A CPAP data–based algorithm for automatic early prediction of therapy adherence

Objective Adherence is a critical issue in the treatment of obstructive sleep apnea with continuous positive airway pressure (CPAP). Approximately 40% of patients treated with CPAP are at risk of discontinuation or insufficient use (< 4 h/night). Assuming that the first few days on CPAP are critical for continued treatment, we tested the predictive value at day 14 (D14) of the Philips Adherence Profiler™ (AP) algorithm for adherence at 3 months (D90). Method The AP™ algorithm uses CPAP machine data hosted in the database of EncoreAnywhere™. This retrospective study involved 457 patients (66% men, 60.0 ± 11.9 years; BMI = 31.2 ± 5.9 kg/m 2 ; AHI = 37.8 ± 19.2; Epworth score = 10.0 ± 4.8) from the Pays de la Loire Sleep Cohort . At D90, 88% of the patients were adherent as defined by a mean daily CPAP use of ≥ 4 h. Results In a univariate analysis, the factors significantly associated with CPAP adherence at D90 were older age, lower BMI, CPAP adherence (≥ 4 h/night) at D14, and AP™ prediction at D14. In a multivariate analysis, only older age (OR 2.10 [1.29–3.41], p = 0.003) and the AP™ prediction at D14 (OR 16.99 [7.26–39.75], p < 0.0001) were significant predictors. CPAP adherence at D90 was not associated with device-derived residual events, nor with the levels of pressure or leakage except in the case of very significant leakage when it persisted for 90 days. Conclusion Automatic telemonitoring algorithms are relevant tools for early prediction of CPAP therapy adherence and may make it possible to focus therapeutic follow-up efforts on patients who are at risk of non-adherence.