Turns inbound EMS/CAD prehospital feeds into a pre-arrival record filed in Epic ASAP before the ambulance rolls in, so the ED activates the right team, room, and orders on wheels-up rather than at the door. Bridges an event-driven NEMSIS ingestion tier to Epic FHIR R4 and HL7v2 A05 pre-registration.
Tags: healthcare, epic, hl7v2, fhir, event-driven, interoperability, connected-devices, api-led
ED Pre-Arrival EMS Workflows
Objective: EMS crews collect a rich field picture (chief complaint, vitals trend, 12-lead ECG, GCS, mechanism, STEMI/stroke/trauma alerts), but it arrives as a verbal radio report the charge nurse scribbles down, forcing the ED to reconstruct it after the patient is on a bed. The goal is to file structured prehospital data into Epic ASAP as a pending pre-arrival encounter the moment the crew commits a destination, so the ED sees an incoming board row with ETA, activation type, and vitals. Done looks like a STEMI or stroke alert paging the on-call team minutes before arrival, the room staged, and the run reconciling with no duplicate patient. This accelerator packages the ED Pre-Arrival EMS Workflows Epic Data Sharing Playbook as a MuleSoft-anchored solution plan.
What's Getting Built
- NEMSIS Ingest (System API) — terminates the EMS ePCR/CAD and field-monitor feed and validates NEMSIS v3.5 against schema ·
POST /nemsis, MLLP LISTEN NEMSIS, ACK on enqueue
- prearrival-buffer (Anypoint MQ queue) — Anypoint MQ buffer that holds pre-arrivals so a Bridges outage delays but never drops the alert ·
PUBLISH prearrival, CONSUME prearrival
- Pre-Arrival Orchestrator (Process API) — normalizes the run, debounces autosaves, and files the pending encounter with activation and ETA ·
CONSUME queue, POST /prearrival, PATCH /prearrival/{id}
- Patient Match (System API) — runs FHIR Patient $match on crew identity and posts serial vitals, falling back to a Doe encounter ·
POST /Patient/$match, POST /Observation, PUT /Encounter/{id}
- Epic Bridges / ASAP Board (target) — files the incoming ASAP row via A05, updates ETA on A08, and reconciles on the arrival A04 ·
MLLP SEND ADT^A05, MLLP SEND ADT^A08, MLLP LISTEN ADT^A04
- Dead Letter Queue (Anypoint MQ queue) — captures schema-invalid or unmappable runs with the raw NEMSIS payload and correlation ID for replay ·
PUBLISH dlq, CONSUME dlq
Prerequisites
- Epic Bridges interface engine configured to accept inbound A05 pre-admit/pre-registration and A08 update messages over MLLP, plus an ADT A04 reconciliation path on physical arrival
- Epic on FHIR client registered for backend OAuth 2.0 client-credentials with scopes for Encounter, Condition, Observation, and Patient search/write
- NEMSIS v3.5 XML or ePCR vendor feed available via HTTPS webhook, MLLP, or message queue from the EMS/CAD dispatch system
- ED ASAP tracking board build that surfaces incoming/pre-arrival encounter status and EMS activation categories to charge RNs
- Anypoint Platform environment with CloudHub or RTF, Anypoint MQ for buffering, and VPN or Anypoint Connector for MLLP to the Bridges endpoint
- Provider directory or destination-facility crosswalk mapping EMS unit and receiving-facility codes to Epic department and location IDs
1. Epic interfaces and standards
This accelerator implements the ED Pre-Arrival EMS Workflows Epic Data Sharing Playbook. Register the connecting app on Epic on FHIR and confirm the customer has the relevant Epic modules licensed before building.
- Bridges HL7v2 ADT A05 pre-admit and A08 update; ADT A04 on arrival reconciliation
- Epic on FHIR R4: Encounter planned/arrived, Condition, Observation vitals and ECG, Patient $match
- Hyperspace/Hyperdrive ED ASAP tracking board and incoming patient activations
- Cogito for EMS-to-door interval reporting
- NEMSIS v3.5 ePCR standard for inbound prehospital payloads
Architecture
2. Business requirements
- File a pre-arrival encounter the instant a crew commits a destination, not when the ambulance backs into the bay; the whole value is lead time.
