Referral leakage is not only an out-of-network contracting story. In US specialty and multispecialty groups, a large share of leakage is simpler and more embarrassing: the referral was ordered, the person never booked, and nobody owned the gap. Classic primary-care referral research — including the ASPN Referral Study indexed on PubMed — showed specialty referral completion is far from automatic. This create-need piece makes that invisible loss concrete for operators and shows what closing the loop looks like without pretending a portal link is a care pathway.

Key takeaways

The ASPN Referral Study (PubMed: specialty referral completion among primary care patients) treated completion as an empirical problem — ordering a referral is not the same as a completed specialty visit.
AHRQ’s CAHPS improvement guide includes rapid referral program strategies precisely because access between primary care and specialty is a known failure point.
Operational leakage happens in the booking window: unread portal messages, unanswered specialty phones, missing prior auth, and no outbound push when the person goes quiet.
Front-desk overflow and portal-vs-phone design decide whether a referral becomes a first appointment — or a chart note that aged out.

The need you are not measuring

Most specialty dashboards celebrate visits completed and relative value units. Few celebrate referrals received → appointments booked ≤ N days → visits attended. The missing middle is where leakage lives.

That gap is expensive in three currencies:

Clinical — delayed specialty care after a primary-care clinician already decided it was needed.
Operational — slots filled by less-appropriate demand while high-priority referrals sit unbooked.
Network — referring clinicians stop trusting your group when their patients report “I could never get through.”

You may already feel the pain as “marketing needs more leads.” Often you do not need more leads. You need to stop losing the ones clinical partners already sent.

What the evidence established long ago

The ASPN Referral Study on specialty referral completion is older literature — and that is the point. Referral completion was already measurable and incomplete when paper and fax still dominated. Digitized ordering did not magically create booked visits; it created faster generation of unfinished work.

AHRQ’s improvement guide strategy on rapid referral programs exists because access between primary care and specialty shows up in patient experience and quality work. Rapid pathways are not a software brand; they are an operating commitment: time-to-booking, clear ownership, and feedback to the referrer.

US Medicare preventive pathways illustrate the same class of problem in another costume: a covered benefit (for example the yearly Wellness visit described by Medicare.gov and CMS’s Wellness Visits page) still fails when outreach and booking fail. Referrals are the specialty analogue — authorized intent without a scheduled encounter.

Where referrals die in 2026 workflows

Map the handoff:

Order placed in the EHR / referral management tool.
Patient instructed to “call the specialist” or “check the portal.”
Specialty phone rings into overflow — see front-desk overflow.
Portal task sits unread — the portal-vs-phone tension in portal vs phone scheduling.
Prior auth / records stall without a chase owner.
Silence — no outbound at day 3 / day 7 / day 14.
Leak — person books elsewhere, or never books.

Each step can be instrumented. Most groups instrument only step 1 and step 7 (if the visit happens).

Leakage stack

From referral order to attended specialty visit

Order

Referral created

Clinical intent captured — still zero appointments.

Reach

Patient contacted / can reach you

Phone answered or portal completed with a real booking.

Clear

Auth & records cleared

Administrative blockers owned, not hoped.

Book

Slot on calendar

Time-to-book measured in days, not anecdotes.

Attend

Visit occurs

Reminders and DNA recovery finish the loop.

Make the gap visible on a weekly board

Referral loop KPI

The four numbers specialty ops should own

T2B

Median days order → booked

Referral workqueue

%Booked

Referrals booked ≤ 14 days

Workqueue

%Silent

No successful contact by day 7

Outreach log

%Attend

Booked referrals attended

Scheduling / EHR

If you cannot produce these four without a hero analyst, you do not have a referral operation — you have a fax museum with a nicer UI.

Close the loop without hiring a call center

Create-need becomes product when the fix is concrete:

Single workqueue for inbound referrals with aging.
Outbound cadence — day 0 portal/SMS identity-safe, day 2 voice, day 7 escalate to coordinator.
Answer capacity — referrals that generate inbound must not hit the same abandoned-call abyss as renewals; overflow design matters (front-desk overflow).
Dual path booking — portal for those who use it; phone for those who will not (portal vs phone).
Referrer feedback — closed-loop status (“booked / unreachable / declined / auth blocked”) so leakage is shared truth, not gossip.

