The Weight of a Card: How Taiwan Uses Its Institutions to Decide Who Pays for 23 Million People

On a Sunday evening, a 76-year-old man who was short of breath was brought by his son to the emergency department of a medical center. He swiped his NHI card (健保卡), and the nurse's screen instantly showed every medication he had taken over the past three months, the report of a chest X-ray taken at a clinic last week, and the diuretic another hospital had prescribed him six months earlier. The triage nurse assigned him triage level 2, and he was wheeled into the observation area. Three days later he was discharged; the out-of-pocket amount on his bill was a few thousand NT dollars, while the true cost of this hospital stay was well over a hundred thousand. Where did the difference in between go? Who decided he should lie in an acute ward rather than a chronic ward? Who decided how much the drug that saved him was worth? And if, after discharge, he could not stand up and needed someone to come to his home every day to bathe him, which pocket would that money come out of?

Clinical textbooks do not say a single word about these questions. But the national board exam tests them — and it tests them every year.

Because medical decisions in Taiwan have never been a standalone question of "what should be done medically"; they are always "what should be done medically" multiplied by "whom the system allows to pay for whom." Every order you write on the ward has an invisible financial line hanging behind it: it decides how many days this bed can be occupied, whether this drug needs prior authorization, and whether this patient will be caught or will fall through after discharge. Understand this line, and you will not merely have memorized a few more regulations — you will suddenly see why Taiwan's entire healthcare system has grown into the shape it has today.

In this volume, we will simply follow the money.

1. Pooling 23 Million People: The Foundation of National Health Insurance

Why It Had to Be "Mandatory," and Why It Had to Be "Single"

🔬 Mechanism, unpacked

Step one, the essence of insurance is "small contributions from the many, to catch the enormous misfortunes of the few." For this to work, the pool must hold healthy people and sick people at the same time. Step two, what happens if people are free to choose whether to insure? Young people who think they won't get sick drop out, and those left in the pool are all high-risk — premiums are forced up, so the next-healthiest also drop out, and premiums rise again. This downward spiral is called adverse selection, and it will strangle any voluntary health insurance to death. Step three, so for insurance to carry the high-risk, the law must force the low-risk to stay in as well — mandatory enrollment is not authoritarianism; it is an actuarial necessity. Step four, push the same logic up one level: if the market has many insurers, the healthy will be siphoned off by cheap plans, and the most expensive patients will be left to the last insurer standing — this is called risk selection. Step five, so Taiwan chose the most extreme solution: a single insurer. The National Health Insurance Administration (NHIA) alone collects all the money and pays all the bills; no one can cherry-pick customers, because there is no second pool to jump into.

This is the single-payer system. Its price is enormous administrative centralization; its benefits come down to three things: extremely low administrative costs (Taiwan's NHI administrative expenses have long hovered around 1% of premium revenue, compared with the 15% or more that is routine for U.S. commercial insurance), extremely strong bargaining power (one buyer facing all sellers — what it sets for drug prices and fee schedules is final), and data so complete it is almost frightening (the entire nation's medical records are concentrated in a single database, something that will come back to bite us later).

The Six Categories of Insured Persons: Classification Is Not Administrative Trivia — It Is a Map of "Who Pays for You"

Taiwan's NHI premium is not shouldered by one person alone; it is split three ways among the insured, the insuring unit (employer), and the government. Which category you belong to decides exactly how much each of the three pays.

Category	Main status	Insured	Insuring unit	Government
Category 1	Civil servants, public office holders	30%	70% (government as employer)	—
Category 1	Private school teachers and staff	30%	35%	35%
Category 1	Ordinary private-sector employees	30%	60%	10%
Category 1	Employers, the self-employed, professionals and technical specialists	100%	—	—
Category 2	Occupational-union members, seafarers employed on foreign vessels (no fixed employer)	60%	—	40%
Category 3	Members of farmers' associations, farmers and fishers (irrigation-association membership ended when the associations became the government Irrigation Agency in Oct 2020)	30%	—	70%
Category 4	Conscripted servicemen, substitute-service conscripts, military academy cadets, inmates of correctional facilities	—	—	100% (fully subsidized by the competent authority)
Category 5	Members of low-income households	—	—	100%
Category 6	Veterans, household representatives of veterans' surviving dependents	Self 0% / dependents 30%	—	Self 100% / dependents 70%
Category 6	Other regional population (without occupation)	60%	—	40%

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Contribution ratios verified 2026-07)

This table looks like an accounting exercise, but it is really saying one thing: Taiwan uses "whether you have an employer" as the main axis for allocating subsidies. If you have an employer, the employer shoulders the larger part for you; if you don't (Category 2 occupational unions, the Category 6 regional population), the government subsidizes only 40% and you carry 60% yourself — this is exactly the institutional reason occupational-union members feel the premium burden most heavily, and it is also the point exam questions love to poke: Category 2 and the Category 6 regional population both pay 60% themselves, double the 30% paid by ordinary employees.

