Abstract
BACKGROUND: Multiple clinical practice guidelines and head computed tomography decision rules exist for emergency department (ED) triage of children with traumatic brain injury (TBI). These vary in structure, aim, target cohort, and outcomes, yet are used clinically in similar populations. We compared important clinical and practical characteristics of all major guidelines in a large, real-world paediatric traumatic brain injury (TBI) cohort. METHODS: Prospective, pragmatic, observational study of children (<18 years of age) presenting with mild-moderate TBI to 16 EDs in Sweden and Norway, including prospective documentation of guideline-specific risk factors and outcomes between April 2018 and May 2024. We assessed the diagnostic accuracy and characteristics of CATCH, CATCH2, CHALICE, PECARN, SNC16, PREDICT, and NICE23, both within guideline-specific application cohorts and across the full study cohort. The primary comparative outcome was significant trauma-related findings on cranial computed tomography (cCT). Secondary outcomes were neurosurgical interventions and guideline-specific endpoints. The study was registered at ClinicalTrials.gov (NCT05964764). FINDINGS: The full cohort consisted of 3012 children (median age, 5.6 years; SD, 4.8). Among these, 0.9% (27/3012) had significant cCT findings, and 2/3012 (0.07%) required neurosurgery. CATCH and CATCH2 could be applied to 31.0% (934/3012) of patients, whereas the remaining guidelines were applicable to >94% of the cohort. In a comparative analysis concerning significant cCT findings, the lowest sensitivity estimates were 74.1% (95% CI: 53.7-88.9) for both PECARN ≥2 years and PREDICT ≥2 years; the highest was 100% (95% CI: 87.2-100.0) for SNC16. Specificity ranged from 41.6% (95% CI: 39.8-43.4) for SNC16 to 78.3% (95% CI: 76.8-79.8) for CHALICE. Mandatory cCT rates varied from 1.2% (PREDICT ≥2 years) to 29.9% (CATCH2). INTERPRETATION: A head-to-head comparison in a real-world, paediatric, TBI cohort highlights key features of established guidelines and decision rules, offering insight into their comparative diagnostic accuracy, practical application and clinical impact. FUNDING: This work was supported by non-commercial state funding from Södra Sjukvårdsregionen, Vetenskapliga Rådet (Hallands Hospital), and Forskning och Utveckling Halland.