Decision tree based diagnostic system for moderate to severe obstructive sleep apnea.

Area of Science:

• Cardiology
• Pulmonology
• Medical Informatics

Background:

• Obstructive sleep apnea (OSA) is a significant health concern linked to cardiovascular issues.
• Current diagnostic methods like polysomnography (PSG) are impractical for widespread screening.
• Existing prediction models perform poorly in Chinese populations.

Purpose of the Study:

• To develop and validate a simple, accurate clinical prediction system for moderate to severe OSA in Taiwan.
• To improve OSA screening efficiency and accessibility.

Main Methods:

• Integrated expert-based feature extraction with decision tree algorithms.
• Employed backward stepwise multivariable logistic regression and four other decision tree algorithms for comparison.
• Utilized non-invasive features: Sex, Age, and Average Systolic Blood Pressure (AveSBP).

Main Results:

• The proposed prediction formula achieved high accuracy (96.9%), with sensitivity of 98.2% and specificity of 93.2%.
• Outperformed all other compared prediction formulas.
• Demonstrated reliability and simplicity for clinical use.

Conclusions:

• The developed prediction formula is a valuable tool for screening moderate to severe OSA in Taiwanese patients.
• It aids in prioritizing patients for PSG and avoids unnecessary testing for low-risk individuals.