Outcome prediction for patients with obstructive lung disease considered for admission to critical care units in England, Wales and Northern Ireland

Objective. To develop an outcome prediction score for patients with obstructive lung disease considered for admission to critical care. Design. Prospective cohort study using multivariate logistic regression for model building followed by score development and bootstrapping to adjust for over fitting. Setting. Critical care and respiratory high dependency units in England Wales and Northern Ireland. Participants. Patients aged 45 years and older with a clinical diagnosis of breathlessness, respiratory failure or change in mental status due to an exacerbation of COPD, asthma or a combination of COPD and asthma. Main outcome measures. The primary outcome was survival at 180 days and the model was constructed to predict this. The secondary outcomes were the accuracy of clinicians' predictions at the time of critical care admission and 180-day health-related quality of life Results. Ninety two critical care units and three respiratory high dependency units took part. Eight hundred and thirty two patients were recruited and the 651 patients without treatment limitations were used to develop the outcome score. Of the 651 patients 450 were intubated and 107 (16.4%) died in critical care, another 66 (10.1%) died in hospital and a further 47 (7.2%) had died by 180 days follow-up, giving a cumulative 180-day mortality of 33.8% (220 deaths). 420 of 518 (81.1%) survivors provided quality of life data and 400 (96.4%) would want ICU again under similar circumstances. A score using length of stay, age, sex, acute physiology, functional capacity, mid-arm circumference, atrial fibrillation, intubation status and diagnosis had an area under the receiver operating characteristic curve of 0.75 after bootstrapping and was well calibrated. Clinicians' predictions had an ROC area of 0.71 and were less well calibrated with a tendency towards pessimism. Conclusions. This study has produced an outpome prediction score with reasonable discrimination and good calibration that has the potential to support clinicians in understanding the prognosis of patients with obstructive lung disease considered for critical care