During my scholarship year, I was fortunate to get involved in a number of different projects. I worked with Dr. Cara O’Brien from the Division of General Internal Medicine to develop and deploy a machine learning model for early detection of sepsis. I led an analysis to better understand the clinical and operational implications of using different sepsis phenotypes in our health system sepsis redesign work. Our findings identified a specific patient population that could stand to benefit from earlier sepsis intervention and our model was designed to target that phenotype. I also helped design and plan for implementation of this sepsis early warning system at Duke University Hospital. I worked with key stakeholders in Duke Health leadership, the Rapid Response Team, and the Emergency Department to integrate this technology into clinical workflow and identify the actions that needed to be taken when a patient is identified as high risk.

Alongside deployment of the sepsis early warning system, I also helped develop and evaluate a new reinforcement learning algorithm to make personalized treatment recommendations for patients with sepsis. Our findings suggest that prescriptive actions recommended by our model may have improved care and open up a discussion about the role that reinforcement learning may one day be able to play in healthcare. Our study highlights the need for further collaborations, both technical and clinical, to thoughtfully incorporate new prescriptive analytic models into clinical practice.

Lastly, I helped develop novel data infrastructure for health data that promises to improve data quality, access, and timeliness for health systems and investigators seeking to derive more meaningful insights from clinical care data. Our team is now working with Duke Health leadership to leverage this resource within our health system to power research, operations, learning health, and medical education.

My time at DIHI has shown me the potential for clinical informatics and data analytics to help improve our diagnostic and prognostic power. The experience has given me an appreciation for not only the challenging process of acquiring health data, but also the difficulty in thoughtfully implementing data-driven insights to improve patient care. I have loved working with such a dynamic, effective, and caring team, and aspire to one day build my own teams with people of such caliber and character. DIHI’s vision for healthcare, focus on rapid innovation, and dogged commitment to “doing what needs to be done” make it a truly unique organization within Duke and inspire students like me to continue playing our part to push open the bounds on healthcare.