Prediction Model for Cognitive Impairment among Disabled Older Adults: A Development and Validation Study

Area of Science:

• Gerontology
• Neurology
• Public Health

Background:

• Disabled older adults face an elevated risk of cognitive impairment.
• Early identification of cognitive impairment is essential for reducing disease burden.
• Existing prediction tools may not adequately address this vulnerable population.

Purpose of the Study:

• To develop and validate a robust prediction model for identifying cognitive impairment in disabled older adults.
• To compare the performance of logistic regression against machine learning models (SVM, Random Forest, XGBoost).
• To create a user-friendly nomogram for practical clinical application.

Main Methods:

• Utilized a large cohort with 2138 participants for model development and two external validation sets (501 and 746 participants).
• Employed logistic regression, support vector machine, random forest, and XGBoost algorithms.
• Developed a nomogram based on significant predictors from the logistic regression model.

Main Results:

• Logistic regression demonstrated superior predictive performance (AUC=0.875) compared to machine learning models.
• The logistic regression model achieved high precision (0.808), specification (0.788), sensitivity (0.770), and F1-score (0.788).
• The final nomogram, incorporating age, daily living activities, instrumental activities of daily living, hearing, and visual impairment, showed strong validation (AUCs: 0.871, 0.825, 0.863).

Conclusions:

• A logistic regression-based nomogram effectively predicts cognitive impairment in disabled older adults.
• The nomogram provides a reliable and accessible tool for early risk identification.
• This model can support timely interventions to mitigate the impact of cognitive decline in this population.