An efficient non-invasive model for predicting cognitive impairment based on comprehensive geriatric assessment:...

Area of Science:

• Gerontology
• Artificial Intelligence in Healthcare
• Public Health

Background:

• Global population aging presents a significant challenge with rising rates of cognitive impairment, including Alzheimer's disease and related dementias.
• There is a critical need for accessible and convenient screening tools to enable early detection and intervention for cognitive decline.
• Comprehensive Geriatric Assessment (CGA) offers a rich dataset for exploring potential risk factors in the elderly population.

Purpose of the Study:

• To develop an efficient and noninvasive risk assessment model for identifying potential cognitive impairment in the elderly.
• To leverage machine learning algorithms applied to Comprehensive Geriatric Assessment (CGA) data for cognitive impairment prediction.
• To identify key noninvasive predictors that can facilitate early screening and intervention strategies.

Main Methods:

• 1410 participants aged 50+ were recruited from geriatric clinics and communities.
• Feature selection combined expert knowledge and machine learning on CGA indicators.
• Multiple machine learning models (Logistic Regression, Naive Bayes, SVM, Neural Networks, Random Forests) were evaluated, with Logistic Regression selected for the final model. Shapley Additive exPlanations (SHAP) were used for model interpretation.

Main Results:

• Thirteen noninvasive predictors were identified, including activities of daily living (Bathing, Homekeeping, Housework), age, caregiver status, sleep duration, hobbies, and cognitive-related difficulties (Focusing Difficulty, UI effect).
• The Logistic Regression model demonstrated strong performance with an Area Under the Curve (AUC) of 0.877, accuracy of 0.815, sensitivity of 0.767, and specificity of 0.827 on the test set.
• SHAP analysis confirmed the clinical relevance of the identified predictors in the context of cognitive impairment.

Conclusions:

• This study successfully identified convenient, noninvasive predictors for cognitive impairment screening.
• An efficient machine learning model, interpretable via SHAP analysis, was developed for predicting cognitive impairment risk.
• The findings provide a foundation for widespread screening, guiding early detection and intervention in high-risk elderly populations.