Fracture Detection on Bone Radiographs: The Impact of an AI Tool on Orthopaedic Night Shifts

We evaluated how using an artificial intelligence (AI)-based diagnostic tool impacts orthopaedists' accuracy in detecting fractures during night shifts without the support of on-site radiologists. We compared diagnostic discrepancies between orthopaedists and radiologists in recorded cases from our emergency department between September 2024 and June 2025. In February 2025, we introduced Gleamer BoneView^® to help orthopaedists with automated fracture detection during shifts without on-site radiologists. Statistical analyses measured the rates of fracture misdiagnosis before and after implementation of Gleamer BoneView^®. Chi-square and Fisher's Exact Tests were employed, and a p-value < 0.05 was considered statistically significant. A total of 28,655 patients were subjected to radiographs resulting in 31/13,813 recalls (0.22%) in the pre-implementation and 27/14,842 recalls in the post-implementation period (0.18%, p = 0.42). Among these, 51 recalls (30 males, age: 39 ± 23 years) were related to fractures: 26 (16 missed fractures, 8 clinical re-assessment, 2 additional CT) occurred in the pre-implementation period (0.19%), and 25 (13 missed fractures, 7 clinical reassessment, 3 additional radiographs, 2 additional CT) in the post-implementation period (0.17%, p = 0.63). Patient management was changed in 9/13,813 patients (0.0006%) in the pre-implementation period and 7/14,842 (0.0005%) after the implementation (p = 0.551). Gleamer BoneView^® use was linked to a non-significant decrease in missed fractures by orthopaedists working without on-site radiologists.