Predicting the course of disease.

The ability to predict the course of disease and the effect of interventions is critical to effective medical practice and health care management. In this analysis, we sought to test whether available clinical data and analytic methodologies can be used to accurately predict the time course of the probability of death after hospital admission and the probability of readmission following discharge for patients with acute myocardial infarction or pulmonary disease. We grouped patients by selected physiologic characteristics and made time-to-event predictions using multiple regression models. These predictions were compared with observed probabilities calculated using the actuarial or life-table method. Predictions made with the Bailey-Makeham model generally replicated observed experience. They accurately accounted for substantial differences in the patterns of death and readmission and accurately delineated the effects of therapies, after adjustment for patient risk. These results were validated by analyses of resampled populations that differed in case mix from the source population. We believe that using such models to project the course of disease and the effects of treatment on that course in defined classes of patients should facilitate the development of practice guidelines for patient care and the management of health care resources.