How Pain Management Practices Speed Up Referral Intake with AI

A faxed referral is the most valuable piece of paper a pain practice receives. It is also the one most likely to sit unread. Referral intake at an interventional pain group is the front of the revenue line. A primary care office, an orthopedic group, or a neurology practice already decided the patient should come to you. What stands between that decision and a booked visit is whether somebody works the document this week.

Demand is not the constraint. In 2023, 24.3% of U.S. adults had chronic pain and 8.5% had pain that frequently limited life or work activities. The referring offices around you are already sending patients your way.

The constraint is what happens after the fax lands. At most pain practices the inbound referral queue is a shared inbox, a fax folder, and one person who works it between phone calls. Referrals that arrive Friday afternoon get looked at Monday. Referrals filed under a satellite department get looked at when somebody remembers that department exists. Every one of them is work the practice has already earned and has not yet billed.

Front office staff know this. They also know the queue is never the most urgent thing in front of them. The patient at the window and the ringing phone both outrank a PDF that will still be there in an hour.

The referral is your growth channel, and it still arrives as a PDF

Pain practices receive referrals through at least three doors. None of them deliver a structured order. Flat PDFs come in over fax from primary care, orthopedics, and neurology. High-volume referring offices want a portal where they can submit and then check status without calling. Patients refer themselves through a web form after a search or a recommendation.

That mix is not going away soon. The 2019 National Electronic Health Records Survey put a number on it. It found that 35% of office-based physicians used only fax, mail, or e-fax to share patient health information outside their organization. Among physicians who did exchange data electronically, 71% said providers in their referral network lacked the capability to exchange at all. The office on the other end of your fax line is usually not withholding a structured record. It has no way to send you one.

So referral intake is a document problem before it is anything else. Somebody has to read a loose PDF and find the patient. Then work out whether a chart already exists, create one if it does not, and file the document where a scheduler will see it. That is the work our AI team does inside athenaOne, in the block we call New Patients and Referrals. Inbound referral fax intake, OCR, chart building, and the write-back that turns a page of paper into a record somebody can act on. Practices with heavy procedure blocks hit the same wall. Referrals stop dying in the fax queue only once somebody owns the document, not just the phone.

Referral faxes only get worked if they land in the right department bucket

Here is the failure that costs a multi-site pain group the most new patients. It is invisible from the outside.

Document workflows in athenaOne are scoped by department. An intake pipeline pointed at the main office picks up what lands there and reports success on all of it, while documents filed under any other department sit untouched. A large share of the new-patient referrals is usually in exactly those other departments. The queue looks healthy while the best referral source piles up somewhere nobody is watching, which is why the number to trust is documents processed against total inbound rather than the job’s own success rate.

The EHR will generally not return every department in one pass. So the fix is dull rather than clever. Walk the departments in order, one after another, and accept that a full sweep takes longer than a single-department pull. That may mean moving from a 15-minute cycle to an hourly one, and clearing every bucket instead of one.

This is the part that gets skipped. Reading a fax is the easy half. The hard half is knowing which departments exist, which ones get referrals, which ones nobody has checked since a location opened, and how often each needs a sweep. That is setup work, and it is where a generic intake tool breaks on a real practice. It is the same reason generic AI fails specialty practices more often than it fails simple ones.

The handoff is explicit. Anything the sweep cannot tie to a patient, a referring provider, and a department goes to a named staff queue. The document is attached, along with the reason it stopped. Nobody has to go hunting for the exceptions. They arrive in one place.

A referral on file is not a patient on the schedule

The referral loop is measured nationally, and the measurement is not flattering. The federal quality measure for closing the referral loop cites an analysis of 103,737 primary care referral scheduling attempts. Only 36,072 of them, or 34.8%, ended in a completed appointment with a report back to the referring clinician.

For a pain practice, the distance between having the referral and having the patient on the schedule is where new-patient growth leaks. The document arrived. The chart exists. Nobody called.

Outbound is the other half of intake. Once a referral is in athenaOne with a chart attached, the same AI team places the call. It offers real slots against the provider’s template and finishes the new-patient registration. Where the referral already carried demographics and coverage, it asks only for the missing fields. Nobody gets walked through an intake they already did. Where nothing came through, the call gets placed the day before the visit so most of the paperwork is done before the patient arrives.

Staff still take the calls that are not bookings. A patient who wants to talk through what the visit involves, or who is unhappy about a wait, goes to a person. The AI books, confirms, and chases paperwork. Anything clinical belongs to clinical staff. The pass itself is where most AI-to-human transfers fail, not the decision to make it.

The rules a pain referral has to clear before it becomes a booked visit

Pain management carries more scheduling rules than almost any segment. The rules attach to the referral rather than to the slot.

Most interventional procedures need authorization, and the authorization has a window. Procedures scheduled outside that window by human schedulers can look correct at booking time. The visit happened. The claim died weeks later. Landing the appointment inside the window at booking time is the whole game. It has to happen while the caller is still on the phone.

Then there is the new-patient clock. A pain practice may count a patient who has not been seen in six months as new again. The intake paperwork has to be redone from scratch. That policy drives the appointment type, the visit length, and which forms attach. A returning patient booked as an established follow-up is a slot that will not hold the visit.

Booking is rarely one write, either. An injection or procedure booked at a satellite clinic may require a case sent to that clinic’s staff. Somebody has to set the room up. The side effect exists at one location and not the others, and nothing in the scheduling template knows about it.

One exception deserves its own rule. Sometimes what the referral asks for does not map to anything the practice offers. The AI does not guess. It also does not tell the referring office the referral is on file. It flags the mismatch, attaches the document, and routes it to staff to decide where it belongs.

Size your own referral gap before you buy anything

You do not need a vendor to tell you how big this is. You need two numbers your practice already has.

What follows is a worked example rather than a benchmark, and your numbers will vary. Assume a pain group receives 120 inbound referrals a month across all departments. Assume that when the queue gets worked promptly, 70 of those become booked new-patient visits, and when it does not, 55 do. The arithmetic is 70 - 55 = 15 visits a month. That is 15 x 12 = 180 new-patient visits a year that turn on nothing except intake speed. Put your own referral count and your own rates in before you believe any of it.

Two figures are worth pulling from athenaOne. The first is time from referral receipt to first outbound contact. The second is referral-to-booked-visit conversion by referring practice. The second one is the more useful. It names the referring offices whose patients never arrive, which is a conversation worth having.

Most practices cannot produce either report easily today, which is its own finding. Referral funnel conversion is one of the reports we build, precisely because the EHR does not surface it well.

Key takeaways

Ask which athenaOne departments your fax pipeline actually sweeps. If the answer is the main office, the referrals you are missing are the ones from your newest locations.
Measure time from referral receipt to first outbound contact. It separates a referral you won from one the patient took somewhere else.
Pull referral-to-booked-visit conversion by referring practice. A referring office whose patients never arrive is a relationship problem, not an intake problem.
Check whether whoever books the visit can see the authorization window at booking time. If they cannot, procedures land outside it and the claim dies after the visit.
Write down your own new-patient clock in months and make sure the intake workflow uses it. Six months and three years produce different appointment types and different forms.
Decide in advance which exceptions go to a person. Unreadable documents, a request your practice does not offer, and any patient who wants to talk rather than book.

Referral intake is the growth channel for a pain practice, and most are running it out of a queue nobody owns. Sweeping every department on a clock, booking inside the authorization window, and handing the exceptions to a named person is the whole job.

Sources: