While there have been increasing efforts to develop more accurate algorithms for clinical decision support (CDS), less attention has been placed on understanding the downstream clinical decisions associated with CDS when deployed in the real world setting. Studies evaluating CDS using electronic health record (EHR) data typically rely on measurements of a priori defined process or outcome metrics with limited insight into the unanticipated downstream actions taken by clinicians<span style="font-size: 10.8333px;">. </span>We demonstrate a data driven approach to detect downstream actions associated with a recently deployed severe sepsis CDS by mining ordering patterns in the EHR. We detected frequent sequences of EHR orders that reflect both anticipated and unanticipated clinical decisions made after exposure to the CDS.

Learning Objective: Describe a data mining approach to identify unanticipated downstream actions taken by clinicians after exposure to clinical decision support alerts.

Authors:

Ron Li (Presenter)
Stanford University School of Medicine

Imon Banerjee, Stanford University School of Medicine
Daniel Rubin, Stanford University School of Medicine
Jonathan Chen, Stanford University School of Medicine

Presentation Materials:

• Abstract/Manuscript - 1