Validation of the Scandinavian guidelines for minor and moderate head trauma in children: protocol for a pragmatic, prospective, observational, multicentre cohort study

Study design

This protocol describes methodology for a prospective multicentre observational cohort study of children aged 0–17 years (definition of children in Scandinavia) with mild and moderate head trauma presenting to 16 EDs in Sweden and Norway, aiming to validate the SNC16 in the intended setting. The study has been registered at ClinicalTrials.gov under the reference NCT05964764. Reporting follows the Standards for the Reporting of Diagnostic accuracy studies32 (see online supplemental file 1).

Eligibility criteria

All children (age 0–17 years) with blunt head trauma within the preceding 24 hours and GCS 9–15 at initial assessment in the ED of a study hospital in Sweden or Norway are eligible for inclusion. Children who have a GCS score of 14 or 15 on arrival in ED (determined by the attending physician) are defined as having a minimal or mild head trauma (minor head trauma). Children with a GCS score of 9–13 are defined as suffering from moderate head trauma, based on the SNC16. Enrolment can be fulfilled after informed oral consent is obtained from at least one guardian (and the child if he/she is sufficiently mature enough to understand the meaning of the consent (normally age above 14 years)).33 Written information is provided to all eligible patients and parents. A patient cannot be included if any of the following criteria are present: patient and/or guardian does not wish to participate in the study, patient is included in other study that may affect the management/treatment in ED, patient suffers from a penetrating head injury, child abuse/non-accidental trauma (NAT) denoted as suspected child abuse where the social services are informed (as these patients will always receive a cCT) or the patient is not a citizen with social security number in the participating country (difficult to follow-up).

Clinical management, including referral for cCT or admission for in-hospital observation, is independent of enrolment in this study as clinicians follow their routine clinical practice, including whatever guideline is used locally. Children with multitrauma injuries are managed via Advanced Trauma Life Support) principles by a trauma team in Scandinavia. These injuries are relatively uncommon and are always admitted to hospital, often with CT scanning, including cCT.

Setting

The study will involve 15 Swedish hospitals and 1 Norwegian hospital. Enrolment of study participants/patients occurs in the general ED and/or paediatric ED at the respective hospital.

Hospitals of different sizes, from university hospitals in larger cities to district and regional hospitals in smaller cities, were invited to participate. It was not possible to control the distribution of participating hospitals due to differing resources allocated for research in different regions and hospitals in Scandinavia. To antagonise introduction of selection bias, efforts were made to obtain a dispersion in participating hospital’s sizes as well as archiving representation in various regions in Scandinavia.

When a hospital was preparing for participation, a set of measures was taken. A physical or digital meeting with the appointed local study coordinator was carried out, containing a structured review of background, methodology and ethical considerations meaningful for the study aims and procedures. Each new hospital applying for participation was approved by the ethical review board in their respective country. When approved, physical and/or digital meetings with nurses and doctors, working both in the ED and in the department with responsibility for management of paediatric head trauma, were carried out. In these meetings, methodology, background and enrolment procedures were reviewed.

Each site has a doctor fulfilling the role of local study coordinator, who usually works in the ED managing these patients. The local study coordinator is responsible for the local maintenance of the study, which involves supervising a functioning system for screening, information and inclusion in the ED and performing the follow-up assessment via medical records and journals. Documents and flyers regarding screening, study information to participants and inclusion routines in the ED, as well as routines for follow-up assessment at 3 months, were provided by the national study coordinator and senior investigator.

Attending doctors and nurses at study sites are aware of the SNC16 but are instructed to manage children in accordance with the hospital’s ordinary guidelines and follow local management routines. A recent study has outlined the current management routines used in Sweden.34 They are also encouraged to not change their local protocols for ordering CT.

Recruitment of patients

Potentially eligible patients are identified in the triage room by the ED nurse or later by the assessing doctor. Oral and written information about the study is provided and informed consent to participate is obtained by the assessing physician or triage nurse.

The primary assessment of children with recent head trauma in EDs is performed according to each hospital’s local regime, involving paediatricians, orthopaedic surgeons, emergency medical doctors, general surgeons, junior doctors or trainees under supervision, depending on type, size and organisation of the department.

In Sweden and Norway, some children, especially with minimal TBI, are managed by nurses in the triage room and never meet a doctor before they are sent home. These patients will also be included, and a notation of nurse-only assessment is made in the electronic case report form (eCRF).

Data collection

Clinical data from the ED, data from the patient’s medical records and follow-up questionnaires are registered in a study database in Entermedic (provided by Entergate AB), as illustrated in figure 1.

Figure 1

Flow chart illustrating screening, inclusion and data collection from time of event to 4 months post-trauma. Long-term follow-up at 5 years is planned. CRF, case report form; ED, emergency department; TBI, traumatic brain injury.

When patient consultation and initial management is complete and the patient has left the ED, either discharged or admitted for in-hospital observation, a web-based hospital-specific questionnaire (questionnaire 1, Q1) is available to be filled in by the managing doctor or nurse. Baseline data, trauma-related factors, medical history and objective physical findings, data on clinical management in the ED and subjective reason for clinical decisions are registered in Q1, as shown in online supplemental file 2. The respondent’s medical specialty if he/she is a doctor, or whether the respondent is a nurse, is registered in Q1. The name of the including clinician is not collected. Assessment of perceived benefit and usefulness of the SNC16 when hypothetically applied to the recently enrolled patient is measured on a 7-grade Likert scale as the respondent takes a position on four statements in the final part of Q1.

Medical records from the region of the enrolling hospital are screened by the local study coordinator or his/her assistant at a time point >30 days after the patient was discharged from ED or hospital (if admitted to ward for observation) for relevant outcomes. These include reassessments, readmissions, CT/MRI reports, observation periods in the ED and/or in-hospital, intracranial complications, neurosurgical interventions and reports, need for ICU admission and/or death after enrolment in the study, as shown in online supplemental file 2. Data are collected in a dedicated eCRF in the research database. In case patients are lost to follow-up via medical records (eg, lives in another region of the country), primary endpoints are retrieved via a caregiver survey, as described below.

Caregivers provide their email and phone number, along with the consent for their child to participate, at the visit in the ED. The follow-up questionnaires (questionnaire 3–5; figure 1 and online supplemental file 2) is sent out at 30, 90 and 120 days after enrolment, by mail or text message. Up to three automatised reminders (with 7 days in between) can be sent, first by email and then the last two by text message. Finally, the principal investigator makes a phone call (if a phone number is registered and correct) to obtain a response.

Radiology

cCT scans are ordered at the discretion of the physician and independent of patient enrolment in the study, guided by local routines and national recommendations. cCT examinations in Scandinavia are interpreted by the on-call radiologist. When the radiologist on call is a resident, a verification of the primary assessment is done by a board-certified consultant. Radiologists are aware of the clinical context as it is included in the referral documentation for a cCT scan. However, they are typically not aware of enrolment in this observational study. According to current clinical practice in Scandinavia, pathological cCT scans are sent to a neuroradiologist and/or neurosurgeon for assessment. Any discrepancies in the primary interpretation are thereby corrected. Consequently, there is no need for a separate neuroradiologist interpretation. Also, this practice mirrors usual care and hence adds to the pragmatic design of the study.

Day and time for cCT’s and MR’s, need for sedation or endotracheal intubation during scan and reported results are collected at the medical records exam (Q2) by the local study coordinator. Also, all results from follow-up CT scans or MR exams are collected in Q2.

Chronic subdural haematomas, hydrocephalus, arachnoid cysts or other malformations are documented as such but considered unintentional findings and are not included in any of the primary endpoints.

cCT and MR findings are categorised as linear skull fracture, depressed skull fractures (>1 bone width), basilar skull fracture, intracerebral haemorrhage, intracerebellar haemorrhage, acute subdural haematoma, chronic subdural haematoma, epidural haematoma, traumatic subarachnoid haemorrhage, cerebral contusion(s), intraventricular blood, cerebral oedema, midline shift (in millimetres), intracranial vessel dissection, extracranial vessel dissection, skull diastasis, signs of brain herniation, sigmoid sinus thrombus, signs of diffuse axonal injury (DAI), traumatic brain infarction, shearing injury, pneumocephalus and/or other trauma related findings in free text.