How is the premium calculated? Salaried workers are insured according to the insured-salary grading table (not calculated dollar by dollar on actual pay, but slotted into a bracket), and the formula is:

General premium = insured amount × general premium rate × contribution ratio × (self + number of dependents)

For calculation purposes, dependents are counted up to a maximum of 3 — a ceiling set for large families. The current general premium rate is 5.17%, effective January 1, 2021 (verified 2026-07); and the National Health Insurance Act sets a statutory cap on the rate: it may not exceed 6% (verified 2026-07).

Mandatory enrollment is not authoritarianism but an actuarial necessity; a single insurer is not about efficiency above all, but about making sure no one can cherry-pick customers.

Second-Generation NHI: When Salary Is No Longer All of Your Income

🔬 Mechanism, unpacked

Step one, first-generation NHI charged premiums only on the "insured amount" (salary). Step two, that was fine in an era when salary equaled income, but when someone lives off dividends, rent, or professional-practice income, his NHI premium is still paid at the lowest bracket. Step three, this created a strong sense of unfairness: an office worker earning NT$50,000 a month might pay a higher premium than an asset holder collecting millions in dividends each year. Step four, the answer given by second-generation NHI (二代健保), implemented on January 1, 2013, was not to redesign the premium base but to "add a second cut" outside of salary — levying an additional supplementary premium on six categories of non-regular income. Step five, so the essence of second-generation NHI is a "patch," not a "rewrite": it kept the original salary-based architecture and merely clawed back part of the income that had slipped through.

The current supplementary premium rate is 2.11% (adjusted on January 1, 2021, in step with the general rate; verified 2026-07). The items subject to withholding and their thresholds are as follows:

Item withheld	Withholding threshold	Cap per payment
Large bonuses	The cumulative portion exceeding 4 times the month's insured amount	NT$10 million
Part-time salary income	A single payment at or above the basic wage	NT$10 million
Professional-practice income	A single payment of NT$20,000 or more	NT$10 million
Dividend income	A single payment of NT$20,000 or more	NT$10 million
Interest income	A single payment of NT$20,000 or more	NT$10 million
Rental income	A single payment of NT$20,000 or more	NT$10 million

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Items and thresholds verified 2026-07)

The other half is often overlooked: the insuring unit (employer) must also pay a supplementary premium — calculated on the difference between the total salaries paid each month and the total insured amounts of its employees, at the same rate. The design intent of this rule is plain: to close off the room for maneuvers like "under-reporting the insured amount and stuffing salary into bonuses."

🦦

PipSo the supplementary premium is "2.11% skimmed off every bit of non-salary income," right? Then my mom's savings interest — a few thousand a year — gets docked too?

🐻‍❄️

BarnabyYou almost fell into the trap. The supplementary premium is withheld only when a single payment reaches the threshold, not on an annual total. For interest, dividends, rent, and professional-practice income, the threshold is NT$20,000 per payment; for bonuses it is the portion exceeding 4 times the month's insured amount; for part-time salary the threshold is the basic wage. Below the threshold, nothing is withheld. And one line people often forget: the cap is NT$10 million — anything above it is no longer charged — so the supplementary premium is actually regressive, which is exactly what it is most often criticized for.

Copayment: Why the Patient Is Made to Feel a Little Pain

🔬 Mechanism, unpacked

Step one, once insurance drives the price down to near zero, utilization inevitably rises — this is called moral hazard, and it does not mean patients are immoral; it describes a purely economic phenomenon: when marginal cost is zero, demand swells to the point where marginal benefit is also zero. Step two, so almost every health insurance scheme keeps a small slice of out-of-pocket payment, so that users feel a price signal at every visit. Step three, but if the out-of-pocket share is too heavy, what it blocks is often not unnecessary care but the necessary care of the poor — this is the greatest ethical tension in copayment. Step four, Taiwan's solution has three layers — "tiered differentiation + exemptions for the vulnerable + annual caps": price differences between levels steer patient flow, status-based exemptions protect the vulnerable, and cumulative caps prevent catastrophic expenditure. Step five, so when you see a copayment question, ask three things first: which level, what status, and was there a referral.

