Using machine learning model for predicting risk of memory decline: A cross sectional study.

Area of Science:

• Neuroscience
• Medical Informatics
• Biostatistics

Background:

• Memory decline is an early indicator of neurodegenerative diseases like Alzheimer's disease (AD).
• Predicting and identifying risk factors for memory decline remains a significant challenge in clinical practice.
• Early detection is crucial for timely intervention and management of cognitive impairment.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting memory decline risk factors in US adults.
• To identify key predictors associated with memory deterioration using advanced analytical techniques.
• To enhance the accuracy of early risk assessment for cognitive decline.

Main Methods:

• Utilized data from 9971 individuals in the National Health and Nutrition Examination Survey (NHANES) 2015-2016.
• Applied the least absolute shrinkage and selection operator (LASSO) for predictor screening.
• Evaluated five ML algorithms: Logistic Regression, ExtraTrees classifier, Bagging classifier, eXtreme Gradient Boosting (XGBoost), and Random Forest (RF).

Main Results:

• The final sample included 4525 subjects, with 7.7% experiencing memory deterioration.
• ExtraTrees classifier and XGBoost models exhibited superior prediction performance with Area Under Curve (AUC) values of 0.915 and 0.911, respectively.
• These models demonstrated consistent accuracy on external datasets (AUC 0.851 and 0.843).

Conclusions:

• ExtraTrees classifier and XGBoost models are highly effective in predicting memory decline.
• These ML models show significant clinical value for identifying individuals at risk of cognitive impairment.
• Further research is warranted to validate these findings and explore clinical implementation.