A Prediction Rule For Mortality in the Medical Intensive Care Unit Based on Early Acute Organ-System Failure

Abstract

We evaluated early acute organ-system failure (AOSF) as a predictor of mortality in medical intensive care unit (MICU) patients. Prospective data were obtained on 825 men admitted to a Veterans Administraion (VA) Medical Center MICU. Clinical criteria were used to diagnose the presence of 7 types of AOSF. Of the 2,364 AOSFs detected, 1,847 (78%) were “early” (i.e., detected within the first 48 hours of MICU stay). A random sample of 550 patients was selected for derivation of a prediction rule for MICU mortality based on age and number of early AOSFs. For each additional early AOSF, the adjusted odds of mortality increased by 3.3 (95% confidence interval: 2.7, 4.0; p < 0.0001). When applied to the cross-validation sample of 275 patients, this rule yielded a sensitivity of 77%, a specificity of 86%, and an overall correct classification rate of 82%. These results suggest that a simple rule based on number of AOSFs detected in the early portion of a patient's MICU stay may be a useful predictor of mortality.