Background Clinical decision rules can help to determine the need for CT imaging in children with head injuries. We aimed to validate three clinical decision rules (PECARN, CATCH, and CHALICE) in a large sample of children. Methods In this prospective observational study, we included children and adolescents (aged <18 years) with head injuries of any severity who presented to the emergency departments of ten Australian and New Zealand hospitals. We assessed the diagnostic accuracy of PECARN (stratified into children aged <2 years and ≥2 years), CATCH, and CHALICE in predicting each rule-specific outcome measure (clinically important traumatic brain injury [TBI], need for neurological intervention, and clinically significant intracranial injury, respectively). For each calculation we used rule-specific predictor variables in populations that satisfied inclusion and exclusion criteria for each rule (validation cohort). In a secondary analysis, we compiled a comparison cohort of patients with mild head injuries (Glasgow Coma Scale score 13–15) and calculated accuracy using rule-specific predictor variables for the standardised outcome of clinically important TBI. This study is registered with the Australian New Zealand Clinical Trials Registry, number ACTRN12614000463673. Findings Between April 11, 2011, and Nov 30, 2014, we analysed 20 137 children and adolescents attending with head injuries. CTs were obtained for 2106 (10%) patients, 4544 (23%) were admitted, 83 (<1%) underwent neurosurgery, and 15 (<1%) died. PECARN was applicable for 4011 (75%) of 5374 patients younger than 2 years and 11 152 (76%) of 14 763 patients aged 2 years and older. CATCH was applicable for 4957 (25%) patients and CHALICE for 20 029 (99%). The highest point validation sensitivities were shown for PECARN in children younger than 2 years (100·0%, 95% CI 90·7–100·0; 38 patients identified of 38 with outcome [38/38]) and PECARN in children 2 years and older (99·0%, 94·4–100·0; 97/98), followed by CATCH (high-risk predictors only; 95·2%; 76·2–99·9; 20/21; medium-risk and high-risk predictors 88·7%; 82·2–93·4; 125/141) and CHALICE (92·3%, 89·2–94·7; 370/401). In the comparison cohort of 18 913 patients with mild injuries, sensitivities for clinically important TBI were similar. Negative predictive values in both analyses were higher than 99% for all rules. Interpretation The sensitivities of three clinical decision rules for head injuries in children were high when used as designed. The findings are an important starting point for clinicians considering the introduction of one of the rules. Funding National Health and Medical Research Council, Emergency Medicine Foundation, Perpetual Philanthropic Services, WA Health Targeted Research Funds, Townsville Hospital Private Practice Fund, Auckland Medical Research Foundation, A + Trust. © 2017 Elsevier Ltd. All rights reserved.