Event: ERS Congress 2025, Amsterdam, Netherlands
Authors: Matthew Moll, Fang Liz Zhou, Majd Mouded, Claire Brulle-Wohlhueter, Lynda M Cristiano, Nina Temam, Julie Barberio, Théophile Reppelin, Ce Jiang, Louise Dry, Muthiah Vaduganathan
Background
Chronic obstructive pulmonary disease (COPD) often coexists with cardiovascular diseases1, which increases morbidity and mortality2 and the need for integrated management strategies.
Objectives
To identify risk drivers of cardiopulmonary events among COPD patients.
Methods
Study cohorts included moderate-to-very severe (MVS) COPD patients, aged ≥ 40 yrs with FEV1/FVC ≤ 0.70 from a large U.S. Electronic Health Record database with linked claims. Machine learning (ML) was used to jointly identify risk factors for two outcomes: cardiovascular (CV) (5-point MACE and arrythmia) and pulmonary (Pul) (e.g., exacerbations). Based on these results, we developed a predictive Cox model and calculated risk scores.
Results
The discovery cohort (index date 1/1/2018) included 6,994 MVS-COPD patients (mean age 70.6 ±9.8 years) who experienced CV and Pul events (15% and 28% respectively) in one-year follow-up. ML identified traditional risk factors (e.g. prior outcome in baseline, disease severity) as well as novel ones (e.g., low albumin, high neutrophils) (Figure 1). In multi-variable models, 20 features were significantly linked to CV and Pul events respectively, with strong performance (C-index: 0.727 for CV, 0.726 for Pul).
Conclusion
In a large real-world dataset, we identified interesting cardiopulmonary risk factors in MVS-COPD, highlighting the importance of risk-based approach for therapy optimization.
