Nonaccidental Pediatric Trauma: Which Traditional Clues Predict Abuse?

Abstract

Objectives: The objective of this study is to determine which of the child abuse clues quoted in the literature predict nonaccidental trauma (NAT): history incongruent with injuries found on examination or imaging, old injuries present, history of trauma denied by caregivers, multiple fractures present, changing history, fractures of varied duration, metaphyseal fracture, ear bruise, neck bruise, different history (second historian), and metaphyseal fracture.

Methods: This is a 4-year retrospective study of all suspected NAT cases referred to our medical center. In addition to the index visit, medical records were searched for visits before the index visit with patient historical or physical findings that might have suggested NAT but were not investigated. The association of diagnostic clues and the outcome were assessed by chi-square and logistic regression analysis.

Results: For 48 months, there were 109 cases of suspected NAT referred for evaluation (age range, 1 week to 15 years). After formal investigation by an abuse specialist, 79.3% of the cases were considered "likely abuse". Those historical or physical findings with a univariate association with a final conclusion of likely abuse included history incongruent with injuries, old injuries present, trauma history denied, changed history, and retinal hemorrhage. In multivariate logistic regression, only a history incongruent with injures remained an independent predictor of likely abuse (odds ratio, 8.65; 95% confidence interval, 1.74-43.07; P = 0.0151). After review of prior records, in only 2 of 109 cases (1.8%) was there a prior visit where NAT could have been suspected; however, history was thought to be congruent with the injury at the original visits.

Conclusion: In this 4-year study of NAT, the clinical clue that best predicted likely abuse after expert investigation was a history that was incongruent with the injuries found on emergency department evaluation. The incidence of possible early recognition from a prior emergency department visit in this group was very low, <2% of cases.