Constructing prediction models for excessive daytime sleepiness by nomogram and machine learning: A large Chinese...

Area of Science:

• Neurology
• Sleep Medicine
• Biostatistics

Background:

• Excessive daytime sleepiness (EDS) is a common yet under-researched symptom in Parkinson's disease (PD).
• Existing predictive models for EDS in PD are limited, necessitating the development of robust, cohort-based tools.
• Understanding risk factors is key to early identification and management of EDS in PD patients.

Purpose of the Study:

• To develop a predictive model for excessive daytime sleepiness (EDS) in Parkinson's disease (PD) patients.
• To utilize a nomogram and machine learning (ML) techniques for enhanced predictive accuracy.
• To identify significant baseline predictors of incident EDS in a PD cohort.

Main Methods:

• A 1-year longitudinal study involving 995 Parkinson's disease (PD) patients without baseline EDS.
• Data collection included demographics, motor, and non-motor symptoms.
• Prediction models were constructed using Cox proportional risk regression and XGBoost machine learning (ML).

Main Results:

• Excessive daytime sleepiness (EDS) developed in 26.13% of patients over one year.
• Significant baseline predictors identified: high body mass index (BMI), late age of onset (AOO), low PDQ-39 motor score, low MMSE orientation score, and absence of freezing of gait (FOG).
• The XGBoost model achieved 71.86% accuracy, with BMI, AOO, PDQ-39 motor score, MMSE orientation score, and FOG as key predictors in descending order of importance.

Conclusions:

• High BMI, late AOO, low PDQ-39 motor score, low MMSE orientation score, and absence of FOG are significant risk factors for EDS in PD.
• The developed predictive model demonstrates relative effectiveness and accuracy in identifying PD patients at risk for EDS.
• These findings can aid in early risk stratification and targeted interventions for EDS in Parkinson's disease.