Forecasting survival in the medical intensive care unit: a comparison of clinical prognoses with formal estimates

Physicians often need to make prognostic judgments. In the present study, the accuracy was explored of survival estimates for patients in the Medical Intensive Care Unit (MICU). Estimates were made by physicians and nurses several times during each patient's stay in the MICU and were compared to those of the APACHE II scale, a widely used quantitative index for critically ill patients. ROC curve and calibration curve analyses were performed to assess the accuracy of these estimates. Results revealed that MICU personnel were fairly accurate discriminators of patients who survived vs. who died, although there was a consistent tendency to underestimate survival. In addition, there was some relationship between the level of physician training and forecasting accuracy, but only within the patient's first 24 hours in the MICU. Finally, the estimates of physicians did not differ significantly from those of the APACHE II scale. Physicians tended to be better calibrated in their predictions, while the APACHE II scale was slightly superior in terms of discrimination.