Skip to main content
Dryad logo

Predictors of abnormal computed tomography findings for paediatric head injury: a retrospective cohort study


Yogo, Naoki et al. (2020), Predictors of abnormal computed tomography findings for paediatric head injury: a retrospective cohort study, Dryad, Dataset,


Objectives: Head injuries in children are common causes for visits to the emergency department (ED). Computed tomography (CT) scans are useful for confirming head injury diagnoses. However, radiation exposure from CT scans might cause lethal malignancies. We aimed to examine predictors for the indication of performing CT scans necessary for diagnosis.

Design: Retrospective cohort study.

Setting: Three EDs in Japan

Participants: Patients aged <16 years with head trauma who underwent CT.

Primary and Secondary Outcome Measures: The primary outcome measure was abnormal CT findings that were evaluated using the area under the receiver-operating characteristic curve (AUC). We derived predictors from three existing CDRs: Canadian Assessment of Tomography for Childhood Head Injury (CATCH), Children’s Head Injury Algorithm for the Prediction of Important Clinical Events (CHALICE), and Paediatric Emergency Care Applied Research Network (PECARN).

Results: Of 1,103 eligible patients, 410 were included in this study. There were 283 (68%) boys, and the median age was 2 years. In total, 35 (9%) patients showed an abnormality, 73 (18%) were admitted, and 3 (0.7%) underwent neurosurgery. We developed a CDR consisting of 6 predictors for identifying children with abnormal CT findings: (1) severe or worsening headache; (2) GCS <15; (3) signs of skull fracture; (4) hematoma; (5) loss of consciousness; and (6) altered mental status. Our CDR had a sensitivity of 74.3%, a specificity of 75.2%, a negative predictive value of 96.9%, and a positive predictive value of 21.8%. The AUC for our rule was not inferior to those for CATCH, CHALICE, and PECARN {0.75 (95% confidence interval [CI], 0.67-0.81) versus 0.64 (95% CI, 0.56-0.73; p<0.05), 0.68 (95% CI, 0.60–0.76; p=0.28), and 0.67 (95% CI, 0.60-0.74); p=0.10}.

Conclusions: Our findings suggest that a CDR, which lowers the frequency of CT in children with head injuries, must be developed and validated.


Study Setting and Population

We conducted a retrospective cohort study involving patients <16 years of age with head trauma admitted to three EDs in Japan. The participating hospitals were district general hospitals with 13,000 to 18,000 emergency patients per year in each hospital. No established CDR is required to be used at any hospitals when assessing patients with head trauma; the decision regarding whether to perform CT scans is made by each physician based on the patient’s clinical characteristics and history.

Inclusion criteria for this study were (1) age <16 years, (2) a history of blunt head injury within 24 hours before admission to the ED, and (3) patients who underwent CT scans for the first time in the ED. Patients transferred from another hospital after undergoing neuroimaging were excluded from this study. Accordingly, 410 patients admitted to the EDs from April 2014 to March 2018 were included (Figure 1).

The primary outcome measure was abnormal CT findings, such as skull fracture and any brain lesion attributed to acute head injury. Staff radiologists at each hospital interpreted the CT scans. If radiologists suggested an uncertainty regarding whether an abnormal finding existed, the scan was considered negative.

Data Collection

Data for demographic variables (age, sex), clinical characteristics (trauma mechanism, clinical history prior to the hospitalization, signs, management of the patient), and outcome information (CT findings) were collected from each patient’s medical record. Abnormal CT findings were defined as any skull fracture and acute brain lesion attributed to acute head injury. For predictive variables in this study, patients with altered mental status were defined as those with agitation, somnolence, or/and repetitive questioning during verbal communication.

The person in charge of each hospital registered the data with the data centre, and then three authors (N.Y., M.G., and M.S.) independently reviewed all data, performed the patient selection, and established the dataset for this study.


ZENKYOREN, Award: 2020-01

ZENKYOREN, Award: 2020-01