Survey Research at the Crossroads: Novel Pathways to Better Data Science

Randomized controlled trials provide strong internal validity, but they often lack external validity when attempting to extend results to a pre-specified external target population. This limitation, known as transportability, arises when the distribution of effect modifiers differs between the trial and target populations of interest. To address this challenge, we develop a novel interaction-based Propensity Score Predictive Inference (PSPI) method that emphasizes the central role of sampling propensity scores for sampling-induced effect heterogeneity, combined with flexible outcome models. We introduce two PSPI variants for estimating average treatment effects and potential outcomes across treatment groups by incorporating a natural cubic spline of the propensity score and modeling high-dimensional covariates using Bayesian Additive Regression Trees. We also introduce a balancing transformation to improve the distributional alignment of propensity scores between trial and target populations. Simulation studies show that PSPI models outperform existing methods, achieving lower root mean squared errors and near-nominal coverage rates, particularly in settings with small sample sizes, imbalanced treatment groups, or covariate shift between trial participants and the target population. We further demonstrate the utility of our approach by transporting the effect of a depression care intervention from the CODIACS Vanguard trial to trial-eligible post-acute coronary syndrome patients with elevated depressive symptoms in the All of Us electronic health records. 

Research on racial and socioeconomic health disparities (HD) in time-to-event outcomes is crucial to
public health. Yet, existing HD decomposition methods can yield biased estimates when censoring, such as dropout or
competing risks, depends on covariates, including demographic, socioeconomic, or behavioral characteristics. Widely
applied Peters–Belson (PB) approaches to time-to-event outcomes use Kaplan–Meier (KM) estimation to quantify
overall HD and Cox regression to predict counterfactual survival. The resulting HD decomposition can be biased:
KM assumes covariate-independent censoring, whereas Cox regression assumes independent censoring conditional on
modeled covariates. To address this limitation, we propose a novel decomposition method that integrates propensity
score (PS) methodology with inverse probability of censoring weights (IPCW). The PS component balances covariates
across groups to estimate counterfactual survival without specifying an outcome model, while the IPCW component
corrects for covariate-dependent censoring by upweighting uncensored individuals to represent those who were censored.
Developed within a semiparametric framework, the proposed IPCW-PS approach reduces bias from traditional
KM-Cox methods. Simulation studies demonstrate substantial bias reduction under dependent censoring. We further
apply IPCW-PS to NielsenIQ Consumer Panel data to examine smoking cessation disparities between White and
Black households and find that smoke-free laws contribute meaningfully to explaining observed disparities in smoking
cessation patterns.