We estimated the relative importance of various client characteristics related to nursing home entry for a national probability sample of Medicare recipients and developed predictive models of nursing home entry that account for the interactive effects among variables. In contrast with previous research, we focused on the characteristics of nursing home entrants, not residents. By using a national sample we ensured that the influence of regional variations in the configuration of long-term care services would not confound estimates of the relative effect of client-related factors. Nine variables emerged as statistically significant predictors: age, being confined to a bed, requiring help to get around, requiring aid getting around, being widowed, never married, welfare as a payment source, insurance as a payment source, and perceived health status. When these factors were controlled for, sex, geographic region, and educational status were not statistically significant.