Abstract
BACKGROUND: Patients with chronic obstructive pulmonary disease (COPD) who are hospitalized multiple times for exacerbations face substantially worse clinical outcomes, including higher mortality, faster lung function decline, and reduced quality of life. Identifying these high-risk individuals is essential for early intervention and improved disease management. However, existing predictive models often lack specificity for this population, particularly in inpatient settings. This study aimed to develop a clinically applicable model-based on routinely available inpatient data-to identify patients at risk of exacerbation-related readmission within 12 months following an index hospitalization. METHODS: This retrospective cohort study included patients hospitalized for acute exacerbations of COPD (AECOPD) at a tertiary hospital in China between January 2021 and December 2023. The primary outcome was defined as an AECOPD-related readmission within 12 months following the index hospitalization. Candidate predictors were selected from demographic, clinical, physiological, and laboratory data. A multivariate logistic regression model was constructed and internally validated using bootstrap resampling. Class imbalance was addressed using oversampling, undersampling, and class-weighting techniques. The study was approved by the hospital's ethics committee (Approval No: 2022-058-01). RESULTS: A total of 1559 inpatients with AECOPD were initially screened. After excluding 272 patients due to incomplete medical records, 1287 patients were included in the final analysis. Seven independent predictors were incorporated into the final model: sex, smoking status, diabetes, coronary artery disease, hemoglobin (Hb) level, forced expiratory volume in one second (FEV1)% predicted, and length of hospital stay (LOHS). The model demonstrated good discriminative ability, with an area under the curve (AUC) of 0.79 (95% CI 0.75-0.83), sensitivity of 76.3%, specificity of 70.2%, and satisfactory calibration. A nomogram and online calculator were developed to facilitate individualized bedside application. CONCLUSIONS: We developed a clinically applicable prediction model to identify hospitalized COPD patients at risk of exacerbation-related readmission within 12 months. The model incorporates routinely available clinical and physiological variables and demonstrated good internal performance. It may support early risk stratification and inform individualized post-discharge management. However, due to the single-center retrospective design and absence of external validation, further studies are needed to confirm its generalizability and real-world clinical utility.