Prospective application of an expert system for the medical history of joint pain.

An expert system with 60 questions about medical history was developed for 32 rheumatologic diseases: 358 outpatients with joint complaints have been examined. The final diagnosis (result of symptoms, signs, and findings) was compared with the computer diagnoses and with the independently assumed diagnoses of the physician. The only source of information available to the physician was the medical history. Misinterpretation of the computer diagnoses occurred in 25.6% of cases compared with 21.5% of the physician. The final clinical diagnosis remained uncertain in 32.6% of cases. The error frequency of the expert system was influenced by the underlying disease, the certainty of the assumed diagnosis by the physician, the user experience in rheumatology, the number of questions asked, and the time of application before or after the doctor-patient contact. Of the errors 44% were produced because of information deficits of the computer using standardized questions. The information of the physician in the diagnostic process is quite different to that of the computer.