Last Friday (30th November) we looked at the paper below in both the physical and virtual journal clubs of St Emlyn’s:
It doesn’t take a great leap of imagination to see how this could be relevant to our own departments: I personally feel an irresistible heartsink when I pick up the next card in the box and the presenting complaint at triage is headache (usually followed by the words “for six weeks”). But these feelings are unfair; we know that some patients presenting to the ED with headache actually have underlying and often serious pathology – the trick (or challenge) is to identify them from the majority who don’t.
Could this paper provide us with insight? Let’s have a look.
The first thing to note is that this paper does not involve true primary research. The data being used was collected for another purpose – in this case, as part of a national survey of ED visits – which, in itself, collected data from a sample of EDs. Patients with headache were retrospectively identified from coding data by diagnosis according to ICD-9 classification. The issue here is accuracy: how good are we at completing our coding accurately? How good are we at diagnosis? Simon has a few things to say about that here, but it’s important to consider how these human factors might influence the outcomes of this piece of research. What’s more, not all of the data was used; the numbers involved were pretty big, so not unreasonably the investigators have used a sub-sample for full analysis.
Both issues were raised on twitter:
#JC_StE I also have a number of issues with the dependency on coding accuracy for diagnosis— Alastair Pickering (@drapickering) November 30, 2012
My head hurts already, but hopefully this should be raising questions for you about how much extrapolation has gone on here to generalise the findings to a population of ED patients/attendances; one of the clues to that is in the introduction; the authors state that the proportion of those presenting to the ED whose primary complaint is headache is 1-3%; in their data, the proportion was 4.5%.
So, 4.5% of ED attendances were due to atraumatic headache; 20.4% of patients received CT or MRI scan. This feels instinctively higher than UK experience on both counts, particularly when patients for whom there were concerns of meningitis or encephalitis were excluded.
The paper demonstrates an increasing trend in the number of scans arranged over the 10-year period, without a corresponding increase in the number of “pathological” scans. While this may reflect better access to scan facilities, the authors reflect whether factors such as threat of litigation and lack of clear clinical guidelines may also impact their findings.
The authors go on to provide a breakdown of odds ratios, expressing the likelihood of intracranial pathology on scan in the presence of various clinical and non-clinical findings. However:
Some interesting adjusted odds ratios in table 2: do they tell us anything new though? #JC_StE— JC_StE (@JC_StE) November 30, 2012
And what of the age cutoff? Why have the authors decided to separate their patients into 18-49yrs and >50 years? The answer isn’t clear, but pathology seemed more prevalent among the older group. Does this reflect generational attitudes towards ED attendance, or unmeasured risk factors; smoking, hypertension, anticoagulant use, all arguably more likely in older patients?
So, the increasing trend in use of radiology in the ED assessment of headache is notable, and likely to be reflected in UK figures. But the feeling on twitter was that otherwise the study adds little to our clinical assessment of patients with headache. It would be practically difficult to develop a sensible clinical decision rule for a collection of disparate presentations. And what of the new NICE guidance on headache? Well, St Emlyn’s has something to say on that too..!
Join St Emlyn’s Twitter Journal Club from 13:00-14:00 GMT/BST every Friday @JC_StE and use the hashtag #JC_StE