A physician in a 25-bed critical access hospital has a patient who is deteriorating in a way her facility cannot manage. There is no neurosurgeon, no interventional cardiology, no intensivist. The patient needs to be somewhere else, tonight.
She picks up the phone.
She calls the regional referral center's transfer line. The transfer line pages the on-call specialist. The specialist calls back twenty minutes later and wants imaging she cannot push to their system and labs she can only read aloud. He says, in the particular phrasing that everyone in emergency medicine knows by heart, that he is not comfortable accepting.
She calls the next hospital. And the next.
A physician in the research literature described this part of the job with unusual bluntness: "Like 50 percent of my job is transferring ED patients." Others describe waits of up to six hours tied to a phone, and transfers that "can take anywhere from hours to days."
Meanwhile the patient is in a bed in a small hospital, getting whatever care is available, while a physician who could help sits eighty miles away, unaware anyone is looking for them.
This is not a bed shortage story, although beds are genuinely short. It is a story about a search that has no index.
What the acceptance data shows
The most useful recent measurement comes from a 2025 study in JAMA Network Open examining 26,020 transfer requests to one academic level I center.
- 61.7 percent were accepted.
- Rural requests fared worse: 56.3 percent versus 62.5 percent for urban, with an adjusted odds ratio of 0.66.
- Acceptance correlated strongly and negatively with the receiving hospital's emergency department boarding (r = -0.73) and inpatient census (r = -0.87).
Read the last two figures carefully, because they are the finding.
Whether a patient is accepted is powerfully predicted by how full the receiving hospital happens to be, and not by the patient's need.
Consider what that means for the sending physician. The single most important variable in the transfer decision, the receiving hospital's current state, is completely invisible to her. She is calling blind into a system whose answer is largely determined by a condition she cannot see and cannot influence.
And rural patients, who by definition have the fewest local alternatives, are refused more often.
The relationship is the actual mechanism
Now the study that should change how transfer is understood, and it is remarkably underdiscussed.
Research published in the American Journal of Accountable Care examined 81 STEMI transfers and measured "relational coordination" between the sending and receiving physicians, a validated construct covering shared knowledge, mutual respect, and communication quality.
Every 40-point rise in relational coordination was associated with a 25 percent shorter time to acceptance.
Same patients. Same clinical urgency. Same hospitals. The variable that moved the clock was the quality of the professional relationship between the two physicians on the phone.
Qualitative research reaches the same conclusion from a completely different direction. A study in Medical Care found that "informal arrangements played a key role in the identification of the receiving hospital," and, strikingly, that hospital quality was not discussed as a factor in choosing where to send patients.
An 81-interview study across seven VHA facilities, published in the Western Journal of Emergency Medicine, found transfer decisions driven by personal relationships and process simplicity rather than quality metrics.
Put plainly: the transfer system runs on who you know. Not officially. Not by design. But measurably, in the acceptance rates and in the clock.
Which produces an entirely predictable inequity. The experienced physician who has worked in the region for fifteen years has a mental map: which specialist at which hospital will take which kind of patient, who to call directly, whose direct line to use rather than the transfer center. That map is personal, unwritten, and unshared.
The locums physician on their second shift starts from zero. So does the new graduate. So does anyone who moved to the region last year. They face the same patients with none of the graph, and no way to acquire it except slowly, through experience they may not have time to accumulate.
EMTALA does not solve this, and understanding why matters
The reasonable objection: does the law not require hospitals with capability and capacity to accept appropriate transfers?
It does. And the operative word is capacity, which is assessed by the person on the phone.
A specialist who says "I am not comfortable accepting" is generally making a clinical judgment, and frequently a legitimate one. Hospitals genuinely do run out of beds. Boarding is genuinely at crisis levels in many places, and the correlation coefficients above (-0.73 and -0.87) show that capacity constraint is real rather than pretextual.
The problem is not that the law is being flouted. It is that the law creates an obligation without creating a mechanism. Enforcement happens case by case, retrospectively, on individual complaints. That is entirely useless to a physician at 11 p.m. with a deteriorating patient and four hospitals left to call.
The gap between legal obligation and operational reality is filled, in practice, by relationships.
