Title: Using design strategies to improve non-experimental study designs
Authors: Elizabeth Stuart - Johns Hopkins Bloomberg School of Public Health (United States) [presenting]
Abstract: Many important research questions can only be answered using non-experimental studies. Propensity scores and related methods are a key tool in the design of non-experimental studies -- allowing the design of the study to proceed without the use of the outcome data, and with clear diagnostics. An overview of these methods is provided, common misperceptions regarding their use, and insights into recent advances in the field, including strategies for using propensity scores with complex survey data, with covariates, or in longitudinal policy evaluation contexts. Examples of their use will come from public health, including suicide prevention. The discussion will also include areas for potential future work in propensity score methods, especially for use in the behavioral and social sciences.