Journal of Biomedical Informatics 36 2003 61 69 www elsevier com locate yjbin Medical errors as a result of specialization Ahmad Hashem a Michelene T H Chi b and Charles P Friedmanc a Healthcare Industry Solutions Group Microsoft Corporation One Microsoft Way Redmond WA 98052 USA Learning Research and Development Center University of Pittsburgh 3939 O Hara Street Pittsburgh PA 15260 USA Center for Biomedical Informatics University of Pittsburgh Forbes Tower Suite 8084 200 Lothrop Street Pittsburgh PA 15213 2582 USA b c Received 8 July 2003 Abstract Errors in medicine result in over 44 000 preventable deaths annually Some of these errors are made by specialized physicians at the time of diagnosis Building on error frameworks proposed in the literature we tested the experimental hypothesis that physicians within a given specialty have a bias in diagnosing cases outside their own domain as being within that domain Thirty two boardcerti ed physicians from four internal medicine subspecialties worked four patient cases each Verbal protocol analysis and general linear modeling of the numerical data seem to con rm the experimental hypothesis indicating that specialists try to pull cases toward their specialty Specialists generate more diagnostic hypotheses within their domain than outside and assign higher probabilities to diagnoses within that domain 2003 Published by Elsevier Inc Keywords Medical errors Diagnostic errors Cost of expertise 1 Introduction Betsy Lehman a Boston Globe health reporter died from an overdose during chemotherapy Willie King had the wrong leg amputated Ben Kolb was eight years old when he died during surgery due to a drug mixup 1 2 And these are just the tip of the iceberg of medical errors 1 2 One important type of medical error occurs at the time of diagnosis The popular press cites cases such as that of Dr Franklin K Yee whose abdominal pain was diagnosed as viral gastroenteritis by a gastroenterologist caused him to be admitted to a coronary care unit by a cardiologist was suspected by a nephrologist to be the result of kidney stones and eventually was found on abdominal surgery to be the result of a ruptured appendix 3 This phenomenon of di erent specialists projecting their specialties on a patient has not been studied systematically In the present paper the role of medical specialization in inducing biases that may un Corresponding author Fax 1 425 936 7329 E mail address ahmad hashem net A Hashem 1532 0464 see front matter 2003 Published by Elsevier Inc doi 10 1016 S1532 0464 03 00057 1 derlie some diagnostic medical errors is investigated empirically 2 The cost of expertise It may seem strange to talk about the costs of being an expert but there is increasing experimental evidence that the bene ts of expertise are not without costs The costs of expertise can be divided into two main categories Those related to accuracy of recall and those related to in exibility 2 1 Accuracy of recall Experts may outperform novices in recalling the details of a problem or text But when the domain knowledge of experts cannot be utilized experts tend to underperform novices For example in a study on recall of random chessboard positions the performance of chess experts was slightly worse than that of novices 4 Similarly in a study on memory for baseball texts participants with high baseball knowledge recalled signi cantly less baseball irrelevant propositions from a 62 A Hashem et al Journal of Biomedical Informatics 36 2003 61 69 text passage describing part of a baseball game than did participants with low baseball knowledge 5 2 2 Reduced exibility Some of the early work on the costs of expertise was done by Ggestalt psychologists 6 8 Luchins argued that experience produced an Einstellung or mental set that limited the search space of subsequent problem solving More recent studies have shown similar results For example Shi rin and Schneider 9 reported that people trained for several thousand trials to detect visual targets among distractors in a consistent mapping condition were at a serious disadvantage when the target and distractor sets were reversed Wiley 10 showed that problem solvers with a large amount of domain knowledge are con ned by their knowledge to one area of the search space a condition that is e cient if the solution happens to fall in this area but which back res if it does not 3 Characteristics of medical expertise There is no reason to suspect that expertise in medicine is di erent from expertise in any other domain In fact research results have emphasized similarities e g 11 15 The general characteristics of expertise are summarized elsewhere 16 17 and will not be repeated here Some characteristics of expertise in the medical domain that bear relevance to the present study will be discussed 3 1 Diagnostic reasoning mode According to Patel and Groen 14 expert clinicians confronted with routine cases use a data driven forward approach in which diagnoses are generated from data by applying a small set of if then production rules without generating intermediate hypotheses and evaluating them Less expert clinicians such as medical students or residents on the other hand tend to use a hypothesis driven backward approach in which reasoning occurs backwards from a hypothesis in an attempt to nd data that elucidates it 18 This assertion is consistent with ndings in other domains such as physics 19 and mathematics 20 and with the general notion that in routine situations experts tend to use highly speci c problem solving structures 21 3 2 Experts working outside their domain Cognitive literature holds abundant evidence that experts excel only at their domain of expertise see 17 for an overview Some of the very few studies that have examined the performance of subspecialist physicians on problems outside their area of specialty were done by Patel and colleagues 22 23 Cardiologists and endocrinologists were asked to read cardiology and endocrinology cases and to think aloud as they were reading them or to recall case information and explain the underlying pathophysiology The general nding is that experts working within their subdomain tend to use forward strategy more and to rely on pathophysiological knowledge less than experts working outside their subdomain No signi cant di erence in diagnostic accuracy was found but it is hard to make any meaningful claim on this issue due to the small sample sizes typically less than 10 participants total working on