Study objective: We evaluate the diagnostic performance of ultrasonography for detection of elbow fracture in pediatric patients with trauma. Methods: PubMed and EMBASE databases were searched for diagnostic accuracy studies that used ultrasonography for detection of elbow fracture in pediatric patients. Bivariate modeling and hierarchic summary receiver operating characteristic (ROC) modeling were conducted to evaluate diagnostic performance. The pooled proportions of the false-negative rate were assessed with a DerSimonian-Laird random-effects model. We performed meta-regression analyses for heterogeneity exploration. Results: Ten articles involving a total of 519 patients were included. The summary sensitivity, summary specificity, and area under the hierarchic summary ROC curve were 96% (95% confidence interval 88% to 99%), 89% (95% confidence interval 82% to 94%), and 0.97 (95% confidence interval 0.95 to 0.98), respectively. The pooled proportion of the false-negative rate of ultrasonography was 3.7%. Among the various potential covariates, ultrasonographic performer (pediatric emergency physician versus others) and presence of extra musculoskeletal ultrasonographic training (trained versus not reported) were associated with heterogeneity of the specificity. Conclusion: Elbow ultrasonography demonstrated high performance in the diagnosis of pediatric elbow fracture, particularly in studies of physicians with extra training in musculoskeletal ultrasonography. Ultrasonography may be performed by trained physicians as a first-line diagnostic tool to diagnose pediatric elbow fracture.