Toby, a 3 year old child, presents to your emergency department with a cough and wheeze. When the triage nurse hands over to you, they are happy with Toby’s observations, but is concerned that he has a number of bruises which ‘look concerning’. How do you assess them?
This is a core question that reaches into the heart of every assessment we make in the paediatric emergency department. Could these injuries be due to abuse?
‘Validation of a Clinical Decision Rule to Predict Abuse in Young Children Based on Bruising Characteristics’ (Pierce, M. et al, JAMA Network Open. 2021;4(4):e215832. doi:10.1001/jamanetworkopen.2021.5832)
What’s it about?
Aim – The authors state their goal as ‘establishing the first clinically sensible evidence-based screening tool to distinguish bruises caused by physical child abuse from those caused by non-abuse’
Study type – A large (N=2161) prospective, observational, cross-sectional study from emergency departments of five urban children’s hospitals in four different states of the USA.
Published in JAMA, this USA-based paper seeks to validate the use of a screening tool developed by the authors for predicting whether a bruise is more likely caused by abuse, depending on the bruise characteristics. The lead author, Dr Mary Clyde (MD) has previously devised ‘Bruising Clinical Decision Rule’ (BCDR) for the same thing in 2009 – the TEN-4 rule. This time around, the authors sought to address some of the self-identified weaknesses of the previous study, such as being of relatively small sample size (N= 95), single centre and of limited detail (the face being counted as one whole area), by refining and validating an updated BCDR in a new, larger study.
Why is this important?
Bruising is a regular part of childhood, with the number of bruises sustained increasing as children get older and more independent. It is, however, also a common injury in children who have been subject to abuse, and in a number of cases is the ‘sentinel injury’ leading to investigation and recognition of child abuse. This is especially the case in the pre-mobile infant, where accidental bruising is extremely rare (0.6-1.3%).
Missing the opportunity to investigate a bruised child directly contributes to poor patient outcomes and potentially allows a child to remain in an abusive environment. Through recognition of their importance, and the use of decision tools to assist with risk stratification, there is a potential to prevent further abuse and make a significant and monumental difference to a child’s life.
Particularly for junior members of the team, finding a bruise, and declaring it as potentially non-accidental can be a daunting thing to do. A decision tool may go some way to help empower healthcare professionals to identify and escalate children at risk, facilitating a timely referral for specialist assessment. It is important to remind those reading this, that it remains the case that it is not possible to accurately age a bruise based on naked eye assessment.
Regardless of presenting complaint, all children (except those who met exclusion criteria) attending the paediatric emergency departments underwent a ‘skin examination’ by their emergency department practitioners. If one or more bruises were identified, the child was enrolled in the study after written informed consent (interestingly two/three of these participants did not have bruising as their presenting complaint). The research team then took a closer look to gain a further level of detail. Parallel to this, all children less than four years of age who were undergoing ‘Physical Abuse Medicals’ – think Child Protection Medicals – had data collected by the research team for comparison. The authors state that a ‘waiver for consent’ was granted for these cases.
Cases were ‘de-identified’ and recorded in a standardised electronic format including history, examination, findings, and relevant investigation results where applicable. The cases were reviewed by a nine member ‘panel of experts’ and labelled as abuse, non-abuse or indeterminate. The authors state that inter-rater reliability of the panellists was ‘high’ and agreement between the members ‘strong’. It is stated that where the outcome/cause was definitely known, the panel were accurate with their classifications.
What’s the point of the tool?
The authors state that the tool is not to diagnose abuse but to function as a screening tool to help recognise potential cases of abuse. The authors are clear to state that the tool is ‘only to inform and never to supplant clinical judgement’.
The wonderful ‘Skin Deep’ team at DFTB has helped us acknowledge the different appearance of skin pathology in different skin tones/pigments. The study went some way to addressing this by having a set of five ‘reference photos’ to categorise the child’s skin tone. The demographics data are difficult to directly compare to the UK, however the study identified 83% of participants as ‘white’, not too dissimilar from the 2011 UK census (86%, obviously fairly old data and wide regional variations). There is some further analysis of result variation by skin tone within the supplements which is worth looking at.
69% were recorded as being ‘non-Hispanic/Latino’, which appears to have been recorded to identify potential cultural protective/at risk demographics. Around 20% of the US population identifies as Hispanic/Latino, however, it is not clear why this ethnicity (and no others) was specifically identified in this study.
The statistics bit
They analysed the number of bruises and the regions of bruising, and compared the groups using a Wilcoxon rank sum test. Within these groups, individual medians, interquartile ranges and differences in proportions were quantified with a 95% confidence interval. Diagnostic accuracy tests were performed on the new decision tool – and comparing this to the previous with sensitivity, specificity, positive predictive value and negative predictive value all calculated, again with 95% confidence intervals. When formulating the new diagnostic tool, a tree fitting of 0.05 was used, with only those that improved the fit included in the tool. Once formulated, it was validated using bootstrap estimates (>10 000 loops) to estimate the properties of the estimator.
In English: Appropriate statistical tests were performed in order to gain valuable information around bruising in children, but also test one decision tool against the other – with the new tool performing well on statistical analysis and hypothetical evaluation.
- Bruising is a common presentation in children and may be an indication of abuse.
- Using this tool could help clinicians assess the need for further investigation, however, clinical judgement is paramount.
- The absence of bruising does not confirm the absence of abuse. Cases should be escalated if there are concerns.
- The tool does not address history or presentation. This is a key part factor in identifying non-accidental injury and should always be explored.
- We need to recognise that bruising in darker skin tones can be more difficult to spot. There should be a thorough skin examination with every consultation.
Will it change my practice?
We don’t feel that any of the individual patterns/characteristics that are ‘more’ associated with abuse (according to this paper) are new or surprising, but further formal research backing up current knowledge and beliefs is always beneficial Having a validated tool likes this will hopefully encourage us to think about, and document, bruising characteristics and their location.
The tool gives a structured approach for the initial examination and management of children with bruising for clinicians with less experience in non-accidental injury. The images used in this article were a powerful tool to help re-enforce more concerning bruise patterns, and acted as a great visual summary of the research. They will definitely be useful for teaching.
Pierce, M.C., Kaczor, K., Aldridge, S., O’Flynn, J. and Lorenz, D.J., 2010. Bruising characteristics discriminating physical child abuse from accidental trauma. Pediatrics, 125(1), pp.67-74.
All fixed – A
Great review of the article – unfortunately I’ve noticed the link to the original JAMA article doesn’t work