As an initial step toward improving admission criteria to the medical intensive care unit (MICU), we examined Acute Physiologic and Chronic Health Evaluation scores and the diagnosis-adjusted mortality rates of 2419 medical patients, including those who received MICU consultation over a 6-month period. There was considerable overlap in the physiologic scores and the predicted mortality rates between those patients who were admitted to the MICU and those who were not. There was no discrete score or mortality rate at which triage to the MICU would have included most MICU patients but excluded most patients who survived without admission to the MICU. While uniform MICU admission criteria would be desirable, current scoring systems may not have the desired sensitivity or specificity to establish such criteria. Using a receiver operating characteristic curve, we demonstrated that diagnosis-adjusted mortality rates could predict as well as Acute Physiologic and Chronic Health Evaluation scores which patients would receive MICU admission and intervention. This suggests that, for different diagnoses, specific clinical variables and laboratory tests may have different predictive importance for MICU admission. Prospective models of clinical variables using receiver operating characteristic curves in various medical diseases may improve triage procedures.