Purpose: The objective of this study is to establish a clinical decision-making rule for mild head injury in young children; these children appear to respond differently to mild head injuries and face different developmental issues than do older children and management guidelines remain unclear.

Methods: Subjects were 97 children less than three years old with mild head injury (ICD9 codes 800-804 and 850-854) and an initial Glasgow Coma Scale score of 14 or 15, consecutively evaluated with a CT scan at a Level 1 Pediatric Emergency Department. Demographic, injury, symptom and treatment characteristics were abstracted from the medical records. Children with penetrating injuries, depressed skull fractures requiring surgery, injuries suspected to be intentional or who had their initial CT scan >24 hours after the injury occurred were excluded from the study.

A classification and regression tree (CART) program was used to identify characteristics that were correlated with intracranial injury (ICI) among children in the study. The information garnered from the tree was used to construct a clinical decision-making rule for the evaluation of very young children with mild head injuries.

Results: Boys accounted for 54% of the injuries in this sample. 46% of the children were less than 12 months, 24% were 12-23 months and 30% were 24-35 months old at the time of the injury. Three-quarters of all injuries occurred from falls.

Almost 25% of the children had evidence of an ICI on the CT scan; more than two-thirds of the ICIs occurred in children less than 12 months old (p=0.03). More than half of the children had a skull fracture; two-thirds of these occurred in children younger than 12 months (p<0.001).

We examined multiple CART models to assess the impact of different misclassification penalties, missing data and various splitting rules. The main parent node on the base CART was age (<24 months); splits also occurred with the presence of scalp lacerations, scalp hematomas, vomiting and the child being unconsolable. No demographic characteristic was a significant predictor. We are continuing to investigate the clinical utility of this tree.

Conclusion: While similarity exists between decision-making rules for older children and that found for this cohort, very young children have unique characteristics that merit further study and may require a separate decision-making process.