Journal Club: COPD: DE-CAF or not DECAF that is the question

Date of JC: 29/06/16

Paper: Validation of the DECAF score to predict hospital mortality in acute exacerbation of COPD

Published: Thorax 2016

Your next patient in majors is a 70 year old smoker who’s been told he has COPD and uses inhalers daily. He continues to smoke and gets a bit breathless when he goes to the front gate (when asked how far he can get before becoming out of puff he answers in the way that all patients with COPD have been instructed to do so: ‘not far’). He has been unwell for a couple of days and is bringing up some mucky stuff off his chest. He is talking sentences, has saturation of 93% in air and has consolidation on his X-ray. He lives half way up a peak district and has no family nearby. its late at night so you opt to admit him after discussion. He is concerned he has ‘another’ bad chest and is worried coming into hospital may mean he die.

smoking old man

Clinical question: In adult patients with COPD being admitted to hospital, does the DECAF score act as a predictor of mortality and could/should it inform practice?

For those of you not familiar with the score the components are (I would recommend reading the derviation paper here too!!):
Dyspnoea (Using the Extended MRC scale)
Eosinopenia (<0.05 10(9)/L) Consolidation on Chest Xray
Acidaemia (pH<7.3) Atrial fibrillation (either present or a history of paroxysmal AF)

Population: Adult patients with a diagnosis of COPD admitted to hospital
Intervention: DECAF score applied to all
Control: There is no ‘control’ however other scoring systems such as CURB65 were compared for predictive value
Outcome: Primary outcome: in-hospital mortality. Secondary outcomes: 30 day mortlaity (compared with other risk scores)

Generalisabity: The score was derived in the North East of England, the internal validation was done at these sites. The 4 external validation sites were not explicitly stated in the write up but they are in the UK and the authors tell us they represent a broad church of COPD experience. The online appendices suggest authors had affiliations to the North East, Cornwall and Sheffield… the patients are probably similar to ours, probably, and we’ll come to that. The score only looked at ADMITTED patients so does miss (hopefully) a large chunk of ED patients with COPD (‘cos I think we’re pretty decent at ambulating and sending these guys home!)

Design: The external validation was done prospectively with specific teams trawling admissions wards and calculating the DECAF score and looking at hospital mortality. They attempted to make the data for the validation cohort consecutive but, (hidden in the online supplement) they only did daily (Monday-Friday) screening, so missed weekends and out of hours patients. The internal validation was done as a retrospective cohort the authors stating that calculation of score was routine practice at the internal sites after the original derivation paper.

A sample size calculation was done and they met the required sample size based on sensitivity prediction of 70%. Have a read of this from St Emlyns on power calculations in clinical trials.

Key results:

They recruited the patients required for the power calculation with 845 patient and 880 patients in the external and internal populations retrospectively.

Median age was 73 and the median DECAF score was 2.

Primary outcome: Area Under the ROC Curve was 0.82 (95% CI 0.79-0.85).
Low risk group Sensitivity 1 (no 95% CI reported)
Highest score (6) has specificity of 1 (again no 95% CI)
Pooled high risk group Specificity 0.93 (Again no 95% CI)

We were impressed with the AUROC, but were disappointed that the 95% CI were not reported on the sensitivity and specificity data, we can’t really know much about the precision of these results and therefore their clinical application. Given the low numbers in the higher scoring ranges the confidence intervals are likely to be wide. There a great collection of critical appraisal resources (largely from St Emlyns) collated at The Bottom Line here. Take some time to listen watch and read the various resources.

In terms of the patients themselves they probably represented admitted patients with COPD, but the numbers recruited per day were low. We think one of the sites is local to us, and have recruited between 0.5 and 1 patients a day. They tell us that 1 in 8 medical admissions is due to COPD in the paper and if one of these sites is ours or at least a hospital representative of ours then we felt that the likely number of admitted patients per day would be far higher, this could represent inclusion bias in the patients recruited. Alternatively if there was only recruitment in daytime hours then perhaps there is a difference between daytime and nocturnal admissions. We don’t know but we are worried about bias here.

I’ve already touched on generalisability to ED patients – this score is primarily looking at hospital mortality and therefore only looked at admitted patients. We don’t admit all patients with COPD and therefore we probably cant apply the secondary analysis of 30 day mortality prediction (no significant difference between DECAF and CURB65 – AUROC 0.76 vs 0.68) to all of our patients.

One of the exclusions were patients with conditions ‘likely to limit; life by a year namely metastatic malignancy. what about primary lung malignancy, how good are we at predicting death? The introduction tells us we are not good at predicting outcome in exacerbations of COPD, are we any better or worse in those with co-morbidity? I doubt it. This potentially may have missed a large number of patients.

The CXR interpretation accuracy and inter-observer reliability (Kappa score) is not discussed but we thought this was pragmatic as a ‘boots on the ground’ junior doc/ANP etc isn’t going to ask for an immediate radiology report on a CXR and therefore we have to accept a degree of error in interpretation and variation in that interpretation.

We think the trial is well conducted, with some concerns about inclusion bias and accuracy. We can see the applicability to risk stratifying in admitted patients, however we don’t think it can be applied across all ED patients with exacerbations of COPD. We await an ED patient validation of this score with particular interest in the 30 day mortality……

You can download the One minute wonder for your department here

Let me know what you think of the paper, of the site and the One minute wonders in the comments below.

Keep well

Keep fighting the good fight.



Leave a Reply

Your email address will not be published. Required fields are marked *