Where the software went
There is a substantial market of transfer technology: transfer center platforms, capacity command centers, patient logistics systems. Several are genuinely good products.
Notice who buys them.
Receiving hospitals buy transfer centers. And receiving hospitals buy them to manage and optimize inflow. That is a legitimate business purpose, and transfer centers deliver real value when they work: one published account from Ochsner reported average time savings of two to six hours per patient, with the operational goal of accepting 80 to 90 percent of appropriate requests within ten minutes and a single call, and satisfaction scores rising substantially.
That is a genuine achievement, and it is achieved for the receiving institution's own referral network.
Look at the market from the sending physician's side and the picture is completely different. She is not a customer of anybody. She does not buy the transfer center software; the receiving hospital does. Her problem, which is finding a willing human across multiple unaffiliated systems, has no product because she is not a purchaser.
Transfer centers are an intake mechanism for tertiary hospitals. The sender needs a search index across institutions, and nobody sells one, because the sender has no budget.
Every technical excuse for this has now expired, which is what makes the persistence notable. Image sharing works. Telemedicine consultation works. Electronic record exchange works. What remains is entirely relational: does a specific human being at a specific institution know who you are and agree to take your patient?
What is at stake clinically
Transfer failure is not an inconvenience metric.
Transferred patients have higher morbidity and mortality than direct admissions, partly because they are sicker and partly because of the delay itself. Work compiled through transfer improvement toolkits suggests better pre-transfer communication may prevent a substantial share of adverse events among transferred critically ill patients, though the evidence base here is preliminary and should be treated as such.
Roughly 1.8 million patients are transferred from emergency departments to other hospitals each year in the United States by secondary estimates, and 44.3 percent of acute myocardial infarction patients present at hospitals without revascularization capability.
And the situation is getting harder rather than easier. Rural hospital closures continue. The Rural Emergency Hospital designation creates facilities that by design stabilize and transfer, which increases transfer volume from exactly the places with the weakest relational networks. Post-pandemic boarding has made refusal routine at receiving centers.
One emergency physician described the position of being unable to transfer a patient who needs it as "soul-crushing and terrifying." That is not overstatement. It is a description of practising medicine while unable to obtain the thing your patient needs, on the phone, repeatedly, while time passes.
What would actually help the sender
The design requirements here are unusually clear because the failure is so specific.
A declared acceptance map, maintained by the accepters. Specialists stating what they will accept, from where, and during what hours, visible to verified sending clinicians. Not a bed board, which measures the hospital. A person map, which measures who will say yes to what.
Parallel rather than serial calling. The current protocol is sequential: call, wait, get refused, call the next. A structured request that reaches multiple potential accepters simultaneously and returns a named yes would collapse the six-hour phone marathon into something bounded.
Trust distance used deliberately. The relational coordination finding says relationships cut acceptance time by a quarter. The co-training graph is the obvious first place to look: the sending physician's former co-residents who are now attendings at tertiary centers are the natural first accepters, and nothing currently surfaces them.
A record of who accepts. Acceptance rates, time to decision, and refusal reasons by receiving institution and condition. This is the first sender-side dataset in a field where all measurement is currently done by and for receivers. It would be immediately valuable to rural hospital associations, state offices of rural health, and regional planning, and it does not exist anywhere.
Transfer centers used for logistics, not gatekeeping. Once a named physician has agreed, the transfer center is genuinely good at everything that follows. The failure is upstream of them.
What you can do now
If you send patients
Build your accepter map explicitly. For each condition you commonly transfer, write down the hospital, the service, the specific people who have accepted before, and any direct numbers you have been given. Most experienced physicians carry a version of this in memory. Writing it down makes it survive your own move and, more importantly, makes it available to your locums colleague on night two.
Give the map to every new person on day one. This is the single highest-value orientation document a rural emergency department can produce, and almost none exist. A new physician's transfer capability currently rebuilds from scratch over months.
Log your transfer attempts. Condition, hospitals called, time to decision, outcome. Nobody collects this and everybody needs it. Six months of logs from one department is a more useful account of regional transfer reality than anything currently published.
Call the person, not only the line. Where you have a relationship, use it. The relational coordination data suggests this is worth roughly a quarter of the acceptance clock.
