Epic and Cerner in Clinical Trials: Integration Patterns for Site Coordinators
Epic and Cerner together cover more than 60% of US hospital EHR market share. For CROs managing multi-site Phase II and III trials, that means the majority of participating sites are running one of two systems—and each has a distinct integration architecture, different API maturity levels, and different coordinator workflow patterns that affect how patient matching tools get used in practice.
Understanding the integration landscape for both platforms isn’t just a technical exercise. It shapes how eligibility rule sets get built, how coordinators receive matching output, and where the friction points appear during site activation.
Epic: FHIR and the MyChart Bedrock API
Epic’s FHIR R4 implementation is accessed through its MyChart Bedrock API (formerly known as FHIR Sandbox). The API exposes patient demographics, conditions, medications, lab results, procedures, and clinical notes as FHIR resources—the core data types needed for eligibility screening. Epic’s SMART on FHIR authorization framework handles access credentials at the site level, requiring each health system to issue an app registration and authorize the integration through their Epic instance.
The practical consideration for CROs: each participating site’s health system IT team controls access, and the approval process at academic medical centers can add 3–8 weeks to site activation. Building site activation timelines with Epic IT review baked in—not treated as a last-minute step—prevents the single most common delay in FHIR-based clinical trial integrations.
Cerner: Millennium FHIR
Cerner Millennium exposes FHIR R4 through its Ignite APIs. Coverage of FHIR resources is similar to Epic, though the specific profile implementations differ. Cerner sites running older Millennium versions may have gaps in FHIR resource completeness, particularly for MedicationStatement and CarePlan resources. CROs should request a data availability assessment from each Cerner site before building eligibility mapping assumptions around medication history completeness.
Cerner also handles clinical notes differently from Epic—note types and documentation structures vary more across Cerner implementations than across Epic deployments, which affects NLP-based screening that relies on specific note types (pathology reports, progress notes, discharge summaries). Site-specific note type inventories are worth collecting during protocol setup.
Coordinator Workflow Integration
Patient matching output is only useful if it reaches coordinators in their existing workflow. Both Epic and Cerner support embedded gadget models—Epic Hyperspace sidebar gadgets and Cerner PowerChart widgets—that allow external tools to surface information inside the EHR interface without requiring coordinators to open a separate application.
The alternative is a standalone dashboard that coordinators access outside the EHR. Adoption rates for standalone tools at busy clinical sites are lower: coordinators managing 4–8 active trials across multiple sponsors are less likely to check a separate application consistently than to interact with a tool that appears in their normal workflow context. For multi-site CROs, the embed model is worth the additional integration work.
What Coordinators Need from the Output
Regardless of whether output is embedded or standalone, the content of the patient matching result needs to serve the coordinator’s next action: deciding whether to approach a patient for a pre-screening conversation. The most useful outputs include the overall eligibility score, a criterion-by-criterion breakdown showing which criteria passed and which are flagged for manual review, and the specific data source behind each determination (e.g., “Exclusion criterion 4: prior nivolumab prescription found, RxNorm 1926866, dispense date 2024-11-03”).
Outputs that provide scores without evidence trails get overridden by coordinator intuition or ignored entirely. The evidence trail is what makes coordinators trust the tool enough to change their workflow.