Thailand’s emergency-patient scheme paid an average of 20,518 baht for each person it covered in its first thirty-one months. One day in Bumrungrad International’s intensive care unit, at the rate the hospital publishes as effective 1 January 2026, is 25,500: the room, the meals and the nursing, before a drug, a scan or a surgeon.
Both of those figures are real. Neither is the one that decides anything. The number that decides is 11,776.
Two sentences that do not agree
Search the scheme in English and the field hands you a coin flip. The strongest page in the field, published in April 2026 and last modified on 30 July, says a foreigner is covered “regardless of nationality or ability to pay”, and attaches no source to it. A second, last modified nineteen days after that, says the opposite in six words: it is available only to Thais, with nothing behind that either. Two pages answer, and they answer opposite. One hedges and never returns to the hedge. Four more, including one written by a hospital that would be doing the treating, do not answer at all.
Three documents do answer it. None of them is the field, and one of them does not agree with itself.
The World Health Organization’s Asia Pacific Observatory published its review of the Thai health system in 2024. Its sentence on the policy reads: “The government adopted a policy ‘Universal Coverage for Emergency Patients’ or UCEP to allow every Thai citizen to access emergency medical services at any health-care facility, both public and private hospitals, from 1 April 2017.” Every Thai citizen. The sentence does not say every patient.
Read on, though. The same review describes the same policy again in its emergency chapter, ninety pages later, as “ensuring free, high-quality emergency care for up to 72 hours at any hospital (public or private)” — no nationality qualifier anywhere in it. The most authoritative document in the field disagrees with itself between chapters, which is a better account of the English-language mess than anyone being careless.
Thai PBS Policy Watch, March 2025, puts the foreigner’s route in a conditional clause: ชาวต่างชาติ และคนต่างด้าว หากมีสิทธิประกันสังคม ก็สามารถใช้สิทธิ UCEP ได้. Foreigners and non-nationals, if they hold social-security rights, can likewise use it.
And the scheme’s own operator, the National Institute for Emergency Medicine, describing its policy to the United Nations Public Service Awards in November 2019, wrote the mechanism down without meaning to give anything away: hospitals recover their money “from healthcare schemes that the patients are entitled to.”
Read them together and what they describe is a plumbing diagram rather than a border. The money has to come from somewhere, the somewhere is a membership, and a man can live in Thailand for twenty years without acquiring one. Which is worse than a rule about nationality, because a rule about nationality is at least visible from the doorway.
One page answers it in its title, and it is the institute’s own: คนต่างชาติมีสิทธิใช้ UCEP ไหม, do foreigners have the right to use it. On 20 August 2026 it did not load. Neither did the health security office’s Thai site, the ombudsman’s site, or the health ministry’s drug-information portal: four sites across five hostnames, every one of them resolving to an address and then answering nothing, tried from three independent routes. Two further hostnames the health security office has been published under do not resolve at all. An ordinary outage covers the first four and an ordinary retirement covers the last two; nothing else is implied. It is recorded because everything below was chosen to survive their absence.
Minute one: a program, not a doctor
The thing that decides is a piece of software.
A patient arriving at a Thai emergency department is screened into one of 25 chief-complaint groups (20 non-traumatic, five traumatic) and measured on blood pressure, respiratory rate, Glasgow coma score and oxygen saturation. Seven findings authorise him automatically: “unresponsiveness, cardiac arrest, hypotension, respiratory failure, oxygen desaturation, stroke, or myocardial infarction.” So does an Emergency Severity Index level of 1 or 2. Anything inconclusive goes to a duty emergency physician, available around the clock, whom the institute appoints.
Read that list again with an aging Western body in mind. Stroke is on it. Myocardial infarction is on it. A fall in a bathroom is not, unless the blood loss has already dropped the pressure. Sepsis from a chest infection is not, unless the breathing has already failed. Nothing about the list is unreasonable. It catches the people who are dying in the next hour, and it was built to.
