Incorporate a Range of Frequently Used Engineering Features from EHRs into the Sentinel Common Data Model in the Sentinel EHR and Claims Linked Data Partner Network

As a component of the Sentinel Common Data Model (SCDM) v8.2 upgrade, the Sentinel Operations Center (SOC) will develop a new Feature Engineering (FE) table to support gathering information from electronic health record (EHR) unstructured data where the source is free-text, but key information has been extracted and standardized to allow querying with the ARIA Tools. 

The main goal of this Innovation Center (IC) project is to leverage and complement the SOC’s SCDM upgrade by applying natural language processing (NLP) pipelines to existing free text source data and populating the FE table.  This project will include multiple sites from both the EHR-claims Development Network (Duke Clinical Research Institute, Kaiser Permanente Washington Health Research Institute, Mass General Brigham, and Vanderbilt University Medical Center), and from the EHR-claims Commercial Network (TriNetX and HealthVerity).  

The IC and FDA will identify a list of 3–5 concepts as candidates for existing NLP pipelines and inclusion in the Feature Engineering table and sites will populate the table with these concepts, using free text from the EHR-claims Commercial Network and Development Network sites’ data. This project will also develop and adopt a set of recommended pre-processing best practices designed to optimize the comparability of NLP-derived data across heterogeneous healthcare systems.