LAUSR

METHODS: In this retrospective cohort study, ED patients with an ICD-10 diagnosis of eCOPD presenting to ED from Jan-May 2017 were screened. Charts were reviewed for inclusion based on OCRS cohort criteria (exclusions: acute MI, shock, primary issue other than eCOPD, intubation on initial presentation). Data collection included demographic variables, medical history, COPD severity, severity of exacerbation, ED disposition, hospital length of stay, and 30-day SAE. SAE included ED re-presentation, mortality, acute MI, acute dialysis, intubation and new NIPPV use. OCRS was calculated for each patient: 1 point each [pCO2 >35, BUN >12, CXR with lung congestion, past Intubation, intervention for PVD, CABG], 2 points each [EKG with ischemia, HR $110 at presentation, SaO2 120 after ED treatment], and 3 points [Hb <10]. We assessed the association between OCRS and SAE and calculated predicted probability of SAE for each subject. We calculated Hosemer-Lemeshow goodness-of-fit (HL) statistics and ROC curves (AUC) using predicted values. We performed multivariate step-wise backward logistic regression analysis for all variables to identify factors associated with SAE. RESULTS: Of the 873 patients screened 247 were included. Overall OCRS was insignificantly higher in patients with SAE (0.4 p1⁄40.17). Patients with ORCS >5 had higher SAE risk compared to ORCS of 0 [OR 11.3 (1.3-98.6)]. Neither OCRS nor SAE differed based on ED disposition (home, ED-observation, hospitalization). OCRS model did not fit well (HL p<0.0001) nor had good discrimination potential for SAE [AUC 0.57 (0.48-0.660)] in our study compared to Stiell et al, (HL p1⁄40.7, AUC 0.79). None of the OCRS variables correlated with SAE (p1⁄40.4). Three variables were significantly related to SAE in multivariate model: Charleson comorbidity index (CCI) [OR 1.3 (1.11.5) each point increases SAE by 30%]; triage pCO2 [OR 1.06 (1.02-1.1) each point increases SAE by 6%]; hospitalization within 1 year [OR 9 (2.6 to 31)]. This model fit (HL p1⁄40.9) and discrimination [AUC 0.77 (0.7-0.85)] were adequate. CONCLUSIONS: OCRS did not reliably predict SAEs in our context. Three variables were significantly associated with 30-day SAE: triage pCO2, CCI, and hospitalization within a year. A model based on these variables may discriminate high risk eCOPD patients. CLINICAL IMPLICATIONS: Further studies are needed to determine utility of OCRS. DISCLOSURES: No relevant relationships by Matthew Doers, source1⁄4Web Response No relevant relationships by Timothy Loftus, source1⁄4Web Response No relevant relationships by Uwe Stolz, source1⁄4Web Response No relevant relationships by Muhammad Zafar, source1⁄4Web Response DOI: https://doi.org/10.1016/j.chest.2019.08.939 Copyright a 2019 American College of Chest Physicians. Published by Elsevier Inc. All rights reserved. hestjournal.org 1015A