The only study of the program in the indexed literature is a 2018 cohort funded by the institute itself. Searched on the policy’s own name, Europe PMC returned five records on 20 August 2026 and this is the one of them that analyses the pre-authorization dataset: 20,206 patients presenting at private emergency departments between April and October 2017, every one of them assessed. 8,430 were authorized. 11,776 were not, which is 58.3 per cent of the people who asked.
| Triage grade → outcome | Authorized | Refused | Share authorized |
|---|---|---|---|
| Critical | Authorized 5,907 | Refused 0 | Share authorized 100% Every patient at this grade was authorized. All 5,907 came through Emergency Severity Index level 1. |
| Emergency | Authorized 2,518 | Refused 5,225 | Share authorized 32.5% 142 authorized at Emergency Severity Index level 1 and 2,376 at level 2. This one grade produced 29.9% of every authorization the program issued, and 44.4% of every refusal. |
| Urgency | Authorized 5 | Refused 5,041 | Share authorized 0.1% |
| General | Authorized 0 | Refused 1,485 | Share authorized 0% |
| Others | Authorized 0 | Refused 25 | Share authorized 0% |
| All grades | Authorized 8,430 | Refused 11,776 | Share authorized 41.7% Columns sum to the cohort of 20,206. |
Source: Counts as printed in Closing the Equity Gap of Access to Emergency Departments of Private Hospitals in Thailand, Emergency Medicine International 2018, Table 5, with the two authorized columns added together and the share derived here · checked 2026-08-20
Read the top and bottom rows first, because they are the ones that behave. Nobody graded Critical was refused: 5,907 for 5,907. Nobody graded General or Others was authorized: 1,510 for nothing. At the two ends the grade is the answer.
Then look at the row in the middle. 2,518 patients graded Emergency were authorized — 142 at Emergency Severity Index level 1, 2,376 at level 2 — and 5,225 patients at that same grade were not. A 32.5 per cent approval rate inside one triage category. Very nearly three in ten of every authorization the scheme issued in those six months came out of that row, and so did 44.4 per cent of every refusal.
So the boundary does not sit below a grade. It runs through the middle of one, and what separated the 2,518 from the 5,225 is an index level the table prints for the authorized and never for the refused.
The study’s authors do not read the result as the machine working. A substantial number of patients, they write, “were denied approval as a result of preauthorization, which raises a concern about patient safety” — and they add: “using the need of hospitalization as a measure of safety, a systematic review did not find sufficient evidence for the validity of 5 triage scales including ESI.”
That 44 per cent is the figure people quote, and the paper’s own phrasing for it — “44% with emergency medical conditions” — is loose enough that the denominator goes missing on the way out. It is 44.37 per cent of the refused, not of the presentations. Restoring the denominator does not make it gentler. It says that of everyone the machine sent away, nearly half held the grade it was simultaneously authorizing other people on.
Thailand’s statutory consumer body tells a Thai reader what to do with that grade. Its five-step account of the right ends on the refusal: กรณีไม่เข้าเกณฑ์ รีบประสานโรงพยาบาลตามสิทธิที่ลงทะเบียนไว้ แต่หากต้องการรักษาต่อที่เดิมต้องรับผิดชอบค่ารักษาเอง — if the criteria are not met, urgently coordinate with the hospital where your entitlement is registered; if you want to go on being treated where you are, you carry the cost yourself. The instruction to a refused patient is to fall back on a registration. Hold that word.
Widen the frame and it gets colder. Across the whole country and thirty-one months, April 2017 to October 2019, the institute puts 446,702 patients through the program and grants 58,484 of them. Thirteen per cent. Of the ones it granted, 84 per cent survived the critical illness, and around 300 complaints were filed across the whole period. Different denominator, different hospitals, different window. The two rates are not comparable and neither corrects the other. They agree on direction, and the direction is the finding.
As of 2019, no appeal. The institute’s own list of unresolved problems, in the document it submitted to the UN that year, includes the sentence: “Procedures of appealing over the results of categorizing medical urgency had not been addressed in any official documents.”
Which fund
Say the program authorises him. The hospital is now owed money by somebody, and the sentence that names the somebody is the one that matters.
