Grant: Operating Grant

Pharmacoepidemiologic research using administrative data is limited due to confounding by indication and unmeasured confounding. This research program involves development and evaluation of new methods for causal inference in pharamcoepidemiology, and improvements in implementation of these methods. Methods for causal inference typically require estimates of the propensity score or a variant thereof to account for measured confounding. In this research program we will develop and evaluate approaches to model the propensity score to optimize causal inference. Methods will be evaluated through simulation studies and two substantive pharmacoepidemiologic problems: the effect of high-dose statin treatment on the development of acute kidney injury, and the effect of proton pump inhibitors on pneumonia. The UK General Practice Research Database will be used as a datas source, and serve as a test-bed for methods work.