Internal vs External Referrals: Design Two Connected Workflows
Compare handoffs inside a group with those to independent offices, preserving different systems, evidence and responsibility.
Define the boundary separately from direction
Record direction and organizational boundary as two different attributes. Define the organizational boundary your network uses for this map, including any partner or affiliated practices whose status is ambiguous:
| Axis | Question it answers | Example |
|---|---|---|
| Direction | Did we send it or receive it? | Primary care sends a patient to cardiology |
| Boundary | Does the receiving party sit inside our organization? | Cardiology is a department in our group, or an independent practice across town |
Direction describes the perspective of the team recording the referral. Neither label determines a measurement clock: define receipt, transmission and subsequent event times separately. Boundary prompts questions about access and responsibility, but the actual arrangement determines what is visible. If you have already mapped direction, see inbound vs outbound referral workflows and outbound referral management for that layer. This article adds the boundary layer on top.
The AHRQ Workflow Assessment for Health IT Toolkit frames health IT implementation around its impact on both clinical and administrative workflow in ambulatory care. The administrative part is what we map here. Clinical decisions about where a patient should go, and any payer or regulatory requirements, are outside this article.
Lane A: map information and responsibility inside the group
For each internal destination, name the receiving system and the people permitted to see its records. Confirm whether the sending team can view an assignment, an appointment and a returned report. Shared ownership of the clinics does not establish any of these permissions or events.
The proposed Lane A map is an administrative design to adapt, not a universal workflow.
| Event | Evidence to establish | Responsibility to agree |
|---|---|---|
| Referral created | Identifiable request in the designated record | Sending team's creation and correction owner |
| Received and assigned | Receiving-system evidence and an accepted task assignment | Receiving queue owner and coverage |
| Booking recorded | Appointment identity, date and source | Receiving scheduling team; local follow-up owner |
| Visit outcome known | Dated outcome from an authorized source | Team permitted to verify and record the event |
| Document returned | Document receipt and routing record | Records owner and designated review route |
Keep administrative follow-up open or closed according to the organization's approved process, with the reason recorded. Document receipt does not establish clinical review. A visible queue item is not an accepted assignment unless that is how the agreed operating process works.
Test the map during a handoff and an absence. Can a covering colleague identify who accepted the task and where the latest evidence lives? If not, record the missing arrangement rather than attributing it to a particular team.
Lane B: map the independent office handoff and the evidence it returns
For an external destination, record the channels and responsibilities agreed with the receiving office. There may be a shared portal, an electronic exchange, a telephone confirmation or returned documents. Do not assume an independent office has no responsibility to provide updates, and do not assume it can see your worklist.
| Event | Possible evidence, when available | What still needs checking |
|---|---|---|
| Transmission | Destination and dated send result | Whether the receiving office registered the request |
| Receipt and assignment | Portal event or documented office acknowledgment | Which team owns the next agreed task |
| Booking | Appointment information with its source and time | Whether it is office-confirmed or patient-reported |
| Visit outcome | Authorized office record or documented report | Whether the source establishes attendance or only scheduling |
| Document return | Identifiable document received and routed | Whether any administrative follow-up remains |
Assign a follow-up owner within your own organization. That person tracks questions across the boundary; they do not take over the receiving office's work or decide clinical questions. Record the external contact and agreed action alongside the local owner.
Preserve evidence quality. An appointment date reported by a patient and a dated confirmation from the receiving office can both be recorded, with attribution, but should not silently become the same type of confirmation. When sources disagree, record the conflict and the person responsible for reconciliation.
Build one exception workbook for both lanes
A common exception view can help a team compare open questions across destinations. It need not be a separate spreadsheet or a single database shared by every organization. Use an authorized worklist with links to the appropriate records, and limit access and copied information to the approved purpose.
The proposed fields are: referral episode, destination and boundary, expected event, available evidence, unknown question, local follow-up owner, receiving contact, next action and review trigger. Set triggers through the approved operating process, not from the illustrative example below.
| Evidence gap | First administrative check | Recovery responsibility |
|---|---|---|
| Receipt not established | Check the expected receiving channel and any returned event | Local follow-up owner contacts the agreed receiving route |
| Assignment unclear | Confirm the accepting team or queue | Sending and receiving operations owners resolve the handoff |
| Appointment outcome unknown | Reconcile the appointment source and subsequent events | Authorized scheduling or records role |
| Returned document missing | Check receipt and request history | Named records follow-up owner |
| Destination information conflicts | Preserve both values and check the authorized destination source | Directory owner; referral-intent questions follow the clinical route |
Synthetic example: a team reviews 40 referral episodes, 25 internal and 15 external, at a defined checkpoint. Three internal episodes lack evidence of the expected assignment and four external episodes lack receipt evidence. All seven have a local follow-up owner, but the next actions differ. These are seven observed evidence gaps in an invented cohort, not seven proven lost referrals. Do not compare lane performance until the event definitions, observation time and evidence access are comparable.
Use two connected maps without assuming shared access
The connection point is an identifiable referral episode and a record of its handoffs. Different organizations may retain separate records. Link or reconcile the events your team is authorized to use, and define how local labels map to the shared milestone. Practical steps, offered as editorial guidance:
- Tag every referral with its boundary (internal or external) at creation, separately from direction.
- Attach the lane-specific evidence definition to each status so staff know what "acknowledged" requires in each case.
- Verify the preferred channel and contact for each destination. A missing response alone does not prove that a fax number is wrong.
- Agree how each owner reviews due exceptions and how cross-team questions are escalated. A group review does not replace an individual follow-up already due.
If you are weighing coordination software for these lanes, the referral coordination software page describes Linear Health's scope. Linear Health automates up to 80% of coordination work, with the team staying in the loop on the rest. Test that against each of your own lanes. You can book a demo and ask the vendor to run your Lane B example, including a missing confirmation.
A practical evaluation method for either lane
Linear Health publishes an Operational AI Evaluation Kit that scores evidence on a 0 to 3 scale from "not addressed" through "claimed," "demonstrated" and "verified" (Operational AI Evaluation Kit). Its method is directly useful for the two-lane question. Three of its checks map cleanly:
- Run a real workflow. The kit asks for one representative case with missing data and at least one exception. For this topic, request two traces: an internal referral where the receiving queue never acknowledges, and an external referral where no confirmation returns.
- Define completion in observable terms. Ask where closed status is recorded and what artifact proves completion in each lane. A booking, attended visit and returned document require different evidence. Test the milestone that the proposed workflow promises.
- Show the exception queue with ownership. The kit flags as a red flag any vendor that cannot show who owns an exception or how an incorrect action is corrected. Ask who owns a Lane B exception when the external office is unreachable.
For integration questions, the kit asks for a field-level read and write map rather than an integration logo. Linear Health lists athenahealth, eClinicalWorks, NextGen and Greenway Health as established integrations; specific write actions and event coverage still need demonstration in your configuration. The kit also states that a high total score does not offset an unresolved privacy, security or accountability issue, which should be sent to the responsible reviewer.
The free Operational AI Evaluation Kit can be used for any vendor, and if you want Linear Health scored against your two lanes you can book a demo and bring the workbook.
Frequently asked questions
Is an internal referral always outbound?
Do internal referrals always share one EHR?
Who owns follow-up after an external referral is sent?
How long should we wait before raising an exception?
Must both lanes use the same workbook?
Sources
Sources accessed November 5, 2026.
Linear Health publishes educational articles about operational AI for healthcare.