Categorizing patients in a forced-choice triad task: the integration of context in patient management.

<h4>Background</h4>Studies of experts' problem-solving abilities have shown that experts can attend to the deep structure of a problem whereas novices attend to the surface structure. Although this effect has been replicated in many domains, there has been little investigation into such effects in medicine in general or patient management in particular.<h4>Methodology/principal findings</h4>We designed a 10-item forced-choice triad task in which subjects chose which one of two hypothetical patients best matched a target patient. The target and its potential matches were related in terms of surface features (e.g., two patients of a similar age and gender) and deep features (e.g., two diabetic patients with similar management strategies: a patient with arthritis and a blind patient would both have difficulty with self-injected insulin). We hypothesized that experts would have greater knowledge of management categories and would be more likely to choose deep matches. We contacted 130 novices (medical students), 11 intermediates (medical residents), and 159 experts (practicing endocrinologists) and 15, 11, and 8 subjects (respectively) completed the task. A linear mixed effects model indicated that novices were less likely to make deep matches than experts (t(68) = -3.63, p =.0006), while intermediates did not differ from experts (t(68) = -0.24, p =.81). We also found that the number of years in practice correlated with performance on diagnostic (r =.39, p =.02), but not management triads (r =.17, p =.34).<h4>Conclusions</h4>We found that experts were more likely than novices to match patients based on deep features, and that this pattern held for both diagnostic and management triads. Further, management and diagnostic triads were equally salient for expert physicians suggesting that physicians recognize and may create management-oriented categories of patients.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.