Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Characteristics That Distinguish Accidental From Abusive Injury in Hospitalized Young Children With Head Trauma.

Bechtel K, Stoessel K, Leventhal JM et al.

Pediatrics 2004;114: 165-168 [abstract]

Reviewed by Margaret K. Winkler MD FAAP FCCM, Associate Professor, University of Alabama at Birmingham

Review posted March 2, 2005

I. What is being studied?

Study objective:

To determine if there are clinical features to help distinguish between accidental and abusive head trauma so that each can be diagnosed more readily and accurately.

There are two primary objectives:

1. To determine the incidence of retinal hemorrhages (RH) in hospitalized children with accidental or abusive head trauma.
2. To describe other clinical findings that may help distinguish between accidental and abusive head injury.

Study design

This is a prospective cohort study of children less than 2 years of age who were admitted to Yale New Haven Children's Hospital between August 1, 2000 and October 31, 2002 with a diagnosis of head injury and who had computed tomography of the head.

II. Are the results of the study valid?

Note: These questions follow from Randolph AG et al. Understanding articles describing clinical prediction tools. Crit Care Med 1998;26:1603-1612. [abstract]

1. Was a representative group of patients completely followed up? Was follow-up sufficiently long and complete?

Unclear. The investigators do not provide a detailed description of the patients, severity of illness, concomitant injuries, or operative procedures which could permit a better comparison to one's own patient population. This was a very select patient population in that only children undergoing a head CT were included Ð they may have missed some older RH's in children chronically abused who presented sub-acutely. Any children who presented with accidental or abusive trauma that did not get a head CT would not have been included in this analysis. This would change the incidence of RH's.

87 children were eligible for the study. However, 5 children did not receive a dilated eye examination before discharge, and therefore were excluded from study. All 82 children who were enrolled in the study completed the study. Patients were followed during their hospitalization with serial neurologic examinations, but length of stay for these patients is not defined. For example, the development of late onset seizures, or readmissions as a result of the head injury are not accounted for.

2. Were all potential predictors included?

No. The authors set out to address the occurrence of RH's in abusive vs. accidental head injury. In addition, they assessed vitreous hemorrhage, skull fracture, abnormal mental status on presentation, seizures, scalp hematomas, need for anticonvulsants, and operative interventions as secondary predictors. However, this was adequate to accomplish their stated aims and a quite thorough evaluation of predictors in abusive head injury.

Additional predictors of abusive vs. accidental head injury that could have been evaluated with this population include the updated risk of mortality scoring models like PRISM III. Time from injury to receiving medical attention could also have been studied as a predictor of abusive injury. Care-givers at the time of the injury should also have been included as a possible predictor.

3. Did the investigators test the independent contribution of each predictor variable?

No. Predictor variables were assessed separately, comparing the rates of each between the two groups (abused vs. accidental). No model was developed that assessed the independent role of each of the predictor variables while controlling for the others. In addition, the investigators also evaluated different types of RH's for their different rates between the two groups (unilateral, bilateral, extending to the periphery, covering the macula, or involving the pre-retinal layer).

4. Were outcome variables clearly and objectively defined?

Mostly. The main outcome measure was the proportion of children in each group with RH's. Secondary measures included: vitreous hemorrhage, abnormal mental status on presentation, seizures, scalp hematomas, need for anticonvulsants and operative procedures.

Abnormal mental status on presentation was defined as unresponsive or poorly responsive to verbal or tactile stimulation, gaze palsies, decorticate or decerebrate posturing and flaccidity. However, it was not clear whether this classification was made before or after any necessary resuscitation had occurred or after any medications (i.e. anticonvulsants) may have been given which could have influenced the mental status of the child.

It is also unclear whether the ophthalmologist was blinded about the etiology of the trauma. However, it does appear as though the retinal exam was performed after the child was classified as accidental or abusive trauma, so the presence or absence of retinal hemorrhages did not influence that classification. The lack of blinding the ophthalmologist may have affected sub-categorization of the sub-types of RH, i.e. pre-macular, vitreous, etc.

III. What are the results?

1. What is(are) the prediction tool(s)?

The investigators did not develop a prediction tool. The investigators performed an analysis of factors that are seen in abusive head injury and accidental head injury and found the following occurred more frequently in abusive injury (p < 0.001): retinal hemorrhages, which are often bilateral, more often extend to the periphery, cover the macula and/or involve the pre-retinal layer; subdural hematomas; abnormal mental status on presentation; and seizures. In addition, they found that children with accidental head injury were more likely to have scalp hematomas.

I think further evaluation of their data is necessary to fully understand the results of the study. Looking at all of the primary outcomes, secondary outcomes and patient characteristics with statistical differences between the two groups in more detail is warranted. Below is a table of relative risk (RR) and likelihood ratio (LR) for all statistically significant differences including 95% confidence intervals (CI).

