GARDE Implementation Overview

Previous GARDE implementation

GARDE supported the Broadening the Reach, Impact, and Delivery of Genetic Services (BRIDGE) randomized controlled trial, which compared the effectiveness of chatbot-based delivery of cancer genetic services compared to standard care (SOC) among 3073 unaffected primary care patients in two large healthcare systems. The study found that chatbots are equivalent to SOC in terms of uptake of pretest cancer genetic services and genetic testing. BRIDGE results suggest that chatbots offer a scalable and secure way to deliver genetic services, allowing genetic counselors to focus on more specialized care.

Initial Integration Requirements

The first level of integration required is between GARDE and data from the EHR, typically through the EHR’s data warehouse (e.g., Epic’s Clarity). When considering eligibility for hereditary cancer genetic testing, GARDE requires demographic (age, sex) and structured family history (relative, diagnosis, age of onset) data to determine who meets criteria. For healthcare systems that use Epic, GARDE provides Clarity queries to extract these data sets in the format GARDE requires (CSV files). For systems that rely on a different EHR, GARDE’s data input format is generic and can be generated using a similar method.

To analyze the population, GARDE’s web-based application is used to select, digest, and analyze the CSV files. Results are then written to another CSV file that includes detailed information about who met criteria for genetic testing and why. GARDE’s results are then available for use. GARDE runs within the institution’s firewall, and EHR data used as GARDE inputs and outputs never leave the institution’s IT premises.

IT Requirements

GARDE can be installed locally on premises, or its installation can be virtual/cloud-based (e.g., AWS, Microsoft Azure). GARDE is based on Java and SpringBot and can be deployed as executable jar files or via Docker on most operating systems (Linux, MacOS, Windows). The GARDE platform consists of three servers as follows:

GARDE Population Coordinator – consists of a web application, REST services, and front-end dashboard/configuration app
OpenCDS Hooks – an open-source clinical decision support server accessed via CDS Hooks standards.
Database (Postgres [default], MySQL, Oracle, MS SQL Server)

While the three servers can run on a single VM (or lightweight laptop), the most secure and performant implementation is to install each on different hardware or virtualized server with adequate memory and CPU, especially when analyzing large datasets with millions of patient records.

Steps to deploy

Task	Person responsible	Steps necessary to accomplish task
Establish IT System Architecture	Enterprise Architect	Determines where/how to deploy GARDE
Allocate and configure hardware/virtual machine (VM)	System Administrator /Cloud Engineer	Prepare servers for GARDE
Install GARDE	System Administrator /Cloud Engineer	Follow step-by-step GARDE install instructions.
Extract Patient Data	Data Architect with Epic Clarity and/or other EHR data.	Identify the population and extract patient demographics and family history.
Map local EHR Data to GARDE	Informaticist / analyst / research coordinator	Map EHR concept to ~90 GARDE concepts.
Test/Refine concept mappings, settings	Informaticist / analyst / research coordinator	Validate GARDE rules are finding the expected patients; adjust the memory and data paging based on performance.
Run GARDE	Informaticist / analyst / research coordinator	Identify patients who are eligible for genetic testing.

The next steps are based on how you will use GARDE’s outputs. If GARDE is used only to identify a population for additional analysis, then nothing else is required. If GARDE will be used to support population-based outreach, pre-test education, and access to healthcare services, additional GARDE tools and interfaces are available including 1) a chatbot for automated outreach and pre-test education 2) Epic registry interface artifacts and instructions for managing and keeping track of outreach activities with GARDE cohorts 3) REDCap dashboard (coming soon!) as an alternative to Epic for patient outreach within research contexts.

Contact information

For additional information about any of these options or collaboration opportunities, email the GARDE team.