Overadjustment bias in systematic reviews and meta-analyses of socio-economic inequalities in health: a meta-research scoping review

Abstract

BACKGROUND: Overadjustment bias occurs when researchers adjust for an explanatory variable on the causal pathway from exposure to outcome, which leads to biased estimates of the causal effect of the exposure. This meta-research review aimed to examine how previous systematic reviews and meta-analyses of socio-economic inequalities in health have managed overadjustment bias. METHODS: We searched Medline and Embase until 16 April 2021 for systematic reviews and meta-analyses of observational studies on associations between individual-level socio-economic position and health outcomes in any population. A set of criteria were developed to examine methodological approaches to overadjustment bias adopted by included reviews (rated Yes/No/Somewhat/Unclear). RESULTS: Eighty-four reviews were eligible (47 systematic reviews, 37 meta-analyses). Regarding approaches to overadjustment, whereas 73% of the 84 reviews were rated as Yes for clearly defining exposures and outcomes, all other approaches were rated as Yes for <55% of reviews; for instance, 5% clearly defined confounders and mediators, 2% constructed causal diagrams and 35% reported adjusted variables for included studies. Whereas only 2% included overadjustment in risk of bias assessment, 54% included confounding. Of the 37 meta-analyses, 16% conducted sensitivity analyses related to overadjustment. CONCLUSIONS: Our findings suggest that overadjustment bias has received insufficient consideration in systematic reviews and meta-analyses of socio-economic inequalities in health. This is a critical issue given that overadjustment bias is likely to result in biased estimates of health inequalities and accurate estimates are needed to inform public health interventions. There is a need to highlight overadjustment bias in review guidelines.