LAUSR

PURPOSE To determine nonocular findings associated with significant retinal hemorrhage on dilated fundus examination in cases of suspected child abuse. METHODS This was a retrospective chart review from May 2014 to August 2021 at a level-1 trauma center. Two hundred seventy-four patients met the following inclusion criteria: (1) children 36 months and younger; (2) concern for child abuse; and (3) had an ophthalmology consultation. Through univariate and multivariate logistic regression, the study produced a screening algorithm for ophthalmic work-up in child abuse. RESULTS One or more abnormal neuroimaging findings had a statistically significant association with retinal hemorrhages and produced the strongest association with a univariate odds ratio of 170 (confidence interval: 10.245 to > 999.999). The multivariate model (P < .0001 with a c-statistic of 0.980) proposes using the following variables for predicting retinal hemorrhage on examination: abnormal neuroimaging, Glasgow Coma Scale score less than 15, altered mental status on examination, seizure activity, vomiting, bruising, scalp hematoma/swelling, and skull fractures. CONCLUSIONS This study elucidates clinical and imaging factors that correlate to retinal findings, validating previously studied variables and introducing new variables to be considered. The authors propose an evidence-based screening algorithm to increase the yield of positive dilated examinations and decrease the burden of potentially unnecessary child abuse ophthalmologic examinations. [J Pediatr Ophthalmol Strabismus. 20XX;XX(X):XX-XX.].