Identification and Estimation of the Average Causal Effects Under Dietary Substitution Strategies.

Abstract

The 2020-2025 Dietary Guidelines suggest that most people can improve their diet by making some changes to what they eat and drink. In many cases, these changes involve simple substitutions. For instance, the Dietary Guidelines recommend choosing chicken instead of processed red meat to reduce sodium intake and switching from refined grains to whole grains to increase dietary fiber intake. The question about such dietary substitution strategies seeks to estimate the average counterfactual outcome under a hypothetical intervention that replaces a food an individual would have consumed in the absence of intervention with a healthier substitute. In this work, we will show the conditions under which the average causal effects of substitution strategies can be non-parametrically identified, and provide efficient estimators for our proposed dietary substitution strategies. We evaluate the performance of our proposed methods via simulation studies and apply them to estimate the effect of substituting processed red meat with chicken on mortality, using data from the Nurses' Health Study.