Empirically assessing the plausibility of unconfoundedness in observational studies

TL;DR

This paper introduces a simple, assumption-light empirical method to assess the plausibility of unconfoundedness in observational studies, focusing on the relationship between covariates, exposure, and outcome.

Contribution

It proposes a novel approach that tests associations among covariates, exposure, and outcome without requiring explicit confounding structure assumptions.

Findings

Method is supported by theoretical proofs.

Simulations validate the approach.

Applied example demonstrates practical utility.

Abstract

The possibility of unmeasured confounding is one of the main limitations for causal inference from observational studies. There are different methods for (partially) empirically assessing the plausibility of unconfoundedness. However, most currently available methods require (at least partial) assumptions about the confounding structure, which may be difficult to know in practice. In this paper we describe a simple strategy for empirically assessing the plausibility of conditional unconfoundedness (i.e., whether the candidate adjustment set of covariates suffices for confounding adjustment) which does not require any explicit assumptions about the confounding structure, relying instead on assumptions related to temporal ordering between covariates, exposure and outcome (which can be guaranteed by design) and selection into the study. The proposed method essentially relies on testing the association between a subset of the covariates included in the adjustment set (those associated with the exposure, given all other covariates) and the outcome conditional on the remaining covariates and the exposure. We describe the assumptions underlying the method, provide proofs, use simulations to corroborate the theory and illustrate the method with an applied example assessing the causal effect of delivery mode and intelligence quotient measured in adulthood using data from the 1982 Pelotas (Brazil) birth cohort. We also discuss the implications of measurement error and some important limitations of the suggested approach.