Conventional statistical methods are not well equipped to appropriately handle confounding results, and they also tend to produce biased or false results. Other statistical methods, called causal inference (CI) are better suited to give patients and doctors accurate and valid study results they can trust and use to make treatment decisions. However, controversies and confusion in the field of CI impede progress. Many researchers are confused as to which of the many CI methods are most appropriate for their specific research question. Others are daunted by the perceived complexity of using these new statistical methods that require unfamiliar software. To address these issues, we propose to conduct a pair of two-day conference and workshops in the Washington, DC, area in February 2016 and 2017 focusing on the selection of appropriate CI methods based on PCOR question, research design, and type of confounding. Our overall goal is to enhance use of CI and increase the value of observational studies.
The projected outputs from this project are a conference series to address conceptual issues related to choosing appropriate CI methods for addressing specific PCOR questions and case studies; and the application of CI in different clinical areas to obtain hands-on experience and practical skills and lessons through plenary, panel presentations, breakout sessions, training, and technical assistance.
Project collaborators include a multi-stakeholder expert advisory committee.