Bayesian propensity scores for high-dimensional causal inference: A comparison of drug-eluting to bare-metal coronary stents.

High-dimensional data provide many potential confounders that may bolster the plausibility of the ignorability assumption in causal inference problems. Propensity score methods are powerful causal inference tools, which are popular in health care research and are particularly useful for high-dimensional data. Recent interest has surrounded a Bayesian treatment of propensity scores in order to flexibly model the treatment assignment mechanism and summarize posterior quantities while incorporating variance from the treatment model. We discuss methods for Bayesian propensity score analysis of binary treatments, focusing on modern methods for high-dimensional Bayesian regression and the propagation of uncertainty. We introduce a novel and simple estimator for the average treatment effect that capitalizes on conjugacy of the beta and binomial distributions. Through simulations, we show the utility of horseshoe priors and Bayesian additive regression trees paired with our new estimator, while demonstrating the importance of including variance from the treatment regression model. An application to cardiac stent data with almost 500 confounders and 9000 patients illustrates approaches and facilitates comparison with existing alternatives. As measured by a falsifiability endpoint, we improved confounder adjustment compared with past observational research of the same problem.

Integrating Quality of Life and Survival Outcomes in Cardiovascular Clinical Trials.

Background Survival and health status (eg, symptoms and quality of life) are key outcomes in clinical trials of heart failure treatment. However, health status can only be recorded on survivors, potentially biasing treatment effect estimates when there is differential survival across treatment groups. Joint modeling of survival and health status can address this bias. Methods and Results We analyzed patient-level data from the PARTNER 1B trial (Placement of Aortic Transcatheter Valves) of transcatheter aortic valve replacement versus standard care. Health status was quantified with the Kansas City Cardiomyopathy Questionnaire (KCCQ) at randomization, 1, 6, and 12 months. We compared hazard ratios for survival and mean differences in KCCQ scores at 12 months using several models: the original growth curve model for KCCQ scores (ignoring death), separate Bayesian models for survival and KCCQ scores, and a Bayesian joint longitudinal-survival model fit to either 12 or 30 months of survival follow-up. The benefit of transcatheter aortic valve replacement on 12-month KCCQ scores was greatest in the joint-model fit to all survival data (mean difference, 33.7 points; 95% credible intervals [CrI], 24.2-42.4), followed by the joint-model fit to 12 months of survival follow-up (32.3 points; 95% CrI, 22.5-41.5), a Bayesian model without integrating death (30.4 points; 95% CrI, 21.4-39.3), and the original growth curve model (26.0 points; 95% CI, 18.7-33.3). At 12 months, the survival benefit of transcatheter aortic valve replacement was also greater in the joint model (hazard ratio, 0.50; 95% CrI, 0.32-0.73) than in the nonjoint Bayesian model (0.54; 95% CrI, 0.37-0.75) or the original Kaplan-Meier estimate (0.55; 95% CI, 0.40-0.74). Conclusions In patients with severe symptomatic aortic stenosis and prohibitive surgical risk, the estimated benefits of transcatheter aortic valve replacement on survival and health status compared with standard care were greater in joint Bayesian models than other approaches.

Assessing Hospital Performance After Percutaneous Coronary Intervention Using Big Data.

BACKGROUND: Although risk adjustment remains a cornerstone for comparing outcomes across hospitals, optimal strategies continue to evolve in the presence of many confounders. We compared conventional regression-based model to approaches particularly suited to leveraging big data. METHODS AND RESULTS: We assessed hospital all-cause 30-day excess mortality risk among 8952 adults undergoing percutaneous coronary intervention between October 1, 2011, and September 30, 2012, in 24 Massachusetts hospitals using clinical registry data linked with billing data. We compared conventional logistic regression models with augmented inverse probability weighted estimators and targeted maximum likelihood estimators to generate more efficient and unbiased estimates of hospital effects. We also compared a clinically informed and a machine-learning approach to confounder selection, using elastic net penalized regression in the latter case. Hospital excess risk estimates range from -1.4% to 2.0% across methods and confounder sets. Some hospitals were consistently classified as low or as high excess mortality outliers; others changed classification depending on the method and confounder set used. Switching from the clinically selected list of 11 confounders to a full set of 225 confounders increased the estimation uncertainty by an average of 62% across methods as measured by confidence interval length. Agreement among methods ranged from fair, with a κ statistic of 0.39 (SE: 0.16), to perfect, with a κ of 1 (SE: 0.0). CONCLUSIONS: Modern causal inference techniques should be more frequently adopted to leverage big data while minimizing bias in hospital performance assessments.