Identification of sleep phenotypes in COPD using machine learning-based cluster analysis

Area of Science:

• Pulmonary Medicine
• Sleep Medicine
• Data Science

Background:

• Disturbed sleep significantly impacts quality of life and predicts adverse outcomes in patients with chronic obstructive pulmonary disease (COPD).
• Objective sleep parameters are crucial for understanding disease progression and patient stratification in COPD.

Purpose of the Study:

• To identify distinct phenotypic clusters in COPD patients using objective sleep parameters.
• To evaluate the association between these clusters and all-cause mortality for improved risk stratification.

Main Methods:

• Longitudinal observational cohort study utilizing nationwide Veterans Health Administration data.
• Unsupervised machine learning (K-means clustering) applied to polysomnography data from 9992 COPD patients.
• Cox regression analysis and Kaplan-Meier estimates used to assess mortality associations.

Main Results:

• Five distinct patient clusters were identified based on age, comorbidity burden, and sleep parameters.
• Mortality rates varied significantly across clusters, from 9.4% to 42% overall.
• Total sleep time and sleep efficiency demonstrated significant associations with mortality in specific clusters.

Conclusions:

• Objective sleep parameters are vital for phenotypic characterization and mortality risk assessment in COPD.
• The identified clusters offer a novel approach to stratify COPD patients based on sleep characteristics.
• Further research into sleep-disordered breathing interventions may improve outcomes for high-risk COPD patients.