AI voice/messaging agents fit as the always-on layer that makes day-0 and after-hours contact real — not as a replacement for prior-auth nurses or clinicians.

1
Define “complete”
Booked appointment within N days, or documented decline / clinically redirected.
2
Age the queue daily
Anything >72 hours without contact attempt is red.
3
Script identity-safe outreach
Specialty name + reason class (“your doctor asked us to schedule”) without over-sharing diagnoses in SMS.
4
Protect booking capacity
Hold new-patient templates for referral traffic; do not let them be eaten by reschedules alone.
5
Report to referring clinicians monthly
Completion rate and median T2B — trust is a metric.

What good looks like in 90 days

You should be able to show:

median time-to-book trending down;
silent referrals (no contact) under a hard cap;
referring practices reporting fewer “we never heard from you” complaints;
DNA rate on new referral visits falling because confirmation is part of the same lane.

Leakage audit (this month)

Sample 50 referrals from last quarterHow many never booked? How many booked elsewhere?
Call your own referral line as a mystery patientHold time and abandonment tell the truth.
Inspect portal-only instructionsIf that is the only path, you chose exclusion.
Name a queue ownerNot “the front desk” — a person.

Why this is a “create need” story

Many specialty leaders believe leakage is a payer-network problem they cannot touch. Some of it is. The portion sitting between order and first booked visit is touchable this quarter with outbound discipline and answer capacity. Until that portion is measured, budgets will keep flowing to top-of-funnel marketing while clinically generated demand dies quietly.

For US market context, see Arbol in the United States. Pair this piece with front-desk overflow and portal vs phone scheduling when you redesign the referral intake path.

A note on numbers and honesty

Older referral-completion studies should not be quoted as if they were your 2026 health-system rate. Use them as proof that completion is a known failure mode, then publish your own loop metrics. That is how operators earn trust with referring clinicians and with their own boards: not by recycling a PubMed abstract as a vanity stat, but by showing the local funnel moving.

If your group cannot yet pull time-to-book, start there. The need you create by showing the gap is the need a closed-loop phone and messaging layer exists to fill.

Specialty types that leak differently

Procedural specialties often leak at prior auth and scheduling complexity. Cognitive specialties leak at phone access and long wait explanations patients abandon. Oncology and other high-anxiety pathways leak when empathy-free portals are the only door. One playbook still works — aged queue, outbound cadence, dual booking path — but scripts and SLAs must differ by service line.

Do not declare victory because orthopedics improved while behavioral health referrals rot. Segment the board.

Referring clinician experience is part of the product

Send a monthly one-page completion report to top referrers: volume received, median time-to-book, percent unable to contact, percent declined. Invite corrections. When referrers see you sweating the loop, they send more. When they only hear patient complaints, they route elsewhere — that is leakage you will mislabel as “market competition.”

After-hours referral calls

People call specialty lines at night with a referral PDF in hand. If those calls become voicemail, day-0 aging starts already late. An after-hours agent that books within published templates or schedules a callback with captured demographics protects the loop while clinicians sleep. That is the same overflow logic as primary care, with higher stakes per missed connection.

Prior auth as a false villain

Staff often blame prior authorization for every unbooked referral. Auth is real friction — and still only explains the subset sitting in “auth pending.” If your silent bucket (no contact attempt) is larger than your auth-pending bucket, you have a phone and outreach problem wearing an auth costume. Split the queue so the right team owns each blocker.

New patient templates are a budget choice

If new-patient specialty templates are unlocked for anyone who calls, referral traffic loses to noisy demand. Protect a daily tranche for referral workqueue bookings. That is not “preferencing”; it is honouring clinical partners who already triaged.

Competitor narrative vs operator narrative

Vendors will sell “referral leakage software” as if dashboards alone stop the leak. Dashboards without outbound contact and answered phones are scoreboards on a sinking ship. The operator narrative is blunter: someone has to speak with the person within days of the order, offer a real appointment, and clear admin blockers with owners. Buy software only if it shortens that human loop — not if it merely charts the failure.

Arbol’s angle in this story is the contact layer: voice and messaging that work when portals fail and when hold queues overflow, with escalation when the request becomes clinical. The dashboard is yours; the answered loop is the product.

Referral leakage: when specialty care is ordered but never booked