Abstract
This issue of the Journal contains a comparison of four commercially available computer programs designed to diagnose diseases that internists might encounter1. All the programs performed equally well. When they were presented with information on known diagnoses from real patients who had been referred to a consultant for diagnostic assistance, the correct diagnosis (or diagnoses) appeared on the lists generated by the programs only half to about three quarters of the time, and relevant diagnoses, as judged by a panel of experts, were generated only about a third of the time.What are we to make of these results? . . .

This publication has 10 references indexed in Scilit: