Industries / Government, education and community / Government / Emergency services / Ambulance and medical transport
Transfers between hospitals
Transfers move patients who are already in one hospital to another that can give care the first cannot, often with specialist teams for critically ill adults, newborn babies or patients on heart-lung machines.
7
Milestones achieved, from 1934 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1934
An ambulance for premature babies
From 1922 the pediatrician Julius Hess ran a station for premature infants at Sarah Morris Hospital, part of Michael Reese Hospital in Chicago, that took babies born anywhere. In 1934 Chicago's Board of Health and the stations at Sarah Morris and Cook County hospitals began a city-wide plan: every premature birth had to be reported by telephone within one hour, and every move, from a home or from another hospital, went by a specially equipped Board of Health ambulance with a nurse trained to care for the baby on the way.
Credit
Julius H. Hess; Chicago Board of Health; Infants' Aid Society of Chicago
1986
Transfers for lack of money
R. L. Schiff, D. A. Ansell and colleagues followed 467 patients sent from other Chicago-area emergency departments to Cook County Hospital, the city's public hospital, and admitted there. 87% had been moved because they lacked adequate insurance, 24% were unstable when sent, and only 6% had given written consent; transferred medical patients died at 9.4%, against 3.8% of medical patients not transferred. The study appeared in February 1986, and that year Congress passed the Emergency Medical Treatment and Labor Act (EMTALA), under which hospitals that take Medicare and offer emergency care must screen and stabilize anyone with an emergency, whatever their ability to pay, and may transfer an unstabilized patient only by an appropriate transfer.
Credit
R. L. Schiff, D. A. Ansell and colleagues; US Congress
2011
Transfer clocks for heart attacks
A patient with the most serious type of heart attack, called STEMI, whose first hospital cannot open the blocked artery must be moved to one that can, and experts recommend no more than 30 minutes from arrival to departure at the first hospital, the door-in to door-out time. In a US registry, 14,821 patients transferred in 2007–2010 waited a median of 68 minutes and only 11% met 30 minutes; those who waited longer died in hospital more often (5.9% against 2.7%). A national study using a new Medicare measure found 9.7% of 13,776 patients within the target.
Credit
T. Y. Wang and colleagues (ACTION Registry–Get With The Guidelines); J. Herrin, H. M. Krumholz and colleagues (Medicare measure)
2015
Heart-lung machines on the road
Extracorporeal membrane oxygenation (ECMO) pumps a patient's blood through an artificial lung outside the body when the lungs or heart fail, and only specialist centers offer it. Since 1996 a team in Stockholm has driven or flown to other hospitals, started ECMO there and brought the patient back on the machine, more than 700 times by 2015. In its 322 transports from 2010 to 2013, 60% by aircraft and up to 13,447 km long, adverse events occurred in 27.3% and no patient died on the way.
Credit
L. M. Broman, B. Frenckner and colleagues, Stockholm
2020
Intensive care by train
On 26 March 2020, with intensive care units in eastern France full of COVID-19 patients, a double-deck TGV high-speed train fitted out as a moving intensive care unit carried 20 patients from Strasbourg and Mulhouse to hospitals in western France, with stretchers fixed above the seats, four patients to a car and a team of six per car: an anesthetist, a junior doctor, a nurse anesthetist and three nurses. In the TRANSCOV study, 130 patients moved by train and 163 by air between 13 March and 10 April 2020 had no difference in 28-day deaths, although train journeys took about 2 hours longer.
Lessonan ordinary passenger train moved ventilated patients with outcomes no worse than aircraft, though the authors note that healthier patients were chosen for the train.
Credit
SNCF; SAMU of Paris (Lionel Lamhaut); TRANSCOV investigators
2021
States lend each other beds
Germany's cloverleaf plan (Kleeblattkonzept), set up in 2020, groups neighboring states into regions that move intensive care patients to one another when one region's units fill; the East cloverleaf, for example, joins Saxony, Thuringia, Brandenburg, Saxony-Anhalt and Berlin. By 6 December 2021, 93 COVID-19 patients from Saxony and Bavaria had been moved to hospitals in northern and western Germany, some on Bundeswehr aircraft. A study by Jena researchers of 48 such transfers involving Thuringia found that the sickest patients gained most when moved from a very small hospital to one of the highest level of care.
