Development and external validation of an interpretable machine-learning model for HFpEF comorbidity risk in COPD patients
A computational model identifies patients with both lung and heart disease who are at risk, helping catch this commonly missed and serious combination.
An XGBoost machine learning model predicting HFpEF in COPD patients achieved AUC 0.898 internally and 0.819 in external validation, with NT-proBNP as the dominant predictor and red blood cell count among the 9-feature panel. The model addresses a clinically important but frequently missed comorbidity overlap; external validation is limited by very small n=69 in the validation cohort.
What the study was
- Study design
- Retrospective ML model development with external validation cohort; XGBoost + SHAP
- Population
- COPD patients; internal cohort n=1,550; external validation n=69
- Sample size
- 1619
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- Respiratory Research
Why it surfaced
Addresses real clinical need (COPD-HFpEF overlap). AUC 0.898 is good; external validation cohort (n=69) is small and limits generalizability. Red blood cell count inclusion adds CBC-adjacent relevance.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.