There aren’t a lot of things in Emergency Medicine that are more important than managing paediatric major trauma. St. Emlyn’s in Virchester is a Paediatric Major Trauma Centre, and it’s virtual catchment area is pretty massive. That means we’re pretty used to the adrenaline surge that you get at 4am when the pager goes off with the words, “Major Trauma, Paediatric Emergency Department”.
Heart racing and hair all over the place, you throw on your scrubs, chuck the Starsky & Hutch flashing green light on top of the car and away you go into the night, saving a child’s life like a real life superhero. You can imagine the frustration you feel when you arrive and the child is actually completely uninjured. Of course, that sense of frustration is nothing compared to the emotional toll and inconvenience this will place on the child and their family, who may have been transported a long distance in the middle of the night for no purpose.
This is over-triage. Under-triage is also a problem and occurs when the major trauma centre is not activated although the child does actually have major trauma. In the UK our national PEM research network, PERUKI, set out to try and find out how big a problem this is with each of the triage tools that regional ambulance services are currently using. This paper is hot off the press (online 1st) at the EMJ:
[DDET What did they do? ] They looked at retrospective data from 4 EDs, two of which were major trauma centres in what is basically a retrospective diagnostic cohort study. They included children who sustained ‘injury and trauma’ and didn’t present by ambulance over an 18-month period up to June 2012. They then collected data on 8 pre-hospital triage tools that are being used around the country (which were essentially the ‘diagnostic tests’). The primary outcome was an ISS > 15. [/DDET]
[DDET What did they find? ] They identified 2,934 patient records that met the inclusion criteria and had an ISS recorded. 4 of these patients had an ISS > 15. The sensitivity of the triage tools varied from 0% (Paediatric Trauma Score) to 100% (the East Midlands, North West and Northern regional triage tools). Specificity varied from 79% (for the North West tool) to 99% (Paediatric Trauma Score). [/DDET]
[DDET What did the authors think it meant? ] The authors say that the scores with <95% sensitivity didn’t have acceptable sensitivity. They thought that all 8 tools had acceptable specificity. This meant that only the 3 scores with 100% sensitivity were considered acceptable. [/DDET]
[DDET What did we think? ] Well, here’s where sample size is so important. 2,934 patients sounds like a really big number, right? You see that and you instinctively think it’s unlikely to be underpowered. But they only had 4 patients with the primary outcome. In a diagnostic study where you want to look at sensitivity, it’s actually the number of cases with the primary outcome that drives the power. So, to examine sensitivity this study wasn’t actually well powered at all really – they only had 4 cases. The triage tool that missed all 4 cases might have picked up most of the next 4,000 cases. The confidence intervals are massive. You can’t really draw meaningful conclusions about the comparison of sensitivity when there were only. 4 cases. We need a larger number.
On the other hand, the statistical power for specificity is driven by the number of patients who don’t have the primary outcome. There were 2,929 of them – so this study is likely to be well powered for that analysis. So we can take something from the analysis of specificity. Clearly, the triage tool used in Essex (specificity 93%, with 95% confidence intervals from 91% to 93%) has better specificity than the North West triage tool (79%, with 95% confidence intervals from 78 to 81%). So we can tell from this that patients in the North West are more likely to be over-triaged than patients in Wessex. No wonder the emergency physicians in Virchester (which is in north west England, btw) are tired. [/DDET]
[DDET Is there anything else we can learn from this? ] Well, there is one more thing to take from this. Actually, among children who self-present to the ED, major trauma is very rare. 0.1% of patients who walk in to the ED with an injury will have an ISS > 15. So major trauma centres need to focus most of their efforts on ambulance arrivals. I’be not seen the data but I’d imagine that this is different in adults. [/DDET]
[DDET Are there any more issues for critical appraisal?] There’s unfortunately one really limiting factor about this study – and that’s the patient population included. A diagnostic study should select the patient population in which a test would be applied in practice. This study wanted to evaluate a pre-hospital triage tool for major trauma. But they evaluated it in patients who did not present by ambulance? By definition, that’s the population we wouldn’t apply the triage tool to – because the triage tools are for use in the pre-hospital environment – i.e. in the ambulance. Are the patients who arrive by ambulance different to those who self present? Well, yes. And that’s actually the biggest problem with generalising the findings from this study. [/DDET]
[DDET What are the take home messages? ] As with all studies you critically appraise, you tend to be able to pick holes when you look hard enough. But there may still be things you can take away. With this study, the key points are:
* Major trauma is vanishingly rare among self-presenting children
* The triage tools being used by paramedics across the country vary in their specificity. The North West tool isn’t very specific and this may be something to focus on. We need more work to decide whether it’s sufficiently sensitive.
Most importantly, however, this study has given us a chance to highlight some important critical appraisal issues, and it shows that PERUKI is starting to yield publications. I think there will be many more to come, and I’m sure that having a national PEM research network will bring some absolutely huge benefits in the long run. Congrats and keep going, PERUKI! [/DDET]