Systematic bias in recording the history in patients with chest pain

Because of its central importance in medical diagnosis, sources of bias in the patient history must be identified. We report here a study of interobserver agreement in chest pain histories. Histories were obtained by physicians, nurse practitioners, and self-administered questionnaires. We used a discriminant rule to classify the histories as being more or less typical of angina pectoris. A sub-group of the subjects underwent coronary arteriography after the histories were obtained. In subjects with positive coronary arteriograms, physicians consistently obtained histories typical of angina pectoris more often than a self-administered questionnaire. There was no significant systematic bias when comparing physician interviews to the questionnaire among subjects with negative arteriograms nor when comparing two physicians' interviews or a nurse practitioner interview to a questionnaire. When compared to physician interviews, nurse practitioner interviews produced histories less typical of angina pectoris. We conclude that there are systematic differences between the histories obtained by physicians, nurse practitioners, and self-administered questionnaires. Questionnaires can produce biased patient histories and should be carefully validated before being used in patient care activities or health care research.