Credit
German federal and state authorities; Bundeswehr; S. Lang, M. Bauer, P. Dickmann and colleagues (Jena University Hospital)
2025
Full hospitals refuse transfers
At the only academic and level I trauma center in a highly rural state in the US Southwest, M. Greenwood-Ericksen and colleagues studied 26,020 transfer requests from January 2019 to May 2023. The center accepted 61.7% overall and 56.3% of requests from rural hospitals. In the weeks when emergency department boarding, patients held in the department while waiting for a ward bed, was worst, the odds of acceptance were 29% lower than when it was least.
Credit
M. Greenwood-Ericksen and colleagues
Watch for
Statewide transfer hubs
The University of Arizona says AZ REACH, a hub begun in 2022 that finds places at larger hospitals for patients at rural, Indian Health Service, tribal and critical access hospitals, has handled more than 20,000 transfer referrals and saved over 27,000 hours of on-site staff time. Indiana's health department is procuring a 24-hour statewide Medical Operations Coordination Center for transfers, ambulance resources and hospital capacity under its Rural Health Transformation Program.
Achieved whenan outside evaluation of a statewide hub publishes times from first call to acceptance and to arrival, and patient outcomes, against the period before it.
Claimed by
University of Arizona, February 2026; Indiana Department of Health, 2026
Open
01
A public transfer request test
A public set of 300 invented transfer requests, written and spoken the way a referring doctor presents a patient to a transfer center, each with an answer key from two emergency or critical care physicians: whether the patient needs a higher level of care, what kind of receiving unit, how urgently, and which transport team (basic ambulance, critical care ground team, helicopter or ECMO team). Score AI models and simple rule lists against it, and count unsafe answers, such as sending an unstable patient without a critical care team, separately.
The briefCredit · unclaimed02
Count the pandemic transfers
One open table of every capacity-driven transfer of critically ill patients between regions or countries in Europe from 2020 to 2022: country, dates, origin and destination, mode (train, plane, helicopter or road), number of patients, adverse events and published outcomes, each figure linked to its source, with conflicts between sources listed.
The briefCredit · unclaimed
Open 01 · the brief
A public transfer request test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of 300 invented transfer requests, written and spoken the way a referring doctor presents a patient to a transfer center, each with an answer key from two emergency or critical care physicians: whether the patient needs a higher level of care, what kind of receiving unit, how urgently, and which transport team (basic ambulance, critical care ground team, helicopter or ECMO team). Score AI models and simple rule lists against it, and count unsafe answers, such as sending an unstable patient without a critical care team, separately.
Why it is still open
The AI studies of transfer requests use private hospital records from one center each: a language-processing model flagged complex cases among 1,895 requests to an academic health center (Lee and colleagues, 2023), and a 2026 study summarized 31 recorded heart attack transfer calls with Whisper and ChatGPT, with physician raters agreeing only slightly. States are now building transfer hubs, and none has published a test set. A search in November 2026 found no public benchmark.
Built with what exists
Synthea, an open generator of synthetic patient records; general AI models such as Claude, Gemini and GPT; speech synthesis for spoken versions and speech-to-text such as Whisper or ElevenLabs Scribe v3 (see the Note "Fewer word errors"); a fixed-answer scoring pass with a model such as Gemini Decision 1 (see the Note "A thousand questions, one pass").
First step
Write 30 requests covering obstetrics, STEMI, stroke and trauma, the conditions the 2025 Southwest study found were given priority, and have two physicians key them.
Done looks like
300 requests across at least 10 conditions, two physicians' keys with their agreement rate, at least three models scored on all four questions, and each model's rate of unsafe answers published with the data.
Open 02 · the brief
Count the pandemic transfers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One open table of every capacity-driven transfer of critically ill patients between regions or countries in Europe from 2020 to 2022: country, dates, origin and destination, mode (train, plane, helicopter or road), number of patients, adverse events and published outcomes, each figure linked to its source, with conflicts between sources listed.
Why it is still open
The figures sit in separate reports and are counted in different ways at different dates. TRANSCOV covers France's first wave only. Germany's count is given as 93 patients by 6 December 2021 in one report and as 115 transfers in two of its regions up to February 2022 in another, and no national total was found. A Dutch study counted 1,213 and then 528 transfers in two periods of 2021, for one region. No pooled count across countries was found in a search in November 2026.
Built with what exists
a browser agent such as Browser Use to collect ministry, army and operator releases (see the Note "Deleting 80% to start over"); the Europe PMC open search interface for studies; an AI model to pull figures into fixed fields, with every figure checked by a person; a spreadsheet published under an open license.
First step
Put the French, German and Dutch figures from the three studies cited here into one table and list where they disagree.
Done looks like
at least 10 countries, every figure linked and checked by a second person, all conflicts listed, and totals by mode and year.