Trauma Surgery

Most epilepsy‑surgery prediction models lack external validation and show high risk of bias

Systematic review included 42 studies and 113 prediction models for epilepsy surgery outcomes. Median AUC was 0.75 and median accuracy 0.76, but only 20.4% underwent external validation. Overall risk of bias was high in 81% of models, with cognitive‑language outcome models performing relatively better. Conclusion: these tools are early and methodologically limited, so apply them cautiously at the bedside. 1

Protocol: individualized intraoperative SBP target versus standard reactive management in older hip fracture patients

Single‑center randomized trial will enroll 180 patients aged 65–85 undergoing hip fracture surgery under general anesthesia. Intervention targets systolic BP within ±10% of baseline versus reactive standard care with a universal MAP ≥65 mmHg. Primary outcome is composite organ dysfunction within 7 days; secondary outcomes include intraoperative hemodynamics and 30‑day mortality. This is a protocol; no outcome data yet, so the strategy remains investigational until trial results are available. 2

VA model predicts protracted recovery or loss of independence with high discrimination

Retrospective VA cohort model predicts long‑term recovery trajectories after major surgery with c‑statistics 0.906–0.908. Top predictors were the 90‑day Care Assessment Needs score and preoperative functional status. Model sensitivity and specificity were both ~82–83%, and machine learning added little incremental performance. Practical use: the calculator can stratify patients for targeted postoperative interventions and shared decision discussions. 3