Predicting Adverse Events Among Patients With COPD Exacerbations in the Emergency Department

Background: COPD exacerbations lead to excessive health care utilization, morbidity, and mortality. The Ottawa COPD Risk Scale (OCRS) was developed to predict short-term serious adverse events (SAEs) among patients in the emergency department (ED) with COPD exacerbations. We assessed the utility of the OCRS, its component elements, and other clinical variables for ED disposition decisions in a United States population.

Methods: We compared the OCRS and other factors in predicting SAEs among a retrospective cohort of ED patients with COPD exacerbations. We followed subjects for 30 d, and the primary outcome, SAE, was defined as any death, admission to monitored unit, intubation, noninvasive ventilation, major procedure, myocardial infarction, or revisit with hospital admission.

Results: A total of 246 subjects (median 61-y old, 46% male, total admission rate to ward 52%) were included, with 46 (18.7%) experiencing SAEs. Median OCRS scores did not differ significantly between those with and without an SAE (difference: 0 [interquartile range 0-1)]. The OCRS predicted SAEs poorly (Hosmer-Lemeshow goodness of fit [H-L GOF] P ≤ .001, area under the receiver operating characteristic [ROC] curve 0.519). Three variables were significantly related to SAEs in our final model (H-L GOF P = .14, area under the ROC curve 0.808): Charlson comorbidity index (odds ratio [OR] 1.3 [1.1-1.5] per 1-point increase); triage venous P_CO2 (OR 1.7 [1.2-2.4] per 10 mm Hg increase); and hospitalization within previous year (OR 9.1 [3.3-24.8]).

Conclusions: The OCRS did not reliably predict SAEs in our population. We found 3 risk factors that were significantly associated with 30-d SAE in our United States ED population: triage [Formula: see text] level, Charlson comorbidity index, and hospitalization within the previous year. Further studies are needed to develop generalizable decision tools to improve safety and resource utilization for this patient population.