Primary Outcomes

Clinical Finding	Relative Risk (95% CI) (of finding in abuse)	Likelihood Ratio (95% CI) (likelihood in abuse compared to accident)
Retinal Hemorrhages	6 (2.6-13.6)	5.7 (2.6-11.4)
Bilateral RH	7 (3.4-8.9)	26.8 (4.7-166)
Pre-RH	7 (3.4-12.6)	47 (3-803)
RH extending to periphery	6 (3.3-11.4)	38 (2-675)

Secondary Outcomes and Patient Characteristics

Clinical Finding	Relative Risk (95% CI)	Likelihood Ratio (95% CI)
Seizures	6.6 (3-12)	8.9 (3.3-24)
Abnormal Mental Status	5.1 (2.8-34)	5.1 (2.2-10.7)
Subdural Hematoma	6.9 (2.3-21)	3 (1.8-3.8)

2. How well does the model categorize patients into different levels of risk?

It doesn't. They did not try to define levels of risk or a true "model." Rather they set out to find characteristics that may help distinguish between the two types of injury.

For retinal hemorrhages overall the likelihood ratio is ~6 implying that patients with abuse are 6 times as likely as accidental head injury patients to have RH's. This in fact is a very strong association, but not predictive alone. However, pre-retinal hemorrhages and retinal hemorrhages extending to the periphery both have very significant likelihood ratios. 95% confidence intervals would suggest that both of these primary outcome measures would be present in less than 6% of the accidental head injury patients.

In addition, children who present with abusive head injury are 9 times as likely to present with seizures as accidental injuries. Abnormal mental status on presentation is also strongly associated with abusive injury (RR = 5.1, CI 2.8-34). However, the large confidence interval makes this a much less reliable predictor of abuse. None of the primary or secondary outcomes are absolutely "predictive", but several are strongly associated with abusive injury.

3. How confident are you in the estimates of risk?

Estimates of risk of abusive head injury were not performed. Characteristics that may help distinguish accidental from abusive head injury were identified, but no actual prediction model with risk estimates was done. Furthermore, no confidence intervals (CI) were provided, which is how we determine precision. For example, the incidence or RH in abuse cases was 60%. This is a point estimate of a rate. The 95% CI is 36% to 80% which is not very precise. Likewise, the 10% RH rate in the accidental cases carries a 95% CI of 5% to 20%.

More importantly, what are we to make of the reported 0% incidence of pre-RH? Can we conclude then that accidental head trauma cases never could have a pre-RH? Calculating the 95% CI here is crucial. (see table above) For the 0% incidence here out of 67 cases, the upper limit of the 95% CI is 6%. So the best we can conclude is that, with 95% confidence, less than 6% of accidental head injured children will have pre-RH.

Therefore, this is not a guaranteed to only be seen in abuse cases, but strongly associated with cases of abusive head injury.

IV. Will the results help me in caring for my patients?

1. Does the tool maintain its prediction power in a new sample of patients?

Unclear. The investigators do not develop a true prediction tool from this study. Several factors are identified that are significantly more common in one type of head injury or the other. These could be helpful in distinguishing accidental from abusive head trauma in young children. Particularly pre-RH's, premacular RH's and RH's that extended to the periphery, seizures, and an abnormal mental status on presentation are all more commonly seen in the abused children.

2. Are your patients similar to those patients used in deriving and validating the tool(s)?

Unclear. A more detailed description of the patient population in this study would help in generalizing their results to my patient population. The investigators do not provide a detailed description of the patients, severity of illness, concomitant injuries, or operative procedures which could permit a better comparison to one's own patient population.

3. Does this tool improve your clinical decisions?

Yes. Differentiating between accidental and abusive head injury in young children can be extremely difficult. However, the classification of an injury being accidental vs. abusive is difficult to make for these injuries. The authors used historical and physical examination findings and an evaluation of the child's psychosocial status rather than type of intracranial injury which is similar to that used by Reece (1). This study points out several clinical features that are more suggestive of one type of injury or the other. The results in this study serve to validate findings in previous studies while underscoring the significance of not relying on one clinical parameter to distinguish between abusive and accidental head injury.

However, this study may only create more confusion in the legal system in that 7 of 67 children with accidental injury also had retinal hemorrhages, all be it that 6 of the 7 were unilateral. It may help with clinical decisions when one finds a pre-retinal hemorrhage or RH extending to the periphery or pre-macular RH or vitreous RH as these weren't found in accidental trauma patients assuming no bias on the ophthalmologist's behalf and taking into account the aforementioned CI's.

4. Are the results useful for reassuring or counseling patients?

Possibly. In the event that head trauma occurred in a child while not in the parent's care, it could be reassuring to the parents to be able to tell them whether the injury is likely or unlikely to be the result of child abuse based on the characteristics outlined in this study.

References

1. Reece RM, Sege R. Childhood head injuries: accidental or inflicted: Arch Pediatr Adolesc Med. 2000;154:11-15 [abstract]

 

Comments

Back to the EB Journal Club Index