How AI Fixes Orthopedic Referral Intake and Stops Leakage

Your orthopedic referral intake process probably runs on a fax machine and one person’s memory. A primary care office sends over a knee referral, it lands in a shared inbox or a paper tray, and somebody has to open it, find the patient in your system, key in the demographics, confirm the plan is in network, and then call to book. Do that forty times a day with two people who are also answering the phone. The referrals nobody got to are not a paperwork problem. That is revenue that already chose your practice and then went somewhere else.

Orthopedics is a referral business. Sports medicine, joint replacement, spine, and hand all fill their schedules with patients another physician sent over. Every one of those arrives as an interruption: a fax, a portal message, a phone call from a referring office, sometimes a patient who says their doctor told them to call you.

The part that gets missed is what happens after the referral arrives. Someone still has to register the patient, verify coverage, decide which surgeon and which visit type fits, and reach the patient before the patient gets tired of waiting. Research on primary care referrals in a large health system found that of 103,737 referral scheduling attempts, only 34.8% ended in a documented completed appointment, and 38.9% of attempts had no appointment date recorded at all. The referral was made. The visit never happened.

That gap is the cheapest revenue in your practice to recover, because the demand already exists. You do not have to market for it, and you do not have to earn the referring physician’s trust twice. You just have to answer faster than the patient’s patience runs out.

The fax queue is still your front door

Orthopedic practices tend to assume referral intake is mostly electronic by now. The federal data says otherwise. In 2025, 40% of hospitals reported often using mail or fax to send a summary of care record, and 35% reported often using it to receive one, according to the Office of the National Coordinator’s tracking of interoperable exchange methods.

The picture in physician offices is sharper. An ONC data brief on office-based physicians found that about one third (35%) used only fax, mail, or e-fax to share patient health information outside their organization, and that among physicians engaged in health information exchange, 71% said providers in their referral network lacked the capability to exchange electronically.

So the fax is not a legacy embarrassment you will retire next quarter. It is the channel a large share of your referring offices actually use, and it will keep being that channel regardless of what your practice buys. The practical question is not how to kill the fax. It is who opens it, how fast, and what happens in the ninety minutes after it lands.

Where orthopedic referrals actually go missing

Referrals rarely disappear in one dramatic failure. They leak in small, boring steps.

A fax arrives at 4:40pm on a Friday and gets stacked for Monday. The demographics on the cover sheet are half legible, so registration guesses at a policy number and the eligibility check fails two days later. The referral says knee but not which side, and it does not say whether imaging is attached, so it sits in a pile waiting for someone to call the referring office. The patient is called twice, both times during their work shift, and after two voicemails the practice moves on.

None of those steps look like a lost patient at the moment they happen. Each one just looks like a task that got deferred. Added together across a week they are the reason your new patient slots have holes in them while your referral queue has a backlog.

The MGMA Stat poll from March 2026 lines up with what orthopedic administrators describe: staff phone time is dominated by eligibility and prior authorization work at 45% and scheduling at 31%, with intake at 9%. When the same two people own the phones and the referral queue, the queue loses. That is the same pressure that shows up in orthopedic call center automation, just arriving on paper instead of a phone line.

What an AI intake agent does with an inbound referral

Take the same Friday afternoon fax and give it to an AI agent that sits on top of athenahealth.

The agent reads the referral, pulls out the patient name, date of birth, referring provider, requested body part, and insurance details, and matches the patient to an existing chart or opens a new registration. It runs eligibility against the plan on the referral. If the policy number is wrong or the plan is out of network, that referral drops into an exception queue with the reason attached instead of failing silently in a batch job on Tuesday.

Then it calls the patient. Not once during business hours, but on the schedule people actually answer, in English or Spanish, with the ability to book directly against your surgeons’ templates and visit-type rules. That booking step is where referral intake meets orthopedic scheduling automation, and it is the handoff most point tools drop.

If the patient asks anything about their condition, their symptoms, or what the surgeon will recommend, the agent does not answer it. It captures the question and routes it to your clinical staff, who own that conversation entirely.

What the agent owns is paperwork and logistics: capture, register, verify, reach, book, document. Everything requiring a clinician’s judgment goes to a clinician, with the context already gathered so they are not starting from a blank screen.

Why this is a capacity argument, not a headcount argument

The honest version of the business case is not that you fire your referral coordinator. It is that your referral coordinator stops being a data entry clerk.

BLS data puts the mean wage for medical secretaries and administrative assistants at $22.50 an hour, about $46,800 a year, across roughly 961,600 workers nationally (May 2025 OEWS). Those are the people currently retyping fax cover sheets. The work is real and it has to happen, but it does not need a trained human doing it at 8:15am while three patients wait at the check-in window and the phone rings.

The practices getting value here are not shrinking their front office. They are pointing it at the exceptions: the referral with a missing authorization, the surgeon whose template needs a human judgment call, the referring office that needs a relationship instead of a fax confirmation. That is work you cannot script, and it is what your staff was hired to do.

Most orthopedic groups we talk to have already bought two or three point tools that each handle a slice of intake and then hand the hard 40% back. The test for anything you add next is simple: does it close the loop, or does it create a new queue somebody has to babysit?

What to measure once intake is automated

Referral intake is easy to declare fixed and hard to prove fixed. Pick a small set of numbers and watch them weekly.

Time from referral received to first patient contact attempt is the single most predictive one. If it moves from two days to two hours, most of the other numbers follow. Track referral-to-scheduled conversion rate next, because that is the metric that ties directly to filled slots. Then add the share of referrals that land in an exception queue and how long they sit there, since that is where automation quietly fails. Coverage problems surfaced at intake belong in the same weekly review as insurance verification for orthopedic practices, because a referral that cannot be verified is not really booked.

On the access side, the CAHPS Clinician and Group Survey measures whether patients got appointments for urgent and non-urgent care as soon as they needed and whether they got a timely answer when they contacted the office. Those items are a reasonable outside check on whether faster intake is showing up in patient experience, not just in your internal dashboard.

Finally, watch new patient slot fill rate by surgeon. Referral intake that works shows up there within a month, or it did not work.

Key Takeaways

Fax is still a primary referral channel: 35% of office-based physicians used only fax, mail, or e-fax to share patient information, and 71% said their referral network partners could not exchange electronically.
Referral completion is the leak, not referral generation. In one large health system study, only 34.8% of referral scheduling attempts ended in a documented completed appointment.
Measure time from referral received to first contact attempt. It is the earliest signal that predicts referral-to-scheduled conversion.
Route every clinical question the agent hears to your staff. The agent should own capture, registration, eligibility, and booking, and nothing beyond that.
Build an exception queue on purpose and watch its depth. Automation that has nowhere to put a bad referral will drop it instead.
Judge any intake tool by whether it closes the loop or creates another queue your team has to monitor.

Referral intake is the least glamorous revenue project in an orthopedic practice and usually the highest yield one. The patients already picked you. The referring physicians already did the selling. All that stands between the referral and the visit is a stack of manual steps that happen too slowly, too late in the day, and too far from the person who could have booked the appointment. Fix the speed of that first touch and the rest of the funnel stops looking broken.

Sources