The Impact of Missing Data on the Results of a Schizophrenia Study

• Activity Number: 605
• Type: Topic Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Denis V Rybin*+ and Gheorghe Doros and Robert A Rosenheck and Robert Lew
• Companies: Boston University/VA Boston Healthcare System and Boston University and Yale Medical School/VA Connecticut Health Care System and VA Boston Healthcare
• Keywords: missing data ; Bayesian ; sensitivity analysis ; Pattern Mixture Model ; clinical trial ; schizophrenia

Abstract

Missing data pose a serious challenge to the integrity of randomized clinical trials, especially in which long-term impact assessment is of great importance but the follow-up rates are low. Sensitivity analysis using Bayesian modeling for missing data offers a potentially useful approach to assessing the sensitivity of the inferences made on the bases of observed data. Methods: Data were obtained from a randomized clinical trial of 369 patients diagnosed with schizophrenia that compared two treatments. Only 223(60%) patients completed 1 year. Bayesian analysis using a Pattern Mixture Modeling was used to either validate the reported results or to identify bias. The analysis was applied to the several outcome measures. Results: Study results showing no difference between treatments on several measures were confirmed against a wide range of patterns of non-random missingness. Robustness of the null conclusions was assessed using sensitivity parameters. Conclusion: Sensitivity of trial conclusions to non-random missingness can be easily carried out using current statistical packages, and at least in this study, show that missing data do not threaten the validity of reported results.