Thirteen percent of individuals have an elevated risk for familial breast / colorectal cancer (Scheuner, 2010) and most are unaware of their risk. National Comprehensive Cancer Network (NCCN) guidelines recommend risk-based genetic testing for these individuals, but the guidelines are not expressed in a computationally shareable form. EHR vendor-provided tools offer proprietary solutions, and as a consequence, are not optimally scalable.
GARDE is a population clinical decision support (CDS) platform based on Fast Healthcare Interoperability Resources (FHIR) and CDS Hooks standards to support interoperability and logic sharing beyond single vendor solutions. GARDE is live at University of Utah Health (UHealth) and New York University Langone Health (NYU). The team is currently exploring deployment options at Intermountain Healthcare and other health systems.
Summary
GARDE screens and identifies patients who meet National Comprehensive Cancer Network (NCCN) criteria for genetic evaluation of familial cancer risk based on their family history in the EHR using both structured data and natural language processing of free-text data. Patients identified by GARDE are imported into an EHR's population health management (PHM) dashboard (e.g., Epic's Healthy Planet module) where genetic counseling staff review individual cases, select, and send bulk outreach messages to patients via chatbot and/or through the patient portal. Additionally, the GARDE-populated dashboard facilitates communicating care decisions with the patient's primary care provider to ensure care is coordinated.
GARDE was made possible through a research grant funded by the Informatics Technology for Cancer Research (ITCR) program of the US National Cancer Institute (NCI), and was devised by extending existing standards and leveraging existing EHR resources. CDS Hooks is a synchronous and individual patient-level CDS standard that was extended to support asynchronous population-level CDS via Web services. HL7 FHIR interfaces were devised to query and extract data on patient populations from existing enterprise data warehouse schemas. Additionally, generic EHR service patterns were devised to read/write patient data through existing proprietary EHR data services to populate the PHM dashboard.
The chatbot outreach interface was made possible through a grant funded by an NCI U01 grant to Broadening the Reach, Impact, and Delivery of Genetic Services (BRIDGE) referenced below.
Technical Details
Scalable Population CDS Platform Architecture
System Components
Population Coordinator - the central application workflow controller and interoperable data pipeline responsible for reading patient data from the EHR, performing population-based CDS operations to identify cohorts, and communicating cohorts/results to a PHM system.
OpenCDS – an open source CDS Hooks-compliant server responsible for computing patient eligibility for PHM cohorts.
EHR Patient Data – an electronic health record/patient data provider used for retrieving data required by the CDS algorithm; typically, an enterprise data warehouse (EDW) or web service API preferably based on FHIR standards.
EHR PHM Tools – tools for loading and managing patient populations, including tracking targeted patients and performing patient outreach. In the figure above, the Population Registry tools store patient populations. Manage Populations tools are used for navigating the registry to review patient statuses, and for performing outreach functions.
Chatbot Services – an automated and interactive patient chatbot that provides education through outreach services. It is not shown in Figure 1 but is a component of the outreach process within the Manage Populations bubble.
Grant Support
NCI U24CA204800 - Scalable Clinical Decision Support for Individualized Cancer Risk Management
NCI U01CA232826 - Broadening the Reach, Impact, and Delivery of Genetic Services (BRIDGE)