Thailand has three public schemes, and they are entered by citizenship or by employment. Nothing in them is entered by being present, privately insured, or sixty-eight with a valid extension of stay.
| Fund | The population the instrument names | Share of Thais in it | Where the retiree stands |
|---|---|---|---|
| Civil Servant Medical Benefit Scheme | The population the instrument names Government employees plus dependents — parents, spouse and up to three children under 20. | Share of Thais in it 7% | Where the retiree stands The instrument names Thai government employees and their families. He is neither. |
| Social Health Insurance | The population the instrument names Private-sector employees, dependents excluded. Tripartite payroll contributions from employer, employee and the state. | Share of Thais in it 18% | Where the retiree stands This is the door the Thai-language explainers point foreigners at, and it opens on employment. The health system review adds that migrant workers reach it only between 18 and 55. The review states the age band for migrant workers specifically; it does not state one for the scheme in general. |
| Universal Coverage Scheme (the gold card) | The population the instrument names Everyone not covered by the other two. Access is through one registered contractor provider, "notably district health system". | Share of Thais in it 71% | Where the retiree stands The reform is described as having brought coverage to the whole Thai population. It is also the registration the handover sentence at hour 73 assumes he has. |
| Health Insurance Card Scheme (Ministry of Public Health) | The population the instrument names Issued under an announcement titled Health Examination and Health Insurance of Foreign Workers B.E. 2562. Premium THB 1,600, plus 500 for the check-up, through the One Stop Service. | Share of Thais in it — | Where the retiree stands The only door with the word foreign written on it, and the word after it is workers. The same review also calls the scheme open to all migrants excluded from social security. Which of those two readings covers a non-working retiree is not settled anywhere in the published record. Reported as unresolved rather than resolved by assertion. Sits outside the three-scheme percentages, hence the dash. |
| Compulsory motor insurance | The population the instrument names Anyone injured by a vehicle, under the Road Accident Victims Protection Act B.E. 2535. Medical costs up to THB 30,000; THB 35,000 preliminary compensation where severe damage results in disability. | Share of Thais in it — | Where the retiree stands Open, and it is the one fund on this table that never asks which scheme he belongs to. It runs out at roughly one published private intensive-care day. Road traffic injury is among the top three causes of medical emergency in Thailand; the review records over 435,000 people injured in vehicle accidents in 2022 alone. |
Source: WHO Asia Pacific Observatory, The Kingdom of Thailand Health System Review (Health Systems in Transition Vol. 13 No. 1, 2024), Table 3.4 and §§5.5.2, 5.5.4, 5.13.3 · checked 2026-08-20
There is a stranger number underneath all this. The 2018 study asked whether which scheme a patient held was associated with whether the program said yes, adjusting for age, sex, mode of arrival, blood pressure, respiratory rate and Glasgow coma score. It was. Gold-card patients had 2.33 times the odds of authorization that civil servants did; social-security patients, 1.77 times.
The authors read that as success, and on their terms it is: the policy pulled emergency access away from the most privileged scheme and toward the least, which is exactly what it was built to do. Read it from the other end and the same number says that a clinical question, is this man’s emergency critical, came back correlated with his payer after the vital signs had been controlled for. Both readings are true of one coefficient. The man with no Thai payer at all is not a row in that regression — and the nearest thing to him was in the data and did not survive it. Of the 19,785 patients whose insurance status the study tabulates, 373, or 1.89 per cent, sit in a category called Others. The model sets three schemes against each other. Others is not one of the three.
Seventy-two hours, or sooner
Everybody in this market argues about hour 72. Almost nobody quotes the sentence, and the sentence has an or in it.
The institute’s own wording is that no fee is charged “within the first 72 hours of their first admission or until their conditions are fit for being transferred to their registered hospitals.” Bangkok Hospital states it correctly on one of its own pages, dated 29 January 2026: the policy runs until patients “are out of critical condition and can be safely transferred, for a period of up to 72 hours”. Of the seven English-language pages scored in the survey behind this file, it is the one that does.
