Recent advances in addressing measurement error and misclassification in epidemiologic research

Measurement error is prevalent across many research fields. It has been known for decades that measurement error arising from various exposures (e.g., nutrition, environment) is a major source of bias in estimating causal effects between exposures and health outcomes in epidemiological research. It is challenging to correct bias due to measurement error, which is crucial for deriving reliable conclusions from complex data. Regression calibration is widely used in practical epidemiologic studies to correct for the bias in causal effect due to measurement error in continuous exposures. However, no systematic discussion exists on how to determine covariates appropriate in measurement error models relating mismeasured exposures to true exposures as well as outcome models relating the exposure to outcome in the regression calibration method. In addition, transportability of the estimated calibration model to the main study is a major concern for the practical use of regression calibration with an external validation study. Over the last decade, Electronic Health Records (EHR) systems have been increasingly implemented at US hospitals. Despite their great potential, a critical challenge for analyzing complex EHR data is the measurement error or misclassification in outcome and exposures. Research on dynamic treatment regimes has sparked extensive interest. Many methods have been proposed in the literature, which, however, are vulnerable to the presence of misclassification in covariates. In particular, although Q-learning has received considerable attention, its applicability to data with misclassified covariates is unclear. This session highlights recent advances in statistical methods designed to address measurement error, which refine data analysis and interpretation. The session will cover innovative methods in selecting the minimal and most efficient covariate adjustment sets for regression calibration under a causal inference framework, in improving transportability of regression calibration with an external validation study, in correcting misclassification effects on Q-learning to optimize dynamic treatment regimes, and in addressing outcome measurement errors in EHR data. The presentation of cutting-edge statistical techniques aims to strengthen research validity and contribute to meaningful improvements in public health. By delving into these innovative methodologies, the session aims to advance the field of epidemiological research, offering researchers valuable tools to enhance the reliability of their findings and contribute to more effective public health interventions. This presentation of innovative statistical techniques is geared towards improving the validity of research outcomes and enriching societal health advancements.