Level II, prognostic study. See the Guidelines for Authors for a complete description of levels of evidence.

Purpose: Hip fractures are a common cause of morbidity and mortality. Automatic identification and classification of hip fractures using deep learning may improve outcomes by reducing diagnostic errors and decreasing time to operation. Methods: Hip and pelvic radiographs from 1118 studies were reviewed and 3034 hips were labeled via bounding boxes and classified as normal, displaced femoral neck fracture, nondisplaced femoral neck fracture, intertrochanteric fracture, previous ORIF, or previous arthroplasty. A deep learning-based object detection model was trained to automate the placement of the bounding boxes. A Densely Connected Convolutional Neural Network (DenseNet) was trained on a subset of the bounding box images, and its performance evaluated on a held out test set and by comparison on a 100-image subset to two groups of human observers: fellowshiptrained radiologists and orthopaedists, and senior residents in emergency medicine, radiology, and orthopaedics. Results: The binary accuracy for fracture of our model was 93.8% (95% CI, 91.3-95.8%), with sensitivity of 92.7% (95% CI, 88.7-95.6%), and specificity 95.0% (95% CI, 91.5-97.3%). Multiclass classification accuracy was 90.4% (95% CI, 87.4-92.9%). When compared to human observers, our model achieved at least expert-level classification under all conditions. Additionally, when the model was used as an aid, human performance improved, with aided resident performance approximating unaided fellowship-trained expert performance. Conclusions: Our deep learning model identified and classified hip fractures with at least expert-level accuracy, and when used as an aid improved human performance, with aided resident performance approximating that of unaided fellowship-trained attendings.

Background: Observational studies have suggested that accelerated surgery is associated with improved outcomes in patients with a hip fracture. The HIP ATTACK trial assessed whether accelerated surgery could reduce mortality and major complications. Methods:We randomised 2970 patients from 69 hospitals in 17 countries. Patients with a hip fracture that required surgery and were ≥45 years of age were eligible. Patients were randomly assigned to accelerated surgery (goal of surgery within 6 hours of diagnosis; 1487 patients) or standard care (1483 patients). The co-primary outcomes were 1.) mortality, and 2.) a composite of major complications (i.e., mortality and non-fatal myocardial infarction, stroke, venous thromboembolism, sepsis, pneumonia, life-threatening bleeding, and major bleeding) at 90 days after randomisation. Outcome adjudicators were masked to treatment allocation, and patients were analysed according to the intention-to-treat principle; ClinicalTrials.gov, NCT02027896. Findings:The median time from hip fracture diagnosis to surgery was 6 hours (interquartile range [IQR] 4-9) in the accelerated-surgery group and 24 hours (IQR 10-42) in the standard-care group, p<0.0001. Death occurred in 140 patients (9%) assigned to accelerated surgery and 154 patients (10%) assigned to standard care; hazard ratio (HR) 0.91, 95% CI 0.72-1.14; absolute risk reduction (ARR) 1%, 95% CI -1-3%; p=0.40. The primary composite outcome occurred in 321 patients (22%) randomised to accelerated surgery and 331 patients (22%) randomised to standard care; HR 0.97, 95% CI 0.83-1.13; ARR 1%, 95% CI -2-3%; p=0.71.Interpretation: Among patients with a hip fracture, accelerated surgery did not significantly lower the risk of mortality or a composite of major complications compared to standard care.

Conservative treatment of humeral shaft fractures has a higher rate of nonunion, while operative treatment is associated with a low incidence of iatrogenic nerve palsy but higher rates of infection.