Accuracy of diagnoses predicted from a simple patient questionnaire stratified by the duration of general ambulatory training: an observational study

Abstract

PurposeTo compare the diagnostic accuracy of diseases predicted from patient responses to a simple questionnaire completed prior to examination by doctors with different levels of ambulatory training in general medicine.Participants and methodsBefore patient examination, five trained physicians, four short-term-trained residents, and four untrained residents examined patient responses to a simple questionnaire and then indicated, in rank order according to their subjective confidence level, the diseases they predicted. Final diagnosis was subsequently determined from hospital records by mentor physicians 3 months after the first patient visit. Predicted diseases and final diagnoses were codified using the International Classification of Diseases version 10. A “correct” diagnosis was one where the predicted disease matched the final diagnosis code.ResultsA total of 148 patient questionnaires were evaluated. The Herfindahl index was 0.024, indicating a high degree of diversity in final diagnoses. The proportion of correct diagnoses was high in the trained group (96 of 148, 65%; residual analysis, 4.4) and low in the untrained group (56 of 148, 38%; residual analysis, −3.6) (χ2=22.27, P<0.001). In cases of correct diagnosis, the cumulative number of correct diagnoses showed almost no improvement, even when doctors in the three groups predicted ≥4 diseases.ConclusionDoctors who completed ambulatory training in general medicine while treating a diverse range of diseases accurately predicted diagnosis in 65% of cases from limited written information provided by a simple patient questionnaire, which proved useful for diagnosis. The study also suggests that up to three differential diagnoses are appropriate for diagnostic prediction, while ≥4 differential diagnoses barely improved the diagnostic accuracy, regardless of doctors’ competence in general medicine. If doctors can become able to predict the final diagnosis from limited information, the correct diagnostic outcome may improve and save further consultation hours.