A prospective observational study to assess the diagnostic accuracy of clinical decision rules for children presenting to emergency departments after head injuries (protocol): the Australasian Paediatric Head Injury Rules Study (APHIRST)

Children with clinically significant intracranial injuries require urgent identification to prevent further damage to the brain. Cranial computed tomography (CT) scans provide rapid and definitive identification of the presence or absence of intracranial injuries, and help guide subsequent management. Positive results allow early intervention and optimise outcomes whilst negative results are reassuring and may allow accelerated discharge and reduce unnecessary admissions.

However, cranial CT scans also have negative effects, particularly in children, who are more vulnerable to radiation-associated cell damage [1]. Radiation from cranial CT scans can cause lethal malignancies with higher risk in younger age groups [1‐4]. Children may require sedation to allow imaging with consequent sedation-associated risks [5, 6]. They also have resource implications for Emergency Departments (EDs) and the health system as a whole [7]. Despite this, the number of cranial CT scans performed for head injuries in children has increased in a number of countries [8‐11]. This increase is likely due to a combination of easier access to CT scanners and more efficient technology and concern amongst physicians of being unable to reliably identify intracranial injury based solely on a child’s clinical condition. One way of increasing clinical sensitivity and specificity (i.e. minimising both missed clinically significant intracranial injuries and unnecessary investigations) is to develop and use clinical decision rules (CDRs).

CDRs help physicians with diagnostic and therapeutic decisions, and can be defined as decision making tools derived from original research (as opposed to a consensus-based clinical practice guideline) which incorporate three or more variables from the history, physical examination, or simple tests. These tools help clinicians cope with the uncertainty of medical decision making and improve their efficiency [12]. Several recent systematic reviews of existing paediatric head injury CDRs have been published [13‐15]. The three CDRs of highest quality and accuracy [15] are the Canadian Assessment of Tomography for Childhood Head Injury (CATCH) from Canada [11], the Children’s Head Injury Algorithm for the Prediction of Important Clinical Events (CHALICE) from the UK [16] and the prediction rule for the identification of children at very low risk of clinically important traumatic brain injury developed by the Pediatric Emergency Care Applied Research Network (PECARN) from the USA [17]. All three CDRs were derived with high methodological standards using large multicentre data sets. However, they differ in key areas, including study population, predictor variables (based on mechanism of injury, clinical history, and clinical examination) (Table 1), inclusion and exclusion criteria (Table 2) and outcomes (including the terminology and definitions used) (Table 3). Most importantly the focus is different in each CDR. CATCH was derived to manage children with minor head injuries presenting within 24 hours, with specific inclusion criteria to be fulfilled before employing the CDR. CHALICE was derived for children with head injuries of all severities, presenting at any point after the injury. Both aim to identify children likely to have significant intracranial injury who warrant a cranial CT scan. PECARN’s CDR focuses on children with minor head injuries presenting within a 24 hour period and aims to identify patients unlikely to have a clinically important traumatic brain injury who can be safely discharged without a CT scan. In addition PECARN has derived different CDRs for children aged less than two years and children aged two years and older. The comparative performance accuracy (as assessed by sensitivity, specificity, negative predictive value and positive predictive value) for each CDR has been presented elsewhere [15]. CATCH and CHALICE CDRs suggest a dichotomous course of action (cranial CT scan/no cranial CT scan) although CATCH stratifies this risk into high and medium categories. The PECARN CDR defines a low risk population in whom cranial CT scans can routinely be obviated.

PECARN’s is the only CDR which has been internally [17] and externally [18] validated. A CATCH validation study has been performed in the derivation setting though results are only available in abstract form at present [19]. Recently the three CDRs have been prospectively validated in the same cohort of 1,009 children presenting to an urban medical center with a designated paediatric ED in the United States. This study showed that baseline physician ordering practice and PECARN outperformed the other CDRs. However, the study population did not reflect the exact population for which each rule was originally developed and the study was underpowered to determine narrow confidence intervals for rare but critically important events [20].

We propose to validate and compare the accuracy of the CATCH, CHALICE and PECARN CDRs using prospectively collected data from 20,000 patients in a multicentre setting in Australia and New Zealand, i.e. outside the countries where these CDRs were derived, and compare their performance against that of our current practice. Triggers for cranial CT use by clinicians in paediatric EDs in Australia and New Zealand are different from the triggers developed in CATCH, CHALICE and PECARN [21]. This study will also help determine which CDR is best suited for use in the Australian and New Zealand setting before incorporating them into local practice.

When applying each CDR, items will be scored as present, absent or unknown. Sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) of each of the CDRs will be calculated using the definitions and parameters set out in the derivation studies as published [11, 16, 17]. In addition, the two CDRs limited to minor head injuries (CATCH and PECARN) will also be applied to patients of all head injury severities to assess their performance in this extended patient group. Likewise, the CHALICE CDR, though derived for all severities of head injury, will undergo separate analysis in minor head injury to allow comparison of performance accuracy of the three CDRs in that population. Performance accuracy will also be calculated in patient subgroups including but not restricted to patients with bleeding diathesis and ventriculoperitoneal shunts. Rates of secondary outcomes such as cranial CT, neurosurgical intervention, ciTBI and CSII and missed ciTBI and CSII will be calculated. Key percentages will be presented with 95% confidence intervals. Data will be entered using Epidata (The Epidata Association, Odense, Denmark) and analysed using Stata 12 (Statacorp, College Station, Texas, USA).

Ideally, all patients with head injuries would receive a cranial CT to determine the presence or absence of significant intracranial injuries. However, this would expose a large number of patients to unnecessary CTs and the associated cancer related risks; therefore, similar to the methodology used in the derivation and validation studies for CATCH [11] and PECARN [17] this study relies on patient follow up by telephone. In doing so we will establish whether a relevant outcome has occurred or not.

CT rates in Australia and New Zealand may be lower [23, 25] than in North America and as reported for the CATCH and PECARN studies [8, 10, 11, 17] and higher than the baseline rates reported from the United Kingdom in the CHALICE study [16]. This highlights one of the potential key strengths of this study as it tests the CDRs in a setting different to that in which each one was derived. Finally, while we were provided with copies of the telephone follow up questionnaires used by the CATCH and PECARN investigators (personal communication, Dr Martin Osmond and Dr Nathan Kuppermann) we reconstructed the predictor variables for the three CDRs solely from the published papers [11, 16, 17]. This may have introduced an element of interpretation in terms of the most precise wording to be used in a clinical emergency setting.

This study will allow the simultaneous comparative application and validation of three major paediatric head injury clinical decision rules outside their derivation setting. In addition to a high recruitment rate, the study will depend on high follow up rates to ensure that our results accurately represent the whole population of children presenting with head injuries.

We have so far recruited more than 10,000 of the planned 20,000 patients. We will complete recruitment by the end of 2014.