Pediatric Emergency Care Applied Research Network head injury clinical prediction rules are reliable in practice
- Deborah Schonfeld1,2,
- Silvia Bressan3,
- Liviana Da Dalt3,
- Mira N Henien1,
- Jill A Winnett1,
- Lise E Nigrovic1
- 1Division of Emergency Medicine, Boston Children's Hospital, Boston, Massachusetts, USA
- 2Division of Emergency Medicine, the Hospital for Sick Children, Toronto, Ontario, Canada
- 3Department of Woman's and Child's Health, University of Padova, Padova, Italy
- Correspondence to Dr Lise E Nigrovic, Division of Emergency Medicine, Boston Children's Hospital, 300 Longwood Avenue, Boston, MA 02115, USA;
- Received 5 August 2013
- Revised 14 December 2013
- Accepted 16 December 2013
- Published Online First 15 January 2014
Objective The Pediatric Emergency Care Applied Research Network (PECARN) traumatic brain injury (TBI) age-based clinical prediction rules identify children at very low risk of a significant head injury who can safely avoid CT. Our goal was to independently validate these prediction rules.
Design Cross-sectional study.
Setting Two paediatric emergency departments located in USA and in Italy.
Patients All children presenting within 24 h of a head injury with a Glasgow Coma Score of ≥14.
Intervention Assessment of PECARN TBI clinical predictors.
Main outcome measure Clinically important TBI defined as head injury resulting in death, intubation for >24 h, neurosurgery or two or more nights of hospitalisation for the management of head trauma.
Results During the study period, we included 2439 children (91% of eligible patients), of which 959 (39%) were <2 years of age and 1439 (59%) were male. Of the study patients, 373 (15%) had a CT performed, 69 (3%) had traumatic findings on their CT and 19 (0.8%) had a clinically important TBI. None of the children with a clinically important TBI were classified as very low risk by the PECARN TBI prediction rules (overall sensitivity 100%; 95% CI 83.2% to 100%, specificity 55%, 95% CI 52.5% to 56.6%, and negative predictive value 100%, 95% CI 99.6% to 100%).
Conclusions In our external validation, the age-based PECARN TBI prediction rules accurately identified children at very low risk for a clinically significant TBI and can be used to assist CT decision making for children with minor blunt head trauma.
What is already known on this topic
The Pediatric Emergency Care Applied Research Network (PECARN) traumatic brain injury (TBI) age-based clinical prediction rules identify children at very low risk of a clinically significant head injury.
These rules were derived and then validated in a large concurrent population, but have not been externally validated to date.
What this study adds
The PECARN TBI clinical prediction rules performed well in our two-centre patient population.
Our findings in combination with the concurrent validation study, suggest that the PECARN TBI rules are ready for prospective implementation.
The number of annual paediatric emergency department (ED) visits due to head injury has dramatically increased over the last decade.1 ,2 Increasing concerns about radiation exposure from cranial CT scans has made the evaluation of children with minor head trauma particularly challenging. Emergency physicians must balance the possibility of missing a clinically significant traumatic brain injury (TBI) with the future risk of malignancy associated with ionizing radiation.3 Clinicians often rely on clinical prediction rules to assist decision making by estimating the risk of a given outcome.4 Only clinical prediction rules that have been derived according to rigorous clinical standards and validated in a wide variety of clinical settings, should be implemented in routine clinical practice.5 ,6
Using a large prospective cohort study of children with minor blunt head trauma, the Pediatric Emergency Care Applied Research Network (PECARN) derived two aged-based TBI clinical prediction rules to identify children who are at very low risk of having a clinically significant injury who may safely avoid cranial CT.7 These rules were validated in a concurrent patient population enrolled at the same study centres after derivation population. To date, the PECARN TBI rules have not been externally validated and, therefore, the broader generalisability has not been adequately examined.
We performed a two-centre cross-sectional study of children presenting to the ED with minor blunt head trauma. Our goal was to evaluate the performance of the PECARN TBI age-based clinical prediction rules in an external patient population.
