BackgroundHousehold settings are high risk for COVID-19 transmission. Understanding transmission factors associated with environmental dwelling characteristics is important in informing public health and building design recommendations. We aimed to develop a directed acyclic graph (DAG) to inform a novel analytical study examining the effect of dwelling environmental characteristics on household transmission of COVID-19. MethodsKey demographic, behavioural and environmental dwelling characteristics were identified by a multidisciplinary team. Using the DAG to visually display risk factors, and using expert knowledge of available datasets we reached a consensus on the factors included and directionality of relationships to build the final conceptual framework. Factors were displayed as nodes and relationships as pathways. ResultsOf 34 potential factors, 16 were included in the DAG, with 13 causal and three biasing pathways. Three variables were not measurable using retrospective datasets. The DAG enabled us to select data sources for the pilot study period and to inform the analysis plan. Key exposure nodes were energy efficiency or dwelling age; dwelling type or number of storeys; and dwelling size. We determined direct and proxy confounders which we could adjust for, potential interactions terms we could test in model building, and co-linear variables to omit in the same model. ConclusionsThe DAG helped identify key variables and datasets. It prioritised key nodes and pathways to formalise complex relationships between variables. It was pivotal in identifying unobserved variables, confounders, co-linearity and potential interactions. It has supported data selection and design of a retrospective pilot study analysis plan.
Read full abstract