Data on excess length of stay (LOS)-the difference between actual LOS and target LOS for the medical problem-were analyzed for three acute care hospitals. For a sample of 2,642 cases, 29.2% of all hospital days were excess days for patients referred to social services. For a sample of 176 social service cases, a stepwise regression yielded two statistically significant predictors of LOS: (1) Medical problem (target LOS) and (2) severity of psychosocial problem (Person-in- Environment scale). The importance of psychosocial severity as a predictor of LOS has clear implications for the role social workers can play in making health care organizations more efficient.