B1119
Title: Measurement error-robust causal inference via synthetic instrumental variables
Authors: Caleb Miles - Columbia University (United States) [presenting]
Brent Coull - Harvard University (United States)
Linda Valeri - Harvard Medical School and McLean Hospital (United States)
Abstract: While measurement error is known to be benign in certain settings, this is often not the case when estimating causal effects. Two scenarios in which it can be malignant are the estimation of (i) the average causal effect when confounders are measured with error and (ii) the natural indirect effect when the exposure and/or confounders are measured with error. Methods adjusting for measurement error typically require external data or knowledge about the measurement error distribution. We propose methodology not requiring any such information. Instead, we show that when the outcome regression is linear in the error-prone variables, consistent estimation of these causal effects can be recovered using what we refer to as synthetic instrumental variables. These are functions of only the observed data that behave like instrumental variables for the error-prone variables. Using data from a study in Bangladesh, we apply our methodology to estimate (i) the effect of maternal protein intake on child neurodevelopment while controlling for lead exposure, and (ii) maternal protein intake's role in mediating the effect of lead exposure on child neurodevelopment. Protein intake is calculated from food journal entries, and is suspected to be highly subject to measurement error.
