Heads Up on Head Injury Algorithms: The Cost of High Sensitivity

By Jeff Russ, MD, PhD – Pediatric/Child Neurology Resident, UCSF

Jeff Russ Headshot BW

Dr. Jeff Russ

Children presenting with head injury are as unremitting in children’s hospitals as the “Frozen” soundtrack, and any physician in a pediatric ED inevitably manages their fair share. The ramifications of missing significant injury to a child’s delicate, developing brain are unnerving. A head CT is central to catching intracranial pathology, but widespread use is not benign, given the risk of malignancy from unnecessary radiation. However, criteria for judiciously navigating this tradeoff remain debated. When is CT appropriate for children with GCS scores of 13-15 and mild symptoms like transient loss of consciousness or vomiting?

A number of guidelines exist to predict when children with a head injury might warrant CT. Physicians may be most familiar with the PECARN algorithm1, endorsed by the American Academy of Pediatrics, but similar guidelines have been developed elsewhere: CATCH2 in Canada and CHALICE3 in the United Kingdom. Though these prediction tools are widely implemented, they have undergone minimal external validation.

Fortunately, a recent study in Lancet4 compared these three tools. First, they replicated the high sensitivities of each tool for predicting the original studies’ (slightly differing) outcomes. Then, to compare performance across the three tools, they measured the ability of each to predict a composite outcome of “clinically important traumatic brain injury.” For this common measure, they found sensitivities of 91.9% for CATCH and 92.5% for CHALICE, while the PECARN algorithm had sensitivities of 100% and 99.2% for children younger and older than two, respectively.

If the goal is a highly sensitive tool, then PECARN appears to win out, since it catches almost every single case of clinically important brain injury. The authors state, “…given the mortality and morbidity associated with missing an intracranial lesion… clinicians therefore prioritize a very high sensitivity4.”

Is there a cost to high sensitivity? By its nature, high sensitivity comes with a high false positive rate. Indeed, while this study showed CATCH and CHALICE to have false positive rates of 29.6% and 21.4%, respectively, PECARN had false positive rates of 40.9% and 48% for children under and over two4. Moreover, the positive predictive value (PPV) of PECARN is only about 2%1,4, indicating that even for children meeting criteria, the vast majority will still have negative imaging findings.

Does this matter? A JAMA study from 2013 attempted to estimate pediatric cancer risk from CT radiation5. Head CTs were the most common type ordered for children, and girls under five were at highest lifetime risk of cancer. The study estimated that about one in 5,000 head CTs in children under five would result in a single case of leukemia, and one in about 600 head CTs in girls under five would result in a solid tumor5.

Here is a back-of-the-napkin calculation using the original PECARN data for children under two1,5: If 10,000 children present with mild head injury and 47% are positive for any predictor, then at least 4,700 could receive a head CT. With a PPV of 2%, about 100 would have a brain injury identified, and 4,600 would have a negative head CT. Four girls with negative scans might then develop a solid tumor, and one child might develop leukemia.

This rough analysis still supports the assertion that using PECARN, the number needed to scan to catch a clinically important brain injury outweighs the number needed to harm with CT-induced malignancies. However, the gap is closer than might be expected, especially for extremely young children.

How might this information affect clinical practice? PECARN divides patients into three tiers. For patients who have predictive features other than altered mental status or skull fracture, clinicians must use their judgment to decide between a CT and observation1. Especially for younger patients, observation may be a more prudent course. Despite the impressively high sensitivity of PECARN, the low incidence of actual brain trauma and the surprisingly high number needed to harm in young children should be carefully weighed.

Clinical judgment and the PECARN algorithm are a physician’s best initial tools for milder head injuries, but if PECARN recommends observation versus CT, consider erring on the side of observation, with clear communication to patients and families regarding risks and benefits, including the use of evidence-based decision aids6 as appropriate.



  1. Kuppermann N, Holmes JF, Dayan PS, et al. Identification of children at very low risk of clinically-important brain injuries after head trauma: a prospective cohort study. Lancet. 2009;374(9696):1160-70.
  2. Stiell IG, Wells GA, Vandemheen K, et al. The Canadian CT Head Rule for patients with minor head injury. Lancet. 2001;357(9266):1391-6.
  3. Dunning J, Daly JP, Lomas JP, et al. Derivation of the children’s head injury algorithm for the prediction of important clinical events decision rule for head injury in children. Arch Dis Child. 2006;91(11):885-91.
  4. Babl FE, Borland ML, Phillips N, et al. Accuracy of PECARN, CATCH, and CHALICE head injury decision rules in children: a prospective cohort study. Lancet. 2017;
  5. Miglioretti DL, Johnson E, Williams A, et al. The use of computed tomography in pediatrics and the associated radiation exposure and estimated cancer risk. JAMA Pediatr. 2013;167(8):700-7.
  6. Hess EP, Wyatt KD, Kharbanda AB, et al. Effectiveness of the head CT choice decision aid in parents of children with minor head trauma: study protocol for a multicenter randomized trial. Trials. 2014;15:253.