Ask what would make it a yes. Refusals frequently turn on a specific missing element: an image, a lab, a stated plan. Asking directly converts a dead end into a task.
If you accept transfers
Publish what you will take. Most sending physicians are guessing about your scope, your hours, and your thresholds. A single page distributed to referring facilities eliminates an enormous amount of wasted calling on both sides.
Answer quickly, including when the answer is no. A fast no lets the sender move to the next option. A slow no costs the patient hours.
Say why, and say who instead. "Not us, but the burn center at X takes these and ask for Y" is a genuinely valuable answer and takes ten additional seconds.
Notice who you are refusing. Rural requests are accepted at 56.3 percent versus 62.5 percent for urban in the published data. If your institution has never examined its own acceptance rates by referring facility type, that number is worth knowing.
If you run a system or a region
Measure sender-side experience. Every existing transfer metric is receiver-side. Nobody knows how many calls a rural physician makes on average, or how many transfers are abandoned entirely. That is the dataset that would change regional policy.
Fund the relationship deliberately. The relational coordination finding suggests that structured contact between sending and receiving clinicians, meeting each other before the crisis, is a measurable clinical intervention rather than a networking nicety.
Frequently asked questions
How often are interfacility transfer requests accepted? In a 2025 JAMA Network Open study of 26,020 requests to one academic level I center, 61.7 percent were accepted. Rural requests were accepted at 56.3 percent versus 62.5 percent for urban requests, an adjusted odds ratio of 0.66.
What determines whether a transfer is accepted? Substantially, the receiving hospital's current state. The same study found acceptance correlated negatively with the receiving hospital's emergency department boarding (r = -0.73) and inpatient census (r = -0.87). Qualitative research consistently finds transfer destination driven by personal relationships and process simplicity rather than hospital quality metrics.
Does the relationship between physicians affect transfer speed? Measurably. Research in the American Journal of Accountable Care examining 81 STEMI transfers found every 40-point rise in relational coordination between sending and receiving physicians was associated with a 25 percent shorter time to acceptance.
Doesn't EMTALA require hospitals to accept transfers? It obliges hospitals with specialized capability and capacity to accept appropriate transfers, and capacity is assessed by the receiving institution. Enforcement is retrospective and complaint-driven, which provides no mechanism for a physician trying to place a patient tonight. The obligation exists; the operational routing does not.
Why don't transfer center platforms fix this? Because receiving hospitals buy them, to manage inflow into their own institution. They deliver real value in that role, with published accounts reporting two to six hours saved per patient. The sending physician, who needs to search across multiple unaffiliated systems for a willing accepter, is not a customer and has no product.
How long do transfers take? Highly variable and frequently long. Physicians in qualitative research describe waits of up to six hours on the phone and transfers taking anywhere from hours to days, with one participant estimating that transferring patients constituted roughly half their job.
The bottom line
A physician in a small hospital has a patient who needs to be somewhere else. Somewhere within a two-hour drive is a specialist who could take that patient and would probably say yes.
Finding them takes hours of serial phone calls, has a roughly 60 percent success rate at any given institution, is measurably worse if the patient is rural, and depends on whether the caller happens to have accumulated a personal map of who takes what.
Every technical barrier has fallen. Images move. Records move. Video works. What remains is a search problem across institutional boundaries, in a system where all the measurement, all the software, and all the purchasing power sits with the receiving side.
The bed is not the bottleneck. The bottleneck is that the sending physician is looking for a specific human being who will say yes, and the only index that exists is the one in her own head, built up over years, and completely unavailable to whoever covers her shift next week.
Part of a series on the missing professional infrastructure of healthcare. Previously: The People Who Have Been Through It
Evidence note: sources include JAMA Network Open (2025) analyzing 26,020 transfer requests to a single academic center; the American Journal of Accountable Care (2022) on relational coordination and STEMI transfer timeliness; Western Journal of Emergency Medicine (2020) qualitative research across seven VHA facilities; Medical Care (2011) on informal arrangements in transfer destination; and the Ochsner Journal (2012) on transfer center performance. Single-center acceptance data may not generalize. The estimate of adverse events preventable through improved pre-transfer communication comes from preliminary evidence compiled in improvement toolkits and should be treated as provisional.
