The Thai Non-O retirement extension does not require medical insurance. The O-A and O-X visas do; the Non-O does not. At 75, with Cigna Senior Plan premiums in the USD 6,469 to 9,430 a year band and the IMG and SafetyWing doors closed for new applicants, the Non-O is the path more expats take than the brochures admit. The premium is now zero. The cost is not.
The cost has moved — it sits on the policyholder’s balance sheet as a reserve. The size of that reserve is the question the brochure has never written down, because the brochure is selling the alternative. The reserve has four sizes, depending on which scenario lands.
The four scenarios, in published components
Build the reserve from components nobody disputes, and the table builds itself. Bangkok Hospital publishes its 2026 cardiac tariffs. Bumrungrad publishes its 2026 ICU rate. Chiang Mai dementia-care facilities publish their monthly rates. Travel Care Air publishes its 2026 medevac guidance. The Alzheimer’s Association publishes survival distributions. The integration is the part nobody publishes.
| Scenario | Floor (low) | Floor (high) | Horizon assumed | Primary cost driver |
|---|---|---|---|---|
| Optimistic, no catastrophe | Floor (low) $30,000 | Floor (high) $60,000 | Horizon assumed 5–10 yrs | Primary cost driver Outpatient and meds at USD 5–12k/yr |
| Central, one major event | Floor (low) $40,000 | Floor (high) $100,000 | Horizon assumed Single year + recovery | Primary cost driver One stroke or cardiac event + 6 mo recovery |
| Adverse, chronic care | Floor (low) $200,000 | Floor (high) $500,000 | Horizon assumed 5–8 yrs | Primary cost driver Chiang Mai dementia at USD 3,100/mo, compounding |
| Tail, multi-event | Floor (low) $350,000 | Floor (high) $700,000 | Horizon assumed 5–10 yrs | Primary cost driver Cancer + cardiac + LTC + repatriation |
The bands are not quotes. They are floors built from published components at 2026 prices, with the medical-inflation compounder taken at Aon’s 8.9% Asia net trend forecast for 2026. FX is held flat at USD/THB 32.7. Both assumptions are charitable. The actual reserve has to ride out both moving.
Single events are bounded. The LTC tail is not.
A stroke at a Bangkok private hospital, self-pay all-in across the surveyed centres, runs USD 15,000 to 60,000. Bangkok Hospital lists coronary angiography plus PCI plus one stent at THB 299,000 to 322,000. On the same hospital’s cardiovascular surgery tariff, off-pump or on-pump CABG runs THB 1,173,000 to 1,437,500, with minimally invasive MICS CABG at THB 1.5 million. Cancer surgery plus chemotherapy or radiation, USD 20,000 to over USD 100,000.
These are large numbers. They are also published numbers. The hospital prints the tariff; the policyholder pays it; the event resolves. The bound on the cost is the hospital’s own price list.
Long-term care does not work that way. Every IPMI policy in the SE Asia individual market excludes nursing-home and care-home stays through the convalescence exclusion. The Cigna Senior Plan’s home-nursing cap at USD 2,500 a year is the highest IPMI ceiling I found and it covers two to three weeks of part-time Bangkok care before it runs out. Beyond that ceiling, the policyholder pays.
The Chiang Mai dementia-care central rate from our own wage-curve ledger is THB 110,000 a month at full care across Vivobene, Care Resort, and Baan Lalisa, about USD 3,100. The band runs THB 78,000 to 150,000. The rate compounds at the Thai minimum-wage cadence plus inflation, which the wage curve projects forward at 6 to 8 percent nominal.
The Alzheimer’s Association puts median survival from diagnosis at four to eight years, with a tail to twenty. At eight years compounding from USD 37,200 at 8.9%, the cumulative is roughly USD 395,000. At the upper care-tier rate of THB 150,000 a month, the same eight-year run lands USD 540,000. These are central-Thailand floors, not Manila or Cebu, where the rate is roughly sixty percent of Thai pricing. The same carriers are also re-pricing across both markets.
This is the structural fact the four-scenario table is built around: single events live inside the hospital’s price list — the LTC tail lives outside it. The reserve is sized for the tail or it is not sized at all.
The probability the tail is the base case
US lifetime risk data from ASPE puts the probability a 65-year-old develops severe long-term services and supports needs before death at 70%. Forty-eight percent receive paid care. Twenty-four percent receive paid care for two or more years. Mean duration of severe need is about 2.2 years. Women’s lifetime severe-need probability is 75%, men’s 64%.
