DAGs of occupation and COVID V1.pdf

Directed acyclic graphs (DAGs) may be used to represent our knowledge (or assumptions) about a data-generating process. A DAG may then be used to identify which variables should or should not be adjusted for in a statistical analysis in order to answer particular questions relating to the effect of an exposure on an outcome. The purpose of this document is to collate directed acyclic graphs (DAGs) that have been used in the study of the effects of occupation on COVID-19-related health outcomes (e.g. infection, severity, mortality). By collating DAGs, we hope to identify points of consensus and of contention. We also hope that this collection might form the basis for critical discussion, which may in turn lead to new proposals.This document is intended to be updated as new DAGs concerning occupation and COVID-19-related outcomes emerge. However, we have not undertaken any sort of systematic searchstrategy in order to identify relevant DAGs, and so this collection is unlikely to be comprehensive. The document represents the beginning of a live and ongoing data collection exercise, rather than one which is complete.We present DAGs in alphabetical order (first author surname) together with limited details relating to the objectives of the study, the design of each study, the methodology for constructing the DAG, and details of how the DAG was used.We then present a table of minimal sufficient adjustment sets for the total effect of occupation on outcome implied by each DAG.The document concludes with some brief comments and observations about the DAGs, and information about how to leave feedback is provided. We have redrawn some of the DAGs using Daggity (daggity.net) for consistency of presentation.