A44-10 Chronic Obstructive Pulmonary Disease Exacerbations: A Nationwide Readmissions Database Analysis

Introduction Using the Nationwide Readmissions Database (NRD), we analyzed risk factors associated with an index admission forCOPD exacerbation and analyzed readmissions for these patients. First, we identified patients with an admission for aCOPD exacerbation and evaluated covariates for associations. To control for confounding we used propensity scorematching. We also evaluated for effect modification. Then, using the cohort formed in the first model, we evaluated various readmission parameters. Methods 1. Index Admissions - An index admission for COPD exacerbation was defined as an admission with ICD-10 “J441” in the first diagnosis field. Covariates were derived from comorbidity indicators based on ICD code substrings and included age, sex, income, and hospital size. We used propensity scoring with 1:1 nearest-neighbor matching (caliper=0.2) and then applied a logistic regression and an interaction model to the matched sample. 2. Readmissions - Using the cohort formed in our first model, we evaluated readmissions including readmissions within 30days, average time between successive admissions, and total average number of readmissions. 3. Model DiagnosticsThe area under the ROC curve, classification tables, and Hosmer-Lemeshow goodness-of-fit tests wereused to assess model performance. Results 1. Index Admissions- aOR for key covariates: Age: 1.02374 per year (95% CI: 1.0234-1.0241). Female: 1.20351 (95% CI: 1.19001-1.2171), Cardiovascular: 2.02058 (95% CI: 1.9880-2.0537), Metabolic: 0.93060 (95% CI: 0.9191-0.9423), Infectious: 0.71772 (95% CI: 0.7071-0.7285);Interaction analysis - The association of cardiovascular conditions diminished with age (interaction aOR ≈ 0.9511). The association of metabolic conditions increased with female sex (interaction aOR ≈ 1.2878). 2. Readmissions -10.57% of admissions were flagged as readmissions within a 30-day window. The average first readmission occurred 13.63 days after the index admission, and the mean total readmissions per patientwas 0.45. 3. Model Diagnostics - The COPD model’s ROC curve had an AUC of 0.6747, and the classification table showed 100% specificity (with 0% sensitivity at a 0.5 threshold). The Hosmer-Lemeshow test was highlysignificant (χ² ≈ 15,010.69) Discussion While the COPD exacerbation model yields clinically plausible associations and demonstrates the utility of propensity score matching, the readmission analysis is hampered by data structure limitations. Future improvements include alternative modeling approaches to address class imbalance and enhanced definitions of readmissions (e.g., planned versusunplanned). This study highlights both strengths and challenges inherent in using large administrative databases for health services research. This abstract is funded by: None