Computer-aided selection of diagnostic tests in jaundiced patients.

A model has been developed for ordering diagnostic tests in jaundiced patients. The system proceeds in two steps: (i) diagnostic hypotheses are calculated for each patient from the results of physical examination and routine biological investigations; (ii) given these hypotheses, the most efficient test (out of 22) for reaching the final diagnosis is selected using four criteria: diagnostic value, risk, financial cost, and time in obtaining the result. This model was tested in 62 patients. In 43 of them (69%), the selected test was sufficient for reaching a diagnostic accuracy of 100%. In this group of patients, a mean of 3.7 (range 1-6) tests per patient was ordered by physicians. In the 19 remaining patients, the selected test was not sufficient for the final diagnosis, thus requiring a multiple choice process. It is suggested that such a system could help physicians to improve the care of patients by more efficient ordering of diagnostic tests.