Abstract
Background
Prognostic studies in AECOPD requiring assisted ventilation often select patients by place of care and predict failure of non-invasive ventilation (NIV), not mortality. Improved mortality prediction for unselected patients requiring ventilation for AECOPD is needed.
Objective
Identify mortality predictors in patients with AECOPD requiring ventilatory assistance.
Methods
Clinical data were collected on consecutive patients hospitalised with AECOPD requiring assisted ventilation (NIV or invasive ventilation) for acidaemic respiratory failure (ARF) during their hospital stay. Independent predictors of in-hospital mortality were identified.
Results
199 received ventilatory assistance: mean (SD) age = 73.9 (9.8) years; FEV1 38.1 (16.1) % predicted and 61.3% were female. 49 (24.6%) patients died in hospital.
Older age, an ineffective cough, and severe stable-state dyspnoea were the strongest mortality predictors. The regression model (table 1) showed excellent discrimination for mortality (AUROC = 0.92, 0.88 to 0.96).

Table 1. Independent predictors of mortality
VariableOdds ratio (95% CI)p value
Age, years1.11 (1.04-1.18)0.001
Ineffective cough5.23 (1.74-15.7)0.003
eMRCD2.08 (1.25-3.43)0.005
[HCO3], mmol/L*0.93 (0.88-0.98)0.008
Stroke disease5.54 (1.49-20.6)0.011
Anxiety / depression0.21 (0.06-0.70)0.012
Recent weight loss3.78 (1.26-11.3)0.017
Time to recognition of ARF, hours1.01 (1.00-1.02)0.020
Neutrophil count, x10^9/L1.10 (1.01-1.20)0.031
Maintenance carbocysteine4.03 (1.06-15.2)0.040
* at time of ventilation commencement


Conclusion
Mortality in patients hospitalised with AECOPD requiring assisted ventilation is high but can be accurately predicted using simple to measure indices.