The design of inpatient copayment rates is itself a lesson in mechanism — the longer you stay, the higher your out-of-pocket share, because what the system wants to push is "acute phase in the acute ward, transfer out once stable":

Ward type	Length of stay	Out-of-pocket rate
Acute ward	Within 30 days	10%
Acute ward	Days 31–60	20%
Acute ward	Day 61 onward	30%
Chronic ward	Within 30 days	5%
Chronic ward	Days 31–90	10%
Chronic ward	Days 91–180	20%
Chronic ward	Day 181 onward	30%

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Acute-ward rates verified 2026-07; chronic-ward rates also verified (NHI Act Article 47, 2026-09))

And the new copayment scheme that took effect on July 1, 2023, is the adjustment most often tested in recent years (verified 2026-07):

Item	Medical center	Regional hospital	District hospital	Primary-care clinic
Emergency copayment	NT$750	NT$400	NT$150	NT$150
Emergency (lower-middle-income households / disability-certificate holders)	NT$550	NT$300	NT$150	NT$150
Outpatient drugs (drug cost NT$100 or less)	NT$10	NT$10	Waived	Waived
Outpatient drugs (drug cost NT$101 or more)	20%, capped at NT$300	20%, capped at NT$300	20%, capped at NT$200	20%, capped at NT$200

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Those exempt from copayment form a must-know list: catastrophic illness (重大傷病), childbirth, care received in mountain and offshore-island areas, low-income households, veterans and household representatives of veterans' surviving dependents, children under 3, Labor Insurance enrollees seeking care for occupational injury or illness, and tuberculosis patients seeking care as ordered by a physician, among others. There are also two further safeguards — a per-admission cap and an annual cumulative cap (amounts announced each year by the competent authority; specific amounts ⚠️ pending verification).

⭐ Board-exam essentials

Chapter 1 Must-Knows

NHI launched on 1995/3/1, integrating the medical benefits of the thirteen existing social insurance schemes; single insurer = the National Health Insurance Administration (NHIA).
The reason for mandatory enrollment is to prevent adverse selection; the reason for a single insurer is to prevent risk selection.
Premiums are split three ways: employee 30% / employer 60% / government 10%; employers and the self-employed pay 100% themselves; Category 2 and the Category 6 regional population pay 60% themselves; Category 5 low-income households are fully covered by the government.
General premium = insured amount × rate × contribution ratio × (self + dependents, dependents counted up to 3); current rate 5.17%, statutory cap 6%.
Second-generation NHI implemented 2013/1/1; supplementary premium rate 2.11%, six withholding items: bonuses (above 4 times the insured amount), part-time salary (at or above the basic wage), professional practice, dividends, interest, rent (single payment of NT$20,000 or more), cap per payment NT$10 million → because of the cap, the supplementary premium is regressive.
The employer-side supplementary premium is charged on "the difference between total salaries and total insured amounts," to prevent under-reporting.
The theoretical basis of copayment is moral hazard; its downside cost is suppressing necessary care among the vulnerable.
Inpatient copayment: acute ward 10% within 30 days, 20% for days 31–60, 30% from day 61 (the longer the stay, the heavier the out-of-pocket share — the incentive is to push transfers out).
New scheme of 2023/7/1: emergency copayment NT$750 at medical centers, NT$400 at regional hospitals; the outpatient-drug cap at medical centers / regional hospitals raised to NT$300.
Exempt from copayment: catastrophic illness, childbirth, mountain and offshore-island areas, low-income households, veterans, children under 3, occupational injury or illness, tuberculosis.

♪ Sing-along text — English rendering of the Mandarin song lyrics

Before the first of March, nineteen ninety-five, Taiwan had thirteen separate social insurance medical benefit schemes, each running on its own; who you were and where you worked decided whether anyone would pay for you when you fell ill, and roughly forty percent of Taiwan's population had no medical insurance of any kind. On that day, everyone was pushed into the same pool. Why did it have to be mandatory, and why did it have to be single? Because the essence of insurance is small contributions from the many catching the enormous misfortunes of the few, and for that to work, the pool must hold healthy people and sick people at the same time. If people are free to choose, young people who think they won't get sick drop out first, those who remain are all high-risk, premiums are forced up, the next-healthiest drop out in turn, and premiums rise again; this downward spiral is called adverse selection, and it will strangle any voluntary health insurance. Push the same logic up one level: if the market has many insurers, the healthy are siphoned off by cheap plans and the most expensive patients are left to the last one — this is called risk selection. Taiwan's solution was to have only one insurer: the National Health Insurance Administration alone collects all the money and pays all the bills, and no one can cherry-pick customers, because there is no second pool to jump into.

Mandatory enrollment blocks adverse selection, a single insurer blocks risk selection — neither is authoritarianism; both are actuarial math.

