High-dimensional propensity score and its machine learning extensions in residual confounding control in pharmacoepidemiologic studies

The use of retrospective health care claims datasets is frequently criticized for lacking complete information on potential confounders. Ultimately, the treatment effects estimated utilizing such data sources may be subject to residual confounding. Digital electronic administrative records routinely collect a large volume of health-related information; and many of whom are usually not considered in conventional pharmacoepidemiological studies. In 2009, a high-dimensional propensity score (hdPS) algorithm was proposed that utilizes such information as surrogates or proxies for mismeasured and unobserved confounders in an effort to reduce residual confounding bias. Since then, many machine learning and semi-parametric extensions of this algorithm have been proposed to exploit the wealth of high-dimensional proxy information properly. 

This workshop will

1. demonstrate logic, steps and implementation guidelines of hdPS utilizing an open data source as an example (using reproducible R codes),
2. familiarize participants with the difference between propensity score vs. hdPS,
3. explain the rationale for using the machine learning extensions of hdPS, and their statistical properties, and
4. discuss advantages, controversies, and hdPS reporting guidelines while writing a manuscript.

Attendees should have prerequisite knowledge of multiple regression analysis and working knowledge in R (e.g., basic data manipulation and regression fitting).