Deciphering key features of social resilience versus social vulnerability in later life: A biopsychosocial model of social asymmetry.

Abstract

Objectives: Confronted with shrinking social networks, older adults exhibit individual differences in social adaptability, reflected as socially resilient versus socially vulnerable. The purpose of this study was to examine key features that reflect this social asymmetry in later life.

Methods: Three datasets were analyzed, with the training set (N = 424) included older adults from China, while two test sets (N1 = 2877, N2 = 2343) were from the United States. Social asymmetry was assessed using residuals from a regression of social network on loneliness, with individuals with positive residuals categorized as socially vulnerable and those with negative residuals as socially resilient. Feature selection was performed with the Boruta algorithm, model building with the gradient boosting machine (GBM) algorithm, and model interpretation with the local interpretable model-agnostic explanations (LIME) algorithm.

Results: Socially resilient older adults were more prevalent than socially vulnerable ones across datasets from various cultural backgrounds. Five key features-depression, anxiety, stress, sleep disturbance, and personality-were found to predict social asymmetry, with area under the curve (AUC) values ranging from 0.76-0.86 across datasets. Older adults with lower levels of depression, anxiety, stress, and sleep disturbance, and typical A or B (versus intermediate) personality, were more likely to be socially resilient.

Discussion: The prevalence of socially resilient older adults indicates a relatively positive trend, and most of the key features are plastic and amenable, such as negative emotions and sleep behavior. Developing emotional regulation strategies and providing sleep hygiene education could improve the social adaptability of older adults.