BMJ Open Respir Res. 2026 Apr 29;13(1):e004016. doi: 10.1136/bmjresp-2025-004016.
ABSTRACT
BACKGROUND: Acute exacerbations (AEs) drive disease progression and healthcare burden in chronic obstructive pulmonary disease (COPD). Although prior AE history is the strongest predictor of future events, it cannot guide risk assessment in AE-naïve patients, highlighting the need for new prediction tools.
METHOD: In this retrospective multicentre cohort study, we enrolled training and validation cohorts from five hospitals. Baseline demographics, symptom scores, spirometry, blood eosinophils and comorbidities were extracted from electronic records. Multivariable logistic regression was used to identify predictors and convert into score-based models.
RESULT: The training and validation cohorts included 310 and 86 AE-naïve patients with COPD, respectively. Independent predictors of 1-year AEs included coexisting asthma, modified Medical Research Council score ≥2, blood eosinophils ≥2%, per cent predicted forced expiratory volume in 1 s <50% and cardiovascular comorbidities (ie, heart failure and ischaemic cerebrovascular events). Three models were developed with areas under receiver operating characteristic curves of 0.727 to 0.750 and 0.717 to 0.728 in the training and validation cohorts, respectively. At the optimal cutoffs, the sensitivity ranged from 76.7% to 88.2% and specificity from 42.0% to 65.7%. Risk stratification separated patients into low-risk, intermediate-risk and high-risk groups in validation cohort, with increasing AE rates (0.10, 0.43 and 0.86 events/year, respectively; P for trend=0.0047).
CONCLUSION: By incorporating baseline AE history, the externally validated, score-based prediction models provide a practical tool to estimate 1-year AE risk in patients with COPD.
PMID:42055736 | DOI:10.1136/bmjresp-2025-004016