Background: Postmortem computed tomography (CT) and magnetic resonance imaging have been gradually introduced to forensic pathology centres over the past two decades, with varying results in comparison to autopsy. Objective: The purpose of this study was to determine the accuracy of postmortem CT in determining a cause of death in children who died of unnatural causes. Materials and methods: This was a prospective recruitment of 30 children (< 18 years) who underwent postmortem CT and a forensic autopsy. A cause of death was independently assigned by two experienced paediatric radiologists and compared to that of the forensic autopsy. Results: A correct cause of death was assigned by reviewers 1 and 2 in 70% (n = 21/30) and 67% (n = 20/30) of cases, respectively. For gunshot injuries and blunt force head injuries, there was 91% (n = 10/11) and 100% (n = 6/6) agreement between forensic autopsy and both reviewers, respectively. No cause of death could be assigned by reviewers 1 and 2 in 27% (n = 8) and 30% (n = 9) of cases, respectively. An incorrect cause of death was assigned by both reviewers in one case (3%). The Cohen Kappa level of agreement between the forensic autopsy and reviewers 1 and 2 was k = 0.624 (95% confidence interval [CI]: 0.45–0.80, P = 0) and k = 0.582 (95% CI 0.41–0.76, P = 0), respectively. There was near perfect agreement between reviewers 1 and 2 (k = 0.905) (95% CI 0.78–1.00, P = 0). Conclusion: Postmortem CT has good diagnostic accuracy for identifying a cause of death related to trauma, but it has poor accuracy for children dying from causes not associated with apparent physical injury.