Medical Education (Introduction to Medical Informatics) (http://www.cpmc.columbia.edu/edu/textbook) LAST REVIEWED: 14 November 1996 classical medical education (4 years) 1. basic sciences - biochemistry, anatomy, physiology (how the body works) 2. disease processes - pathology, pathophysiology (how the body fails and what to do about it) 3. clinical rotations - medicine, surgery, specialties (real patient contact) 4. focused clinical rotation - narrow down areas 5. continuing medical education (CME) so that practitioners can remain current two kinds of knowledge (as in AI) factual knowledge - eg, anatomy of the heart problem solving knowledge - eg, diagnosis recent trends more facts more diagnostic and therapeutic procedures more costs more complex cases in academic centers more specialization first two years lecture format memorization of facts evaluation based on recall second two years interaction with real patients (best way to learn) limited number of cases of each type skewed towards complex cases some liability problems potential problems with traditional approach lack of emphasis on life-long learning skills does not match independent study of usual lib arts major emphasis on rote recall too much biomed knowledge to handle this way resistance to change departmental organization of curriculum (w/o coordination) shift to ambulatory care: able to observe patients less than in hospital the new stategy is to: focus more on problem solving skills earlier learn techniques for acquiring information (but taken too far, this approach is just as bad) beyond this => life-long CME (continuing med educ) 1984 report by General Professional Education of Physicians include information science and computers in curriculum computer-assisted instruction (CAI) or learning interactive session - force involvement immediate feedback on performance, privacy individualized to the student (pace, weak areas) objective testing of skills = more reproducible entertaining can experiment w/o danger to patients use of CAI in clinical medicine focus on diagnosis and therapy rather than on facts encourage experimentation and exploration greater scope (see outpatient cases) prototypic cases that are not complex includes rare cases better measure of competency than multiple choice test but no data to show that CAI is superior to other forms except early data (Friedman, 1978) showed that it may improve test ordering and treatment selection examples 1967 - OSU Tutorial Evaluation System (TES) standard questions, but immediate evaluation corrective feedback 1970 - MGH case simulation work up (diagnosis) and therapy decide which info to collect; interpret data; choose tx; feedback loop 1970 UI Computer-Aided Simulation of Clinical Encounter natural language interface to Dx & treat computer 1974 survey 362 different programs work to share a program = work to rewrite it 1972 NLM network (access via terminal) for OSU/MGH/UI programs advent of PCs reduce cost, improve transfer 1987 Computer-Based Examination (CBX) NBME part III supposed to test diagnosis, therapy (previously done with rubbing out choices) can ask for history, PE, tests, therapy, consults thousands of images (X-ray, pathology, ...) patient's condition evolves over time 1986 TIME (Tech Innovation in Med Educ), Lister Hill videodisc simulation of patient interaction freeze scene -> student takes an action voice input w/ specified set of inquiries and commands patient "thoughts" - increase awareness of emotion decision points - cases are not deterministic 1989 HeartLab (Bergeron, BWH) heart auscultation gives sound and visual diagram of it diagrams to show where to listen can use for case analysis; review; lab (i.e., create your own murmur by varying physical parameters) 1986 GUIDON applied teaching rules to the MYCIN knowledge base can plug in other EMYCIN-based systems (PUFF) teaching rules: determine what technique [e.g., present a conclusion; give a summary; ask for hypothesis] to apply MYCIN knows about infectious disease but not how to teach it types of CAI drill-and-practice - defined set of questions branching questions - next question depends on previous answers (same topic, or move on) simulations - learn by doing static - predefined case, multiple paths through dynamic - case changes based on student actions (eg, therapy could cause complications) follow course of patient over time can practice without hurting patient uses physiologic models, production rules, ... response choices constrained - pick from list of answers cues user to possible answers detract from realism but program is easier to code (fewer branches) and faster to use intermediate - one menu of answers to questions program specific, not situation-specific unconstrained - natural language harder to write get discouraged when answers not recognized structure structured - steer inquiry force focus on desired topic eg, drill-and-practice unstructured - free to explore different areas can waste time or miss main point feedback why answer is right or wrong summaries references help mixed-initiative systems: allow independent choice, but constrain if obvious errors; two types: (a) coaching - intervene only for severe mistakes (b) tutoring - aggressive guidance difficult: need to know medicine and student audiovisual support (eg, videodisc) more realism (eg, ER simulation) somewhat expensive, slow on network authoring systems 50-100 hours to generate 1 hour of instruction authoring system lets expert focus on content typical structure: frame-based, slot-filler reasons for lack of success previously primitive interfaces difficulty in replacing a book; a book = cheaper, more portable and easier to underline than a computer screen! future progress is slow need an advocate in each institution credit for developing CAI (tenure) cost of equipment: lack of financial support integration of CAI programs not-invented-here suggestions for success Medline use: computer-based literature access word processing and email rapid retrieval of recorded facts ("pearls"): computer must be more than a simple page turner integration with expert systems simulations that go beyond the book model reproducible assessment integrate w/ computer-based patient record with access to knowledge sources related reading: Barnett GO. Information technology and medical education. JAMIA 1995;2(5):285-91.