Evaluation of the diagnostic accuracy of automatic polygraphy analysis versus manual analysis in obstructive sleep apnea.

Abstract

Introduction: The high prevalence of obstructive sleep apnea (OSA) and limited healthcare resources have led to an increase in automated diagnostic testing in clinical practice. However, there is limited scientific support for this approach, which may pose risks for patients. Previous studies comparing automatic and manual OSA diagnostic methods often involve small sample sizes and lack standardized protocols.

Aim: To evaluate the diagnostic accuracy of the apnea-hypopnea index (REI) obtained through automatic versus manual polygraphy (PR) analysis and its performance across OSA severity levels.

Method: This prospective study included 3,144 subjects who underwent both automatic and manual PR analysis.

Results: A low diagnostic concordance (44.52%) was observed between automatic and manual PR analyses, indicating a poor diagnostic capacity of automated analysis. Sensitivity (S), specificity (E), positive predictive value (PPV), and negative predictive value (NPV) were assessed for OSA diagnosis at different severity levels. While high S and NPV were noted for excluding OSA, E and PPV were low. For mild OSA, S and PPV were particularly low, as were E and PPV. The accuracy was similarly poor for moderate and severe OSA, though severe OSA showed higher E and NPV. Finally, for very severe OSA, S remained low, contrasting with high E, PPV, and NPV.

Conclusions: Automated polygraphy shows limited diagnostic accuracy, especially in mild and moderate OSA. Despite good sensitivity to rule out disease, its low specificity may lead to misclassification. Manual analysis remains essential for reliable diagnosis.