Biomarkers in mTBI

Currently, biomarkers are used clinically to manage adults with mTBI in Scandinavia.28 Although the evidence seems promising, biomarkers are not yet used clinically in children with these injuries.35 Given the importance of avoiding cCT scans in this patient group and the cost of in-hospital observation, a reliable biomarker would be welcome. We aim to investigate potential improvement in diagnostic ability when adding a biomarker or a combination of biomarkers as an adjunct to the guidelines (SNC16 and other international CDRs). Currently, the most promising biomarkers seem to be glial fibrillar acidic protein,36 S100B37–39 and ubiquitin C-terminal hydrolase L1.40 Due to the practical difficulty of venous sampling in this patient group, both capillary and saliva samples may be theoretically advantageous. Venous, capillary and saliva samples will be taken at admission to the ER, in a subgroup of patients where management can be especially difficult (typically a subgroup of mTBI patients with lower risk of complications). Venous and capillary samples will be transferred to the biomedical laboratory in the hospital, where it will be allowed to clot for 30–60 min and then centrifuged. Serum is then aliquoted and stored in −80°C until later batch analysis in a central laboratory. Saliva samples will be directly aliquoted and frozen until later batch analysis.

Outcome measures

To calculate the diagnostic accuracy parameters for SNC16, the primary outcome measure was defined by a composite variable ‘clinically important intracranial injury’ (CIII), positive within 1 week from head injury. This composite consists of death, neurosurgery, admission to hospital ward 2 days or more due to head injury or intubation 1 day or more due to pathological traumatic CT findings. This definition differs from the primary outcome measure of neurosurgical intervention used by the SNC16 guideline. The composite variable above was deemed more suitable as it aligns better with international research. Also, the rate of neurosurgical intervention in this group is very low, which would require an unreasonable number of patients in order to achieve adequate confidence intervals for diagnostic accuracy. Definitions for the separate variables of this composite are defined in the secondary outcome measures.

Statistical methods

Relevant group characteristics on demography, trauma mechanism, clinical findings, risk factors, management in ED and outcomes will be presented descriptively. The mean and SD, as well as the median with IQR, will be reported for each continuous variable. Categorical data will be reported as numbers and percentages. Primary and secondary endpoints are considered dichotomous, as are inclusion/exclusion criteria and predictor variables. If data on an item are missing, the patient will be excluded from that specific analysis.

Specifically descriptive data will be, but not exclusively, presented for

1. Rate of cCTs and MRs.

2. Rate of neurosurgery.

3. Rate of hospital admission.

4. Observation rate.

5. Rate of death.

6. Rate of CDR-specific outcome.

7. Rate of patients with poor outcome on GOS-E PEDS and persistently affected PCSI-P at 3 months.

Entermedic (Entergate AB, Halmstad, Sweden) is used for data management and IBM SPSS Statistics (V.29) for all statistical analyses.

Accuracy of the SNC16 CDR

To assess diagnostic performance of SNC16 to predict the primary composite endpoint, sensitivity, specificity, likelihood ratio (LR), negative predictive value (NPV) and positive predictive value (PPV) will be presented with point estimates and 95% CIs. CI will be calculated as one-sample proportions and reported as Clopper-Pearson exact intervals. Accuracy to detect the secondary outcomes ‘neurosurgery’ and ‘significant cCT-findings’ will be specifically reported. Three different analyses will be performed with three different definitions for test positive/negative, as shown in table 1.

Gold standard for detection of intracranial trauma-related pathology is cCT. SNC16 uses both cCT and observation with varying duration depending on risk categorisation.

Rates that will be reported are as follows:

1. Application rate.

2. Mandatory cCT rate.

3. cCT or admission rate (ie, optional cCT)

4. Admission rate.

5. Missed patients with positive outcome.

Comparative analysis

Patients (the study cohort) will be analysed in different strata named application cohorts (n=6) and a comparison cohort (n=1). An application cohort is a subgroup of the study population defined by respective CDRs inclusion and exclusion criteria (table 2). This means there are six application cohorts, one each for PECARN; CATCH; CHALICE; PREDICT, NICE23 and SNC16. Predictors in the respective CDRs are shown in table 3. Furthermore, similar as in other studies, a comparison cohort of all patients with minor head injury, defined as GCS 13–15, will be created and all CDRs will be applied in this cohort.13 42 Most importantly, a comparative analysis will also be carried out in the entire cohort, for pragmatic reasons. Diagnostic accuracy parameters will be calculated when applying the prespecified CDR’s on its respective, intended cohort. For each CDR, both rule specific primary (and when defined also secondary) outcome (table 4) as well as the in-study defined primary and secondary outcome will be assessed. Application rate will be assessed and reported. For PECARN and CATCH, which, according to their respective inclusion criteria are used in patients with GCS of 14–15 or 13–15, respectively, accuracy will be calculated both in a rule specific cohort and in the entire validation cohort. Rates for cCT, admission and missed patients (including clinical characteristics), as stated under primary objective, will also be reported for each of the CDRs when relevant. In the PREDICT rule, children with head injuries can be included in the algorithm if assessed within 72 hours from trauma. In this Scandinavian cohort, the upper time limit for enrolment is 24 hours from trauma, which means that the application cohort for PREDICT is restricted to this time frame.

