Background: Severe exacerbations of Chronic Obstructive Pulmonary Disease (ECOPDs) are associated with significant morbidity and mortality. Current guidelines emphasise using ECOPD history to guide preventive treatments but offer limited guidance for stratifying the risk of a first severe ECOPD. PRECISE-X aimed to develop and validate an individualised risk prediction model for the first severe ECOPD using variables routinely captured in electronic medical records.

Methods: We created a cohort of newly diagnosed COPD patients in UK’s Clinical Practice Research Datalink (2004–2022) to develop a risk prediction model for the first severe ECOPD during the next 5 years (primary outcome) and the next 12 months (secondary outcome). Candidate predictors were identified through clinical expertise and data-driven variable selection. Internal-external validation was performed across practice regions to evaluate the out-of-sample performance of the model in terms of discrimination (c-statistic), calibration, and net benefit.

Findings: The study included 219,015 patients (mean age 66.0; 42.4% female). The predicted 5-year risk of experiencing a first severe ECOPD was 29.5%. The final model included four mandatory predictors (sex, age, Medical Research Council (MRC) dyspnoea score and forced expiratory volume in one second) and 22 optional predictors. In internal-external cross-validation, the average out-of-sample c-statistic was 0.815 (95%CI 0.804–0.826) for 5-year prediction and 0.767 (95%CI 0.757–0.776) for 1-year prediction. The model demonstrated good calibration across regions and showed positive net benefit across the entire plausible range of risk thresholds.

Interpretation: The risk of a first severe COPD exacerbation can be predicted with high accuracy using routine clinical data, addressing a critical gap in COPD management.