Soon after joining DIHI in July 2018, I had the unique opportunity to co-lead the pilot implementation of Sepsis Watch™ across Duke University Health System. As I enter my tenth year in the healthcare provider space, this experience has set Duke apart as a truly innovative health system with extraordinary staff who rise to the challenge to deliver collaborative, impactful patient care.
Sepsis Watch™ was the accumulation of a three year effort to bring machine learning into the hospital setting to predict a patient’s risk of becoming septic at some point in the next 36 hours. With guidance from Dr. Cara O’Brien and other clinical leaders at Duke, the DIHI project team developed a sepsis definition based on criteria captured in the electronic medical record in real-time. The team then incorporated that definition into a platform to predict and facilitate management of sepsis in the emergency department, which was identified as having the highest volume of sepsis events in the hospital. My role focused on integrating the Sepsis Watch™ solution into an effective yet minimally burdensome workflow for the front line staff involved and then on implementing that solution across the emergent care delivery settings within Duke Health.
In the months leading up to the pilot’s go live this past fall, the project team partnered with the Rapid Response Team (RRT) nurses from the Cardiac Intensive Care Unit and others to finalize the team members and mechanics for the new Sepsis Watch™ workflow. Beginning on November 5th, 2018, RRT nurses reviewed new patients as they entered the Duke University Hospital (DUH) Emergency Department who either met sepsis criteria or were at high risk of meeting sepsis criteria, as evaluated every five minutes by our model and displayed on the Sepsis Watch™ application. The RRT nurse then contacted the ED physician in charge of that patient’s care, as identified through the patient’s bed location on the app. Upon a positive diagnosis of sepsis by the ED physician, the RRT nurse would interact with the application to virtually “move” the patient into a bundle treatment pool on the Sepsis Watch™ application, which then converted to a semi-automated 3-hour and 6-hour SEP-1 bundle compliance tracking mechanism. As a final step, the RRT nurse utilized the app to help close gaps on bundle compliance as the patient transitioned from the ED to the floor.
A Governance Committee was established to steer the project through its six month pilot period. Sitting in on this committee, RRT nurses and ED physicians provided first-hand feedback on the workflow and led iteration of design improvement updates to further refine the Sepsis Watch™ application into an even more efficient and effective platform, which was updated based on feedback in February 2019. The DIHI team continued to learn from our clinical teammates within Duke University Hospital through the end of the pilot phase on May 5th, 2019. We then applied those learnings to lead successful implementations of the Sepsis Watch™ application and tailored workflows at Duke Raleigh Hospital (DRAH) and Duke Regional Hospital (DRH), as part of a Care Redesign partnership with Duke Health’s Performance Services. We are now continuing to evolve these site-specific solutions at DUH, DRAH, and DRH to facilitate seamless surveillance of new patients at risk of sepsis in their respective emergency departments.
It was humbling to observe the natural leadership roles taken on at each level of the Sepsis Watch™ project – the sheer energy and flexibility of the RRT nurses and ED staff, the support and proactivity of Duke Health Technology Solutions, and the operational effectiveness of Performance Services. We are excited to observe findings from the Center for Medicare and Medicaid Services which indicate a 2X improvement in 3-hour bundle compliance for Duke University Hospital during the Sepsis Watch™ pilot period when compared to the prior two year average compliance rate. We are currently evaluating the pilot phase to confirm Duke’s ascension into a national leader for treatment of sepsis in the acute setting.