Background: High concentration oxygen therapy in hypercapnic chronic obstructive pulmonary disease (COPD) is associated with increased mortality. In ward-based patients with acute exacerbation of COPD and hypercapnia, this study examines oxygen prescription and the association between high concentration oxygen therapy and adverse outcome.
Aims:To investigate the association between over-oxygenation and in-hospital adverse events.Methods: Inpatients with acute exacerbation of hypercapnic COPD at a tertiary Australian hospital over a 1-year period were retrospectively identified. Oxygen prescription and therapy was determined based on chart review. Over-oxygenation was defined as ≥10% of nursing chart observations recording oxygen delivery with oxygen saturation above 92%. A composite adverse outcome was defined as medical emergency team response, recommencement of non-invasive ventilation or death. The association between over-oxygenation and adverse outcome was assessed using survival analysis and conditional logistic regression modelling.
Results:The study cohort comprised 106 unique patients and 157 admissions. Oxygen prescription was recorded in 132 (84%) and over-oxygenation occurred in 97 (62%) admissions. Over-oxygenation was higher in non-respiratory ward admissions (76% vs 57%, P = 0.03) and those without any form of oxygen prescription (84% vs 58%, P = 0.01). During follow up, 23 (22%) patients experienced an adverse event. Cox proportional hazards modelling found weak evidence for increased risk of an adverse event in over-oxygenated patients (hazard ratio 2.5; 95% confidence interval 0.8-7.3, P = 0.10). Conditional logistic regression, after matching on age, Charlson comorbidity category and length of follow up, found a similar estimate of association (hazard ratio 2.6; 95% confidence interval 0.8-8.7, P = 0.12).Conclusions: Over-oxygenation to hypercapnic COPD inpatients is common and rates of oxygen prescription are suboptimal. We found weak evidence of association between over-oxygenation and adverse events, likely due to low statistical power. Larger prospective studies are needed.