These are US numbers. They are the only published actuarial set with this breakdown, and the expat in Thailand carries the same risk distribution because they carry the same body. The Genworth and CareScout cost data is US-specific (the 2025 CareScout survey put a private US nursing-home room at USD 10,798 a month, USD 129,575 a year). The probability of needing the room is portable. Only the per-month rate changes by geography.
Combine the US lifetime risk with the Alzheimer survival tail and the implied reserve sizing logic is uncomfortable. A 75-year-old self-insurer faces something close to a one-in-four probability of needing two-plus years of paid care over the remaining horizon. The central reserve case is not the optimistic band. The central reserve case is the adverse band.
The withdrawal-rate compounder
Morningstar’s 2025 retirement income research puts the safe withdrawal rate for a 75-year-old on a 20-year horizon at 5.3%, against the base 30-year SWR of 3.9%. The shorter horizon is what unlocks the higher rate.
That rate breaks under chronic-care draw. A USD 300,000 fund at 75 draws USD 15,900 a year at SWR. Chiang Mai dementia care draws USD 37,200 a year in year one and compounds. The fund has to support 2 to 3 times its SWR draw rate in the adverse scenario, against a medical trend running 4 to 7 times faster than US CPI. Aon puts APAC net at 8.9%; WTW puts Philippines at 18.3% for 2025, the second-highest in the region.
The compounder is the part the reserve cannot beat with equity allocation. A diversified portfolio earning 6% real is below the 8.9% medical compounder. Every year the diagnosis is delayed, the fund grows a little; the cost of the diagnosis grows more.
FX is the second compounder
USD/THB has been broadly range-bound between 29 and 37 across 2013 to 2024.
| Date | THB per USD | Range | Basis | Note |
|---|---|---|---|---|
| Date 2013 | THB per USD 30 | Range 29–30 | Basis triangulated | Note dipped sub-30, the band's strong-baht end |
| Date 2022 | THB per USD 36.5 | Range 36–37 | Basis triangulated | Note weak-baht end of the band |
| Date 2026-05 | THB per USD 32.68 | Range 30.85–33.5 | Basis sourced | Note May spot; 52-week range; down ~2% on the year. Re-read 24 May ~32.7 and 26 May ~32.5 (range 32.40–32.68), within range; value held. |
| Date 2026-08 | THB per USD 33.06 | Range — | Basis sourced | Note Publish-clock re-verification (GIGA-PLAN 0.4). Up ~1.2% on the May reading and still inside the ~29–37 band the direction_note describes — a decade of range, not a trend. FRED DEXTHUS, the series source of record, would not serve from this environment on 8 Aug 2026, so the reading was taken from the ECB reference set and cross-checked against a second independent feed. |
Source: Federal Reserve EXTHUS (Thai baht to U.S. dollar spot) · latest 33.06 THB per USD (2026-08) · as of 2026-08-08
The reserve held in USD does not earn a tailwind from this; it earns no tailwind at all. The per-month dementia rate in baht is what compounds, and the dollar value is whatever the spot rate gives on the day the bill is paid.
USD/PHP has weakened from 45 in 2010 to 61.7 in May 2026.
| Date | PHP per USD | Range | Basis | Note |
|---|---|---|---|---|
| Date 2010 | PHP per USD 45.1097 | Range — | Basis sourced | Note annual average |
| Date 2013 | PHP per USD 42.4462 | Range — | Basis sourced | Note Annual average, and the local minimum of the modern series — the peso was stronger in 2012–13 than in 2010. Added 2026-08-08 because three sidecars cite a 2013 baseline the ledger did not carry. |
| Date 2015 | PHP per USD 45.5028 | Range — | Basis sourced | Note annual average |
| Date 2026-05 | PHP per USD 61.7 | Range 55.1–62 | Basis sourced | Note mid-May spot; 52-week range. Re-read 24 May ~61.6 and 26 May (BSP ref 61.45, market ~61.57), within range; value held. |
| Date 2026-07 | PHP per USD 61.596 | Range — | Basis sourced | Note Monthly average; end-July 61.432. First full month re-verified against BSP directly rather than a rate-aggregator mirror. |
| Date 2026-08 | PHP per USD 60.981 | Range — | Basis sourced | Note Month-to-date average through 7 August 2026, not a full month — the peso has firmed roughly 1% off the July mean. Supersede with the full-month average once BSP publishes it. |
Source: Bangko Sentral ng Pilipinas — Table 12, Philippine Peso per US Dollar Exchange Rate (monthly average and end-of-period) · latest 60.981 PHP per USD (2026-08) · as of 2026-08-08
A reserve held in USD against PHP-priced care has earned a 37% nominal tailwind over 16 years. That tailwind has been eaten by 18.3% Philippine medical inflation. The dollar bought more pesos and the pesos bought less care. The net for a US-pension self-insurer in Manila or Cebu is roughly the same as in Bangkok in real terms.
