Mednition Reveals Highly Accurate AI-Enabled Sepsis Diagnostic

Mednition has developed an AI sepsis model that achieves the highest published Area Under the Curve (AUC) for diagnosing Emergency Department (ED) sepsis using the Sepsis-3 criteria. Developed on the KATE AI platform, the model could potentially revolutionize sepsis detection and treatment, according to the company.

The AI model achieves a 99% AUC, indicating exceptional accuracy in distinguishing between septic and non-septic patients. Notably, the model demonstrated a 95% sensitivity (TPR) and a 96% specificity (TNR) on a retrospective cohort of 540,884 patients with 14,676 positive sepsis cases across 16 hospital sites. These results, available on mednition.com/research,¹ confirm the KATE AI platform’s ability to effectively identify patients with sepsis using the latest academic definition of sepsis. This new research builds on the team’s prior successes for sepsis screening in Emergency Department triage. That sepsis model, used before any diagnostic labs are available, achieved a 94% AUC.²

Too often, model performance data is partially disclosed in publications that mask an AI model’s true performance. Moreover, published research has been plagued with low model AUC scores that require a tradeoff between sensitivity and specificity. By publishing the results to fully disclose the model performance in sensitivity, specificity, and model AUC, Mednition aims to redefine the minimum standards necessary for clinical healthcare leadership to transparently evaluate clinical artificial intelligence (AI).

“We are thrilled to announce the development of this additional sepsis model,” Mednition President/Co-Founder Christian Reilly said. “Early and accurate screening and diagnosis of sepsis is critical for improving patient outcomes, and our combined models for sepsis have the potential to transform the way we detect and manage this life-threatening condition. By leveraging the power of AI, we can provide clinicians with the information they need to make faster, more informed decisions, ultimately leading to better patient care.”

The KATE AI platform was recently awarded U.S. Food and Drug Administration Breakthrough Device Designation for the sepsis screening model at triage and awarded “Best In Show” in the 2025 HIMSS Global Health Conference & Exhibition Emerge Innovation Experience in the Hospital Systems Toughest Challenges category.

Mednition was founded in 2014 to help clinicians improve healthcare delivery, combining the power of EHR-integrated artificial intelligence and clinical expertise to address critical healthcare challenges. KATE AI, the company’s flagship solution, is designed specifically to empower emergency nurses, reduce clinical risk, and improve care quality. The company is funded by a select group of private investors and major healthcare financial institutions, including Concord Health Partners (AHA Innovation Development Fund LP), Wildcat Capital Management and Moneta Ventures. The company is based in Burlingame, Calif.

References
¹ Ivanov O, Reilly C. Detection of sepsis-3 in the emergency department using machine learning. mednition.com/research. Preprint posted online March 4, 2025.
² Ivanov O, Molander K, Dunne R, Liu S, Brecher D, Masek K, Lewis E, Wolf L, Travers D, Delaney D, Montgomery K, Reilly C. Detection of sepsis during emergency department triage using machine learning. arXiv. Preprint posted online April 15, 2022. doi:10.48550/arXiv.2204.07657