The Riyadh Intensive Care Program applied to a mortality analysis of a teaching hospital intensive care unit

A computerised system of prediction of death using the Riyadh Intensive Care Program was applied retrospectively over a 17-month period to data collected prospectively on 1155 patients admitted to our intensive care unit. Variables which enable organ failure scores to be generated were recorded daily to make these predictions. Consultant medical opinion predicted that outcome was hopeless in 55% (115/209) of the patients who died. The predictive power of the computer demonstrated a sensitivity of 14.8% and a specificity of 99.8%. It is possible that the occurrence of three false predictions of death in the latter part of the series may have been related to a change in our antibiotic policy. We would be unhappy to recommend the general use of a computerised program for prediction of death without careful explanation of its significance and dangers.