EHR integration end to end: SMART on FHIR adverse event grading, blinding and MRN search

In week 9 of 2026 (February 23 to March 1), coordinators gained SMART on FHIR sign-in to their EHR for adverse event grading: they authorize the platform with their own clinical account, and it reads patient records under that clinician's access and no wider. Blind grading became stricter, with a rater's grades hidden from admins and adjudicators until the rater submits. Patients can now also be found by Medical Record Number.

SMART on FHIR sign-in for adverse event grading

SMART on FHIR is the authorization standard EHR vendors implement for third-party clinical applications: OAuth 2.0 sign-in plus a defined set of FHIR resources the application may read. HL7 maintains it. This release took the connection end to end for Epic, and applied the same clinician-level configuration to Oracle Health (formerly Cerner) and athenahealth.

The key change is the kind of access requested. The platform asks for user-level scopes, not system-level ones, so it sees what the signed-in clinician is authorized to see, and nothing more:

FHIR resource	Access
Patient	Read
Encounter	Read
Observation	Read
Condition	Read
MedicationRequest	Read
AllergyIntolerance	Read
Procedure	Read
DiagnosticReport	Read

These are the FHIR resources requested from Epic in this release, alongside the sign-in and identity scopes. The sign-in path was also hardened:

• Epic's authorize request carries the required audience parameter, and its token exchange uses the client authentication Epic requires
• Each Epic connection is tied to a stable identity taken from the ID token, with Epic's user-info endpoint as a fallback
• Redirect addresses are normalized so a trailing slash cannot break the callback
• Disconnecting an EHR works cleanly, and cookie contents are no longer written to logs

For the vendor list as it stands today, see which EHRs Burna AI integrates with; for the standard itself, what SMART on FHIR is.

Since then: the Oracle Health connection was completed on March 31, 2026; EHR endpoints moved to organization-level configuration with dynamic discovery in April 2026; and in August 2026 write-back arrived, filing a signed CTCAE grading summary to Oracle Health as a chart document, with filing into Epic enabled per site once Epic's Vendor Services approval is in place.

Blinding gated on submission

Multi-rater grading depends on rater independence, and a reviewer who can see a rater's work in progress can be swayed by it, even without meaning to be. The boundary is now hard: admins and adjudicators see a rater's CTCAE grades only after that rater formally submits.

What they can see before then is progress, not content. The case list shows an hourglass for work still in progress, a checkmark once a rater has submitted, and a count of graded events. On the case detail page, an adjudicator sees the raters as Rater A and Rater B rather than by name.

Underneath, grading cases and grading events were separated in the data model, and each event now records the user who created it. That gives per-rater progress tracking and a per-rater audit trail without exposing unsubmitted work.

Medical Record Number search

Coordinators can now find a patient by Medical Record Number as well as by first or last name, from one search field. Name and MRN are combined into a single search index that refreshes whenever a patient's name or MRN changes, and existing patients were backfilled so they are searchable straight away.

The MRN appears in the patient list, the add-patient form and the patient details drawer. A coordinator often has the MRN from the source EHR in hand first; searching on it removes a manual cross-reference step.

Organizations: one settings page, invitations and research-team roles

Organization management moved into dedicated flows:

• Organization settings in one tabbed page: Current Organization, and Switch or Create Organization
• An invitation route that checks the session, sends the invitee to sign in when needed, and makes the invited organization active on acceptance
• Leaving or deleting an organization returns the user to their personal workspace, which cannot itself be left or deleted
• Member roles named for research teams: owner (site PI), admin (principal investigator), sub-investigator, member (coordinator or nurse) and contributor (CRA or monitor)

Graders now see only grading cases in their active organization, and grading lookups are indexed by rater and organization. Whether one organization can reach another's data is among the first questions a security review asks; see Burna AI's security architecture for IT reviewers.

Under the hood: one link between a workflow run and its record

Grading, attribution and clinical decision runs used to be stored as separate fields on the visit. They now link to the record they belong to through one relationship, so every workflow type is queried and audited the same way, and a new workflow type needs no change to the visit itself.

Smaller changes

• A patient created during an encounter is assigned to that encounter's visit in the same step
• Visits filter by type and by review status, and Grade Note shows a "Grading..." state while it runs
• Adjudication activity now streams into the analytics pipeline
• Sign-in and sign-out land on the right pages

Since then (July 2026): these screens lived in the web provider app, which was retired in favor of app.burna.ai.

The AI platform for accelerated clinical research, from first-in-human trials through postmarket surveillance. Burna AI performs automated adverse event grading and attribution in real time, providing clear rationale and audit-ready evidence, including generation of IRB, sponsor and FDA reports. Burna AI delivers CTCAE v5 and v6 grades and MedDRA LLT codes on every encounter; enabling standardization, cross-site consistency and accuracy of clinical research data. AI suggests, clinicians decide. Human-in-the-loop, always.

To see the EHR connection on your own systems, book a 15-minute conversation.