We agree with Dr Redelmeier that it is the extreme estimates of mortality risk that may have the greatest impact on the clinical management of individual patients. This was one of the reasons for presenting our sensitivity and specificity findings using a 90% decision criterion. However, there are other potential uses for predictive models in which evaluating intermediate mortality risk is very important, for example, incorporating severity scores or mortality risk estimates into prospective payment programs, stratifying patients in clinical trials, or comparing patient outcomes following a given surgical procedure. ROC curve analysis allows an overall assessment and comparison of predictive accuracy, which was the purpose of our study. Nevertheless, using a 90% decision threshold, we demonstrated that the specificity for predicting death approached 100% for physicians, nurses, and the APACHE II model. Clinical predictions over 90% and below 10% comprised approximately 35% of the total predictions.Predictive
Kruse JA, Carlson RW. Who Will Die in the ICU? APACHE II, ROC Curve Analysis, and, of Course, Cleone-Reply. JAMA. 1989;261(9):1279–1280. doi:10.1001/jama.1989.03420090043025
Customize your JAMA Network experience by selecting one or more topics from the list below.