A Multivariable Prediction Model for Mild Cognitive Impairment and Dementia: Algorithm Development and Validation.

Area of Science:

• Gerontology
• Computational Neuroscience
• Public Health

Background:

• Mild cognitive impairment (MCI) and dementia present diagnostic challenges, impacting individuals and healthcare systems.
• Early detection is crucial for timely intervention and mitigating the burden of dementia.
• Machine learning (ML) offers advanced data analysis capabilities for predicting cognitive decline.

Purpose of the Study:

• To evaluate the predictive accuracy of various ML models for identifying MCI and dementia onset.
• To utilize the Korean Longitudinal Study of Aging (KLoSA) dataset for this assessment.
• To identify key sociodemographic and health factors influencing cognitive impairment prediction.

Main Methods:

• Analysis of KLoSA data (2018-2020) from 4975 older adults (≥60 years).
• Application of multiple ML models (logistic regression, XGBoost, random forest, etc.) for prediction.
• Evaluation of model performance using Area Under the Receiver Operating Characteristic Curve (AUC) and Shapley values for feature importance.

Main Results:

• Random forest was the best model for MCI prediction (AUC 0.6729).
• XGBoost demonstrated superior performance for dementia prediction (AUC 0.8185).
• Key predictors for MCI included pain, widowhood, living alone, and exercise; for dementia, education level, exercise, and social engagement were significant.

Conclusions:

• ML algorithms, particularly XGBoost, show promise for predicting cognitive impairment in older adults.
• Current models require further refinement for robust MCI and dementia prediction accuracy.
• Sociodemographic and health factors are vital for early cognitive condition identification and intervention strategies.