Accessible moderate-to-severe obstructive sleep apnea screening tool using multidimensional obesity indicators as...

Area of Science:

• Medical Informatics
• Public Health
• Biostatistics

Background:

• Obesity indicators are linked to adiposity and its distribution.
• Assessing moderate-to-severe obstructive sleep apnea (OSA) risk can benefit from multidimensional obesity indicators.
• Automated models can improve the efficiency of OSA risk assessment.

Purpose of the Study:

• To develop and validate machine learning models for predicting moderate-to-severe OSA.
• To utilize multidimensional obesity indicators as features for OSA risk prediction.
• To compare the performance of various machine learning algorithms in OSA detection.

Main Methods:

• Trained and validated logistic regression and five machine learning algorithms on clinical and community datasets.
• Employed 19 obesity indicators as input features for the models.
• Utilized Light Gradient Boosting Machine (LGB) for its superior performance.

Main Results:

• The Light Gradient Boosting Machine (LGB) model demonstrated superior calibration and clinical utility compared to other algorithms.
• The LGB model achieved considerable accuracy in predicting moderate-to-severe OSA in both clinical and community settings.
• The developed model effectively utilized 19 obesity indicators for prediction.

Conclusions:

• Machine learning, particularly the LGB algorithm, can accurately predict moderate-to-severe OSA using multidimensional obesity indicators.
• The developed model shows feasibility for real-world application in effectively detecting undiagnosed moderate-to-severe OSA.
• Deployment of this model via a user-friendly interface can enhance OSA screening in clinical and community settings.