Emergency Medicine

AI ECG alerts boosted treatment for detected high-risk hyperkalemia in the ED

This physician-level pragmatic RCT tested a real-time AI-enabled ECG alert for dyskalemia in 14,989 ED patients at two hospitals. The alert flagged patients at risk for potassium ≥6.0 or ≤3.0 mmol/L via EHR pop-ups to intervention physicians only. Overall 3-hour treatment rates for hyperkalemia and hypokalemia were similar between alert and control groups (HRs 1.05 and 0.91). Among patients the AI identified as hyperkalemic, alerts increased hyperkalemia-directed treatment from 41.6% to 69.1% (HR 2.23). AI ECG alerts therefore meaningfully intensified treatment in high-risk hyperkalemic subsets, without yet shifting population-level dyskalemia care. 1

Prehospital AI outperformed shock index for predicting trauma mortality

This study developed a real-time Prehospital-AI model using 21 prehospital variables from 204,189 Korean trauma patients. The ensemble model achieved an AUROC of 0.923 with sensitivity 0.780 and specificity 0.880, outperforming shock index (AUROC 0.712). External validation across four South Korean trauma centers yielded AUROCs of 0.925–0.956, and 0.895 in an Australian Level 1 center. These results show AI-based risk prediction can more accurately flag high-risk trauma patients for aggressive resuscitation and definitive-center transport than conventional metrics. Further multinational implementation studies are needed before embedding such models broadly into EMS and ED triage pathways. 2

2015 field trauma triage update reduced undertriage in Ontario’s mature trauma system

This population-based cohort evaluated 281,268 injured adults in Ontario from 2009–2020 around the 2015 field trauma triage guideline update. Only 19.2% presented directly to trauma centers, and over half of those had minor injuries, indicating significant overtriage. Population-level undertriage and overtriage were 63.5% and 12.3%, respectively, across the study period. Implementation of the updated criteria produced a 15.2% immediate reduction in undertriage (rate ratio 0.85; 95% CI 0.77–0.94). There was no instantaneous change in overtriage, suggesting improved delivery of severely injured patients to trauma centers without worsening resource overuse. 3

Normobaric hyperoxia improved 3‑month outcomes in acute ischemic stroke without clear safety signal

This meta-analysis pooled eight RCTs (804 patients), mainly from Chinese populations, testing normobaric hyperoxia within 24 hours of ischemic stroke onset. Normobaric hyperoxia for at least 2 hours increased 3‑month functional independence and reduced overall disability versus room air or low-flow oxygen. It also lowered 90‑day mortality and improved NIHSS scores at 72 hours in the primary analysis. Symptomatic intracranial hemorrhage and pneumonia were not increased, including in patients receiving endovascular therapy. These data suggest potential early neuroprotective benefit from normobaric hyperoxia, though generalizability beyond the studied populations remains uncertain. 4