Study design and setting
We conducted a cross-sectional study of children with minor blunt head trauma presenting to the ED for evaluation at one of two paediatric EDs located in Boston, Massachusetts (USA) and Padova (Italy). Padova Children's Hospital is a tertiary care academic hospital with approximately 25 000 paediatric ED visits per year. Boston Children's Hospital is a tertiary care academic hospital with a level 1 trauma centre designation and an annual ED census of approximately 65 000 paediatric ED visits per year.
The study protocol differed by participating centre. The PECARN TBI rules were published in October 2009. In Boston, we performed a prospective cross-sectional cohort study (April 2011–July 2013). Clinicians in Boston had an available head trauma guideline based on the PECARN TBI rules (institutional guideline released in August 2011). The PECARN TBI rules were introduced into clinical practice in May 2010 in Padova. In the Italian ED we performed a prospective (June 2010–November 2010)8 cross-sectional study, immediately followed by retrospective (December 2010–May 2011) data collection. The institutional review board of each participating centre approved the study protocol.
We included all children with blunt head trauma and an initial Glasgow Coma Score ≥14, who presented to the ED within 24 h of injury. In Boston we included children <18 years of age and in Padova children <15 years of age (reflecting current patient triage practices at this site). We excluded children with a trivial injury mechanism (eg, ground-level falls or running into stationary objects with no signs of TBI other than scalp abrasions and lacerations), as well as those with any of the following: neurological comorbidities, bleeding disorders or suspected child abuse. Patients with neuroimaging performed prior to ED physician evaluation were excluded. For the prospective cohorts, we also excluded children who had any neuroimaging performed prior to study form completion.
Paediatric emergency medicine or general paediatrics attendings (Boston and Padova), paediatric emergency medicine fellows (Boston) or senior residents (Padova) completed all study forms for the two prospective cohorts. For the retrospective cohort, a trained researcher (SB) performed a standardised medical record review and chart abstraction but was not blinded to the patient's clinical outcome. Study forms were completed for all patients to capture the presence or absence of each of the six PECARN age-based TBI predictors including the composite predictors (table 1). Children who arrived in the ED with an initially normal mental status which declined within 1 h of ED arrival were classified as having an abnormal mental status.
For the retrospective and prospective Padova cohort, children who were discharged from the ED without a CT scan were contacted by telephone for clinical follow-up approximately 2 weeks after initial evaluation. For the Boston prospective cohort, clinical follow-up was limited to complete hospital medical record review for the 2 weeks from the initial ED evaluation to determine whether or not the study patient had any neuroimaging performed or any clinical interventions for the management of their head injury. For the purposes of this study, children who had either cranial MRI or CT performed were included in the CT group.
Our primary outcome was the presence of a clinically important TBI (ie, gold standard for the clinical prediction rule) defined as death, intubation >24 h, neurosurgery or two or more nights in the hospital for management of the head injury.7 Our secondary outcome measure was a positive CT defined as any of the following traumatic findings: intracranial haemorrhage or contusion, traumatic infarction, sigmoid sinus thrombosis, diffuse axonal injury, pneumocephalus, midline shift or signs of brain herniation, diastasis of the skull, and/or skull fracture.7
We described the data with population proportions with 95% CIs. The age-based PECARN TBI prediction rules were applied to the study population to stratify children into three risk groups (very low, intermediate and high) based on published classifications (table 1). Children with no TBI predictors were classified as very low risk. Children with one or more high-risk TBI predictors were classified as high risk and those with no high risk but one or more intermediate-risk predictors were classified at intermediate risk. Children who were missing one or more predictors with no high-risk predictors could not be classified. We calculated the performance of the two age-based PECARN TBI rules in our study population for the primary (clinically important TBI) and the secondary (positive CTs) outcome measures. We report the sensitivity, specificity, negative predictive values (NPV) and positive predictive values for the age-based rules independently and combined together.
We used the Statistical Program for the Social Sciences for all analyses.9
During the study period, 2750 eligible children presented to two participating EDs of which 2439 (91%) were enrolled in our study (figure 1). The vast majority of patients were missed during overnight hours when research co-ordinators were not available to assist with enrolment. Missed eligible patients had a similar age distribution and CT rate as those children included in the study (data not shown).
