Xiaofang Yan, Department of Bioinformatics and Biostatistics, University of Louisville

"Estimation of Average Treatment Effects Among Multiple Treatment Groups by Using an Ensemble Approach"

In observational studies, generalized propensity score (GPS) based statistical methods, such as stratification, inverse probability weighting (IPW), and doubly robust (DR) method, have been proposed to estimate average treatment effect (ATE) among multiple treatment groups. In this article, we investigate the GPS-based statistical methods to estimate treatment effects in two aspects. The first aspect of our investigation is to obtain an optimal GPS estimating method among the three choices (i.e., multinomial regression model, generalized boosted method (GBM) and random forest) such that the method is optimal in balancing covariates. The optimal GPS estimating method is obtained by using the rank aggregation approach. We further examine whether the optimal GPS based stratification, IPW, and DR methods would improve the performance for estimating ATE. It is well known that DR method is unbiased if either the GPS or the outcome models are correctly specified. The second aspect of our investigation is to examine whether the estimates of ATE based on DR method could be improved if we ensemble outcome models. To that end, bootstrap method and rank aggregation method are used to obtain the ensemble optimal outcome regression model from the three possible outcome models (i.e., multiple regression model, GBM, and random forest), and the resulting outcome model is incorporated into the DR method, resulting in an ensemble DR method. Extensive simulation results indicate that the ensemble DR method provides the best performance in estimating the ATE regardless of the method used for estimating GPS. We also notice that a smaller covariate balance score does not necessarily ensure a better performance in estimating ATE. We illustrate our methods using the MarketScan healthcare insurance claims database to examine the treatment effects among three different bones and substitutes used for spinal fusion surgeries. We draw conclusions based on the estimates from the ensemble DR method coupled with the optimal GPS estimating method.

Stay connected TwitterFacebookLinkedInYouTubeInstagram