I recently had a bit of fun at the Telford trauma conference talking about the top 10 papers of 2012 and in passing mentioned the #dogmalysis of ATLS shock categories published in Resuscitation last year (with thanks to Cliff Reid for blogging on it). Anyway, the idea of clinical signs as predictors of outcome is very topical at the moment as those of us in UK trauma centres are being measured and funded against our ability of identifying all patients with an ISS of 15 within 5 minutes of arrival…, it’s a bit of a problem and something worthy of a complete blog post later.
Over the years we have used many indicators in the clinical examination to try and predict outcome and a ‘set of vitals’ or a ‘set of obs’ on this side of the Atlantic is a standard request on all trauma patients. The problem is that a single set of obs is not really that much use unless grossly abnormal. As clinicians we are also interested in patient trajectory, are they getting better or getting worse and what does this mean in terms of outcome.
So basically it’s a good question to ask whether a change (a delta if we are feeling posh) is better than a single set of obs in determining trauma outcome. Thankfully an excellent group of individuals have examined the TARN database to ask just this question in this paper from the EMJ. My conflict of interest is that I know all the authors and I think they are super.
[learn_more caption=”What kind of study was this?”] This is a retrospective cohort study looking back at data up to 2006. Unfortunately that’s quite old data and certainly in the UK predates trauma centre care for most of the country.[/learn_more] [learn_more caption=”Who was studied?”] Again a retrospective data set from the TARN database. Over 29000 patients were included with some sensible exclusions (CPR, children etc.). Head and spinal injuries were excluded on the basis that these injuries can affect vital signs. Whilst I understand this from a stats perspective it can limit the generalisability.
Also, although 29000+ patients sounds like a lot there were only 107 deaths, so the event rate for this important outcome is quite low.[/learn_more] [learn_more caption=”What did they look at?”] A few things.
- Shock index (HR/SBP)
- Delta SBP
- Delta RR
- Delta HR
The outcome is death at 48 hours. Death is a great outcome in that it is easy to measure, but in many ways I don’t like it. I’m interested in those patients in whom I can make a difference. If patients are going to die in any case there is less need to identify them. It is fiendishly difficult to differentiate out the potential survivors in this sort of study so we will live with this as an outcome for now, but bear in mind that it might not represent the population of true interest to the EP.[/learn_more][learn_more caption=”……and the major results are???”] The abstract suggests that there is a difference in the performance of the parameters they have looked at. They suggest that if you get significant changes in the RR or Shock index are associated with increased mortality, but there is a caveat. Firstly the changes need to be pretty big, for example when looking at the resp rate the change needs to be in the order of 8 breaths per minute. Secondly, we need to look at the overall performance of the test. The authors do this using ROC curves and in particular the area under the ROC curve which is a measure of the overall performance of the test. In this case the best performing test has an area under the curve of 0.65 (for change in shock index), which would be considered to be poor by most standards. Other tests such as change (delta) SBP perfomed even less well.
In the paper there is an emphasis on identifying patient who ‘will die’ by 48 hours, i.e. in ensuring a high specificity, with the authors looking at specificities of around 90 and 95%. This is fine, but at these levels the sensitivity is low…., really low……, for example to get a specificity of 95% for delta RR you only get a sensitivity of 13% (and that’s the best sensitivity available for a specificity of 95%)[/learn_more] [learn_more caption=”So in summary?”] An interesting paper from a great research group. However, I’m a little curious as to the usefulness of the findings in clinical practice within the UK practice of (in general) short scene to ED times.
- It looks as though we are still seeking reliable methods of identifying patients’ prognostic signs at scene and in the ED.[/learn_more]
Before you go please don’t forget to…
- Subscribe to the blog (look top right for the link)
- Subscribe to our PODCAST on iTunes
- Follow us on twitter @stemlyns
- Like us on Facebook
- Find out more about the St.Emlyn’s team
4 thoughts on “JC: Delta signs for shock trauma. St.Emlyn’s”
Pingback: Initial Stablization | EMchatter
Under ‘and the major results’, line 11, I think you mean high specificity (SPin/ rule in), not high sensitivity (SNout/ rule out). What was interesting for me was the fact that delta HR were absolutely worthless. Thanks for the mention Simon though. We used death as an outcome mainly because the three other papers we could find that have sort of looked at this also did so. I will give myself a gold star and raise a glass of Pimms for making it onto St. Emlyn’s blog.
Thanks for that (I have corrected), and thanks for the comment about heart rate. I think I was surprised too that the effect of changes in any parameter were less than I expected.
I think death is a very common outcome measure in studies of this type because it is tangible, recognisable and fairly easy to measure! I have done some work with Lee Wallis about this sort of question in the past (in relation to major incidents) where we looked to identify ‘life saving procedures’ as an important outcome for triage scores. I wonder if that is something that we might use as a indicator in the future, well, I would report both to be honest as neither is perfect but they have slightly different implications. Interested to know what you think.
Pimms? Lovely, the bank holiday weekend is surely a good time to consider it in England. I’m a Number 1 fan myself with orange, mint and cucumber.
Also worth saying that this is good research. We should not forget that finding that parameters are not as predictive as I thought is arguably just as important as reinforcing what I previously thought to be true!