Predicting the Outcomes of Nursing Home Patients

We propose a system of nursing home reimbursement based on attaining achievable outcomes. The crux of the system rests on our ability to predict patient outcomes from one point in time to the next. Using three waves of data collected at 3-month intervals on approximately 250 patients, we were able to predict patient functioning in six domains (physiologic, activities, affective, cognitive, social, and satisfaction) with R2 values ranging from 0.51 to 0.93. Predictions of discharge (better, worse, or dead) were less accurate, with R2 values of 0.36 to 0.39.