Modern referral platforms react to events, not just nightly batches
Once referrals are modeled as stateful work, the natural next step is event-driven orchestration. A system can validate, assign, notify, or escalate work when a state changes instead of waiting for manual polling or batch scripts.
Event-driven referral orchestration
Loading diagram...
The AIW-I actor split matters beyond imaging AI itself. It shows the healthier modernization pattern: new automation services consume events, publish governed outputs, and leave ordering, archive, reporting, and queue authorities explicit instead of hiding them behind one black-box service.
Automation should accelerate, not conceal
The broken-path queue is part of the product. If automation hides exceptions instead of surfacing them, the organization loses trust in its own statuses.
Task - FHIR v4.0.1
Official workflow resource used to represent operational work generated from referral state changes.
Read the Task resourceSubscription - FHIR v4.0.1
FHIR event-notification resource used when systems need to react automatically as referral states change.
Read the Subscription resourceCross-Enterprise Document Workflow (XDW)
IHE workflow profile for sharing and tracking cross-organizational work states over time through a versioned workflow document.
Read the XDW profileException reconciliation deserves its own visible workflow
Modern platforms still need a deliberate path for the referrals that do not fit the happy path. Common examples include duplicate orders, incomplete requests, temporary or incorrect patient identities, and requests that must be redirected or withdrawn after review.
The right design move is not to force those cases through the same automation branch. It is to move them into a reconciliation state machine where staff can clarify, merge, cancel, or re-release the work with a trustworthy audit trail.
Exception reconciliation lifecycle
Loading diagram...
Encounter-Based Imaging Workflow supplement
IHE radiology supplement relevant to encounter-driven workflows, including edge cases where patient and workflow reconciliation matter.
Read the EBIW supplementContained HISO 10011.3 referral implementation guide
Contained legacy implementation guide for representing referral states and status transitions, including non-happy-path handling.
Read HISO 10011.3 (contained guide)Modernization stays trustworthy only when observability, replay, and rollout controls are explicit
Teams often talk about automation as if the hard part is only connecting systems. In production, the harder work is operational safety: making retries harmless, dead-lettering broken events, replaying history when consumers recover, and turning new automation branches on gradually enough that you can still roll back.
Observable automation path
Loading diagram...
Controls that make referral automation operable in production
| Control | Why it matters |
|---|---|
| Idempotent handlers | Prevent duplicate bookings, messages, or closures when the same event is retried. |
| Dead-letter and replay path | Keeps broken events visible and recoverable instead of silently dropping them. |
| Feature flags or staged rollout | Lets teams limit blast radius while proving a new branch behaves correctly. |
| Telemetry by queue and state | Shows whether automation is genuinely reducing delay or just moving failures elsewhere. |
AuditEvent - FHIR v4.0.1
Official FHIR audit resource for recording security-relevant workflow events across applications, services, and users.
Read the AuditEvent resourceProvenance - FHIR v4.0.1
Official FHIR resource for linking state changes back to the agents, entities, and activity that produced them.
Read the Provenance resourceKnowledge Check
Test your understanding with this quiz. You need to answer all questions correctly to mark this section as complete.