OBJECTIVES:  Missing data has presented challenges to health economic analyses, especially for a long-term observational study with repeated measures of clinical and economic outcomes.  The aim of this analysis was to compare the total health care costs between symptom remission and non-remission from a long-term, observational study using mixed-effects models with and without multiple imputations (MI) of missing data. METHODS: Data (N=2282) used for this analysis were from a 3-year observational study of patients treated for schizophrenia in the United States between July 1997 and September 2003. Costs of mental health services were obtained at enrollment and at 6-month intervals during the 3-year follow up.  Cohorts of remitters versus non-remitters at enrollment were created using established criteria. Total costs for remitters and non-remitters were compared using mixed-effects models with and without MI based on Markov chain Monte Carlo with multivariate normality assumption (MI-MCMC) or fully conditional specification with predictive mean match method (MI-FCS).  All analyses on costs were adjusted for patient's demographics and comorbidities. RESULTS: The majority of the patients were male (61.6%) and non-remitters (73.8%) with a mean age of 42 years. Out of 2282 patients, 41.2% had at least 1 visit (out of 7 visits) with missing costs data.  Without MI, the total healthcare costs were estimated to be $8689.6 for the non-remitters and $6730.0 for the remitters with a difference of $1959.7 (95% CI: $790 - $3129.4) over a 6-month period (p=0.001). The estimated differences in total costs between remitters and non-remitters were $1763.3 over the 6-month period with the MI-MCMC method (p=0.004) and $1483.9 with the MI-FCS method (p=0.009).    CONCLUSIONS: Significant differences in total costs between remitters and non-remitters were obtained from this study using mixed-effects models with and without MIs.  Further analysis will be conducted to explore MI for estimation of other costs and examine missingness mechanisms.