Objectives: To compare the risk of death or emergy transfer (ET) to an acute care hospital for acute and subacute rehabilitation inpatients, to identify risk factors, and to determine whether the risk factors vary by level of care. Design: Proportional hazards regression analysis of retrospective cohort data. Explanatory variables included level of care, age, sex, medical stability at admission, and diagnosis. Setting: An acute inpatient rehabilitation hospital and a skilled nursing facility-based subacute rehabilitation program. Patients: Patients were selected from a database of all acute and subacute inpatients discharged between January 1992 and August 1994 (N = 4,755). Readmissions and pediatric patients were excluded ( n = 896). Patients who had complete data ( n = 3,185) were included in the survival analysis; a separate univariable analysis was performed for patients who lacked complete data ( n = 674). Outcome Measure: Rate of death or emergency transfer during rehabilitation. Results: The overall risk was greater for subacute patients than for acute inpatients, 20.6% vs 11.6%, odds ratio = 2.0, CI = 1.6 to 2.5. The survival analysis indicated that age, level of care, sex, and stability were associated with the risk of death or emergency transfer (2.1 ≤ z ≤ 4.7; p < .05 for each). However, the difference in risk between subacute and acute care was greatest among younger patients. Conclusions: Careful consideration of risk factors is critical for proper assignment of patients to rehabilitation levels of care. However, research is needed to test these findings in other settings and to improve identification of high-risk patients.