Propensity score methods in the context of covariate measurement error
Propensity score methods are commonly used to estimate causal effects in non-experimental studies. Existing propensity score methods assume that covariates are measured without error but covariate measurement error is likely common.
This talk will discuss the implications of measurement error in the covariates on the estimation of causal effects using propensity score methods and investigates Multiple Imputation using External Calibration (MIEC) to account for covariate measurement error in propensity score estimation. MIEC uses a main study sample and a calibration dataset that includes observations of the true covariate (X) as well as the version measured with error (W). MIEC creates multiple imputations of X in the main study sample, using information on the joint distribution of X, W, other covariates, and the outcome of interest, from both the calibration and the main data.
In simulation studies we found that MIEC estimates the treatment effect almost as well as if the true covariate X were available. We also found that the outcome must be used in the imputation process, a finding related to the idea of congeniality in the multiple imputation literature. We illustrate MIEC using an example estimating the effect of neighborhood disadvantage on the mental health of adolescents, where the method accounts for measurement error in the adolescents' report of their mothers' age when they (the adolescents) were born.The talk will also discuss the use of SIMEX to account of measurement error, and the implications of measurement error in multiple covariates.
Elizabeth A. Stuart, Ph.D. is Professor in the Department of Mental Health at the Johns Hopkins Bloomberg School of Public Health, with joint appointments in the Department of Biostatistics and the Department of Health Policy and Management, and Associate Dean for Education at JHSPH. She received her Ph.D. in statistics in 2004 from Harvard University and is a Fellow of the American Statistical Association. Dr. Stuart has extensive experience in methods for estimating causal effects and dealing with the complications of missing data in experimental and non-experimental studies, particularly as applied to mental health, public policy, and education. She has published influential papers on propensity score methods and generalizing treatment effect estimate to target populations and taught courses and short courses on causal inference and propensity scores to a wide range of audiences, including at JHSPH (both in person and online), the US Food and Drug Administration, and at conferences. Her primary areas of application include estimating the effects of health care interventions (such as an accountable care model) on mental health service utilization, treatments for children with autism spectrum disorders, and the evaluation of school-based preventive interventions. She also has extensive experience with policy evaluation, through previous employment at Mathematica Policy Research, and co-directs the JHSPH Center for Mental Health and Addiction Policy Research. Dr. Stuart has received research funding for her work from the National Institutes of Health, the US Institute of Education Sciences, and the National Science Foundation and has served on advisory panels for the National Academy of Sciences and the US Department of Education. She is Methods Editor for the Journal of Research on Educational Effectiveness and serves as Chair of the Patient Centered Outcomes Research Institute's (PCORI's) inaugural Clinical Trials Advisory Panel. Dr. Stuart was recently recognized with the mid-career award from the Health Policy Statistics Section of the American Statistical Association and the Gertrude Cox Award for applied statistics.