Clinician Versus Computer in the Choice of 11 Differential Diagnoses of Jaundice Based on Formalised Data

Abstract

The information on 20 jaundiced patients available within 48 hours of their admission to general hospitals in the London area was presented, firstly to a computer-assisted model for the diagnosis of jaundice, and secondly to each of 5 clinicians. When the diagnoses produced by each were compared with the final diagnoses on each patient the computer was correct in 70% of cases and the clinicians in between 45 and 65%. When second and third choices were taken into account the clinicians’ scores rose more than that of the computer. When the clinicians’ choices of diagnosis were compared with those of the computer, irrespective of the accuracy of either, a good correlation with the computer was obtained by 2 clinicians and a poorer one by the other 3. When the items of information most helpful to the clinicians in reaching their diagnoses were compared with those items most helpful to the computer, the correlation between clinician and computer was poor in all cases.