The reserve has to be sized in real terms against the destination’s cost curve, not the home currency’s drift. The brochures conflate the two routinely.
What “self-insurance” actually is in the industry literature
IRMI’s definition names self-insurance as a method “rarely used by individuals” because individuals “rarely have funds large enough to cover large uncertain risks.” Catastrophic loss, in IRMI’s reference work, is “rarely self-insured.”
That is the industry reading. The self-insurer past 75 in Thailand is doing what IRMI says individuals rarely do, against a risk distribution IRMI says is rarely self-insured against. The naming is honest about what the structure is: a position taken against the carrier’s optionality, with the policyholder’s balance sheet absorbing the variance the carrier used to absorb.
The behavioural failure modes follow. A reserve held in liquid cash loses 8.9% a year in real terms. A reserve invested for growth has drawdown timing risk. A reserve held in joint accounts or transferred for estate-planning reasons becomes hostage to a relative’s life. A reserve sized at the optimistic floor is the central-case loss.
The break-even with IPMI
The Cigna Senior Plan premium at 70 to 79 (USD 6,469 to 9,430 a year, before the age-band step at 75) buys a USD 1,000,000 annual benefit cap, narrow LTC carve-outs notwithstanding. Compounded at 8.9% across a 15-year horizon, that premium stream sums to roughly USD 175,000 to 260,000 in nominal dollars.
The adverse reserve scenario is USD 200,000 to 500,000.
Cigna Senior Plan 70–79 band ($6,469–9,430/yr) compounded at Aon 2026 APAC 8.9% trend = ~$175–260k nominal in 15 yrs. The adverse reserve scenario at the same horizon is $200–500k. The IPMI premium is the price of optionality; the reserve is the price of holding it yourself.
These two numbers are close enough that the trade-off is not “self-insurance is cheaper.” It is “self-insurance is cheaper if the tail does not land, and ruinous if it does.” The IPMI policy is paying for the optionality the reserve has to carry instead. The premium is the price of not running the four-scenario table against your own balance sheet every year.
The premium is also climbing at 8.9% a year. So is the reserve requirement. They are climbing at the same rate because they are pricing the same underlying. The choice is which side of the table holds the risk.
The repatriation line
If the reserve does not survive the holding period, the US State Department does not fund repatriation. Neither does the UK Foreign Office. Thailand-to-USA body export runs about USD 3,060 for the air leg per the US Embassy guidance, USD 10,000 to 20,000 all-in. Thailand-to-UK runs about GBP 3,400 for the air leg, GBP 5,000 to 12,000 plus all-in. Manila-to-USA is similar in structure.
A reserve sized to the optimistic floor that gets eaten by an adverse-band diagnosis exits the country in a cargo crate at USD 15,000 if the family is solvent enough to pay it, and at a Bangkok or Manila crematorium if the family is not. That is not a moral observation. It is the published default behaviour of two sovereign governments and three private funeral logistics chains.
The line goes in the reserve or the family carries it.
What the four-scenario table changes
The Non-O extension is the rational choice for many expats at 75 because the IPMI premium curve from 75 to 85 is steeper than the reserve curve they think they are accepting. The premium is visible and the reserve is not. The brochure prices one side honestly and the table prices the other.
The four-scenario floor is the table the brochure was always implying. It is the price of optionality the carrier was selling. The reserve is the price of holding the optionality yourself, in 2026 prices, in cash, against a medical trend that compounds faster than the cash. The structural recommendation that follows is the same one this site keeps reaching: the decision had to be made earlier than 75. By 75, the choice is between two priced options, both worse than the choice that was available at 55.
The piece writes the reserve down so it can be carried with eyes open. It does not change which scenario lands. The lifetime risk distribution is what it is, the per-month Chiang Mai rate is what it is, the Aon trend is what it is. The arithmetic does not ask the policyholder to be more or less brave. It asks the policyholder to be more accurate about what was already being signed when the Non-O extension was filed without an insurance proof attached.