QJM Advance Access originally published online on April 15, 2009
QJM 2009 102(6):389-399; doi:10.1093/qjmed/hcp036
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Predicting mortality for patients with exacerbations of COPD and Asthma in the COPD and Asthma Outcome Study (CAOS)
From the 1Sheffield Thoracic Institute, Northern General Hospital, Herries Road, Sheffield S5 7AU, 2Health Services Research Unit, London School of Hygiene and Tropical Medicine, London, 3Heartlands Hospital, Birmingham, 4Health Services Research Unit, London School of Hygiene and Tropical Medicine, London, 5University of Aberdeen, 6Intensive Care National Audit and Research Centre, London, 7Nuffield Department of Anaesthesia, Oxford and 8Intensive Care National Audit and Research Centre, London, UK
Address correspondence to Dr M.J. Wildman, Sheffield Thoracic Institute, Northern General Hospital, Herries Road, Sheffield S5 7AU, UK. email: martin.Wildman{at}sth.nhs.uk
Received 7 June 2008 and in revised form 6 March 2009
 |
Abstract |
|---|
Introduction: Decisions about the intensity of treatment for patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are influenced by predictions about survival and quality of life. Evidence suggests that these predictions are poorly calibrated and tend to be pessimistic.
Aim: The aim of this study was to develop an outcome prediction model for COPD patients to support treatment decisions.
Methods: A prospective multi-centre cohort study in Intensive Care Units (ICU) and Respiratory High Dependency Units (RHDU) in the UK recruited patients aged 45 years and older admitted with an exacerbation of obstructive lung disease. Data were collected on patients’ characteristics prior to ICU admission, and on their survival and quality of life after 180 days. An outcome prediction model was developed using multivariate logistic regression and bootstrapping.
Results: Ninety-two ICUs (53% of those in the UK) and three RHDUs took part. A total of 832 patients were recruited. Cumulative 180-day mortality was 37.9%. Using data available at the time of admission to the units, a prognostic model was developed which had an estimated area under the receiver operating characteristic curve (‘c’) of 74.7% after bootstrapping that was more discriminating than the clinicians (P = 0.033) and was well calibrated.
Discussion: This study has produced an outcome prediction model with slightly better discrimination and much better calibration than the participating clinicians. It has the potential to support risk adjustment and clinical decision making about admission to intensive care.