Identification of Predictors of Adaptability in Older Adults Based on the Roy Adaptation Model Using Machine Learning

Area of Science:

• Gerontology
• Nursing Theory
• Health Psychology

Background:

• The Roy Adaptation Model emphasizes physical, psychological, and social factors in later life adaptation.
• Limited empirical research exists on predictive models integrating these domains for older adults.
• This study addresses the need for predictive evidence on adaptation in older women.

Purpose of the Study:

• To identify key predictors of adaptive classification in older adult women.
• To utilize functional and subjective well-being measures for predictive modeling.
• To explore the interplay of physical and psychosocial factors in adaptation.

Main Methods:

• A predictive study design was employed with older women in community exercise programs.
• Data collected using the Senior Fitness Test (SFT), SF-12, and WHO-5 questionnaires.
• Random Forest multiclass classification models were trained and evaluated using robust metrics and interpretability techniques.

Main Results:

• The Random Forest model achieved 74% accuracy and a 0.73 macro-F1 score under oversampling.
• Key predictors identified include the SF-12 physical and mental components and the 2 min step and chair sit-and-reach tests.
• Model robustness analyses showed performance variations, particularly for the 'High' adaptation class.

Conclusions:

• Physical and psychosocial factors jointly contribute to adaptive processes in older women, supporting the Roy Adaptation Model.
• Exploratory evidence supports the integrated use of SFT, SF-12, and WHO-5 for assessing adaptation.
• Further external validation and longitudinal studies are recommended before clinical application.