Appropriately interpreting clinical decision support and predictive models is vital when caring for patients in an acute care setting. We implemented a state machine to aid in interpretability of machine learning based risk scores that provide automatic real-time risk assessment for patients with sepsis. This study prospectively compared the discriminatory capacity of two sepsis predictive models for early sepsis identification in the emergency department after implementation within a state machine

Learning Objective 1: Determine whether discretizing machine learning models into sepsis risk strata within state machine can aid in prediction.

Learning Objective 2: Compare the discriminatory capacity of two state machines after implementing different predictive models when classifying risk of sepsis.


Samuel McDonald (Presenter)
UT Southwestern

Karen Wang, UT Southwestern
DuWayne Willett, UT Southwestern
Mujeeb Basit, UT Southwestern

Presentation Materials: