Why referrals disappear before an appointment is scheduled

The referral population has to exist before it can be managed

I would choose one specialty and identify how referrals enter it. Internal orders, faxed requests, telephone requests, and electronic referrals may follow different intake paths. An analysis limited to records already in Epic cannot identify a fax that never became a referral record. I would compare the intake source with created records for a bounded period, preserving duplicates and rejected requests as explained outcomes rather than silently removing them.

The specialty should help define what counts as a complete request and which cases need clinical review before scheduling. A missing attachment, an unclear requested service, and a request sent to the wrong specialty need distinct resolution paths. Treating all three as a patient outreach task creates calls that cannot result in an appointment.

I would follow the handoffs between records and teams

In Epic, I would review the relationship among the order, referral record, appointment request, and eventual appointment. Those records serve related but different purposes. I would not assume that a referral status establishes whether a scheduling request exists or whether the patient completed the visit. The application and operational teams should help reconcile the identifiers and define the state transitions used locally.

Referral/authorization workqueues can support the work that prepares a referral for scheduling. Appointment request workqueues can support scheduling the resulting request. I would check whether the organization has made the handoff explicit, particularly where one team handles clinical triage and another handles outreach. A transfer should leave a usable request with an accountable next owner.

For a parent-child referral workqueue structure, I would inspect parent membership separately from the distribution into child queues. I would also distinguish a deferred referral from a completed one. A temporary disappearance from the Active view does not establish that the next team received a schedulable request.

The initial evidence request would be small and traceable

I would ask for the referral receipt and creation dates, source, destination specialty, status history, linked orders, appointment requests, scheduled encounters, outreach attempts, and closure reasons. I would also request representative intake documents and the relevant workqueue configuration. The sample should include old open records, recently closed records, and records that moved through without difficulty.

For rule-driven selection, I would request the relevant CER records with their contexts, criteria, property parameters, and references to the configuration that uses them. Then I would test specific referrals. A rule inventory explains intended selection, but it does not establish that an incoming document created a record or that a staff member made the promised handoff.

The resolution depends on the missing step

An intake failure may require a receipt reconciliation and a clear transcription responsibility. A clinical-review delay needs an escalation path owned by the specialty. An appointment request that lacks scheduling information needs a correction at the handoff. A patient who declined or received care elsewhere needs an accurate outcome, not endless repeat outreach.

I would document these paths with the teams and keep the patient-facing language consistent. Closure should mean the organization knows the outcome, not simply that an aging record was removed from view. I would preserve legitimate patient choice and avoid calling every external appointment a preventable revenue loss. The purpose is a reliable referral process, not an inflated conversion report.

The test should cross the specialty boundary

I would walk an internal referral and an external referral through intake, triage where required, scheduling, and final disposition. Additional scenarios would include a duplicate request, a referral returned for missing information, a patient who cannot be reached, a patient who declines, and a referral redirected to a different specialty.

For each scenario, I would confirm the receiving team sees the correct task, the record retains relevant history, and the outcome appears in reporting. I would also retest an older referral after the change. A workflow that handles only newly created records can leave the existing backlog outside the improved process.

The measures need to follow a mature referral cohort

For conversion, I would use referrals received during a defined period and allow enough follow-up for scheduling and visit completion. I would report clinical-review delays and patient-choice outcomes separately. A monthly count of appointments divided by that same month’s new referrals mixes different populations and can make performance look better or worse without a real change.

I would estimate financial opportunity only after validating clinical capacity, patient intent, reimbursement, and whether the activity was actually incremental. A cleaner referral pipeline is valuable operationally even when every recovered scheduling opportunity does not produce additional collectible revenue.