Capturing the well-being exposome in poly-environmental scores.

In this study we use aggregated weighted scores of environmental effects to study environmental influences on well-being and happiness. To this end, we split a sample of Netherlands Twin Register (NTR) participants into a training (N = 4857) and test (N = 2077) sample. In the training sample, we use elastic net regression to estimate effect sizes for associations between life satisfaction and two sets of environmental variables: one based on self-report socioenvironmental data, and one based on objective physical environmental data. Based on these effect sizes, we create two poly-environmental scores (PES-S and PES-O, for self-reports and objective data respectively). In the test sample, we perform association analyses between different measures of well-being and the two PESs. We find that the PES-S explains ∼36% of the variance in well-being, while the PES-O does not significantly contribute to the model. Variance in other well-being measures (i.e., different life satisfaction domains, subjective happiness, quality of life, flourishing, psychological well-being, self-rated health, depressive problems, and loneliness) are explained to varying extents by the PESs, ranging from 6.36% (self-rated health) to 36.66% (loneliness). These predictive values did not change during the COVID-19 pandemic (N = 3214). Validating the PES-S in the UK biobank (N = 40,614), we find that the UK biobank PES-S explains about ∼12% of the variance in happiness. Lastly, we examine if there is any indication for gene-environment correlation (rGE), the phenomenon where one's genetic predisposition influences exposure to the environment, by associating the PESs with polygenic scores (PGS) in a sample of Netherlands Twin Register (NTR) and UK Biobank participants. While the PES and PGS were not correlated in the NTR sample, they were correlated in the larger UK biobank sample, indicating the potential presence of rGE. We discuss several limitations pertaining to our dataset, such as a potential influence of common method bias, and reflect on how PESs might be used in future research.