Predictors for Nonaccidental Trauma in a Child With a Fracture-A National Inpatient Database Study.

Abstract

Despite heightened awareness and multidisciplinary efforts, a predictive model to help the clinician quantify the likelihood of nonaccidental trauma (NAT) in a child presenting with a fracture does not exist. The purpose of this study was to develop an evidence-based likelihood of NAT in a child presenting with a fracture. Using the 2012 Kids' Inpatient Database, we identified all available pediatric inpatients admitted with an extremity or spine fracture. Children with a fracture were subcategorized based on the diagnosis of NAT. Multivariate analysis using multiple logistic regression was used to generate odds ratios and create a predictive model for the probability of NAT in a child with a fracture. Of the 57,183 pediatric fracture cases, 881 (1.54%) had a concurrent diagnosis of NAT. Of these children, those presenting with multiple fractures had the highest rate of NAT (2.8%). The overall mortality rate in patients presenting with fractures and abuse was 1.8%, which was twice as high as patients without abuse (odds ratio [OR] = 2.0). Based on multivariate analysis, younger age (OR = 0.5), black race (OR = 1.7), intracranial injury (OR = 3.7), concomitant rib fracture (OR = 7.2), and burns (OR = 8.3) were positive predictors of NAT in a child with a fracture. A weighted equation using regression coefficients was generated and plotted on a receiver operative characteristic curve, demonstrating excellent correlation and probability of NAT (area under curve = 0.962). (Equation - ln (P/(1 - P)) = -1.79 - 0.65 (age in years) + 0.51 (black race) + 1.97 (rib fracture) + 1.31 (intracranial injury) + 2.12 (burn)). Using a large, national inpatient database, we identified an overall prevalence of 1.54% of NAT in children admitted to the hospital with a fracture. Based on five independent predictors of NAT, we generated an estimated probability chart that can be used in the clinical workup of a child with a fracture and possible NAT. This evidence-based algorithm needs to be validated in clinical practice. Prognostic study, Level III (case-control study).