Using machine learning to identify predictors of self-perceptions of aging among older adults in England

Abstract

Objectives

This study aimed to identify the strongest predictors of self-perceptions of aging (SPA) from an extensive set of 55 variables, including demographic, psychological, social, and economic factors.

Methods

Data were drawn from the English Longitudinal Study of Ageing, comprising over 7,000 adults aged 50 and older. Two advanced machine learning models, Random Forest and extreme gradient boosting (XGBoost), were used for data analysis. This approach allowed for a comprehensive evaluation of the relative importance of each predictor.

Results

Psychological factors emerged as the strongest predictors of positive SPA, outweighing health-related and demographic variables. Sense of control was identified as the strongest predictor, followed by pleasure derived from daily life experiences and the perception of life as worthwhile. Other significant predictors included components of emotional well-being (e.g., anxiety and happiness), autonomy, self-realization, and the quality of interpersonal relationships.

Discussion

The results indicate that a sense of competence, autonomy, and relatedness forms a vital foundation for positive self-perceptions of aging. These factors are enhanced by both hedonic and eudaimonic experiences, which contribute to the emotional and existential richness of the aging process. The findings highlight opportunities for targeted interventions and the refinement of existing theoretical models.