METHODS: Nine features of diagnostic difficulty were identified from the reasoning literature, a systematic review of the clinical literature, and interviews with Family Physicians. Seven realistic patient scenarios were formulated containing different numbers and combinations of these features. Patients presented in ways frequently encountered in Family Practice. We determined the existing evidence base through systematic literature reviews. We thus identified ‘critical cues' as cues with likelihood ratios >1.5 or <0.67. By statistically modelling the probability estimates of multiple experts in a separate study, we determined the most likely diagnoses and additional ‘critical cues' for the scenarios. 21 Family Medicine residents, 21 Family Physicians with 1-3 years in practice and 42 Family Physicians with ≥10 years in practice diagnosed and managed all seven scenarios, presented on computer. Participants initially saw a detailed patient description and complaint, then requested further information. Their information search, diagnosis and management were recorded.

RESULTS: The correct diagnosis was given in 37% of the cases and the appropriate management decision in 47% of the cases. 95% of inappropriate management was associated with incorrect diagnoses. The number of critical cues requested was a significant predictor of accuracy in 5 scenarios: one additional critical cue increased the odds of obtaining the correct diagnosis by between 1.6 (95% CI 1.32-2.03) and 11 (95% CI 3.55-34.39), depending on the scenario. Experience was another significant predictor of accuracy (χ2=6.95, df=2, p<.05): Family Physicians with ≥10 years in practice were 9% more likely to diagnose accurately than residents (adjusted odds ratio 2.15, 95% CI 1.21-3.82). Residents requested significantly more cues than Family Physicians with ≥10 years in practice (F=7.62, df=2, p<.01). There were no differences in the number of critical cues requested (p=.55).

CONCLUSIONS: Diagnostic accuracy in these difficult scenarios was limited, as expected. Furthermore, diagnostic accuracy determined appropriateness of management. The amount of critical information selected was the strongest predictor of accuracy. The small increase in accuracy with experience was unrelated to critical cue selection. Future research should investigate the role of information interpretation as a possible contributor to differences in accuracy between residents and experienced Family Physicians.