Predictive Model for Operative Intervention after Blunt Abdominal Trauma in Children with Equivocal CT Findings: A Pilot Study

Gastrointestinal Perforation with Blunt Abdominal Trauma in Children

Gastrointestinal tract perforation is uncommon in children, accounting for <10% of cases of blunt abdominal trauma. Diagnosis of bowel perforation in children can be challenging due to poor diagnostic imaging accuracy. Intra-abdominal free air is found only in half of the children with bowel perforation. Ultrasound findings are nonspecific and suspicious for perforation in only two-thirds of cases. A computer tomography (CT) scan has a sensitivity and specificity of 50% and 95%, respectively. Surgical decisions should be made based on clinical examination despite normal CT results. Management of bowel perforation in children includes primary repair in 50-70% and resection with anastomosis in 20-40% of cases.

Validation of the predictive model for operative intervention after blunt abdominal trauma in children with equivocal computed tomography findings: a multi-institutional study

BACKGROUND

We recently developed a preliminary predictive model identifying clinical and radiologic factors associated with the need for surgery following blunt abdominal trauma (BAT) in children. Our aim in this study was to further validate the factors in this predictive model in a multi-institutional study.

METHODS

A retrospective chart review of pediatric patients from five pediatric trauma centers who experienced BAT between 2011 and 2020 was performed. Patients under 18 years of age who had BAT and computed tomography (CT) abdomen imaging were included. Children with evidence of pneumoperitoneum, and hemodynamic instability were excluded. Fisher's exact test was used for statistical analysis of the association between the following risk factors and need for laparotomy: abdominal wall bruising (AWB), abdominal pain/tenderness (APT), thoracolumbar fracture (TLF), presence of free fluid (FF), presence of solid organ injury (SOI). A predictive logistic regression model was then estimated employing these factors.

FINDINGS

Seven hundred thirty-four patients were identified in this multi-institutional dataset with BAT and abdominal CT imaging, and 726 were included. Of those, 59 underwent surgical intervention (8.8%). Univariate analysis of association between the studied factors and need for surgical management showed that the presence of TLF (p < 0.01), APT (p < 0.01), FF (p < 0.01), and SOI (p < 0.01) were significantly associated. A predictive model was created using the 5 factors resulting in an area under the curve (AUC) of 0.80. For the motor vehicle collisions (MVC) group, only FF, SOI, and TLF are significantly associated with the need for surgical intervention. The AUC for the MVC group was 0.87.

CONCLUSIONS

A clinical and radiologic prediction rule was validated using a large multi-institutional dataset of pediatric BAT patients, demonstrating a high degree of accuracy in identifying children who underwent surgery. FF, SOI, and TLF are the most important factors associated with the need for surgical intervention.

LEVEL OF EVIDENCE

Level III.