Of 2439 included patients, 959 (39%) were <2 years of age and 1439 (59%) were male. Overall, 371 (15% of study population) had a CT performed and 2 (0.1%) patients underwent MRI, for an overall neuroimaging rate of 15%. Of the 373 children that had neuroimaging, 69 (18%) had a positive CT. Overall, 19 patients (0.8% of entire study cohort) had a clinically important TBI. Of those with a clinically important TBI, two children required neurosurgery and all 19 children were hospitalised for two or more nights for management of their head injuries. We completed clinical follow-up for 578 patients (81% of children in the Padova cohorts who did not have an initial neuroimaging) and none had a clinically important TBI.
CT rates differed between participating centres (304/1657 (18%) Boston vs 69/782 (9%) Padova, p<0.001), although positive CT (59/304 (19%) Boston vs 10/69 (15%) Padova, p=0.34) and clinically important TBI (11/1651 (0.7%) Boston vs 8/782 (1%) Padova, p=0.35) rates were similar.
We successfully applied the PECARN age-based TBI clinical prediction rules to 2428 children (99.5% of study patients). The CT rate varied by the PECARN TBI risk strata: 139/197 (70%) of high-risk patients, 199/918 (22%) of intermediate-risk patients and 34/1313 (3%) of very low-risk patients had a CT performed (p<0.001).
We assessed the performance of the PECARN TBI prediction rules for the primary outcome (table 2) and secondary outcome measure (table 3) overall and for children <2 years of age and ≥2 years. None of the children with a clinically important TBI were classified as very low risk (table 4). In the 372 children to whom we could apply the PECARN TBI rules and who had a CT performed, we measured the ability of these rules to identify children with a positive CT: sensitivity 97.1% (95% CI 90.0% to 99.2%), specificity 10.5% (95% CI 7.6% to 14.5%) and NPV 94.1% (95% CI 80.9% to 98.4%). Two children with a positive CT were classified as very low risk. Both of these children had an isolated skull fracture without intracranial injury and were hospitalised overnight without requiring other interventions.
We report the first published external validation of the PECARN TBI clinical prediction rules in a cohort of 2439 children with minor blunt head trauma who presented to one of two EDs located in the USA or Italy. In our study cohort, the two age-based rules had excellent sensitivity and NPV for clinically important TBI as well as positive cranial CT. Children in the PECARN TBI very low risk group can safely avoid a CT scan, as the risk of significant head injury is quite low.
Our results were similar to the previously published concurrent validation study.7 In that study of 8627 children, only two of the 88 children with a clinically important TBI were classified as very low risk. These children were ≥2 years of age and required two or more nights of hospitalisation for management of their head trauma. Our external validation study adds to these results in two ways. First, our study was conducted independently from the rule derivation at two paediatric EDs located in different countries, increasing generalisability. Second, the PECARN TBI rules include two composite predictors (altered mental status, severe injury mechanism). Our validation study was the first to collect the PECARN TBI predictors as composite variables rather than individual components.
In this study the PECARN TBI rules classified two children with a positive CT as very low risk. Both children had an isolated skull fracture and were hospitalised overnight but required no acute neurosurgical or social work interventions. In particular, most children with non-depressed skull fractures without intracranial injury (ie, an isolated linear skull fracture), do not require hospitalisation as the risk of requiring neurosurgical intervention is quite low. Despite this fact, current admission rates for children with isolated skull fractures remain high (approximately 80%).10 Although CT scans may not be necessary for some types of TBI, clinicians must maintain a high level of suspicion for non-accidental trauma, especially in the youngest children.11 ,12
PECARN TBI rules can assist clinical decision making by providing an accurate assessment of the risk of a significant head injury. Children at very low risk for TBI may not need a cranial CT as ionizing radiation from CT has been associated with an increased lifetime risk for radiation-associated malignancies.13–16 For children who are not at very low risk, observation can also be an important management strategy, allowing selective CT use for those children whose symptoms progress or fail to improve during the period of observation. Observation has been associated with a significant reduction in cranial CT rate with no change in rate of clinically important TBI.17 ,18 Importantly, children with significant head injuries very rarely have delayed presentations (more than 4–6 h after injury),19 mitigating the risk of missing a significant TBI when clinicians choose to observe a child prior to CT decision making.
