Propensity score adjustment of a treatment effect with missing data in psychiatric health services research

摘要: Background: Missing values are a common problem for data analyses in observational studies, which frequently applied health services research. This paper examines the usefulness of different approaches to tackle incomplete data, focusing whether Multiple Imputation (MI) strategy yields adequate estimates when complex analysis framework. Methods: Based on study originally comparing three forms psychotherapy, simulation with missing scenarios was conducted. The considered model comprised propensity score-adjusted treatment effect estimation. were handled by complete case analysis, MI approaches, as well mean and regression imputation. Results: All point estimators methods lay within 95% confidence interval derived from set. Highest deviation observed analysis. A distinct superiority could not be demonstrated. Conclusion: Since there no clear benefit one method deal over another, researchers faced well-advised apply imputation compare results order get an impression their sensitivity.