Development Workshop.
This work was supported by a writing conference funded by the Medallion Fund and the Josiah Macy, Jr. Foundation. The conference was entitled “A 2020 Vision of Faculty Development Across the Medical Education Continuum” and was held at Baylor College of Medicine on February 26–28, 2010.
Dr. Stead is director of HealthStream and co-inventor of two medical record products—one licensed to McKesson, and one licensed to Informatics Corporation of America—from which he receives royalties through Vanderbilt University. Dr. Shortliffe is author of a textbook on biomedical informatics, for which he receives royalties from Springer.
Presented in part at the national symposium on faculty development in medical education reform mentioned above.
1 Ledley R, Lusted L. Reasoning foundations of medical diagnosis; symbolic logic, probability, and value theory aid our understanding of how physicians reason. Science. 1959;130:9–21.
2 Berstam EV, Hersh WR, Johnson SB, et al. Synergies and distinctions between computational disciplines in biomedical research: Perspective from the Clinical and Translational Science Award programs. Acad Med. 2009;84:964–970.
3 Stead WW. Health care culture in the United States. In: Grossman C, Goolsby A, Olsen L, McGinnis JM, eds. Engineering a learning healthcare system: A look at the future. 2008 Roundtable on Value & Science-Driven Health Care. Washington, DC: The National Academies Press; 2010 (in press): Chapter 3.
4 Miller GA. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychol Rev. 1956;63:81–97.
5 Stead WW. Contending with the changes (Chapter 4). In: McClellan MB, McGinnis JM, Nabel EG, Olsen LM, eds. Evidence-Based Medicine and the Changing Nature of Health Care. 2007 IOM Annual Meeting Summary. Washington, DC: National Academies Press;2008:18–20.
6 Srinivasan M, Keenan CR, Yager J. Visualizing the future: Technology competency development in clinical medicine, and implications for medical education. Academic Psychiatry. 2006;30:480–490.
7 Stead WW, Starmer JM. Beyond expert-based practice. In: McClellan MB, McGinnis JM, Nabel EG, Olsen LM, eds. Evidence-Based Medicine and the Changing Nature of Health Care. 2007 IOM Annual Meeting Summary. Washington, DC: National Academies Press;2008:94–105.
8 Miller BM, Moore DE, Stead WW, Balser JR. Beyond Flexner: A model for continuous learning in the health professions. Acad Med. 2010;85:266–372.
9 Beaumie K, Reeves TC. Reframing research on learning with technology: In search of the meaning of cognitive tools. Instructional Science. 2007;35:207–256.
10 Denny JC, Smithers JD, Armstrong B, Spickard A 3rd. Where do we teach what? Finding broad concepts in the medical school curriculum. J Gen Intern Med. 2005;20:943–946.
11 Humphreys BL, Lindberg DAB, Schoolman HM, Barnett GO. The Unified Medical Language System: An informatics research collaboration. JAMIA. 1998;5:1–11.
12 Denny JC, Bastarache L, Sastre EA, Spickard A 3rd
. Tracking medical students' clinical experiences using natural language processing. J Biomed Inform. 2009;42:781–789.
13 Stead WW. Clinicians and the electronic health record as a learning tool. In Olsen L, Aisner D, McGinnnis JM, eds. The Learning Healthcare System. Workshop Summary: IOM Roundtable on Evidence-based Medicine. Washington, DC: National Academies Press;2007:268–274.
14 Stead WW, Patel N, Starmer JM. Closing the loop in practice to assure the desired performance. Trans Am Clin Climatol Assoc. 2008;119:185–195.
15 Kupersmith J, Francis J, Kerr E, et al. Advancing evidence-based care for diabetes: Lessons from the Veterans Health Administration. Health Aff. (Millwood) 2007;26:w156–w168.
16 Wang X, Hripcsak G, Markatou M, Friedman C. Active computerized pharmacovigilance using natural language processing, statistics, and electronic health records: A feasibility study. JAMIA 2009;16:328–337.
17 Hanauer DA, Rhodes DR, Chinnaiyan AM. Exploring clinical associations using ‘-omnics' based enrichment analyses. PLoS ONE 2009;4:e5203.
18 Eddy DM. Linking electronic medical records to large-scale simulation models: Can we put rapid learning on turbo? Health Aff. (Millwood) 2007;26:w125–w136.
19 Berg M, Aarts J, van der Lei J. ICT in health care: Sociotechnical approaches. Methods Inf Med. 2003;42:297–301.
20 Peled JU, Sagher O, Morrow JB, Dobbie AE. Do electronic health records help or hinder medical education? PLoS Medicine. 2009;6:e1000069.
21 Patel VL, Kaufman DR, Arocha JF. Emerging paradigms of cognition in medical decision-making. J Biomed Inform. 2002;35:52–75.
22 Eddy DM. Probabilistic reasoning in clinical medicine: Problems and opportunities. In: Kahneman D, Slovic P, Tversky A, eds. Judgment under uncertainty: Heuristics and biases. Cambridge: Cambridge University Press; 1982.
23 Gigerenzer G. Adaptive thinking: Rationality in the real world. New York: Oxford University Press; 2000.
24 Greiner AC, Knebel E, eds, Committee on the Health Professions Education Summit. Health Professions Education: A bridge to quality (Quality Chasm Series). Institute of Medicine of the National Academies. Washington: The National Academies Press, 2003.
25 Morrow JB, Dobbie AE, Jenkins C, Long R, Mihalic A, Wagner J. First-year medical students can demonstrate EHR-specific communication skills: A control-group study. Fam Med. 2009;41:28–33.
26 Designing the informatics component of an IOM chasm health professions core competencies curriculum. Vanderbilt Center for Better Health. American Medical Informatics Association (AMIA) Design Session, AMIA Academic Forum, and American College of Medical Informatics. Vanderbilt University, Nashville, TN, 2008. Available at: https://www.mc.vanderbilt.edu/vcbh/ds/081001amia/index.html
. User id: 081001_amia, Password: grant.
27 Frisse ME, Musen MA, Slack WV, Stead WW. How should we organize to do informatics? Report of the American College of Medical Informatics Debate at the 1997 American Medical Informatics Association Fall Symposium. JAMIA. 1998;5:293–304.
28 Stead WW and Lin, HS, eds. Committee on Engaging the Computer Science Research Community in Health Care Informatics. National Research Council. Computational technology for effective health care: Immediate steps and strategic directions. Washington D.C.: National Academies Press; 2009.
29 Stead WW. Electronic Health Records. In: Rouse WB, Cortese DA, eds. Engineering The System Of Healthcare Delivery: Volume 153 Studies in Health Technology and Informatics. Amsterdam: IOS Press;2010:119–143.