Using Artificial Intelligence to assess the impact of social, physical, and financial health and personality on subjective well-being in a representative, multinational sample of older European and Israeli adults.

Background: Subjective well-being (SWB) is an important outcome influenced by other aspects of health and personality. However, we know little about the independent effects of multiple health and personality dimensions on SWB in large, representative international samples. Artificial Intelligence (AI) models are particularly well-suited to detect multi-factor patterns in complex topics such as SWB. Methods: This study involved a representative sample of 37 991 older adults from 17 European countries and Israel. Machine-learning algorithms, general additive modelling, low-degree polynomials (i.e. splines), and regression analyses were used to determine the independent effects of the Big 5 personality traits on social, physical and financial health factors, and the impact of all of these on an aggregate measure of SWB. Results: Loneliness, overall physical health, and making ends meet were the strongest social, physical and financial health predictors of SWB, respectively (absolute value (|t|s) = 29.77-51.53). Neuroticism was a consistent, negative determinant of health (|t|s = 2.82-11.42), but reduced the adverse impact of poor physical health on SWB (|t|s = 4.57-5.98). Extraversion was linked to better social and financial health (|t|s = 2.96-7.74), but also to higher body mass index (Student's t test (t) = 13.52), while openness to experience was positively associated with social and physical health (|t|s = 3.02-7.86), but negatively related to income (t = -19.76). Conclusions: All adverse health factors and neuroticism were linked to lower SWB, while SWB was positively associated with the other health measures and personality traits. Some traits had unexpected effects on health outcomes, and some had moderating effects on the links between these outcomes and SWB, suggesting that the links between personality, health and SWB depend on the types of health considered. Future multivariate modelling is recommended to clarify the mechanisms for these and other observed relationships.