Article
Using directed acyclic graphs (DAGs) to guide trial protocol design investigating the effect of nature-based prescribing on physical and mental health
Search Medline for
Authors
Published: | August 30, 2022 |
---|
Outline
Text
Background/research question: Loneliness is a common condition impacting productivity and wellbeing. Nature-based social prescribing (NBSP) may promote nature contact, activate and strengthen social structures, and improve longer term mental and physical health.
In six studies (three observational and three randomized controlled trials), the RECETAS consortium aims at evaluating how NBSP can reduce feelings of loneliness and improve quality of life (QoL) in urban contexts. Observational studies bear the risk of confounding-by-indication, and trials investigating sustained intervention effects have the risk of post-randomization bias.
Indicators for NBSP (e.g., low socioeconomic status, low education, recent immigration, isolation, old age, and unlikely users of nature and outdoor spaces) are also risk factors for the outcomes (loneliness, poor health status, and decreased QoL) and simultaneously are affected by the NBSP interventions. This bears the risk of time-dependent confounding. Therefore, proper adjustment for time-dependent confounding using causal inference methods such as g-methods, are needed to draw causal conclusions from such study.
In this study, we aim at identifying potential baseline and time-dependent confounding, structure based as well as other biases using causal diagrams, and proposing analytic pathways to account for the biases.
Methods: Our framework is based on directed acyclic graphs (DAGs), which are a visual and structured approach to identify variables directly or indirectly influencing the action and outcome of interest and their relation. This approach may help to identify biases, such as confounding and selection bias, and draw conclusions on the required statistical analytic framework essential for a causal interpretation of the results.
Experts in social science, epidemiology, and causal inference discuss and create DAGs for several nature-based interventions and their effect on loneliness, quality of life, physical, and mental health. Each potential intervention and corresponding outcome of interest will be discussed separately
Preliminary/expected results, outlook: One DAG per intervention and outcome will be presented. These DAGs will be used to identify open backdoor paths and to identify causal analytic frameworks that allow for drawing causal conclusions from the studies. The findings will guide and improve the design of planned studies and assure sufficient data collection to enable the application of adequate analytic methods.
Competing interests: We have no conflicts of interests