The premium is not shouldered by one person alone but split three ways among the insured, the insuring unit, and the government, and which category you fall into decides exactly how much each of the three pays. Civil servants pay thirty percent themselves, with the government paying seventy percent as their employer; private school teachers and staff pay thirty percent, the school thirty-five percent, and the government thirty-five percent; ordinary private-sector employees pay thirty percent, the employer sixty percent, and the government ten percent; but employers themselves, the self-employed, and professionals and technical specialists must pay the full amount. Category Two occupational-union members pay sixty percent, with the government subsidizing forty percent; Category Three farmers and fishers pay thirty percent, with the government subsidizing seventy percent; in Category Four, conscripted servicemen and inmates of correctional facilities are fully subsidized by the competent authority; Category Five low-income households are fully covered by the government; in Category Six, veterans themselves are fully subsidized by the government while their dependents pay thirty percent, and the other regional population pays sixty percent with the government paying forty. The whole table is really saying one thing: Taiwan uses whether you have an employer as the main axis for allocating subsidies — people with an employer have the larger part carried by that employer, while people without one are subsidized only forty percent, and this is the institutional reason occupational-union members feel the heaviest premium pressure. The premium is calculated as the insured amount times the rate, times the contribution ratio, times the number of people — yourself plus your dependents — with dependents counted up to three; the current general rate is five point one seven percent, and the statutory cap is six percent.

Mandatory enrollment blocks adverse selection, a single insurer blocks risk selection — neither is authoritarianism; both are actuarial math.

Second-generation NHI took effect on New Year's Day, two thousand thirteen, and it dealt with an awkward problem: first-generation NHI charged premiums only on salary, so when someone lived off dividends, rent, or professional-practice income, his premium was still paid at the lowest bracket. Second-generation NHI did not rewrite the premium base; it chose to add a second cut outside of salary, levying an additional supplementary premium on six categories of non-regular income, at a current rate of two point one one percent. Large bonuses are counted on the portion exceeding four times the month's insured amount, part-time salary has a threshold of a single payment at or above the basic wage, and professional practice, dividends, interest, and rent are withheld only when a single payment reaches twenty thousand NT dollars — and the chargeable cap on each payment is ten million dollars. Because of that cap, the supplementary premium is actually regressive, which is exactly what it is most often criticized for. The other half is often forgotten: employers must also pay a supplementary premium, calculated on the difference between the total salaries paid each month and the total insured amounts of their employees, with the aim of closing off the maneuver of under-reporting insured amounts and stuffing salary into bonuses.

The theoretical basis of copayment is moral hazard: when insurance presses the price down to near zero, utilization inevitably swells to the point where marginal benefit is also near zero, so a small slice of out-of-pocket payment is kept to let people feel a price signal. But an out-of-pocket share that is too heavy often blocks not unnecessary care but the necessary care of the poor, so Taiwan balances it with three layers: tiered differentiation to steer patient flow, status-based exemptions to protect the vulnerable, and annual cumulative caps to prevent catastrophic expenditure. The design of inpatient copayment is itself a lesson in mechanism: in an acute ward you pay ten percent within thirty days, twenty percent from day thirty-one to sixty, and thirty percent from day sixty-one on, so the longer you stay the more you pay, because what the system wants to push is acute phase in the acute ward, transfer out once stable. The new scheme of the first of July, two thousand twenty-three, set the emergency copayment at medical centers to seven hundred fifty dollars and at regional hospitals to four hundred, kept district hospitals and clinics at one hundred fifty, and raised the out-of-pocket cap on outpatient drugs at medical centers and regional hospitals to three hundred dollars. Know the exemption list by heart: catastrophic illness, childbirth, care in mountain and offshore-island areas, low-income households, veterans and household representatives of veterans' surviving dependents, children under three, Labor Insurance care for occupational injury or illness, and tuberculosis care as ordered by a physician.

2. How One Dollar Becomes Ninety Cents: Global Budgets, Point Values, and the Evolution of Payment Systems

That is the prisoner's dilemma showing its face inside NHI, and its source is something called the global budget payment system.

From Fee-for-Service to Global Budget: Change the Incentive, Change the Whole Pattern of Behavior

🔬 Mechanism, unpacked

Step one, in its early years NHI used fee-for-service (FFS): provide one service, claim one set of points, collect one payment. Step two, the incentive in this design is crystal clear — the more you do, the more you get — so it systematically drives up service volume (supplier-induced demand), and medical spending bolts like a runaway horse. Step three, there are two ways to hit the brakes: control price, or control total volume. Controlling price gets offset by rising volume (you cut the unit price, I do twice as much), so what really works is controlling the total. Step four, hence the global budget payment system: first agree on how much to spend in a year, then go back and divide it up. Step five, the key design is this — the total is fixed, but the points are not. How many points providers claim floats, so "how much each point is worth" becomes the result of a division: payment per point = the sector's annual global budget ÷ the sector's total points actually claimed. Step six, so once the whole profession pushes volume together, the denominator grows and the point value is diluted — you claim 100 points but receive only NT$87. That is the slide.

