EHR-based Genomic Discovery and Implementation [Funded Extension]

Mayo Clinic Rochester, Rochester MN

Investigators

Linked publications, trials & patents

Abstract

In phase IV of eMERGE, a single, Network-wide, combined risk assessment and management protocol was implemented across the various groups represented in the 25,000 patients recruited at 10 sites in the United States (US). Polygenic risk scores (PRS) were returned for 10 conditions, family history was ascertained using the MeTree or backup survey and monogenic basis was assessed for Tier 1 conditions. In this Supplement application we propose to finalize analysis-ready datasets, ascertain near term outcomes within 1 year after return of results (RoR), assess response of providers and participants to the return of Genome Informed Risk Assessment (GIRA). Our specific aims for the extension year are: Aim 1. Create a curated dataset of eMERGE IV participants for analyses. Using data from Common Data Platform (CDP), surveys, and the EHR, finalize an analysis ready data set for assessment of outcomes and genomic discovery. Along with genomic data from Broad and Invitae, phenotype dataset will be placed in ANVIL and we will deploy Tanagra and 12b2 to analyze data. Aim 2. Ascertain near-term outcomes after return of GIRA through automated and manual abstraction. We will use novel phenotyping methods described in Aim 1. We will compare the accuracy of EHR algorithms with and manual abstraction. Outcomes will include process, intermediate and clinical outcomes and will be assessed Network-wide as well in a condition-specific manner. We will explore phenotype abstraction using Large Language Models. Aim 3. a) Clinical informatics and EHR integration Continue to develop an ecosystem to store GIRA in the EHR and develop computable representation of the PRS report to enable automated generation of clinical decision support (CDS). b) Investigate ELSI aspects including assessing responses to PRS return among patients and providers. Aim 4. Genomic discovery using the EHR. We will compare predictive value of the PRS, family history and monogenic conditions for disease, stratifying by self reported race/ethnicity. We will also compare the predictive value of SDOH to PRS and jointly model both to improve risk assessment. Genomic data from Broad and Invitae will be placed on AnVIL will enable numerous analyses including computing updated PRS and also PRS for conditions that were not selected for clinical deployment, e.g., abdominal aortic aneurysm and deep vein thrombosis.