Collected data include inclusion and exclusion criteria, predictor/risk variables and outcomes used by the SNC16, PECARN, CATCH, CHALICE, PREDICT and NICE23 guidelines. Included variables and definitions of resembling variables are somewhat dissimilar between these CDR’s, as described by Babl et al42 and Lyttle et al43 for three of them. A description of how collected variable components (online supplemental file 2) are used to compose individual CDR variables are presented in online supplemental files 3–5. Decision trees for application of respective CDR are shown in online supplemental file 6.

CIs for specificity in regard to the in-study defined primary outcome between the included CDRs will be compared with detect statically significant differences with χ² test.

Biomarkers

Clinical variables may be susceptible to clinical judgement and are positive predictors (ie, the presence of a risk factor leads to an intervention). In contrast, biomarkers are objective and provide negative prediction for avoiding an intervention, at least in adults.28 Additionally, as biomarker results are continuous variables, it may be possible to use different cut-offs to tailor for different outcomes (eg, different cut-offs for negative and positive prediction). Biomarker results will be analysed primarily in an intermediate-risk group. This group (specifically, the ‘medium-risk’ and ‘low-risk’ mTBI groups from the SNC16 guideline would be included) would be the target population for such a test, as children with minimal head injury will be discharged irrespective of biomarker results and more severely-injured children will generally receive a cCT scan and/or admission, irrespective of any potential biomarker results. This intermediate-risk group, however, is more difficult to manage, with current guidelines (including the SNC16) recommending observation and/or cCT scans, despite the low risk of complications. This is, therefore, the group of patients where a potential biomarker will have the greatest clinical impact. Analysis will be performed to achieve sensitivities, specificities, NPV and PPV, using predefined cut-offs (if available) and derived cut-offs. Receiver operating characteristics curves will also be constructed to examine different cut-offs. Analysis concerning comparisons between venous, capillary and saliva samples will be made including, but not limited to, Bland-Altman analysis. Biomarker results will also be analysed with respect to clinical outcome and long-term effects after mTBI.

Management of missing data points

Data needed for the primary outcome measure, the composite variable named ‘CIII’ (death, neurosurgery, admission to hospital ward 2 days or more due to head injury or intubation 1 day or more due to pathological traumatic CT findings) will all be retrieved primarily from medical journals in Q2. In those cases where outcome data are missing in the medical records, these data points will be retrieved from the parent assessment form or by a directed phone call to the caregiver from the principal investigator. The same procedure will be used for each secondary outcome, although some of these will primarily be found in the parent report (Q3) and not in the medical journals (death at 3 months, outcome assessed by GOS-E PED and outcome assessed by PCSI-P). When data variables necessary to categorise and/or risk stratify the patient in a CDR is missing, this patient will be excluded from the specific analysis.

A sensitivity analysis using multiple imputation for missing data points will be presented as anappendix to the publication of the results.

Eligible but not included patients

Consecutive sampling is intended in this prospective multicentre study. Realistically, we do however anticipate a number of eligible patients to be missed for inclusion. This introduces a risk of selection bias in the sample. To assess to magnitude of bias introduced, a period (1–4 weeks) of consecutive sampling will be performed at centres of different sizes and locations (urban/rural). During this period, 100% inclusion is aimed for and screening logs will be reported and all missed patients characterised with as much data as possible. This will generate a representative subsample and enable a sensitivity analysis on gender, age, time of assessment in ED and SNC16 risk classification with χ² test.

Sample size estimation

As there is no hypothesis to be tested, is sample size based on the calculated precision of the guidelines’ sensitivity. For validation studies, there is limited empirical evidence to suggest which sample size (and more important number of events) that is needed. Vergouwe et al44 and Collins et al31 recommend at least 100 events to provide a more accurate estimate of performance. The estimated incidence of significant intracranial complications in children with mild head injuries is between 0.8% and 1.2% in earlier studies, while the presence of traumatic CT findings is between 2% and 4%.14 15 When including moderate head injuries (which has a significantly higher risk of intracranial complications), the incidence of events has been reported to increase slightly and is estimated to 1.9% for significant intracranial injuries and 5%–7% for traumatic CT findings.

