Identity Review Queue
Category: Healthcare Standards
Node id: identity-review-queue
Maturity: production
Risk tier: C
Standards: HIPAA, GDPR
Bridge the EMPI queue of undecided matches to the platform human-review inbox: sync opens a review for each pair the thresholds could not settle, apply writes the reviewer decision back to EMPI. The EMPI link remains the auditable record of the decision.
What it does
- Opens one review per undecided pair, carrying the evidence and the score
- Idempotent: a pair already under review is not raised twice
- Writes approve/reject back as the EMPI MATCH / NOT_MATCH decision
- Leaves unanswered reviews queued rather than guessing
Use cases
- Routing doubtful patient matches to a data integrity reviewer
- Routing doubtful clinical act duplicates to a clinical reviewer
- Closing the loop from a reviewer decision to the identity record
Runtime
- Performance tier:
variable
Input ports
| Name | Type | Required | Description |
|---|---|---|---|
links | json | Yes | Queued POSSIBLE links, as emitted by the EMPI node in list_possible mode. Include the records so the review carries the evidence. |
Output ports
| Name | Type | Required | Description |
|---|---|---|---|
result | json | No | What the run did: reviews opened, or decisions applied, and what was skipped and why. |
reviews | json | No | Reviews opened by sync, each linked to the EMPI pair it is about. |
decisions | json | No | Decisions applied by apply, in the form the EMPI link accepted. |
count | number | No | Reviews opened (sync) or decisions applied (apply). |
Settings
Basic
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| Operation | operation | enum | Yes | sync | sync, apply | sync opens a review for each undecided pair; apply writes resolved decisions back to EMPI. |
| Entity Label | entity_label | string | No | record | What the pair describes, used in the review title a person reads: patient, clinical act, and so on. | |
| Priority | priority | enum | No | medium | low, medium, high, urgent, critical | Priority of the reviews this run opens. |
| Required Role | required_role | string | No | Role a reviewer must hold to decide these pairs. Leave empty to let any reviewer in the organization take them. |
Advanced
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| Workflow ID | workflow_id | string | No | empi-review-queue | Recorded on each review so the inbox can group and filter by the workflow that raised it. | |
| Escalation Hours | escalation_hours | number | No | 24 | How long a pair may sit undecided before it escalates. A queue nobody works is how a doubtful match becomes a permanent one. | |
| EMPI Base URL | empi_base_url | string | No | Override the EMPI service base URL (default from EMPI_SERVICE_URL or http://empi-service:8090). |
Errors
| Code | Description |
|---|---|
REVIEW_STORE | The node was executed without a human-review store (no platform database) |
REVIEW_INPUT | The links input is missing or is not a list |
EMPI_REQUEST | The EMPI service returned a non-2xx status or was unreachable |
Runtime notes
- Requires the platform database: the node is registered without a review store and wired at server boot, and reports a clear error when it is missing.
- Requires a signing secret — EMPI_JWT_SECRET, or the platform JWT_SECRET_KEY / JWT_SECRET — and a pipeline org context.
- apply acts only on pairs EMPI still reports as undecided, so a decision already recorded is never written twice.
- apply sends the reviewer and review ids with every decision, and leaves a review with no human reviewer (none, or "system") queued with reason no_reviewer: EMPI refuses service-applied decisions that name nobody.
- A decision EMPI refuses with 409 is reported under pending and the run continues: identifier_conflict (the records carry different identifiers of the same system; a note is left once on the review) or link_not_possible (decided elsewhere). Any other EMPI error fails the run.
- sync records the pipeline execution id on each review it opens; outside an execution it generates one and marks it with request_metadata.execution_id_source=generated.
- Only approve and reject are propagated; modify, escalate and request_info leave the pair queued.
See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.
¿Algo no está claro? Envíe sus comentarios sobre la documentación por correo electrónico..