Background: Spiral fractures of long bones have long been cited as indications of non-accidental trauma (NAT) in children; however, fracture types are only loosely defined in the literature, and intraobserver and interobserver variability in defining femoral fracture patterns is rarely mentioned. We sought to determine reliability in classifying femoral fractures in young children using a standard series of radiographs shown to physicians with varied backgrounds and training and to determine if a quantitative approach based on objective measurements made on plain radiographs could improve definition of these fractures.
Methods: On 50 radiographs, the fracture ratio—fracture length divided by bone diameter—was determined and radiographs were reviewed by 14 observers, including pediatric orthopaedic surgeons, emergency room physicians, and musculoskeletal radiologists, who classified the fractures as transverse, oblique, or spiral. A second review of the images in a different order was carried out at least 10 days after the first.
Results: Overall, intraobserver agreement was strong, whereas interobserver reliability was moderate. Experience level did not correlate with either result. Complete agreement among all observers occurred for only 5 fractures: 3 transverse and 2 spiral. An average fracture ratio near 1.0 appeared to be predictiveof a transverse fracture and a ratio of >3.0, a spiral fracture; ratios between these 2 values resulted in essentially random classification.
Conclusions: The ability to reproducibly classify femoral fractures in young children is highly variable among physicians of different specialties. These results support the belief that fracture morphology has little predictive value in NAT because of the wide variability in what observers classify as a spiral fracture of the femur. Caution should be used in the use of descriptive terms such as spiral, oblique, or transverse when classifying femoral fractures, as well as when evaluating children for possible NAT, because of the variability in classification.
Level of Evidence: Level III—diagnostic study.