Correlates of Diagnostic Accuracy in Patients with Nonspecific Complaints

OBJECTIVE: To investigate diagnostic accuracy in patient histories involving nonspecific complaints and the extent to which characteristics of physicians and structural properties of patient histories are associated with accuracy. METHODS: Six histories of patients presenting to the emergency department (ED) with nonspecific complaints were provided to 112 physicians: 36 ED physicians, 50 internists, and 26 family practitioners. Physicians listed the 3 most likely diagnoses for each history and indicated which cue(s) they considered crucial. Four weeks later, a subset of 20 physicians diagnosed the same 6 histories again. For each history, experts had previously determined the correct diagnoses and the diagnostic cues. RESULTS: Accuracy ranged from 14% to 64% correct diagnoses (correct diagnosis listed as the most likely) and from 29% to 87% correct differential diagnoses (correct diagnosis listed in the differential). Acute care physicians (ED physicians and internists) included the correct diagnosis in the differential in, on average, 3.4 histories, relative to 2.6 for the family practitioners (P = 0.001, d = .75). Diagnostic performance was fairly reliable (r = .61, P < 0.001). Clinical experience was negatively correlated with diagnostic accuracy (r = -.25, P = 0.008). Two structural properties of patient histories-cue consensus and cue substitutability-were significantly associated with diagnostic accuracy, whereas case difficulty was not. Finally, prevalence of diagnosis also proved significantly correlated with accuracy. CONCLUSIONS: Average diagnostic accuracy in cases with nonspecific complaints far exceeds chance performance, and accuracy varies with medical specialty. Analyzing cue properties in patient histories can help shed light on determinants of diagnostic performance and thus suggest ways to enhance physicians' ability to accurately diagnose cases with nonspecific complaints.