Estimation of causal effects of binary treatments in unconfounded studies

Abstract

Estimation of causal effects in non-randomized studies comprises two distinct phases: design, without outcome data, and analysis of the outcome data according to a specified protocol. Recently, Gutman and Rubin (2013) proposed a new analysis-phase method for estimating treatment effects when the outcome is binary and there is only one covariate, which viewed causal effect estimation explicitly as a missing data problem. Here, we extend this method to situations with continuous outcomes and multiple covariates and compare it with other commonly used methods (such as matching, subclassification, weighting, and covariance adjustment). We show, using an extensive simulation, that of all methods considered, and in many of the experimental conditions examined, our new ‘multiple-imputation using two subclassification splines’ method appears to be the most efficient and has coverage levels that are closest to nominal. In addition, it can estimate finite population average causal effects as well as non-linear causal estimands. This type of analysis also allows the identification of subgroups of units for which the effect appears to be especially beneficial or harmful.

Medical foster home is less costly than traditional nursing home care

“…In the second stage, we treated the unobserved potential outcome for each unit as missing data 29 and multiply imputed these missing potential outcomes to compare the mortality and costs for Veterans in MFH to Veterans in CNH. 30,31 Specifically, for all of the Veterans in MFH, we imputed the mortality and costs that would have been observed if they were in a CNH, and for all of the patients that were in a CNH, we imputed the mortality and outcomes if they were in a MFH. This imputation method that utilized the propensity score was shown to have good operating characteristics in comparison with matching and weighting.…”

“…This imputation method that utilized the propensity score was shown to have good operating characteristics in comparison with matching and weighting. 31 It is important to note that the first stage was implemented without viewing the outcome data.…”

“…The first stage comprised propensity score estimation and declaration of the procedures that would be used for data analysis. In the second stage, we treated the unobserved potential outcome for each unit as missing data and multiply imputed these missing potential outcomes to compare the mortality and costs for Veterans in MFH to Veterans in CNH . Specifically, for all of the Veterans in MFH, we imputed the mortality and costs that would have been observed if they were in a CNH, and for all of the patients that were in a CNH, we imputed the mortality and outcomes if they were in a MFH.…”

“…Specifically, for all of the Veterans in MFH, we imputed the mortality and costs that would have been observed if they were in a CNH, and for all of the patients that were in a CNH, we imputed the mortality and outcomes if they were in a MFH. This imputation method that utilized the propensity score was shown to have good operating characteristics in comparison with matching and weighting . It is important to note that the first stage was implemented without viewing the outcome data.…”

Persistent pain after motor vehicle collision: comparative effectiveness of opioids vs nonsteroidal antiinflammatory drugs prescribed from the emergency department—a propensity matched analysis

“…Using these methods, we estimated both average (group) effects and individual effect of opioid versus NSAIDs prescription on the likelihood of having moderate to severe musculoskeletal pain 6 weeks after an MVC. [12] In addition, we used innovative analytic techniques to estimate outcomes for individuals and groups had they received the other treatment option (opioid instead of NSAID, or vice versa). [7] By comparing these observed and estimated counterfactual (unobserved) outcomes, we determined both the overall average differential treatment effect of opioids versus NSAIDs and differential treatment effects according to individual characteristics.…”