Oral

In public health research, it is not always feasible to conduct randomized controlled trails due to practical or ethical concerns, so investigators usually rely on observational data to estimate treatment effects. Due to the lack of randomization in observational studies, selection bias can be evident in the unbalanced distributions of observed covariates between the treatment conditions (i.e., treatment vs. control); as a result, the section bias may impair the validity of casual inference based on the results from observational studies. Rosenbaum and Rubin (1983) coined propensity score analysis to help investigators balance the distributions of observed covariates so as to increase the validity of causal inference for treatment effects based on the observational studies. Since then, propensity score analysis has become increasingly popular in evaluating treatment effects. However, there still exist both methodological and practical issues in propensity score analysis to be further investigated, such as how to parameterize propensity function for continuous treatments, how to improve propensity score matching quality, what the implications of over adjustment for many covariates in estimating propensity scores are, and how to apply propensity score analysis effectively. Therefore, investigators are more or less plagued with these issues when using propensity score analysis. To help investigators deal with those issues, the proposed special interest session will discuss those issues, introduce some new developments in propensity score analysis, illustrate the new techniques through empirical data, and demonstrate the use of propensity score analysis in public health research. We expect this proposed session will promote dialogues on propensity score analysis between statistical methodologists and applied investigators as well as among statistical methodologists themselves and, therefore, advance the application of propensity score analysis in public health research.

Session Objectives: Describe the issues of propensity score analysis. Identify appropriate methods of propensity score analysis. Demonstrate the use of propensity score analysis in real research.

CE Credits: Medical (CME), Health Education (CHES), Nursing (CNE), Public Health (CPH)