Luma Health's referral workflow documentation describes three connected but distinct mechanisms: inbound referral reminders, EHR-synced referral records, and Fax Transform document processing. The practical research question is not whether the platform automates referrals, but where each mechanism owns the record, what it writes back, and which boundaries remain undocumented. The vendor's product pages state that referral status changes made in Luma are not written back into the EHR, while a separate marketing page describes a scheduled appointment closing the referral in the EHR. Those statements describe different integration contexts and should not be treated as one universal workflow. This analysis separates documented capabilities from promotional claims and identifies the unresolved questions a software evaluation should test before procurement.

Record Ownership Differs Across the Three Referral Mechanisms

The inbound referral product page describes Luma as the system that sends referred patients reminders to schedule, tracks whether patients scheduled, and records whether a patient never followed up [1]. In this configuration, Luma holds the referral record and drives outreach. The Athena integration page adds a different ownership model: if an organization integrates Luma with its EHR, Luma can regularly sync referral details from the EHR into the Luma account, and Luma recommends this as a best practice so staff do not have to manually create and modify referrals [2]. Here the EHR remains the source of referral creation, while Luma becomes the operational layer for reminders and status tracking.

Fax Transform introduces a third ownership pattern. The product page states that Fax Transform's agentic AI finds the right next steps from any fax or document, then coordinates to make them happen, and that next steps can include creating a referral in the EHR [4]. In this flow, the fax is the input, Luma's AI performs classification and routing, and the EHR receives the created referral. The eClinicalWorks integration page confirms this by listing configurable referral fax workflow settings such as default provider, default facility, and whether the document should have a reviewed or unreviewed status in the EHR [6]. Each mechanism therefore has a different authoritative system at different stages, which affects audit and reconciliation design.

Status Write-Back Is Explicitly Asymmetric

The most consequential documented boundary appears in the Athena integration specification. Under bulk referral management, the page states that referral status changes are not written back into your EHR [2]. This means an operations team can mark a referral incomplete or expired inside Luma without that state appearing in the clinical record. The same page separately describes automatic status updates when a patient calls from a reminder and automatic detection of existing patient profiles matched to synced referrals [2]. Those updates occur within Luma's referral record, but the documentation does not state that the call-derived status or the profile match is written back to the EHR.

A different Luma marketing page describes a four-step flow in which a scheduled appointment is reflected on the EHR schedule and the referral is automatically closed in the EHR [3]. That statement is not accompanied by the same integration-specific caveat found in the Athena specification. The apparent conflict is best understood as a difference between a scheduled appointment event, which originates in the EHR schedule, and a manual Luma status change, which does not. An evaluation team should ask Luma to specify which status transitions propagate to the EHR, which remain local, and whether the closed-referral behavior depends on the appointment being booked through a particular scheduling path.

Questions to Send to Luma Support Before Procurement

Request a written list of every referral status transition that propagates to the EHR, distinguishing manual changes, automatic incomplete or expired changes, click-to-call updates, and scheduled appointment closures. Ask whether the closed-referral behavior described on the marketing page applies to the organization's specific EHR integration and scheduling path [3]. For Fax Transform, request the default confidence thresholds by document type, the supported document type list, and the conditions under which a fax creates a referral versus an order or staff message [4]. Confirm whether SFTP-based syncing requires a Statement of Work and what additional configuration or cost that entails [1]. Finally, ask whether Luma's referral records follow the HL7 SDOH Clinical Care Task.status model or a proprietary status scheme [8].

Inputs, Outputs, and Configuration Boundaries

The referral management specification lists several input methods: manually adding a single inbound referral, bulk upload via CSV, an online referral form shared with external providers, and bulk upload of referring providers with Luma Support [1][2]. Outputs include referral reminder messages using pre-written templates in English and over 30 additional languages, automatic reminders on a default cadence during set daytime hours, and status changes such as incomplete or expired [1]. The documentation also notes that certain syncing methods, such as Secure File Transfer Protocol, may require a Statement of Work [1]. That is a deployment boundary, not a product limitation, but it affects implementation cost and timeline.

Fax Transform adds configuration controls that are absent from the inbound referral specification. The product page describes AI confidence thresholds set by document type, pool or team routing, duplicate detection, redaction mode, and outbound faxes for missing information [4]. The eClinicalWorks page lists workflow settings for document types, action type, default provider, default facility, and document status in the EHR [6]. These settings indicate that Fax Transform is a configurable document workflow engine rather than a fixed referral pipeline. The documentation does not specify which document types are supported out of the box, how confidence thresholds are calibrated, or what happens when a fax matches multiple patient records beyond the stated fuzzy matching behavior [4].

A Validation Scenario for Separating Documented Behavior from Assumption

A useful hypothetical validation scenario would test the boundary between Luma-owned status and EHR-owned status. An evaluation team could create a synced referral in a test EHR environment, allow Luma to send the first reminder, then manually mark the referral incomplete in Luma. The team would then inspect the EHR referral record to confirm whether the incomplete status appears. Next, the team could book the patient through Luma's scheduling path and check whether the EHR referral closes automatically, as the marketing page describes [3]. This two-part test would reveal whether the write-back asymmetry documented for bulk status changes also applies to individual status changes and scheduled appointments.

The same scenario should include a Fax Transform input. The team could submit a test referral fax, configure a low confidence threshold for that document type, and observe whether the fax enters a staff review queue or proceeds automatically [4]. The eClinicalWorks specification indicates that the reviewed or unreviewed EHR status is itself configurable [6], so the test should record which setting was active. This scenario is a suggestion for evaluation design, not a claim that Luma has been independently tested. It isolates the exact points where the documentation stops describing behavior and the buyer must obtain a written answer from Luma support.

Unresolved Questions and Evaluation Risks

Several questions remain unanswered by the supplied documentation. The referral pages do not state whether the automatic incomplete and expired statuses are written back to the EHR, even though manual bulk status changes are explicitly not written back [1][2]. The pages also do not specify whether the click-to-call status update and the existing patient profile match are visible to EHR users or only to Luma users [1][2]. For Fax Transform, the documentation does not define the default confidence thresholds, the supported document type taxonomy, or the exact conditions under which a fax becomes a referral rather than an order, eligibility check, or staff message [4].

The broader integration page describes Luma connecting referral workflows with existing systems and using shared data to keep access, intake, communication, referrals, and follow-up working together [5]. That is a platform-level claim, not a per-field data mapping. An evaluation should request the specific integration specification for the organization's EHR, because the Athena and eClinicalWorks pages show that referral fax workflow settings differ by system [2][6]. The HL7 SDOH Clinical Care implementation guide provides a general Task.status model in which the referral source sets status to requested and the target updates it through the workflow [8]. Luma's documentation does not state whether its referral records follow that model, so conformance should not be assumed.