Up to. The window closes on whichever edge arrives first, and neither edge is set by how much help the patient still needs. Treatment that works fast closes it early. Treatment that does not work at all leaves him inside it when the clock runs out.
hours from first admission
| When · hours from first admission | What | Duration / until | Note | Flag |
|---|---|---|---|---|
| When · hours from first admission 0 | What The covered window | Duration / until 72 (to 72) | Note Opens at first admission. Closes on whichever comes first: fitness for transfer, or hour 72. | Flag — |
| When · hours from first admission 0 | What Pre-authorization | Duration / until — | Note A program grades the patient, not the treating doctor. 11,776 of 20,206 requests refused in the published cohort. | Flag warn |
| When · hours from first admission 72 | What Hour 72 | Duration / until — | Note The outer edge. The inner edge has no fixed hour and can arrive on day one. | Flag deadline |
| When · hours from first admission 96 | What Median stroke stay, countrywide cohort | Duration / until — | Note Four days across 1,222 acute ischaemic stroke patients, mean age 65, admitted 2008 to 2010; a prospective multicentre study, not a registry analysis. | Flag — |
| When · hours from first admission 146 | What Mean ischaemic stroke stay, claims data | Duration / until — | Note 6.1 days across the 192,414 ischaemic strokes inside a Thai national claims cohort of 386,484 first-ever strokes, 2017 to 2020. | Flag warn |
| When · hours from first admission 238 | What Mean haemorrhagic stroke stay | Duration / until — | Note 9.9 days across the 78,081 haemorrhagic strokes in the same cohort. | Flag — |
Source: NIEM UCEP executive summary (UNPSA, 2019); Nilanont et al., J Stroke Cerebrovasc Dis 2014; Kumluang et al., BMJ Open 2023;13:e072259 · checked 2026-08-20
Stroke is one of the seven findings that authorise automatically. It is also, in a Thai national claims cohort of 386,484 first-ever strokes between 2017 and 2020, a mean admission of 6.1 days across the 192,414 ischaemic cases and 9.9 days across the 78,081 haemorrhagic ones. The event most certain to get a man through the gate is the event most certain to outlast what is behind it. Those are public-system means and the authors note they run short against international figures of ten to forty days, which only sharpens the point: even the compressed version is double the window.
The people who clear the program are not people who go home. The study followed a slice of its cohort out of the emergency department: 3,620 records, of which 1,375 have no outcome recorded at all, a monitoring gap the authors flag themselves. The window is unclear in the paper itself, whose methods describe reviewing 3,355 records from April to June against a table that totals 3,620 and states no window. Of the 2,245 that do carry one: 1,000 went to intensive care, 648 to a ward, 367 were moved on to a bigger hospital, 195 died in the department. Twenty went home. Fifteen declined treatment. The guarantee attaches, by design, to exactly the admissions that will still be running at hour 73.
And the institute knows the boundary is undefined. In the same UN submission, under issues for improvement, it asks itself: “What is the scope of the 72 hours-period? If the patients recover from the criticality before 72 hours of admission, does the policy cover them?” That is the operator of the scheme, in its showcase document, unable to state where its own product stops.
Hour seventy-three
Every official sentence about where he goes next reaches for the same word, and the word is registered.
The institute: no fee is charged until the patient’s condition is “fit for being transferred to their registered hospitals.” The World Health Organization’s review of the same policy: during the covered period “the patient can be transferred to their registered hospital if ongoing inpatient treatment is required.” The consumer council’s instruction to the patient the screen has just refused, three paragraphs back: coordinate with the hospital where the entitlement is registered.
A registration is not an address. It is a membership artefact, and the health system review’s own table of the three schemes says so in the column headed access to service. Gold-card members reach care through a “registered contractor provider, notably district health system”. Social-security members reach it through “registered public and private competing contractors”. Of the three, the civil-service scheme is the one that requires no registration at all, and it is the scheme with 7 per cent of the population in it.
So hour 73 does not throw a foreigner out of a hospital. It hands him to a sentence with no object. He can be moved to a public hospital, and he will be, but he arrives as a self-paying stranger rather than as a returning member, which is a different transaction wearing the same ambulance.
Meanwhile the hospital he is lying in has been recovering about half of what it would otherwise have charged him. The fee schedule pays a private hospital roughly 50 per cent of hospital charge, which researchers funded by the institute give as one of the three reasons private hospitals accepted the policy at all. The pilot it replaced paid about a third. Charge is not cost, and what the hour actually costs a hospital to deliver is not on any public record, so nothing here says the hour is loss-making. What is on the record is a published rate, a discount against it, and a clock. The party holding the clock is the party taking the discount.
At hour 73 the discount ends, and the room he is in starts appearing on a bill at the rate the hospital publishes.
