Breaking the Cycle: Predicting 30-Day Readmission in Chronic Obstructive Pulmonary Disease Using Electronic Health Record Models
Claire Bernard¹, Oliver Bennett², Eleni Kosta³
Keywords:
COPD, Readmission, Predictive Model, Internal Medicine, EHRAbstract
Background: COPD is a leading cause of hospital readmissions worldwide, with 30-day rates often exceeding 20%. Accurate predictive models could reduce preventable readmissions.
Objective: To develop an EHR-based predictive model for 30-day readmission in COPD patients.
Methods: A retrospective cohort of 3,214 COPD patients hospitalized in France (2018–2022) was analyzed. Variables included age, comorbidities, prior admissions, medications, and spirometry. Logistic regression and random forest models were trained and validated. Model accuracy was measured with AUC, sensitivity, and specificity.
Results: Thirty-day readmission occurred in 21.6% of cases. The random forest model outperformed logistic regression with AUC 0.87 (95% CI: 0.84–0.90), sensitivity 82%, and specificity 78%. Independent predictors included ≥2 admissions in prior year (OR: 2.4), comorbid heart failure (OR: 1.9), and low FEV1 (<50% predicted, OR: 2.7).
Conclusion: EHR-based models can predict COPD readmission with high accuracy. Incorporating such tools into discharge planning could guide targeted interventions, reduce costs, and improve patient outcomes.
Downloads
