Factors affecting subjective cognitive decline: an automated machine learning approach.

Area of Science:

• Machine Learning
• Cognitive Science
• Health Informatics

Background:

• Subjective Cognitive Decline (SCD) poses a growing public health challenge.
• Early detection of SCD is crucial for timely intervention and management.
• Existing screening methods may lack the precision needed for widespread application.

Purpose of the Study:

• To develop and validate a machine learning-based screening model for Subjective Cognitive Decline (SCD).
• To identify key psychological and clinical factors associated with SCD risk.
• To compare the performance of various machine learning algorithms for SCD prediction.

Main Methods:

• A retrospective cohort study utilized data from the 'Active Health' screening app.
• Machine learning models including Logistic Regression, Naive Bayes, SVM, Decision Tree, and Neural Networks were trained and validated.
• Model performance was evaluated using AUC, accuracy, sensitivity, specificity, precision, recall, and F1 score, with SHAP values for interpretability.

Main Results:

• The Support Vector Machine (SVM) model exhibited strong predictive performance with an AUC of 0.82.
• Key predictors for SCD included Information Overload (IO), Self-Perception (SP), Energy Level (EL), Depressive Emotion (DE), Gender (SEX), Risk Decision (RD), and Short-Term Memory (STM).
• The SVM model outperformed other evaluated algorithms in predicting SCD risk.

Conclusions:

• An SVM-based model was successfully developed for effective SCD risk screening.
• The developed model offers a promising tool for early identification of individuals at risk for SCD.
• Machine learning approaches, particularly SVM, show significant potential in enhancing cognitive decline screening.