To examine the effectiveness of an education intervention for reducing physician diagnostic error in identifying pediatric burn and bruise injuries suspicious for abuse, and to determine case-specific variables associated with an increased risk of diagnostic error. This was a multicenter, prospective, cross-sectional study. A convenience sample of pediatricians and other front-line physicians who treat acutely injured children in the United States and Canada were eligible for participation. Using a web-based education and assessment platform, physicians deliberately practiced with a spectrum of 300 pediatric burn and bruise injury image-based cases. Participants were asked if there was a suspicion for abuse present or absent, were given corrective feedback after every case, and received summative diagnostic performance overall (accuracy), suspicion for abuse present (sensitivity), and absent (specificity). Of the 93/137 (67.9%) physicians who completed all 300 cases, there was a significant reduction in diagnostic error (initial 16.7%, final 1.6%; delta -15.1%; 95% CI 13.5, 16.7), sensitivity error (initial 11.9%, final 0.7%; delta -11.2%; 95% CI 9.8, 12.5), and specificity error (initial 23.3%, final 6.6%; delta -16.7%; 95% CI 14.8, 18.6). Based on 35,627 case interpretations, variables associated with diagnostic error included patient age, sex, skin color, mechanism of injury, and size and pattern of injury. The education intervention substantially reduced diagnostic error in differentiating the presence versus absence of a suspicion for abuse in children with burn and bruise injuries. Several case-based variables were associated with diagnostic error, and these data can be used to close specific skill gaps in this clinical domain.