Taiwan's global budget did not arrive all at once; it was rolled out sector by sector (verified 2026-07):

Sector	Implementation date
Dental outpatient global budget	July 1998 (ROC year 87) — the first brave enough to try it
Traditional Chinese medicine outpatient global budget	July 2000 (ROC year 89)
Western-medicine primary care global budget	July 2001 (ROC year 90)
Hospital global budget	July 2002 (ROC year 91) — full implementation from this point

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Taiwan uses an expenditure cap, not an expenditure target. The difference between the two is the core of exam questions: under an expenditure cap, the global budget is a hard ceiling, and excess service volume is absorbed through point-value dilution (the risk falls on providers); under an expenditure target, when volume overshoots, the point value is only partly discounted and the insurer still has to pay somewhat more (the risk is shared by both sides). Taiwan chose the former, which amounts to outsourcing "total volume control" to the self-discipline of the medical profession — and that is exactly the source of the young attending's confusion: self-discipline is unstable in game-theoretic terms, because those who restrain themselves don't get the benefit of their restraint.

The true genius and the true cruelty of the global budget system are the same thing: it never tells you that you can't do something; it simply turns the cost of "everyone doing more together" into everyone's point value falling together.

Diagnosis-Related Groups: Turning "One Hospital Stay" into a Unit of Pricing

🔬 Mechanism, unpacked

Step one, the settlement unit of fee-for-service is "one service," so the incentive is to do more items. Step two, what if you enlarge the settlement unit to "one hospital stay"? — no matter how many tests you order or how many days the patient stays, the same diagnosis-related group pays the same amount. Step three, the incentive flips immediately: cost becomes the hospital's own loss, so hospitals find ways to shorten length of stay and cut unnecessary tests. Step four, but a flipped incentive always has side effects: grabbing mild cases while avoiding severe ones (cream skimming), diagnostic upcoding (DRG creep — coding patients into groups that pay more), and readmissions caused by premature discharge. Step five, so the larger the settlement unit, the stronger the incentive to economize, but the stronger the incentive for risk selection as well — this spectrum runs from fee-for-service, to per diem, to per case (DRG), all the way to capitation, with the economizing incentive rising monotonically and the risk-selection incentive rising monotonically too.

The rollout history of Taiwan's version of diagnosis-related groups (Taiwan Diagnosis Related Groups, Tw-DRG) is a policy story in itself (verified 2026-07):

January 1, 2010: phase one introduced 164 DRGs.
July 1, 2014: phase two introduced 254.
Phase three (originally planned to expand further to more than a thousand groups): postponed after strong backlash from the medical profession and the Legislative Yuan; the NHIA has since studied expanding DRG coverage again, but no full rollout had been announced as of September 2026, so check the NHIA's latest announcements.

Tw-DRG has exclusion clauses and outlier handling: excluded cases (cancer, psychiatric, rare-disease, stays over 30 days, ECMO use, and others) are paid fee-for-service, while high-cost outliers above the upper threshold receive the fixed DRG payment plus 80% of the amount above that threshold (cases below the lower threshold are paid as actually claimed), precisely to suppress the side effect of "severely ill patients being bounced from hospital to hospital."

Pay-for-Performance: Paying for "Doing It Right," Not for "Doing More"

Fee-for-service pays for quantity, DRGs pay for efficiency — so who pays for quality? The answer is pay-for-performance (P4P). NHI has piloted P4P since November 2001 (verified 2026-07), initially targeting a few chronic diseases and conditions requiring long-term follow-up — the logic being that outcomes for these diseases lie not in any single clinic visit but in whether the patient is continuously followed up and whether the indicators hit their targets, so payment should be tied to process indicators and outcome indicators rather than to the number of visits. Today the care-quality improvement programs for diabetes, asthma, schizophrenia, hepatitis B and C, tuberculosis, breast cancer, cervical cancer, and others all fall into this category (the complete disease list of the early pilots and each program's start year ⚠️ pending verification).

The three must-know pitfalls of P4P: indicator selection bias (only what can be measured gets tracked; what can't be measured gets ignored), patient selection when risk adjustment is inadequate (physicians tend to enroll well-controlled patients to protect their indicators), and the ceiling effect (institutions already at the top lose the incentive to keep improving).