With the use of a composite variable (similar but not entirely the same as the PECARN rules) defined as death, neurosurgery, admission to hospital ward 2 days or more due to head injury or intubation 1 day or more due to pathological traumatic CT findings and an estimated prevalence of 1.5% in an unselected TBI cohort including patients with GCS 13, an event rate >70 and an presumed positive guideline indicator (which means that the guideline gives advice if to perform a cCT scan and/or admit for clinical observation) of 40%, 4787 children would be required to obtain a sensitivity of 99% and a lower 95% CI of >95%. With addition of a 5% margin final sample size is calculated to 5026 patients. We aim to include 5300 children to allow for 5% lost to follow-up (of the primary endpoint), but not continue more than 4 years with active enrolment (during the COVID-19 epidemic, active recruitment was essentially halted for a period of approximately 1 year).

Logistic regression models

We anticipate sensitivity for SNC16 to detect neurosurgery and the in-study defined composite primary outcome to be high, with CI size depending on cohort size. A multiple logistic regression model will be developed with the aim to improve specificity for the SNC16 rule, and thereafter cross-validated in the cohort.

Logistic regression models to predict clinical outcome at 3 months on GOS-E PEDS (all age groups) and persistent PCS (5–18 years), defined as modified binary Zemek score 2+ on PCSI-P will also be tested.10 11

For construction of an updated guideline, risk factors will be analysed using multiple logistic regression and presentations of diagnostic abilities.

Patient and public involvement

Patients or patient advisors have not been directly involved in the planning or conduct of the study. Results will be disseminated via public media channels to inform patients and parents about results. However, patient representatives will be involved in the process of an updated clinical guideline, which will in part be based on the results of this study.

Ethics and dissemination

This a non-interventional observational validation study of the SNC16, carried out in accordance with the Declaration of Helsinki.

Ethical application is approved by the ethical review board in Lund, Sweden (EPN Lund Dnr 2017/238, with approved amendments ref EPN Lund Dnr 2018/670; EPN Lund Dnr 2020-05876; EPN Lund Dnr 2021-01580; EPN Lund Dnr 2022-01686-02; EPN Lund Dnr 2023-00412-02) and EPN Norway EPN Norway reference number 1085.

There is no change or impact on head trauma standard treatment regimen in participating hospitals during the study period. We believe that there are no ethical issues connected to the study as no intervention takes place. Informed verbal consent is obtained from a caregiver, and when applicable from the child if able to understand the study information. Written information concerning study details is presented before inclusion. Clinical management in the ED will be fully independent of consent or decline to participate from the patient and/or caregiver. Data are managed in accordance with regulations and are presented only on group level where no individuals can be identified.

Patients and caregivers may choose to withdraw their consent for participation in the study at any time during the study period. Personal data from patients who withdraw consent will be deleted from the study database.

For the substudy on biomarkers, written informed consent is required from both caregivers before blood sampling can be performed. Blood samples are then sent to the lab and stored in freezers, with no results are returned to the managing clinician or the caregiver. Participation in the biomarker substudy does hence not affect management in ED. Anticipated risks in this substudy are (1) pain during venepuncture and (2) risk for postpuncture haematoma. The procedure venepuncture is, however, a standard procedure in paediatric EDs, and only skilled personnel will perform the procedure after application of topical anaesthesia. Separate ethical approval and biobank permission will be sought for this substudy.

Results will be submitted for peer-review and publication in applicable scientific journals and on scientific congresses concerning TBI.

Data statement

During the enrolment period is data stored in the database Entermedic, with password access limited to principal investigator, senior investigator and site coordinators (with restricted access to his/her own site), in accordance with applicable laws and regulations. Datasets for analysis will be pseudonymised and the code key stored separately. Pseudonymised datasets and all aspects of methodology (including copies of the questionnaires) will be available on reasonable request from the corresponding author for at least 10 years from end of inclusion. Data management is in accordance with the European Union general data protection regulation (GDPR).

Study status

The study is presently active and recruiting children. The original time line of the study has been extended due to poor recruitment during the COVID-19 epidemic.