Using artificial intelligence to identify the top 50 independent predictors of subjective well-being in a multinational sample of 37,991 older European & Israeli adults

Abstract

Abstract Subjective well-being (SWB) is widely recognized as an important health outcome, but its complexity, myriad predictors, and analytic requirements pose significant challenges to identifying the relative order and impact of SWB determinants. This study involved a representative sample of 37,991 older adults from 17 European countries and Israel. An aggregate index of SWB was developed and compared across countries, and machine-learning algorithms were used to rank-order the strongest 50 (of an initial 94) SWB predictors from 15 categories. General Additive Modeling (GAM) and low-degree polynomials (i.e., splines) were used to determine the independent effect sizes and significance levels for each of these top-50 SWB predictors. Of the 18 countries included in this study, Denmark had the highest mean SWB, while Greece had the lowest. The two top-ranked SWB predictors (loneliness, social activity satisfaction) were social factors, which also had the highest overall group ranking, followed by physical health, demographics, financial status and personality. Self-reported health was the strongest health-related predictor, neuroticism was the strongest personality predictor, and women reported higher SWB than men. SWB decreased with age, and increased with income up to 350,000 euros/year, after which it declined. Social factors were of primary importance for subjective well-being in this research, while childhood experiences and healthcare status exerted the smallest effects. The vast majority of the top 50 SWB predictors were statistically significant, with the notable exceptions of body mass index and most health behaviors, which may impact SWB indirectly through their effects on physical health. Future multivariate modeling is recommended to clarify the mechanisms for these and other observed relationships.