This study used hierarchical linear models to analyze post-discharge data among 17,130 veterans who received inpatient VA mental health services between 1995 and 2001, in order to determine whether changes in mental health service affected inpatient satisfaction. In models without covariants, half the satisfaction measures increased while the other half indicated no significant change. After adjusting for changes in patient characteristics, however, nine measures indicated no change and seven showed significant declines. Demonstrating the importance of risk-adjustment, satisfaction appeared to increase over time prior to risk adjustment, but declined after adjusting for changes in patient characteristics that were associated with patient satisfaction.