Over the past two decades, Vanderbilt University Medical Center (VUMC) engineered a vast array of knowledge assets driving and supporting quality patient care, and reducing costs through clinical decision support (CDS).1 These knowledge assets are critical for effective CDS, and VUMC’s past transition from a home-grown electronic health record system (EHR) to a commercial EHR provided unique challenges and opportunities to extract, understand, catalog, evaluate, and transfer decades of CDS knowledge assets.

Much of VUMC’s legacy CDS knowledge assets (KA) were created to deliver knowledge-based CDS for a single dedicated system. Consequently, their respective system defines how their content is encoded, stored, and utilized. Such compartmentalization of KAs led to a set of problems, including

1. complicated searching of KAs across systems
2. problems retrieving KAs based on basic metadata in a consistent manner (e.g. owners, review dates, clinical venue, asset dependencies, etc.)
3. duplicative logic and challenges associated with keeping all updates synchronized
4. managing obscured dependencies across KAs with updates to components.

As part of the knowledge transfer process, we created an iterative approach to support EHR to EHR CDS knowledge migration and on-going enterprise-wide maintenance.

To support an EHR to EHR CDS content migration and ongoing KM efforts, we integrated a vendor solution for KM as part of a custom built tool chain (COMET). Relying on this infrastructure, as part of Phase 1, we cataloged approximately 20,000 KAs, including orderables, order sets, protocols, event-condition-action rules, and many supporting assets. We are currently in Phase 2 and are working on implementing Phase 3.