Outcome-Wide Longitudinal Designs for Causal Inference: A New Template for Empirical Studies

Abstract

In this paper, we propose a new template for empirical studies intended to assess causal effects: the outcome-wide longitudinal design. The approach is an extension of what is often done to assess the causal effects of a treatment or exposure using confounding control, but now, over numerous outcomes. We discuss the temporal and confounding control principles for such outcome-wide studies, metrics to evaluate robustness or sensitivity to potential unmeasured confounding for each outcome and approaches to handle multiple testing. We argue that the outcome-wide longitudinal design has numerous advantages over more traditional studies of single exposureoutcome relationships including results that are less subject to investigator bias, greater potential to report null effects, greater capacity to compare effect sizes, a tremendous gain in the efficiency for the research community, a greater policy relevance and a more rapid advancement of knowledge. We discuss both the practical and theoretical justification for the outcome-wide longitudinal design and also the pragmatic details of its implementation, providing publicly available R code.

The Role of Financial Fragility and Financial Control for Well-Being

“…An outcome-wide approach was applied, in which multiple outcomes were considered for a single exposure (VanderWeele 2017a ). This approach has been theoretically argued (VanderWeele 2017a ; VanderWeele et al 2020b , a ) and empirically confirmed as effective in presenting temporal associations across the whole spectrum of outcomes (Białowolski et al 2019 ; Chen et al 2019 ; Chen and Vanderweele 2018 ; Kim et al 2020 ; Pawlikowski et al 2019 ; Steptoe and Fancourt 2020 ; Węziak-Białowolska et al 2018 ). It has the advantage of avoiding cherry-picking of significant results only, limits the risk of p-hacking (Head et al 2015 ; Lakens 2015 ) and seems to facilitate reporting of the so-called ‘negative’ or ‘non-significant’ findings, which has been already proven problematic in scientific publishing due to resistance of journal editors to publish negative results (Fanelli 2010 , 2012 ; Ioannidis 2005 ; Matosin et al 2014 ).…”

“…Because of the longitudinal panel data employed, in contrast to many cross-sectional analyses, our approach attempted to offer more plausible evidence for causality by ensuring a logical temporal sequence of cause and effect from the data (VanderWeele et al 2020a , b ; VanderWeele et al 2016a , b ). Specifically, since two exposure variables were measured temporarily prior to outcomes, the intended causal links were more plausible.…”

Commentary: Developing best-practice guidelines for the reporting of E-values

“…This could be done in an online supplementary table, with some comment given in the main text. In outcome-wide settings in which numerous outcomes, 10 , 11 and thus numerous E-values, are assessed, a single table perhaps reporting the three largest covariate–outcome associations (properly inverted for protective associations) for each outcome could once again be helpful. In principle, a similar approach might also be used for the exposure, by providing the largest exposure–covariate relative risk associations across covariates.…”

Religious-service attendance and subsequent health and well-being throughout adulthood: evidence from three prospective cohorts

“…There has been growing interest in examining multiple categories of outcomes within the same study, which can potentially reduce publication bias and provide a broader picture of the complex religion–health dynamics. 24 , 25 Finally, religious involvement is often characterized by a cyclic structure across a lifetime, often with a return to religious community in later life. Prior work has, however, seldom compared the religion–health associations across life stages.…”