Using administrative data to predict important health outcomes. Entry to hospital, nursing home, and death.

This paper assesses our ability to use administrative data for developing indicators of health status. Traditionally, measures of health status have been derived from interviews. Here indicators from administrative data and from interviews are compared, i.e., their ability to predict important health outcomes for a large representative sample of elderly residents of Manitoba, Canada. Indicators of health status derived from an administrative data system and from health interviews are shown to provide roughly similar predictions of nursing-home entry. Administrative data provide significantly better predictions of death and future hospital entry than do variables from interview data.