Brief digital sleep questionnaire powered by machine learning prediction models identifies common sleep disorders.

Area of Science:

• Sleep Medicine
• Computational Health
• Public Health Screening

Background:

• Common sleep disorders such as insomnia, delayed sleep phase syndrome (DSPS), insufficient sleep syndrome (ISS), and obstructive sleep apnea (OSA) significantly impact public health.
• Accurate and efficient screening tools are crucial for early identification and management of these prevalent conditions.

Purpose of the Study:

• To develop and validate an abbreviated Digital Sleep Questionnaire (DSQ) capable of identifying common sleep disturbances.
• To assess the diagnostic accuracy of the DSQ using machine learning models compared to physician diagnoses.

Main Methods:

• The DSQ was administered to a large cohort of community volunteers (n=2113).
• Machine learning (ML) models, specifically ElasticNet, were trained and validated using physician diagnoses as the reference standard.
• Model performance was evaluated using metrics including sensitivity, specificity, accuracy, and Area Under the Receiver Operating Curve (AUC).

Main Results:

• The DSQ effectively identified key indicators associated with physician-diagnosed insomnia, DSPS, ISS, and OSA.
• Validated ElasticNet models demonstrated high sensitivity (80-83%) and high AUC (0.80-0.85) for diagnosing sleep disorders.
• Model accuracy in agreement with physician diagnoses ranged from 68-73%.

Conclusions:

• The abbreviated DSQ is a brief, engaging tool for efficient large-scale screening of common sleep disorders.
• ML-powered DSQ demonstrates potential for accurate classification of sleep disturbances, aiding in the improvement of population sleep, health, productivity, and safety.