⭐ Board-exam essentials

Chapter 2 Must-Knows

Fee-for-service (FFS) incentive = do more, get more → supplier-induced demand → runaway costs.
Global budget payment system: set the total first, then allocate; payment per point = sector global budget ÷ sector total claimed points → floating point value.
Implementation sequence: dental 1998/7 → Chinese medicine 2000/7 → Western-medicine primary care 2001/7 → hospitals 2002/7 (full implementation).
Taiwan uses an expenditure cap (not an expenditure target) → the risk of excess volume is borne by providers through point-value dilution.
Point value < 1 = point-value dilution / discount; collective volume-pushing is a prisoner's dilemma, and self-discipline is game-theoretically unstable.
Tw-DRG: phase one 2010/1, 164 groups; phase two 2014/7, 254 groups; phase three postponed.
The larger the settlement unit (FFS → per diem → DRG → capitation), the stronger the incentive to economize, and the stronger the incentive for risk selection.
DRG side effects: cream skimming, diagnostic upcoding (DRG creep), premature discharge and readmission; mitigated by exclusion clauses and outlier add-on payments (80% of costs above the upper threshold).
P4P (pay-for-performance) piloted since 2001/11; pitfalls are indicator selection bias, patient selection due to inadequate risk adjustment, and the ceiling effect.

♪ Sing-along text — English rendering of the Mandarin song lyrics

On the slide at a regional hospital's monthly meeting there was only one number, zero point eight seven, meaning that for every point of medical service the hospital had claimed last month, NHI would pay back only eighty-seven cents. The more you do, the less you get. This is the global budget payment system showing its face. In its early years NHI used fee-for-service — provide one service, claim one set of points, collect one payment — and the incentive was crystal clear: the more you do, the more you get, so service volume was systematically driven up; this is called supplier-induced demand. There are two ways to hit the brakes, controlling price or controlling total volume, and controlling price gets offset by rising volume — you cut the unit price, I do twice as much — so what really works is controlling the total. The global budget system is designed to first agree on how much to spend in a year and then go back and divide it up; the key is that the total is fixed but the points are not, so how much each point is worth becomes a division problem: the sector's annual global budget divided by the sector's total points actually claimed. Once the whole profession pushes volume together, the denominator grows, and the point value is diluted.

The total is fixed, the points float — when everyone does more together, together they drive their own point value down.

Taiwan's global budget did not arrive all at once: dental outpatient care went first in July nineteen ninety-eight, ROC year eighty-seven, followed by Chinese medicine outpatient care in July two thousand, Western-medicine primary care in July two thousand one, and hospitals in July two thousand two, at which point it was fully implemented. Taiwan uses an expenditure cap rather than an expenditure target, and the difference between the two is the core of exam questions: under an expenditure cap, the global budget is a hard ceiling and excess service volume is absorbed through point-value dilution, so the risk falls on providers; under an expenditure target, when volume overshoots, the point value is only partly discounted and the insurer still pays somewhat more, so the risk is shared by both sides. Taiwan chose the former, which amounts to outsourcing total volume control to the profession's self-discipline, and self-discipline is unstable in game-theoretic terms, because those who restrain themselves don't get the benefit of their restraint — this is the prisoner's dilemma.

The total is fixed, the points float — when everyone does more together, together they drive their own point value down.

Diagnosis-related groups change something else: they enlarge the settlement unit from one service to one hospital stay, so no matter how many tests you order or how many days the patient stays, the same diagnosis-related group pays the same amount. The incentive flips immediately — cost becomes the hospital's own loss — so hospitals shorten length of stay and cut unnecessary tests. But a flipped incentive always has side effects: grabbing mild cases while avoiding severe ones, diagnostic upcoding that codes patients into groups that pay more, and readmissions caused by premature discharge. So the larger the settlement unit, the stronger the incentive to economize, and the stronger the incentive for risk selection too; this spectrum runs from fee-for-service to per diem, to per case, and on to capitation, with both incentives rising monotonically together. Taiwan's diagnosis-related groups introduced phase one, with one hundred sixty-four groups, on New Year's Day, two thousand ten, and phase two, with two hundred fifty-four groups, in July two thousand fourteen, while phase three was postponed because of strong backlash from the medical profession and the Legislative Yuan. The system includes exclusion clauses and outlier handling: excluded cases such as cancer, psychiatric cases, and stays over thirty days are paid fee-for-service, while cases whose costs exceed the upper threshold receive the fixed payment plus eighty percent of the excess, precisely to suppress the side effect of severely ill patients being bounced from hospital to hospital.

