The Duke Institute for Health Innovation (DIHI) was founded a decade ago as a "living lab" for developing new models that can transform and improve patient care. One pervasive problem that hospitals have long struggled to overcome has been the identification and treatment of a dangerous condition known as sepsis, in which a patient's immune system overreacts to an infection and begins attacking the body's own organs. Sepsis is extremely hard to detect and can be fatal if not treated quickly. DIHI researchers realized that if doctors and nurses could easily tell which patients were at risk, they could intervene sooner—and save countless lives.
The ITI Institute recently visited the DIHI team in North Carolina to learn firsthand how they used creative thinking, collaboration, and AI-powered technology to improve patient outcomes. Sepsis is the number one reason patients must be readmitted to the hospital and the leading cause of hospital deaths, accounting for fully one-third of mortalities at hospitals.
Speed is a critical component of treatment: Just an hour delay in treatment drives the mortality rate up by 6.7 percent. But hospital nurses and doctors, responsible for a full spectrum of care while triaging emergencies, do not have the bandwidth to monitor dozens of patients minute-by-minute for the unique combination of sepsis risk factors. DIHI recognized that this was not a problem of ability but of capacity, so the team developed a cutting-edge tool that could crunch the data for them.
Sepsis Watch uses generative AI to analyze patients' information and predict in close to real time when their symptoms are likely to lead to sepsis. The model was trained on Duke-specific, privacy-compliant electronic health records and translates patient risk into an easy-to-understand, color-coded chart. This allows nurses to quickly spot the most critical patients so they can get treatment.
The results already speak for themselves: After Duke hospitals began using the tool, patient deaths dropped by 27 percent.
While the technology behind Sepsis Watch is groundbreaking, the tool is only effective if nurses and physicians understand how to use it and how it can help them and their patients.
The Duke team knew that buy-in from medical workers would be essential to the success of Sepsis Watch—and to saving lives. Dr. Mark Sendak, a population health and data science lead at DIHI, told ITI that his team conducted tests and interviews which incorporated analysis by anthropologists and other social scientists to ensure Sepsis Watch fit within the hospital workflow.
Typically, physicians lead the decision-making about patient treatments. But Sepsis Watch empowers nurses to monitor patient risk and alert doctors when there's a potential case of sepsis. Deploying the tool required hands-on training and frank conversations between medical teams to ensure that everyone was aligned on their roles and responsibilities.
DIHI's mission is to scale ideas beyond Duke through strategic partnerships across academia, medicine, government, and industry. The ITI Institute is excited to be part of that effort and to highlight the team's life-saving work. Reaching more communities across the country will require investment, collaboration, and public-private partnership. Working together, we can all help make DIHI a true "living lab."
