Employing the Power of AI and Interoperability to Reduce Sepsis Mortality

September, 2019

Reduce Sepsis Mortality

Dr. Penny Cooper, DHSc, is a data scientist at Augusta Health

Augusta Health is a community hospital nestled in Virginia’s beautiful Shenandoah Valley focused on improving the health and well-being of our community. The statewide sepsis mortality rate in Virginia is 12.7%, while at Augusta Health it is 4.76%. We have been able to achieve a significantly lower sepsis mortality rate by employing the power of artificial intelligence (AI) and communication technology. So far, we have saved nearly 300 lives that could have been lost to sepsis.

Mortality from sepsis increases by as much as 8% for every hour that treatment is delayed. Our nurses are highly trained and are skilled at detecting early symptoms of sepsis based on standard indicators. They are also very busy. Aware of how many patients our nurses care for and the many tasks nurses juggle at once, our leadership team wanted to provide our nurses with additional support. Among the resources provided was a program to identify symptoms of sepsis sooner.

We formed a Sepsis Team and Taskforce in 2016 to take advantage of AI and communication capabilities and give our nurses an extra set of eyes to automatically review patient data and alert them as soon as a patient shows signs of being septic. Our goal was to detect and treat sepsis as early as possible and save lives.

Automated Process for Sepsis Risk Detection and Alerts

We began by using the four traditional criteria to identify sepsis – a temperature greater than 38°C, heart rate greater than 90, respiratory rate greater than 20, and an abnormal white blood cell count. In addition, we monitored mean arterial pressure and shock index. Our Sepsis Team and Taskforce then began a retrospective study of the data to determine which variables had the highest correlation to Sepsis.

We used the results of that study as a benchmark for developing an automated process that analyzes and compiles real-time data from medical systems each hour and assigns each patient a score. If the score is above a specific threshold, a sepsis alert is sent automatically to team members caring for that patient. Team members receive the alert with contextual information on their mobile communication device, which might be a hands-free Vocera Badge or a smartphone running a Vocera app.