Fee-for-service pays for quantity, diagnosis-related groups pay for efficiency — so who pays for quality? The answer is pay-for-performance. NHI began piloting it in November two thousand one, targeting chronic diseases and conditions requiring long-term follow-up, because outcomes for these diseases lie not in any single clinic visit but in whether the patient is continuously followed up and whether the indicators hit their targets, so payment is tied to process and outcome indicators rather than to the number of visits. Pay-for-performance has three pitfalls of its own: indicator selection bias lets whatever cannot be measured be ignored, inadequate risk adjustment makes physicians inclined to enroll well-controlled patients to protect their indicators, and institutions already at the top run into a ceiling effect and lose the incentive to improve.

3. The Drug-Price Black Hole: A Problem with the Wrong Name

The Drug-Price Gap Is Not a Black Hole — It Is the Shadow of Negotiation

🔬 Mechanism, unpacked

Step one, NHI reimburses hospitals for drugs at the reimbursement price listed in the Pharmaceutical Benefit and Reimbursement Schedule (formerly called the "drug price standard"). Step two, but when hospitals buy drugs from manufacturers they can negotiate, and the actual purchase price is often below the reimbursement price. Step three, the difference in between is the drug-price gap — it is not illegal; it is the lawful return on bargaining power, but it causes the "reimbursement price" to drift further from the "true market price" year after year. Step four, so the NHIA regularly conducts a drug price survey, collecting the actual transaction prices at healthcare institutions on the ground and then lowering reimbursement prices accordingly to claw the gap back. Step five, this is what the media call the "drug-price black hole" — but the name is wrong: no one stole any money; it is a structural gap between the reimbursement price and the market price. Step six, the real policy cost shows up at the next step: year-after-year price cuts drive low-margin old drugs out of the Taiwan market, so the cure for the "drug-price black hole" directly creates another problem — drug shortages.

Besides the drug price survey, NHI uses another valve to control drug spending: the Drug Expenditure Target (DET) — a target value for the year's drug spending is set in advance; if actual drug spending exceeds the target, a drug price adjustment is triggered the following year to recover the overspend (verified 2026-07). Structurally, this is identical to the global budget: draw a box first, and absorb whatever spills over it yourself.

Worth noting: the drug price survey mechanism itself has come under review in recent years — in March 2026 the President directed that a three-year suspension of the drug price survey be studied while the drug supply system is comprehensively reviewed, and the MOHW said it would amend the relevant regulations accordingly (verified 2026-09; price cuts for off-patent drugs continue; as of September 2026 the amended Drug Price Adjustment Regulations had not yet been published, so check the authority's latest announcements), driven precisely by the pressure of drug shortages and old drugs exiting the market.

How Many Gates Must a New Drug Pass to Enter NHI Coverage?

🔬 Mechanism, unpacked

Step one, marketing authorization and reimbursement are two different things — a drug is first approved by the Taiwan Food and Drug Administration (TFDA) and granted a drug license, which only proves that it "may be sold," not that "NHI will pay for it." Step two, after the manufacturer submits a reimbursement application to the NHIA, the NHIA commissions a health technology assessment (HTA), whose core question is not "does it work?" but "how much additional health does this extra money buy?" — that is, cost-effectiveness, commonly measured by the incremental cost-effectiveness ratio (ICER), the additional cost required for each additional quality-adjusted life year (QALY) gained. Step three, the assessment results go to the NHI Drug Benefit Items and Reimbursement Schedule Joint Committee — under NHI Act Article 41 its formal members are the insurer with representatives of relevant agencies, experts, the insured, employers, and healthcare providers, while drug suppliers and patient groups may be invited to express views; in short, seated at the table are payer representatives, the medical profession, the pharmacy profession, experts and scholars, and patient groups, and its deliberations are public; this is the most concrete embodiment of second-generation NHI's spirit of "broader participation and information transparency." Step four, if the efficacy evidence is insufficient or the financial impact too large, the choice is not necessarily a binary "cover / don't cover"; there are intermediate options such as temporary reimbursement, conditional reimbursement (restricted indications, prior authorization, discontinuation if ineffective), and drug reimbursement agreements / risk sharing (managed entry agreements). Step five, so the essence of modern coverage decisions is sharing risk under uncertainty, not waiting for complete certainty before deciding.

⭐ Board-exam essentials

Chapter 3 Must-Knows

Drug-price gap = NHI reimbursement price − the institution's actual purchase price; it is a lawful negotiated margin, and "drug-price black hole" is a misnomer.
Drug price survey → lower reimbursement prices based on actual transaction prices; the side effect is low-margin old drugs leaving the market → drug shortages.
Drug Expenditure Target (DET): if the annual drug-spending target is exceeded, a price adjustment is triggered the following year to recover the overspend.
Marketing authorization (TFDA) ≠ reimbursement (NHIA): the former asks about safety and efficacy, the latter about cost-effectiveness.
Coverage-assessment tools: HTA (health technology assessment), ICER (incremental cost-effectiveness ratio), QALY (quality-adjusted life year).
Drug Benefit Items and Reimbursement Schedule Joint Committee: members include payers, the medical profession, the pharmacy profession, experts, and relevant agencies (patient groups may only be invited to express views), with public deliberations → the concrete institution behind second-generation NHI's "broader participation, information transparency."
Intermediate options: temporary reimbursement, conditional reimbursement (restricted indications / prior authorization), risk-sharing agreements — used to share financial risk when the evidence is uncertain.