What the state pays for the life it saves
Divide the money by the people and the guarantee has a price on it.
Around 1.2 billion baht across 58,484 granted patients, April 2017 to October 2019, from the operating institute's own published totals. One published day in a Bangkok private intensive-care unit, effective January 2026, is 25,500.
That covers everything from a stabilised faint to a ventilated three-day stay, so it is a mean over unlike episodes rather than a tariff. It is also 2017-to-2019 money against a January 2026 rate, which is a real objection and worth removing rather than managing. Weight the Thai consumer price index across the months the payments ran and the covered period sits at 112.4, against 122.4 for 2025. Restated, the mean payment is 22,348 baht and the published intensive-care day is still 1.14 times it. Two things about that method, both stated because they cut against it: the index stops at 2025 while the rate starts in January 2026, though it fell 0.13 per cent in that last year; and a headline consumer index is not a medical price index, so if Thai medical prices outran the headline the restated payment would be larger and the gap narrower. This is the version most generous to the scheme.
The period-clean anchor points the same way. In the national payment guideline for the current fiscal year, the public emergency line pays 8,350 baht per adjusted relative weight in-zone, 9,600 across zones, with 728 priced items appended to the announcement.
Set that beside the illness. In Thai claims data a first-ever stroke costs the system a mean of 37,179 baht for the entire first year, and 40,111 if the patient is in his seventies. Deflated the same way, 40,512 and 43,706. A whole year of Thai stroke care, drugs and readmissions and rehabilitation together, is under two days at the published rate of the private intensive-care bed the ambulance would take a foreigner to.
That same UN submission carries seven unresolved issues and five policy recommendations. Recommendation three reads: “Establish a central fund to perform reimbursement to hospitals when some healthcare funds are not affordable.” The document is dated 27 November 2019. Whether the fund for the patients whose fund cannot pay now exists is not a question any reachable Thai government source could be asked on 20 August 2026.
Three hundred and ninety-nine pages
The National Health Security Office publishes a guideline every year telling providers how to claim. The one for fiscal year 2026 was issued in October 2025 and runs to 399 pages. It names entitled populations line by line, service by service. On the row headed ผู้มีสิทธิ, the entitled, against the public-facility emergency line it names two: national health security entitlement, and the entitlement of insured persons with disability.
Page 13 is worth reading in full rather than in part, because it prints this year’s rule beside last year’s in two columns. Under fiscal 2569 the emergency line is assessed on the institute’s own criteria for sorting patients in an emergency room, หรือตามตามที่กระทรวงกำหนด, or as the ministry prescribes. Under fiscal 2568, in the column immediately to its left, that same sentence ended หรือตามดุลยพินิจของแพทย์ที่รักษา: or at the discretion of the treating physician. A named route through a doctor’s judgement was in the rulebook a year ago and is not in it now. This is the public-facility line rather than the private one, and it is one revision on one slide, so it is a direction of travel rather than a finding about the whole scheme. But somebody wrote the doctor out on purpose, and dated it.
Extract the whole document and count. ผู้ป่วย, patient, appears 544 times. ประกันสังคม, social security, 46. ต่างชาติ and ต่างด้าว, the two ordinary Thai words for a foreigner, appear zero times between them across all 399 pages. The phrase สัญชาติไทย, Thai nationality, appears exactly once, and it is a parenthesis inside the social-security line: ผู้ประกันตนที่ไม่ได้มีสัญชาติไทย, insured persons who do not hold Thai nationality.
This office runs one of the three funds, so a rulebook about its own beneficiaries naming its own beneficiaries proves nothing about intent, and none is claimed. What it shows is how the machinery is shaped. In four hundred pages of a document about who gets paid for treating whom, a non-Thai enters the text exactly once, through the employment door, as somebody’s employee.
He is sixty-eight. He does not have an employer. The stroke does not check first.
Everything he can establish in advance sits on the private side of the fence: what a Thai hospital will put a price on before he needs it, the insurance mandate that rides on one of the retirement visas, what the roads do to men his age. The guarantee sits on the other side, and it is written about somebody else.
Seventy-two hours is how long two other parties have agreed to argue about who pays for his bed. He was never one of the parties.