Decoding emotional resilience in aging: unveiling the interplay between daily functioning and emotional health

Background: EPs pose significant challenges to individual health and quality of life, attracting attention in public health as a risk factor for diminished quality of life and healthy life expectancy in middle-aged and older adult populations. Therefore, in the context of global aging, meticulous exploration of the factors behind emotional issues becomes paramount. Whether ADL can serve as a potential marker for EPs remains unclear. This study aims to provide new evidence for ADL as an early predictor of EPs through statistical analysis and validation using machine learning algorithms.

Methods: Data from the 2018 China Health and Retirement Longitudinal Study (CHARLS) national baseline survey, comprising 9,766 samples aged 45 and above, were utilized. ADL was assessed using the BI, while the presence of EPs was evaluated based on the record of "Diagnosed with Emotional Problems by a Doctor" in CHARLS data. Statistical analyses including independent samples t-test, chi-square test, Pearson correlation analysis, and multiple linear regression were conducted using SPSS 25.0. Machine learning algorithms, including Support Vector Machine (SVM), Decision Tree (DT), and Logistic Regression (LR), were implemented using Python 3.10.2.

Results: Population demographic analysis revealed a significantly lower average BI score of 65.044 in the "Diagnosed with Emotional Problems by a Doctor" group compared to 85.128 in the "Not diagnosed with Emotional Problems by a Doctor" group. Pearson correlation analysis indicated a significant negative correlation between ADL and EPs (r = -0.165, p < 0.001). Iterative analysis using stratified multiple linear regression across three different models demonstrated the persistent statistical significance of the negative correlation between ADL and EPs (B = -0.002, β = -0.186, t = -16.476, 95% CI = -0.002, -0.001, p = 0.000), confirming its stability. Machine learning algorithms validated our findings from statistical analysis, confirming the predictive accuracy of ADL for EPs. The area under the curve (AUC) for the three models were SVM-AUC = 0.700, DT-AUC = 0.742, and LR-AUC = 0.711. In experiments using other covariates and other covariates + BI, the overall prediction level of machine learning algorithms improved after adding BI, emphasizing the positive effect of ADL on EPs prediction.

Conclusion: This study, employing various statistical methods, identified a negative correlation between ADL and EPs, with machine learning algorithms confirming this finding. Impaired ADL increases susceptibility to EPs.