Background: We examined the visibility of fractures of hand and forearm in whole-body CT and its influence on delayed diagnosis. This study is based on a prior study on delayed diagnosis of fractures of hand and forearm in patients with suspected polytrauma. Methods: Two blinded radiologists examined CT-scans of patients with fractures of hand or forearm that were diagnosed later than 24 h after admission and control cases with unremarkable imaging of those areas. They were provided with clinical information that was documented in the admission report and were asked to examine forearm and hands. After unblinding, the visibility of fractures was determined. We examined if time of admission or slice thickness was a factor for late or missed diagnoses. Results: We included 72 known fractures in 36 cases. Of those 65 were visible. Sixteen visible fractures were diagnosed late during hospital stay. Eight more fractures were detected on revision by the radiologists. Both radiologists missed known fractures and found new fractures that were not reported by the other. Missed and late diagnoses of fractures occurred more often around 5 pm and 1 am. Slice thickness was not significantly different between fractures and cases with fractures found within 24 h and those found later. Conclusions: The number of late diagnosis or completely missed fractures of the hand and forearm may be reduced by a repeated survey of WBCT with focus on the extremities in patients with suspected polytrauma who are not conscious. Level of evidence: III
Background/Objectives: Multimodal large language models (MLLMs) offer potential for automated fracture detection, yet their diagnostic stability under repeated inference remains underexplored. This study evaluates the diagnostic accuracy, stability, and intra-model consistency of four MLLMs in detecting hand fractures on plain radiographs. Methods: In total, images of hand radiographs of 65 adult patients with confirmed hand fractures (30 phalangeal, 30 metacarpal, 5 scaphoid) were evaluated by four models: GPT-5 Pro, Gemini 2.5 Pro, Claude Sonnet 4.5, and Mistral Medium 3.1. Each image was independently analyzed five times per model using identical zero-shot prompts (1300 total inferences). Diagnostic accuracy, inter-run reliability (Fleiss’ κ), case-level agreement profiles, subgroup performance, and exploratory demographic inference (age, sex) were assessed. Results: GPT-5 Pro achieved the highest accuracy (64.3%) and consistency (κ = 0.71), followed by Gemini 2.5 Pro (56.9%, κ = 0.57). Mistral Medium 3.1 exhibited high agreement (κ = 0.88) despite low accuracy (38.5%), indicating systematic error (“confident hallucination”). Claude Sonnet 4.5 showed low accuracy (33.8%) and consistency (κ = 0.33), reflecting instability. While phalangeal fractures were reliably detected by top models, scaphoid fractures remained challenging. Demographic analysis revealed poor capabilities, with age estimation errors exceeding 12 years and sex prediction accuracy near random chance. Conclusions: Diagnostic accuracy and consistency are distinct performance dimensions; high intra-model agreement does not imply correctness. While GPT-5 Pro demonstrated the most favorable balance of accuracy and stability, other models exhibited critical failure modes ranging from systematic bias to random instability. At present, MLLMs should be regarded as experimental diagnostic reasoning systems rather than reliable standalone tools for clinical fracture detection.