AI/ML Implementation & Enabling Infrastructure
DIHI has developed a diverse portfolio of Augmented Intelligence & machine learning models and is responsible for the underlying infrastructure to facilitate their development and implementation into Duke Health. DIHI applies devops principles to the entirety of its infrastructure design to achieve reproducible, monitored, automated, and tested deployments of innovations at-scale for Duke.
Sepsis Watch™: the implementation of a Duke-specific early warning system for sepsis
Sepsis Watch is a deep learning model that leverages real-time EHR data to improve detection and treatment of sepsis in the hospital. It was successfully integrated into routine clinical care in November 2018 and has reshaped how local machine learning projects are executed.
Reducing In-Hospital Mortality
A machine learning model to predict in-hospital mortality, run at the time of admission to the hospital, for all adult patients.
Duke Virtual Assistant (Chatbot)
A Virtual Assistant to help patients understand whether they should go to the ER immediately rather than sending the email they are about to send to their physician.
Detection and Treatment of Patient Deterioration
Develop and implement a machine learning risk score for patient deterioration.
Using Machine Learning in Emergency Department Patient Flow
Develop a clinical dashboard built on existing DIHI databases and predictive modeling to assist in providing more efficient dispositions for emergency department patients by predicting clinical trajectory and need for admission.
Machine Learning Models to Recognize Early Clinical Deterioration in DUHS Pediatric Inpatients
Develop machine learning methods that will identify children at risk for clinical deterioration.
Machine-learning algorithm to predict ICU readmissions in the cardiothoracic surgical population
Develop predictive models to identify rapidly and accurately CTS patients at high-risk (and high-cost) for postoperative clinical deterioration necessitating readmission from the SDU back to the ICU.
Duke Health Data Pipeline
A foundational, fully-automated data curation tool enabling data liquidity which accelerates quality improvement, learning health, research and innovation projects.