Using Conditional Linear Regression Trees to Identify Potential Construct Validity and Measurement Error Differences in Health Outcomes from a U.S. National Survey

Abstract Accurate measurement is critical for the assessment of health outcomes and behaviors across population subgroups. Estimated differences in the prevalence of a given health outcome or risk factor can be affected by whether the outcomes are measured or self-reported. Using data from the National Health and Nutrition Examination Survey (NHANES), which contains both measured and self-reported health information, we used conditional linear regression trees (CLRT) to assess differential measurement error across several chronic health conditions and risk factors, where measurement error is defined as differences between self-reported and measured (e.g., gold standard or objective) health outcomes or behaviors. Using data on four health conditions/behaviors (e.g., diabetes, hypertension, high cholesterol, and current smoking) from NHANES (2015 through March 2020), we assessed differences between self-reported and measured outcomes by various sociodemographic characteristics (e.g., age, sex, race/ethnicity, education, marital status, health insurance, and poverty). We used CLRTs via the “rpms” package in R to identify demographic subgroups with larger differences between self-reported and measured health conditions/behaviors. Identifying subgroups where measurement error may be larger or smaller could help inform future work on improving the estimation of the prevalence of chronic conditions and health behaviors.