A Method to adjust for measurement error in multiple exposure variables measured with correlated errors in the absence of an internal validation study

摘要: Difficulty in obtaining the correct measurement for an individual’s longterm exposure is a major challenge epidemiological studies that investigate association between exposures and health outcomes. Measurement error biases disease outcome. Usually, internal validation study required to adjust error; it challenging if such not available. We propose general method adjusting where multiple are measured with correlated errors (a multivariate method) illustrate using real data. compare results from those obtained ignores (the naive correlations true univariate method). It found ignoring leads bias underestimates standard error. A sensitivity analysis shows magnitude of adjustment sensitive error, sign, correlation errors. conclude can be used outcome-exposure case absence study. The also useful conducting on sign correlation.