Abstract
Background: It is estimated that a seventh of the world’s population have obstructive sleep apnoea (OSA). Even if all were to be offered the highly efficacious Continuous Positive Airway Pressure (CPAP) therapy, 46-83% of patients do not adhere to CPAP, even in trials. CPAP adherence in clinical practice is unclear and relationships with sleep centre treatment pathways unknown. Clinical practice statements are based on small studies reporting adherent and non-adherent behaviour developing from treatment onset. We addressed these evidence gaps using a large multi-centre clinical dataset using sleep centre treatment pathway changes during the COVID-19 pandemic as a natural experiment.
Methods: Five UK sleep centres that telemonitor CPAP-usage data were recruited. Objective CPAP-usage data from the first three treatment months were collected from 100 patients who started CPAP pre-pandemic (April 2019) and 100 patients peri-pandemic (September 2020), per centre. CPAP adherence was defined using accepted criteria (mean CPAP use ≥4 hours/night for ≥70% of nights, except median use for Nights 1-3). Growth mixture modelling (GMM) and logistic regression were performed using the entire (1000-patient) dataset.
Findings: Three months after treatment started, 34% of patients were treatment-adherent in 2019 and 42% in 2020 (p=0.24). GMM identified six distinct CPAP-usage behaviours over Month 1, each with a different likelihood of CPAP non-adherence at Month 3. Four involved changing CPAP use (54% of patients), two comprised consistent use (46%). Logistic regression with an additional 200-patient dataset demonstrated that early behaviours strongly predicted Month 3 CPAP adherence, contributing 86% to the full model.
Interpretation: This behavioural classification can explain why current practice produces poor CPAP adherence; the different groups are not differentiated and managed appropriately. Guidelines and practice should be changed to precision medicine based on specific behaviour from Week 2 to improve CPAP adherence from current levels.
Funding: Imperial College London Masters grant.
Declaration of Interest: We do not have any competing/conflicting interests with regards to the manuscript submitted.
Ethical Approval: The study was approved by Imperial College London Research Governance Department and London Queen’s Square Research Ethics Committee (23/LO/0142, appendix p2-4).
Methods: Five UK sleep centres that telemonitor CPAP-usage data were recruited. Objective CPAP-usage data from the first three treatment months were collected from 100 patients who started CPAP pre-pandemic (April 2019) and 100 patients peri-pandemic (September 2020), per centre. CPAP adherence was defined using accepted criteria (mean CPAP use ≥4 hours/night for ≥70% of nights, except median use for Nights 1-3). Growth mixture modelling (GMM) and logistic regression were performed using the entire (1000-patient) dataset.
Findings: Three months after treatment started, 34% of patients were treatment-adherent in 2019 and 42% in 2020 (p=0.24). GMM identified six distinct CPAP-usage behaviours over Month 1, each with a different likelihood of CPAP non-adherence at Month 3. Four involved changing CPAP use (54% of patients), two comprised consistent use (46%). Logistic regression with an additional 200-patient dataset demonstrated that early behaviours strongly predicted Month 3 CPAP adherence, contributing 86% to the full model.
Interpretation: This behavioural classification can explain why current practice produces poor CPAP adherence; the different groups are not differentiated and managed appropriately. Guidelines and practice should be changed to precision medicine based on specific behaviour from Week 2 to improve CPAP adherence from current levels.
Funding: Imperial College London Masters grant.
Declaration of Interest: We do not have any competing/conflicting interests with regards to the manuscript submitted.
Ethical Approval: The study was approved by Imperial College London Research Governance Department and London Queen’s Square Research Ethics Committee (23/LO/0142, appendix p2-4).