At one orthopedic group we talked with this year, an inbound referral gets touched by three separate teams before anyone calls the patient. Medical records pulls the fax off the queue and works out that it's a referral. If the patient isn't in the system yet, someone types in enough demographic information to create a placeholder chart, because the document needs somewhere to live. Then it gets tasked over to a referral inbox, where a different team picks it up and starts working it. That group handles 7,000 to 8,000 referrals a month, and the first outbound call to the patient goes out about a week after the fax lands.

Every one of those steps exists for the same underlying reason.

Referral management is still manual in specialty care because the referral isn't a record. It arrives from an outside organization, usually as a fax, before the patient exists in your system. Nothing in the EHR is designed to hold it, so staff build a placeholder chart to attach it to, read the document to find out what it actually is, and track everything after that in a spreadsheet. Each step depends on a person to move the referral, and on the same person to write down that it moved.

That's the short version. Here's what's underneath it.

Three reasons referral management is still manual

1. The referral arrives before the patient exists in your system

Your EHR is organized around the patient. Every object in it hangs off a patient ID, whether that's the appointment, the note, the order, or the claim. It's a clean design and it works well for the thing it was built for, which is documenting care your organization delivers.

A referral shows up before any of that is true. Someone else's physician made the decision, in someone else's system, about a person who may have never been to your group. So the document arrives with no patient to attach to, and the workaround is to manufacture one. Coordinators call it a shell patient, and it's exactly what it sounds like, a chart with a name and a date of birth and nothing else, created so a real document has an address.

The reason this is still normal in 2026 has less to do with EHR capability than with what sits between organizations. Referrals cross an organizational boundary, and that boundary is where health IT is weakest. As of 2023, about 70% of U.S. hospitals reported at least sometimes finding, sending, receiving, and integrating patient information from outside sources, which means roughly three in ten still weren't doing all four. Hospitals are the better-resourced end of that curve. Independent specialty groups and the primary care offices sending them work sit further down it, which is why the fax machine is still load-bearing infrastructure.

2. A fax can hold five documents, and only a person can tell them apart

Referrals don't arrive as data. They arrive as paper that got scanned, and that's a bigger problem than it sounds like, because a fax is rarely one thing.

A nine-page fax might contain the referral order, a progress note, an imaging report, a copy of the front and back of an insurance card, and two pages that belong to a completely different request. Plenty of inbound faxes aren't referrals at all. So before anyone can work a referral, a person has to open the packet, read enough of it to classify what's in there, and separate the pages that matter from the pages that don't. The first task of the day isn't working referrals. It's sorting mail.

Then comes extraction. Patient name, date of birth, insurance, referring provider, reason for referral, and in orthopedics, body part, which is the field that determines everything downstream. Body part decides subspecialty, subspecialty decides which physicians are eligible, and that decides which locations and which open slots are even possible. That information is usually sitting in a sentence in the middle of a scanned clinical note, in whatever format the sending office happens to use.

And a fair amount of the time, the information simply isn't there. A nationally representative survey of American physicians found that about 69% of primary care physicians believed they routinely sent patient history and the reason for referral, while only about 35% of specialists said they received it. That was 2008 data. When researchers repeated the measurement using 2019 responses and published the comparison in Annals of Family Medicine, the gap had barely moved, even though EHRs had spread to nearly every practice in the country in the years between.

3. Status tracking lives in a spreadsheet next to the EHR

Because there's no referral object in the EHR, there's nowhere to put the fields a referral needs. Status, owner, source, urgency, follow-up date, and number of contact attempts are all properties of a record, and a PDF attached to a chart doesn't have properties.

So teams build a second system alongside the first. It's usually a spreadsheet, sometimes a shared inbox being used as a work queue, often a notepad, and always partly in the head of whichever coordinator is working that referral. It works, in the sense that referrals do get processed. It just doesn't connect to anything, which produces three specific problems.

Aging is invisible. You can see when a referral arrived, because the fax server logged it, but nothing records when it last moved. A referral that's been sitting for six days looks the same as one that came in this morning.

Urgency gets discovered rather than sorted. A STAT referral surfaces because somebody happened to open that one, not because the queue put it on top.

Handoffs drop things. When a coordinator is out, or leaves, the in-flight context leaves too. Referrals that go missing during a staffing transition tend to resurface weeks later, if they resurface at all.

There's a fourth consequence that shows up in every leadership conversation we have, and it's the one COOs feel most. Arrival is the only step in the entire workflow that a system participates in. The fax lands, so it gets counted. Everything after that is people talking to other people, so it leaves no trace. That's why groups can tell you exactly how many referrals came in last month and can't tell you their average time to schedule, their conversion rate, or which referring partners are actually driving surgical volume. It isn't a reporting discipline problem. You can't measure a lifecycle that was never modeled.

