Where can I find this paper?
What is this paper about (what is the research question)?
Are clinicians better at predicting intra-abdominal injuries in children with blunt torso trauma than a derived clinical prediction rule?
Summary of the Paper
Design: Secondary analysis of some existing PECARN group data from a prospective cohort study of children with blunt torso trauma
Objective: to compare the test characteristics of clinician suspicion with a derived clinical prediction rule to identify children at very low risk of intra-abdominal injuries undergoing acute intervention
Outcome: test characteristics for clinician suspicion, measured against presence or absence of need for acute intervention for intra-abdominal injury.
Comparison: test characteristics of a derived clinical prediction rule from the same population.
Participants: 12044 patients recruited between May 2007-January 2010 and eligible to participate in the parent study (http://www.ncbi.nlm.nih.gov/pubmed/23375510) underwent secondary analysis.
- Inclusions: children <18 years old with blunt torso trauma presenting to participating PECARN Emergency Departments
- Exclusions: injury >24h prior to attendance; pre-existing neurological disorders affecting examination findings; pregnancy; transfer from another institution.
3016/9252 deemed low risk (<1%) for clinician suspicion had CT abdomen performed; 35 patients subsequently had acute intervention. Of the remaining patients with clinician suspicion ≥1%, 168/2667 had an acute intervention.
Negative clinician suspicion had the following test characteristics;
- sensitivity 82.8% (95% CI 77.0-87.3)
- specificity 78.7% (95% CI 77.9-79.4%)
- NPV 99.6 (95% CI 99.5-99.7%)
- LR- 0.2 (95% CI 0.2-0.3)
Low risk on the prediction rule had the following test characteristics;
- sensitivity 97.0% (95% CI 93.7-98.6)
- specificity 42.5% (95% CI 41.6-43.4%)
- NPV 99.9 (95% CI 99.7-99.9%)
- LR- 0.1 (95% CI 0.0-0.2)
A clinical prediction rule had a significantly higher sensitivity for identifying intra-abdominal injury undergoing acute intervention, but a lower specificity. The higher specificity of clinician suspicion did not translate into clinical practice as clinicians frequently obtained abdominal CT scans in patients they considered to be at very low risk.
On the study design
This was a secondary analysis of data collected as part of an original PECARN study on abdominal trauma in children. It’s always worth remembering that while secondary analysis can reveal some very useful information and trends, this was not the original purpose for which the study group was recruited or the study powered (although the authors tell us this study was preplanned, and the standardised data collection forms used to collect information about clinician decision making supports this).
The study has an issue in that the “gold standard” abdominal CT was not applied to all patients, only those deemed to be at risk of injury. This means there is a large portion of patients who had no imaging and no intervention who may still have had intra-abdominal injury although without a need for clinical intervention the significance of this is doubtful.
Good attempts were made to follow subjects up to ensure no clinically important outcomes were omitted.
What were the results and what does this mean?
There is an important distinction in this paper between the presence of an abdominal injury and one requiring intervention (specified as death, therapeutic intervention at laparotomy, angiographic embolisation, blood transfusion for anaemia or administration of intravenous fluids for at least two nights). This composite reference standard is pragmatic but we could argue about whether intra-abdominal injuries not requiring intervention are also clinically relevant or not, considering the comparative risks of radiation exposure with abdominal CT.
It is worth noting that not all of the 12044 subjects enrolled had CT abdomen performed. 11919 were deemed to have no suspicion of injury, which we must doubt given the fact that neither clinician suspicion nor clinical prediction rule achieved 100% sensitivity.
The study found that in patients with intra-abdominal injury requiring intervention, the clinician correctly identified the risk as ≥1% in 82.8% (95% CI 77.0-87.3) of cases, and in patients who did not have intra-abdominal injury requiring intervention, the clinician correctly identified that the risk was <1% in 78.7% (95% CI 77.9-79.4%) of cases. Unfortunately this shows that clinician judgement alone is neither sensitive nor specific enough to support decision making in isolation. This is borne out in a high CT abdomen rate in the population, despite a high proportion of low risk patients.
The decision rule, which determined risk as “not low” in the presence of any one of:
- no evidence of abdominal wall trauma or seat belt sign
- GCS >13
- no abdominal tenderness
- no evidence of thoracic wall trauma
- no complaints of abdominal pain
- no decreased breath sounds
- no history of vomiting after the injury
had better sensitivity (so the absence of these signs performs better as a predictor of the lack of need for CT and intervention) but poorer specificity (i.e. the presence of any sign does not accurately predict a need for intervention).
Of note there were three patients whose injuries were not identified by clinician prediction or derived clinical prediction rule, so neither predictor achieved 100% sensitivity.
What can we take from this paper into clinical practice?
We as clinicians rely a lot on clinical judgement but that alone is a poor predictor of the need for intervention for intra-abdominal injury, especially when compared with this non-validated derived prediction rule. Following validation the prediction rule may have some diagnostic utility, especially when combined with observation.
More questions to ask
- How will this decision rule perform when validated?
- How would the rule perform if the specificity of clinician judgement was incorporated?
St Emlyns – RCR Guidelines on imaging in paediatric trauma Imaging in Paediatric Trauma – RCR Guidelines – St.Emlyn’s
Follow us on twitter: @PEMLit