OBJECTIVE: To examine the predictive validity of the Self-Reported Functional Measure (SRFM), a new measure derived from the FIMtrade mark instrument, for health care utilization in multiple sclerosis (MS) and spinal cord injury (SCI). DESIGN: Prospective cohort study using a mailed survey in 1995 and administrative records from 1996 and 1997. SETTING: Veterans Health Administration hospitals and outpatient clinics. PATIENTS: A total of 6361 veterans with SCI and 1789 veterans with MS. MAIN OUTCOME MEASURES: SRFM score was compared with subsequent outpatient visits, hospitalizations, hospital lengths of stay (LOSs), and residence peri-hospitalization. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for these variables. RESULTS: A total of 3836 subjects (47.6%) were hospitalized during 1996-1997, and all but 874 (10.7%) had 1 or more outpatient visits. SRFM score predicted inpatient, but not outpatient health care utilization. Persons in the lowest SRFM quartile were over 90% (OR = 1.91, 95% CI = 1.71-2.13) more likely to be hospitalized compared with those in the highest SRFM quartile; also, they were over 2 times (OR = 2.18, 95% CI = 1.85-2.57) more likely to have a LOS greater than 7 days, were over 2 times (OR = 2.41, 95% CI = 1.62-3.58) more likely to die in hospital, and were nearly 3 times (OR = 2.86, 95% CI = 2.00-4.08) more likely to be discharged to an institution. CONCLUSIONS: SRFM had excellent predictive validity for hospitalization, LOS, and discharge destination among patients with MS or SCI.