Predicting Depression Among Chinese Patients with Narcolepsy Type 1: A Machine-Learning Approach

Area of Science:

• Neurology
• Psychiatry
• Computational Medicine

Background:

• Depression is a prevalent comorbidity in narcolepsy type 1 (NT1).
• Accurate prediction of depression is crucial for effective management in NT1 patients.
• Machine learning (ML) offers novel approaches for identifying predictive factors.

Purpose of the Study:

• To identify factors predicting depression in Chinese NT1 patients.
• To evaluate the efficacy of ML models in depression prediction for NT1.
• To develop a data-driven tool for personalized risk assessment.

Main Methods:

• Recruited 203 drug-free NT1 patients (aged 5-61) diagnosed via ICSD-3 criteria.
• Assessed depression, sleepiness, and impulsivity using validated scales (CES-DC/SDS, ESS/ESS-CHAD, BIS-11).
• Employed Logistic Regression (LR), Random Forest (RF), and Support Vector Machine (SVM) models for prediction, evaluating performance with AUC, accuracy, precision, recall, F1, and DCA.

Main Results:

• Logistic Regression identified hallucinations and motor impulsivity as significant depression predictors in NT1.
• Support Vector Machine (SVM) achieved the highest performance: AUC 0.653, accuracy 0.659, sensitivity 0.727, F1 score 0.696.
• SVM effectively integrated sleep-related and psychosocial data for prediction.

Conclusions:

• ML models, particularly SVM, show significant potential for predicting depression in NT1 patients.
• The findings support the development of personalized, data-driven tools for depression risk stratification.
• Further validation in diverse populations and inclusion of additional psychological variables are recommended to enhance predictive accuracy.