Statistical prediction of rehabilitation in elderly patients with hip fractures.

A statistical investigation using multivariate techniques was made of the rehabilitation outcome of 103 consecutive elderly hip fracture patients, all admitted from their own homes. Cluster analysis revealed two main groups of variables, i.e., background variables (age, general medical condition, type of fracture, sex, living with someone) and functional variables (prefracture ability to visit someone and to shop, ability to walk and to manage activities of daily living two weeks postsurgery). A stepwise linear discriminant analysis was used, to estimate the probability of a given patient returning home. At discharge from hospital, certain factors, i.e., ability to walk two weeks postsurgery, living with someone, general medical condition, and type of fracture, were found to be most important for direct return home. At one year after hip fracture, the prefracture ability to visit someone and the age of the patient were of most prognostic significance for having returned and remained at home. The accuracy of the prediction tested by a "jackknife" procedure showed an over all increasingly correct classification during the follow-up year, more than 80% correct for all patients at discharge and 86% correct at one year. This approach to the analysis of soft data concerning hip fracture rehabilitation may prove of prognostic value also in sociomedical systems elsewhere.