- Surface EMS activation type (STEMI, stroke, trauma, sepsis, cardiac arrest, ROSC) as a first-class field so the board triggers the right team page automatically.
- Carry an ETA and update it on every A08 so the ED can stage rooms; treat a stale ETA as worse than no ETA.
- Keep prehospital data provisional by flagging every field as EMS-sourced so clinicians never mistake a field vital for a triage vital.
- Reconcile the pre-arrival encounter to the real registration on arrival without spawning a duplicate patient or duplicate encounter.
3. Ingesting the prehospital feed
- Accept NEMSIS v3.5 XML from the ePCR/CAD system over HTTPS webhook or MLLP; validate against the NEMSIS schema first and reject malformed runs to the DLQ.
- Assign a correlation ID from the ePCR incident number so every subsequent A08 update and the final A04 reconciliation thread to one run.
- Buffer in Anypoint MQ so a Bridges outage delays but never drops a pre-arrival; the ED can lose lead time but must not lose the alert.
- Extract vitals as time-series (serial BPs, HR, SpO2, EtCO2, GCS) rather than a single snapshot; the trend is the clinical signal.
- Debounce rapid ePCR autosaves; collapse updates within a short window so you page the team once, not five times.
4. Mapping and normalization
- Map NEMSIS eSituation chief complaint and provider impression to FHIR Condition and to the HL7v2 A05 reason-for-visit; keep the EMS free-text alongside coded values.
- Normalize NEMSIS vitals to LOINC-coded FHIR Observations, tagging each with a performer of EMS and a status of preliminary.
- Cross-map EMS activation categories to the ED tracking-board activation codes the ASAP build expects; an unmapped activation must alert, never silently pass through.
- Translate receiving-facility and unit codes to Epic department and location IDs via a managed crosswalk, not hardcoded values.
- Use DataWeave to render both the A05 (PV1/DG1/OBX) and the FHIR bundle from one canonical model so the two paths cannot drift.
5. Orchestration and matching
- Run FHIR Patient $match on name/DOB/sex when the crew has identity; on low-confidence or no-match, file the pre-arrival as an unidentified Doe encounter rather than force a wrong link.
- File the encounter as planned/incoming via A05, then flip to arrived on the A04 the ADT feed emits when registration happens at the door.
- Post serial vitals as FHIR Observations against the pre-arrival encounter so the trend is visible on the board before the patient exists in triage.
- Emit the team-activation trigger only after the encounter and activation code are confirmed filed; a page with no board row to click is a dead end.
- On reconciliation, merge or supersede the Doe encounter into the registered one and carry the correlation ID forward for closed-loop reporting.
6. Error handling and operations
- Alert on the absence of an expected arrival A04 within an ETA-plus-buffer window; a pre-arrival that never reconciles is the failure mode dashboards miss most often.
- Route schema-invalid or unmappable runs to a dead-letter queue with the raw NEMSIS payload and correlation ID intact for replay.
- Track EMS-notify-to-board-visible latency as a live SLI; the blueprint is worthless if the row appears after the patient does.
- Fail closed on activation pages: if the board write fails, escalate to a fallback notification channel rather than dropping the STEMI/stroke alert.
- Log every A05/A08/A04 with correlation ID so a charge nurse dispute about what EMS sent is answerable from one query.
7. Security and compliance
- Encrypt PHI in transit and at rest.
- Audit-log every access to and modification of patient data, and carry a correlation ID end to end.
- Enforce minimum-necessary access for integration operators and support users.
- Restrict the inbound NEMSIS endpoint by mutual TLS and an allowlist of EMS/CAD source systems; a spoofed pre-arrival can trigger a false trauma activation and pull a team off real patients.
- Treat unidentified-patient pre-arrivals as PHI from creation and apply break-the-glass-equivalent access controls to the Doe encounter until reconciliation.
- Sign and verify each A08 update against the original run correlation ID so an attacker cannot inject a spoofed ETA or downgrade an activation.
- Segregate the field-monitor 12-lead/ECG feed on its own credential and scope so a compromised device credential cannot write encounters.