The cost of a manual, fragmented referral process

The first cost lands on the patient, in the gap between the referral being sent and anyone reaching out. In the group above, that gap is about a week, and the reason is straightforward volume-to-staffing math rather than anything the team is doing wrong. But the patient doesn't know any of that. They know their doctor said someone would call. While your team works the backlog, the patient is doing their own research, calling other groups, or deciding it can wait.

Then it becomes leakage, and the difficult part is that most of it is invisible. Peer-reviewed estimates put the share of outpatient specialty referrals that are never completed at roughly 30% to 50%. Some of that is clinical, some is patient choice, and some is patients who went somewhere with an appointment available sooner. Without referral-level tracking there's no way to tell those apart, and no way to answer the question a referring partner asks most often, which is what happened to the patient they sent you. That question gets asked in the relationships that generate your volume, and "let me look into it" is an expensive answer to give twice.

The third cost is staffing, and it compounds. Manual referral coordination scales more or less linearly with volume, so more referrals means more coordinators, and the group described above already knows that math because its first outbound call slipped to roughly a week out on volume-to-staffing pressure alone. MGMA's 2026 Regulatory Burden Report found that 40% of medical groups have hired multiple full-time administrative staff per physician just to keep up with payer rules, appeals, and reporting, and referral coordination sits in that same category of work. The researchers studying cross-institutional referral loops list clinician burnout as a direct consequence, which shows up operationally as turnover, and every departure carries undocumented process knowledge out the door with it, because in a spreadsheet-based workflow a good deal of the process only exists in somebody's memory. Then you train the replacement on a system nobody ever wrote down.

What should you automate first?

Automate intake and document processing before anything else. Not because it's the most visible problem, but because everything downstream inherits it. You can't route, prioritize, track, or report on data that was never structured at the front door, so a group that automates its queue before it automates extraction ends up with a very organized view of records that are still missing half their fields. Reading the packet, classifying the document, pulling out demographics and insurance and body part, and matching the patient against the EHR instead of hand-building a shell chart, that's the layer everything else stands on.

Second, put every referral in one place with real fields. One queue, regardless of whether the referral came in by fax, portal, email, or phone, where status and source and urgency and follow-up date are attributes of the referral itself rather than columns in a file on somebody's desktop. This is what ends the spreadsheet, and it's also what makes reporting stop being a monthly manual exercise, because once the work happens on the record, the data comes out as a byproduct.

There's peer-reviewed evidence for what that change alone does. In a quality improvement study at a tertiary hospital, embedding a structured referral order in the record and routing it to a centralized coordination team took documented referrals from fewer than two per week to a sustained average above 800, with about 80% of referred patients contacted within nine minutes of entry. The referrals had been happening all along. They just weren't records, so nobody could see them, count them, or work them systematically.

Communication comes third, and the order matters. Automated updates to patients and referring partners only work if the underlying status is trustworthy, so the sequencing isn't a preference. A status trigger firing off stale data will tell a partner their patient is scheduled when nobody has called them yet.

Worth being straightforward about the limits. Two things in this workflow stay manual for a while, and those are insurance verification and getting the patient onto a schedule. Payer rules vary by plan and by procedure, capacity is a real constraint, and no software resolves either by itself. What changes is that a referral waiting on authorization stops being indistinguishable from a referral nobody has touched, and your team can work the queue by what's actually blocking rather than by what happens to be on top.

This is the work Hatch does for orthopedic and specialty groups. Proliance Surgeons deployed it inside their value-based care surgical line, where referrals come in by fax, email, portal, and aggregator feed across 25 care centers running nine different EMRs. After implementing smart intake and referral parsing, triage time per case dropped from roughly 15 minutes to about 3, and document processing time fell 40%. That line moves approximately $25 million in annual surgical revenue, and one coordinator runs it. Read how they do it.

Sources

1. ASTP/ONC, Interoperable Exchange of Patient Health Information Among U.S. Hospitals: 2023, Data Brief No. 71, May 2024. (Government)

2. Savoy A., et al., Consultants' and referrers' perceived barriers to closing the cross-institutional referral loop, 2023. (Peer-reviewed)

3. Communication Gaps Persist Between Primary Care and Specialist Physicians, Annals of Family Medicine, 2022, using 2019 survey data. (Peer-reviewed)

4. O'Malley A.S., Reschovsky J.D., Referral and consultation communication between primary care and specialist physicians, Archives of Internal Medicine, 2011. (Peer-reviewed, original source of the 69% / 35% figures)

5. Improving Referral and Continuity of Care Through Structured Outpatient Disposition Planning Enabled by Electronic Referrals, quality improvement study. (Peer-reviewed)

6. MGMA, 2026 Regulatory Burden Report, April 2026. (Industry association)