Estimation of causal effects of multiple treatments in healthcare database studies with rare outcomes

The preponderance of large-scale healthcare databases provide abundant opportunities for comparative effectiveness research. Evidence necessary to making informed treatment decisions often relies on comparing multiple options outcomes interest observed in a small number individuals. Causal inference with treatments and rare is subject that has been treated sparingly the literature. This paper designs three sets simulations, representative structure our database study, propose causal analysis strategies such settings. We investigate compare operating characteristics types methods their variants: Bayesian Additive Regression Trees (BART), regression adjustment multivariate spline generalized propensity scores (RAMS) inverse probability weighting (IPTW) multinomial logistic or boosted models. Our results suggest BART RAMS lower bias mean squared error, widely used IPTW deliver unfavorable characteristics. illustrate using case study evaluating robotic-assisted surgery, video-assisted thoracoscopic surgery open thoracotomy treating non-small cell lung cancer.