Latent Propensity Score Approach for Post-market Evaluation of Regular Breast Pump Usage
*Yi Huang, University of Maryland 

Keywords: Causal inference, measurement error, finite mixture model, propensity score subclassification, causal effect estimation, post-market.

Many postmarketing studies are based on observational data where confounders (i.e. covariates) are often measured with error. However, this unobserved measurement error thread the validity of the strong ignorability assumption which the standard propensity score methods rely on for valid inference on the average causal effect (ACE) estimation. The naive approach is to ignore the error and use the observed covariates in current propensity score framework for ACE estimation. However, under extended causal inference framework allowing this unobserved measurement error in covariate, we showed that the naive approach typically produces biased ACE inference. In this talk, I propose a newly developed joint likelihood based latent propensity score approach for ACE estimation, which will take into account the uncertainty from the effect of covariate’s measurement error on propensity score subclassification. Its performance will be evaluated by extensive simulations studies. We applied this method in post-market evaluation of the impact of long-term breast pump usage on infant’s comorbidity index during 1st year, using the recently collected Infant Feeding Practice Study (II) data (FDA and CDC). In summary, we extend the standard propensity score framework allowing covariate measurement error for general need of the post-market evaluations of FDA regulated medical products, including device, drug, and biologics. This is a joint work with Drs. Xiaoyu Dong, Andrew Raim, Karen Bandeen-Roche, and Cunlin Wang. Thank the support from FDA Critical Path Grants for this work.