1ST RECETAS CAUSAL INFERENCE WORKSHOP

 

On March 4, 2022, RECETAS held the 1st Causal-Inference Workshop. Partners of all six study sites were invited.

RECETAS aims at evaluating nature-based social prescribing (NbSP) within Europe, Latin America, and Australia. Among other aspects, we will evaluate the causal effect of NbSP on loneliness. The purpose of the workshop was to present the planned causal analyses, discuss data that might be needed for causal inference analyses, and learn about data and their format that are planned to be collected.

For causal analyses, it is important to either have a randomized controlled trial or – in the case of observational studies – to have sufficient data to control for confounding. For the assessment of a sustained intervention effect, for both the experimental and the observational setting, it is important to control for post-treatment-start time-varying confounding that may occur when the intervention impacts factors that again impact adherence (intervention-confounder loop). To identify these intervention-confounder loops, causal diagrams, more specifically, directed acyclic graphs (DAGs), are a helpful visual and analytical tool to discuss causal relations on the pathway between the intervention (NbSP) and outcome (loneliness) and to identify variables that are relevant for a valid causal analysis.

After Ursula Rochau welcomed everyone to the workshop, Felicitas Kuehne introduced the causal concept as background for the planned causal analyses, Uwe Siebert presented the concept and logic of causal graphs and specifically DAGs, and Laura Coll Planas introduced the logic model concept. These presentations lead to a lively and fruitful discussion including all participants regarding variables that are planned to be and should be collected at the different study sides.

The focus of the discussion was on defining factors leading to participating in or abandoning the NbSP-programs, mode of action of the programs, and whether these factors also have an influence on loneliness. Furthermore, it was debated how information overload in the questionnaire can be avoided, and how the feasibility studies may help to identify factors leading to discontinuation of the programs.

The results of the workshop will help us defining the causal diagram, which will be applied to identify factors that are crucial for drawing causal conclusions on the effect of NBSP-programs on loneliness. As next steps, we will create a draft of a causal graph, discuss the causal relations, and agree on a final causal graph. This graph will then be used to identify factors that need to be collected in the trials.

Aller au contenu principal