Truace
Subscribe

Identification of clinical phenotypes and development of a predictive model for rapid lung function decline in COPD patients

Background Chronic obstructive pulmonary disease (COPD) exhibits significant clinical heterogeneity. This study integrates phenotype identification with longitudinal lung function analysis to develop a prediction tool for rapid decline in a large Chinese cohort. Methods This study included 38,863 COPD patients from a national multicenter screening program. Latent class analysis identified clinical phenotypes. Among 1,215 patients with baseline and 24-month follow-up spirometry, rapid decline was defined as annua…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1496· cite & verify
Identification of clinical phenotypes and development of a predictive model for rapid lung function decline in COPD patients
Miner spirometry (9253624711) by National Institute for Occupational Safety and Health (NIOSH) from USA. Public domain

In brief

Researchers drew on a national multicenter screening program in China covering 38,863 COPD patients. They used latent class analysis to sort patients into three clinical phenotypes. They then trained six machine learning models on the 1,215 patients who had spirometry at baseline and again 24 months later, to predict rapid decline in lung function. CatBoost performed best, using region, fuel type, income level, wheezing, hip circumference and baseline FEV1.

Chronic obstructive pulmonary disease varies widely between patients, so a simple screen that picks out those likely to lose lung function quickly could help target follow-up and intervention. The authors present the tool for screening. The model's overall discrimination was moderate, and the report gives no evidence from use in routine care, so its practical value in clinics is still unproven.

Main points

  1. Latent class analysis of 38,863 patients produced three phenotypes: GOLD Stage 2 non-smoking females (25.8%), GOLD Stage 1 smoking males (44.8%), and GOLD Stage 3 severe smokers (29.4%).
  2. Phenotype 2 had the highest rate of rapid decline at 55.8%, and the authors name it a key target group for intervention.
  3. The best of six machine learning models, CatBoost, reached an AUC of 0.712 using six variables.
  4. SHAP analysis ranked baseline FEV1 as the strongest predictor.
  5. At a 0.365 threshold the model had 95.5% sensitivity and 86.4% negative predictive value, with significant net benefit.

The gain

A CatBoost model using six simple indicators screened for rapid FEV1 decline with high sensitivity and a useful negative predictive value among Chinese COPD patients with follow-up spirometry.

The rundown

Rapid decline was defined as an annual FEV1 decline of at least 40 mL or an FEV1 % predicted decline above 1.5%. Candidate predictors were narrowed with LASSO regression and the Boruta algorithm before six machine learning models were built.

The models were evaluated for discrimination, calibration, and clinical utility. SHAP analysis was used to interpret the final CatBoost model, which kept six variables: region, fuel type, income level, wheezing, hip circumference, and baseline FEV1.

The study found that the group with the mildest airflow limitation by GOLD stage, GOLD Stage 1 smoking males, had the highest share of rapid decliners. The authors suggest this group should be targeted for intervention. The tool's high sensitivity at the chosen threshold favors catching likely rapid decliners over avoiding false alarms.

Sources

  1. Peer-reviewedAnnals of Medicine2026-10-01

The debate