Quality control of the ambulatory polygraphy using automatic analysis

Area of Science:

• Sleep Medicine
• Biomedical Engineering
• Data Science

Background:

• Home-based sleep studies require quality assurance due to potential variations.
• Automatic analysis offers a simplified approach for auditing sleep study data.
• Establishing a standardized reading method for home polygraphy is crucial for diagnosing sleep apnea-hypopnea.

Purpose of the Study:

• To develop a predictive model for visual reading of home sleep polygraphy using automatic analysis of saturation and respiratory signals.
• To establish a reliable reading standard for patients with clinical symptoms of sleep apnea-hypopnea.

Main Methods:

• Analysis of 189 home sleep studies from 218 patients.
• Two definitions of hypopnea were used: duration >or= 10s with signal decrease >30%, or associated with desaturation >or= 3% or >or= 4%.
• Multiple linear regression analysis was performed using automatic respiratory disturbance index (RDIa) and automatic desaturation index (DI3%a or DI4%a).

Main Results:

• Initial Bland-Altman analysis showed large limits of agreement between visual RDI and automatic RDIa/DI%a.
• Multiple linear regression models (RDIa + DI3%a or RDIa + DI4%a) showed acceptable agreement with visual RDI (r(2) = 96.2% and 97%, p < 0.001).
• The 95% confidence intervals for the differences were +/- 10.1 or +/- 8.8 events/hour, suggesting studies outside these limits require revision.

Conclusions:

• A predictive multiple regression model utilizing automatic oximetry and respiratory signal analysis can establish a standard for visual reading of home polygraphy.
• This approach enhances the reliability and consistency of sleep study interpretations.
• The model provides a framework for quality control in home sleep apnea diagnosis.