Respiratory Rehabilitation Index (R2I): Unsupervised Clustering Approach to Identify COPD Subgroups Associated with...

Area of Science:

• Pulmonary Medicine
• Data Science in Healthcare
• Rehabilitation Science

Background:

• Chronic obstructive pulmonary disease (COPD) presents heterogeneous endotypes and clinical manifestations, complicating outcome prediction.
• Existing univariate markers may fail to capture complex interactions crucial for prognosis in COPD.
• Unsupervised clustering offers a data-driven approach to integrate multifactorial patient data for improved prediction.

Purpose of the Study:

• To apply unsupervised clustering to pre-rehabilitation characteristics of COPD patients.
• To identify patient subgroups predictive of discharge outcomes following pulmonary rehabilitation.
• To explore the utility of clustering for personalized COPD rehabilitation strategies.

Main Methods:

• 126 COPD patients undergoing pulmonary rehabilitation were analyzed.
• Admission assessments included forced oscillation technique, spirometry, and the six-minute walk test (6MWT).
• K-means clustering identified patient subgroups (R2I) associated with 6MWT improvement (discharge vs. admission).

Main Results:

• K-means clustering (Ncl = 2) yielded an optimal respiratory rehabilitation index (R2I).
• The R2I significantly associated with the minimal clinically important difference in 6MWT outcomes.
• Patients with R2I=1 (severe impairments) showed significantly higher post-rehabilitation functional improvement (p=0.032).

Conclusions:

• Unsupervised clustering effectively identifies distinct COPD patient subgroups.
• These subgroups exhibit significant differences in pre-rehabilitation characteristics.
• The findings support the development of personalized pulmonary rehabilitation strategies based on identified patient clusters.