Mednition Introduces Machine Learning Powered Early Sepsis Detection Model that Doubles 1hr Sepsis Bundle Compliance

Sepsis Upgrade of KATE™ Triage Solution Achieves 50% Improvement in Early Sepsis Detection Accuracy Over Prior Best Practice Screening Protocol

BURLINGAME, Calif. (Sept. 1, 2020) – Mednition, makers of machine learning powered software solutions for healthcare, today announced the availability of an advanced model for early sepsis detection for use by emergency department (ED) clinicians at triage that shows 50% more accurate results than the prior best practice screening protocol.

Initial results detected sepsis more accurately than their prior protocol which was based on the Surviving Sepsis Guidelines of two SIRS vitals and a suspected source of infection. The new sepsis model is now integrated into KATE, the company’s real-time clinical decision support service. 

This enhanced sepsis detection capability, now in use at Adventist Health White Memorial (AHWM), is the result of an initiative Mednition launched 10 months ago with the clinical leadership at the hospital. AHWM is a 353-bed, nonprofit, faith-based teaching hospital that provides a full range of inpatient, outpatient, emergency and diagnostic services to communities in and near downtown Los Angeles.   

Sepsis is a leading cause of death in the US. Each year approximately 1.7 million people are diagnosed with it, resulting in more than 275,000 deaths. Someone dies from sepsis in the US about every two minutes. In 2016, the CDC attributed sepsis to be the highest inpatient cost, consuming more than $27 billion in hospital costs annually.

“Accurately addressing the challenge of early sepsis detection is one of the Holy Grails of medicine,” states Mara Bryant, Operations Executive at AHWM, a 2019 Malcolm Baldrige National Quality Award winner.  “Every hospital strives to improve outcomes and reduce mortality.  This delivers better care to our patients, supports our clinicians, and drastically reduces risk for our hospital.”

“KATE is catching patients with sepsis at the door, without using lab results, that would have been otherwise missed,” states Dr. Stephen Liu, ED Medical Director at AHWM. “Our nurses are more accurately identifying and initiating care for patients with Sepsis by using KATE than our prior screening protocol.  Mednition got it right, they’ve applied proven machine learning technology to one of medicine’s most vexing problems and they’re delivering it in real-time.”

As part of Mednition’s flagship machine learning powered solution, KATE, the new model is designed specifically to improve sepsis identification using only information gathered during the ED triage process (no labs). During development, the machine learning model had a Sensitivity of 74.7% (True Positive Rate) and Specificity of 95.9% (True Negative Rate) with an Area Under the Curve (AUC) of 0.96 on a retrospective data set of 228,929 patients, 3,683 of which were diagnosed with sepsis. The model was tested prospectively at AHWM, where early clinical observations on 5,283 patients showed a Sensitivity of 82.2% and Specificity of 93.6% with a 29% improvement in compliance with 1hr antibiotic administration over a 42 day period in July and August of 2020. This result improved sepsis detection over their prior screening protocol, which had a Sensitivity of 54.8% and Specificity of 96.0% over the same time period. 

“These results demonstrate our commitment and capability to advancing clinical accuracy with machine learning,” said Christian Reilly, co-founder of Mednition.  “The enhancement of KATE for early sepsis detection used the same clinical data science methodology as the KATE triage service.  Having a repeatable clinical experimentation approach and deep collaboration between clinicians and data scientists allows us to rapidly improve condition detection across the full spectrum of healthcare diagnoses.” 

About KATE and Mednition

The SaaS-based KATE is available immediately on a subscription basis, designed with emergency clinicians in mind, requiring no capital expense and is simple to use and easy to implement. It supports all popular EHR and clinical messaging software, requiring no change to clinical workflow, no downtime and minimal training.  It is the only solution architected to read, extract and understand the entirety of EHR, recognize potential under and over triage acuity assignment, and communicate with nurses in real time.  Underlying KATE is the Mednition Clinical Data Engine of more than 10 billion de-identified patient clinical data points, providing deep clinical insight for every patient.

For more information about the Mednition early sepsis detection solution and to get an online  demonstration of its capabilities, contact clinical.sales@mednition.com or  VISIT MEDNITION ONLINE AT EN20X, Sept. 9 – 11.

Mednition was founded in 2014 with a passion to help clinicians improve healthcare delivery and save lives. The company is funded by a select group of private investors and major healthcare financial institutions, including Concord Health Partners with participation by its limited partner the American Hospital Association Innovation Fund. The company also has partnered with the Emergency Nurses Association to help bring machine learning capabilities into ED nursing training. The company is based in Burlingame, CA. For more information, please visit Mednition.

 

Media Contact

Rich Moore

Big Sprocket LLC

415-608-7441

rich.moore@bigsprocketllc.com