Sample payload — HL7v2 ADT A05 pre-arrival pre-admit from EMS
MSH|^~\&|MULE_EMS|MEDIC12|EPIC|GENERAL_ED|20260820T142455||ADT^A05|EMS-INC-88213|P|2.5.1
EVN|A05|20260820T142455|||^EMS^PREARRIVAL
PID|1||||DOE^JANE^^^^^U||19XX0101|F
PV1|1|E|ED^^^GENERAL^^^^^ED ASAP||||||||||||||||EMS-INC-88213|||||||||||||||||||||||||20260820T145000
DG1|1||R07.9^Chest pain unspecified^I10|STEMI ACTIVATION|20260820T142000
OBX|1|NM|8867-4^Heart rate^LN||118|/min|||||F
OBX|2|NM|8480-6^Systolic BP^LN||96|mm[Hg]|||||F
OBX|3|NM|2708-6^SpO2^LN||94|%|||||F
OBX|4|ST|EMS-ACTIVATION^EMS activation type^L||STEMI|||||F
OBX|5|ST|EMS-ETA^Estimated arrival^L||20260820T145000|||||F
Testing and validation
- Munit-test the NEMSIS-to-canonical DataWeave against real ePCR export samples including a run with only partial vitals and no patient identity.
- Verify a Patient $match miss produces a Doe pre-arrival encounter, not a failed message or a mislinked patient.
- Assert the A05 lands as an incoming row on the ASAP board in a test Epic environment with the correct activation code and ETA.
- Fire a sequence of A08 updates and confirm ETA and vitals refresh on the same encounter via the correlation ID with no duplicate rows.
- Simulate a Bridges MLLP outage and confirm messages hold in Anypoint MQ and replay in order once the endpoint recovers.
- Send a malformed and a schema-valid-but-unmappable-activation run and confirm both land in the DLQ with payload and correlation ID.
Phased rollout
Phase 1 — Connect
- Stand up the NEMSIS System API and register the Epic FHIR backend client and Bridges A05 interface in a non-production environment.
- Wire one EMS agency ePCR/CAD feed to the ingestion endpoint over mutual TLS and confirm end-to-end connectivity.
Phase 2 — Map and test
- Build and Munit-test the canonical model and the A05/FHIR DataWeave maps against sampled real runs.
- Validate pre-arrival rows, activation codes, ETA updates, and Doe-then-reconcile flows on the Epic test ASAP board with the ED charge team.
Phase 3 — Production
- Go live with one pilot EMS agency and one ED, monitoring notify-to-board latency and reconciliation completeness daily.
- Onboard additional agencies via the facility/unit crosswalk only after the arrival-A04 reconciliation rate holds steady.
Success metrics
- EMS-notify-to-board-visible latency, target well under the shortest transport time
- Pre-arrival encounters reconciled to a registration on arrival, percent with no duplicates
- Team activations paged before patient arrival vs at or after the door
- Unmatched or Doe pre-arrivals as a share of total runs
- DLQ rate and mean time to replay for rejected NEMSIS runs
Troubleshooting
- Pre-arrival appears on the board but no team page fires. — Likely cause: The EMS activation category did not map to a tracking-board activation code the ASAP build recognizes. Fix: Check the activation crosswalk for the inbound value, add the mapping, and confirm the blueprint alerts rather than passes unmapped activations through.
- Every A08 update creates a new board row instead of updating the existing one. — Likely cause: The correlation ID from the ePCR incident number is not being carried onto the update messages. Fix: Verify the incident-number-to-correlation-ID assignment persists across the run and that the A08 targets the same patient/visit identifiers as the A05.
- Pre-arrival encounters linger and never flip to arrived. — Likely cause: The arrival A04 from the door-registration ADT feed is not being matched back to the pre-arrival encounter. Fix: Confirm the reconciliation subscriber reads the ADT A04 stream and matches on correlation ID or EMS visit number, and alert on encounters past ETA-plus-buffer.
- A field vital shows as a triage vital in the chart. — Likely cause: The Observation performer and status were not tagged as EMS-sourced and preliminary during mapping. Fix: Set the FHIR Observation performer to the EMS agency and status to preliminary, and OBX to EMS-attributed, so clinicians can distinguish field from triage data.
Deliverable: A deployable Anypoint application: NEMSIS-to-Epic System API, an EMS Pre-Arrival Process API that files prehospital data as a pending encounter, and a Bridges-fed ED tracking board notification, packaged with DataWeave maps, Munit tests, and runbooks.