Patient and parental preference should play a significant role in CT decision making for children with minor blunt head trauma. In a recent survey of parents of children with head trauma, the majority preferred observation to immediate CT after a standardised education about the risks of ionizing radiation.20 In a second survey of parents of ED patients, approximately half were aware of the risks of ionizing radiation from diagnostic CT scans.21 Additionally, most parents preferred disclosure of risk before proceeding with imaging. Further study is needed to better understand the optimal methods of risk communication as well as parental preferences for the ED management of minor head trauma for their children.
Our study has the following limitations. First, our study was conducted in two academic paediatric EDs, raising the possibility that our findings may not be generalisable to other settings, particularly general EDs that care for children. Although additional impact studies are needed, the implementation of the PECARN TBI might may have a greater impact on CT utilisation in general than paediatric EDs since current CT rates for children with blunt head trauma are higher at general than at paediatric EDs.22 Second, the data was abstracted from the medical record for some study patients. However, the Padova retrospective cohort followed the prospective implementation study and treating clinicians were previously trained to collect and record the age-based PECARN TBI predictors. Third, of the five children who developed altered mental status within 1 h of initial assessment, two had a clinically important TBI. These injuries might have been missed injury if clinical decision making occurred immediately on arrival to the ED. We considered these patients to be high risk by the PECARN TBI rules, further highlighting the utility of clinical observation prior to CT decision making to minimise the risk of missing a significant head injury. Fourth, not all patients had a CT scan performed. As children with a positive CT may not require specific interventions (eg, children with an isolated skull fracture10), we likely underestimated the rate of our secondary outcome measure (positive CT). Fifth, we did not perform clinical follow-up for all patients discharged from the ED without a CT scan although we did review all medical records after the initial ED visit for repeat visits within 2 weeks of initial injury. In the previous PECARN TBI derivation study where clinical follow up was performed, only 1 out of the more than 40 000 children studied had a clinically important TBI diagnosed after initial ED discharge,7 suggesting that risk of missing our primary outcome measure was quite low even if the patient presented again to another ED. Sixth, although we were unable to apply the PECARN TBI rules to those children missing clinical predictors, we only excluded 11 children (0.5% of study patients) for missing data, minimising the potential impact or bias on our study findings. Last, few children had a clinically significant TBI which limited our ability to assess rule sensitivity. Future multicentre validation studies are needed to further demonstrate the accuracy and generalisability of the PECARN TBI prediction rule across different domains, such as general EDs.
In our external validation study, the PECARN TBI age-based clinical predication rules performed well by accurately identifying children who are at very low risk for a clinically important TBI who can safely avoid a CT scan. These can be used to assist with clinical decision making for children with minor head trauma. Prospective implementation of the PECARN TBI rules must be carefully studied to determine the impact on CT rates as well as missed injuries.
The authors would like to thank the research coordinators who collected patient information in the emergency department (John Andrea BA (Biddeford, ME), Paola Berlese MD (Padova, Italy), Brittany Kronick BA (New York, NY), Teresa Mion MD (Padova, Italy), Kaitlin Morris BA (Boston, MA), Sabrina Romanato MD (Padova, Italy) Sandy Wong BA (Boston, MA)). This study was supported by the Boston Children's Hospital House officer Development Award as well as the Hood Childhood Research Grant.
Contributors DS and LEN were responsible for the conception and design of the study and drafted the manuscript. DS, SB, MNH and JAW acquired the data. DS, SB, LDD and LEN were responsible for the analysis and interpretation of data. All collaborators critically revised the manuscript and granted final approval of the submitted manuscript.
Funding Hood Childhood Research Grant, Boston Children’s Hospital House-officers Grant.
Competing interests None.
Ethics approval Boston Children's Hospital, University of Padova.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Please contact the corresponding author to discuss data sharing.