Development and Validation of Prediction Models for Severe Obstructive Sleep Apnea Based on Periodic Health...

Area of Science:

• Occupational Health
• Sleep Medicine
• Medical Diagnostics

Background:

• Obstructive sleep apnea (OSA) poses significant occupational health risks, including reduced productivity and increased accident rates.
• Traditional screening methods relying on questionnaires often miss OSA due to a lack of subjective symptoms.
• There is a need for objective, questionnaire-independent screening tools for severe OSA in occupational settings.

Purpose of the Study:

• To develop and validate a prediction model for severe obstructive sleep apnea (OSA).
• To utilize readily available periodic health examination (PHE) data for objective OSA screening.
• To create a model independent of subjective symptoms, enhancing diagnostic accuracy.

Main Methods:

• A multivariable prediction model was developed following TRIPOD guidelines, analyzing data from 671 patients.
• Eight predictors from routine PHE data (age, sex, obesity, hypertension, diabetes, dyslipidemia, polycythemia, liver dysfunction) were used.
• Severe OSA was defined by apnea-hypopnea index (AHI) ≥ 30 or oxygen desaturation index (ODI) ≥ 30.
• Internal validation used bootstrap samples; external validation involved 100 university employees' oxygen saturation data.

Main Results:

• The prediction models achieved areas under the receiver operating characteristic curve of 0.67 (AHI-based) and 0.72 (ODI-based).
• Internal validity was found to be generally acceptable.
• External validation demonstrated high performance: the AHI model showed 1.00 sensitivity and 0.95 specificity; the ODI model showed 0.50 sensitivity and 0.97 specificity.

Conclusions:

• Two validated predictive models for severe OSA were developed using periodic health examination data.
• These models offer a practical, objective screening method for occupational physicians and clinicians.
• The findings support the integration of these models into routine health assessments for early OSA detection.