Multidimensional prediction of continuous positive airway pressure adherence

Abstract

Objective

Continuous positive airway pressure (CPAP) is the standard treatment for obstructive sleep apnea (OSA). Unsatisfactory adherence to CPAP is an important clinical issue to resolve. Cluster analysis is a powerful tool to distinguish subgroups in a multidimensional fashion. This study aimed to investigate the use of cluster analysis for predicting CPAP adherence using clinical polysomnographic (PSG) parameters and patient characteristics.

Patients/methods

Participants of this multicenter observational study were 1133 patients with OSA who were newly diagnosed and implemented CPAP. Ward's method of cluster analysis was applied to in-laboratory diagnostic PSG parameters and patient characteristics. CPAP adherence was assessed during 90- and 365-day periods after CPAP initiation in each cluster. We adopted the Centers for Medicare and Medicaid Services criterion for CPAP adherence, i.e., CPAP use ≥4 h per night for 70 % or more of the observation period. Logistic regression analysis was performed to stratify clusters according to CPAP adherence.

Results

Five clusters were identified through cluster analysis. Clustering was significantly associated with CPAP adherence at 90- and 365-day periods after CPAP initiation. Logistic regression revealed that the cluster with features including apnea predominant sleep-disordered breathing, high apnea-hypopnea index, and relatively older age demonstrated the highest CPAP adherence.

Conclusion

Cluster analysis revealed hidden connections using patient characteristics and PSG parameters to successfully identify patients more likely to adhere to CPAP for 90 days and up to 365 days. When prescribing CPAP, it is possible to identify patients with OSA who are more likely to be non-adherent.