Identifying the top determinants of psychological resilience among community older adults during COVID-19 in Taiwan: A random forest approach

Abstract

Resilience in the context of the ongoing COVID-19 pandemic has emerged as a critical public health concern for the elderly population. However, the extent to which a structured model can effectively determine resilience among older adults remains uncertain. In this study, we sought to uncover the most influential factors of resilience using the Promotive and Protective Factors and Processes (PPFP) model. Our sample comprised 936 community-dwelling older adults aged 50 years and above. Through the implementation of random forest analysis and traditional logistic regression, we investigated potential determinants of resilience. Our results demonstrated the efficacy of random forest (RF) analysis, with the area under the receiver operating characteristic curve (AUC) ranging from 0.806 to 0.890 for different resilience score thresholds. Notably, the determinants of psychological resilience that emerged as most significant included stress, depression, self-rated socioeconomic status, exercise habits, and cognition. The application of the PPFP framework in our study offered substantial benefits to healthcare practitioners, enabling them to identify both internal and external variables that warrant intervention in bolstering resilience. Furthermore, the utilization of the random forest algorithm proved invaluable in machine learning, particularly for ranking the importance of resilience-related factors.

Keywords

• COVID-19
• Machine learning
• Older adults
• Promotive and Protective Factors and Processes (PPFP) model
• Resilience
• Random forest (RF)