Abstract / Summary
BackgroundHospitalization for acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is associated with substantial morbidity and healthcare utilization in community hospitals.Because in-hospital mortality may be infrequent in single-center chart reviews, composite morbidity/utilization endpoints may better capture clinically meaningful adverse hospital trajectories. MethodsWe conducted a single-center retrospective chart review at a community hospital using unique-patient index admissions only (earliest admission per medical record number (MRN)), yielding 163 patients.The primary outcome was a composite adverse hospital course defined as any ICU admission, endotracheal intubation, prolonged length of stay (LOS > 5 days; cohort median), 30-day readmission, or in-hospital mortality.Baseline comparisons used Mann-Whitney U tests and chi-square/Fisher's exact tests.Multivariable logistic regression was used for the composite outcome.Thirty-day readmission (post-discharge) occurred in 21 cases; therefore, the primary readmission analysis used a pre-specified parsimonious model (age, congestive heart failure (CHF), creatinine, and discharge against medical advice (AMA)) with Firth penalized logistic regression as a sensitivity analysis to mitigate overfitting.LOS was modeled using negative binomial regression.AUROC and calibration were assessed as internal/apparent performance (no external validation). ResultsThe composite adverse outcome occurred in 105/163 (64.4%) index admissions.ICU admission occurred in 53/163 (32.5%), noninvasive ventilation (NIV) in 58/163 (35.6%), intubation in 9/163 (5.5%), in-hospital mortality in 3/163 (1.8%), and 30-day readmission in 21/163 (12.9%).In multivariable analysis of the composite outcome, CHF (adjusted odds ratio (aOR) 3.11, 95% CI 1.15-8.43)and hypercapnia (PaCO₂ per 10 mmHg; aOR 1.94, 95% CI 1.25-3.00)were associated with higher odds of adverse course.In the parsimonious standard logistic readmission model, CHF (aOR 4.16, 95% CI 1.54-11.20)and creatinine (per 0.5 mg/dL; aOR 1.23, 95% CI 1.05-1.45)were associated with readmission; estimates were attenuated in the Firth penalized model.In negative binomial LOS modeling, NIV (adjusted incidence rate ratio (aIRR) 1.58, 95% CI 1.03-2.42)and intubation (aIRR 2.97, 95% CI 1.36-6.49)were associated with longer LOS.Apparent discrimination was AUROC 0.7192 for the composite model and AUROC 0.722 for the parsimonious readmission model. ConclusionIn this community hospital chart review of 163 unique AECOPD index admissions, an adverse hospital course was common and largely reflected utilization and prolonged recovery rather than mortality.CHF and hypercapnia were associated with adverse composite outcomes, and CHF and renal dysfunction were associated with 30-day readmission in parsimonious modeling.Performance metrics represent internal/apparent estimates without external validation; larger, externally validated studies are needed before clinical deployment of prediction tools.