♪ Sing-along text — English rendering of the Mandarin song lyrics

The director of pharmacy at a district hospital stared at this year's drug price adjustment notice: an old antibiotic in use for twenty years had been cut by another eight percentage points, and the supplier said that if it were cut again they would stop making it; on another page of the same notice, a newly approved targeted therapy had been added to coverage at over a hundred thousand dollars a month, with half a page of conditions attached. These two things are really one and the same: the money NHI can spend on drugs this year is fixed, and for a new drug to come in, something has to make room. NHI reimburses hospitals for drugs at the reimbursement price in the Pharmaceutical Benefit and Reimbursement Schedule, but hospitals can negotiate when buying from manufacturers, and the actual purchase price is often below the reimbursement price; the difference in between is the drug-price gap. It is not illegal — it is the lawful return on bargaining power — but it makes the reimbursement price drift further from the true market price year after year. So the NHIA regularly conducts a drug price survey, collecting the actual transaction prices at healthcare institutions and then lowering reimbursement prices accordingly. This is what the media call the drug-price black hole, but the name is wrong: no one stole any money; it is a structural gap between the reimbursement price and the market price.

The drug-price black hole isn't stolen money — it's the gap between the reimbursement price and the market price, a crack that gets measured all over again every year.

The real cost shows up at the next step: year-after-year price cuts drive low-margin old drugs out of the Taiwan market, so the cure for the drug-price black hole directly creates another problem — drug shortages. Besides the drug price survey, NHI controls drug spending with the drug expenditure target: a target value for the year's drug spending is set in advance, and if actual spending exceeds it, a drug price adjustment is triggered the following year to recover the overspend. Structurally this is identical to the global budget — draw a box first, and absorb whatever spills over it yourself.

The drug-price black hole isn't stolen money — it's the gap between the reimbursement price and the market price, a crack that gets measured all over again every year.

How many gates must a new drug pass to enter NHI coverage? First, be clear that marketing authorization and reimbursement are two different things. A drug is first approved by the Food and Drug Administration and granted a license, which only proves that it may be sold, not that NHI will pay for it. After the manufacturer submits a reimbursement application to the NHIA, the NHIA commissions a health technology assessment, whose core question is not whether it works but how much additional health the extra money buys — that is, cost-effectiveness — commonly measured by the incremental cost-effectiveness ratio, the additional cost required for each additional quality-adjusted life year. The assessment results go to the Drug Benefit Items and Reimbursement Schedule Joint Committee, where the NHIA decides jointly with payer representatives, the medical profession, the pharmacy profession, experts and scholars, and relevant agencies, while patient groups and drug companies may by law only be invited to express views, with public deliberations — the most concrete embodiment of second-generation NHI's spirit of broader participation and information transparency. If the efficacy evidence is insufficient or the financial impact too large, the choice is not necessarily a binary of cover or don't cover; there are intermediate options such as temporary reimbursement, conditional reimbursement, and drug reimbursement agreements. The essence of modern coverage decisions is sharing risk under uncertainty, not waiting for complete certainty before deciding. So the same drug can be both safe and effective and yet not cost-effective, and not covered does not mean it can't be used — only that you pay out of pocket.

4. Why You Pay NT$250 More at a Medical Center: The Healthcare Network and Tiered Care

The Healthcare Network: Draw the Map First, Then Talk About Tiers

🔬 Mechanism, unpacked

Step one, medical resources naturally concentrate where population and income are dense — this is the inevitable result of the market, and it is also the geographic shape of health inequality. Step two, so Article 88 of the Medical Care Act authorizes the central competent authority to divide the country into medical care regions, establish a tiered healthcare system, and draw up a healthcare network plan. Step three, the point of dividing is to cut an unanswerable question like "does the country have enough doctors?" into an auditable one like "does this sub-region have a community hospital?" Step four, so the country is divided into 6 primary medical regions (Taipei, Northern, Central, Southern, Kaohsiung-Pingtung, Eastern), 17 secondary medical regions, and 50 sub-regions (verified 2026-07), with the planning goal that every secondary region has a regional-level hospital and every sub-region has a community hospital. Step five, only with a map can you talk about who is responsible for catching whom — this is the premise of tiered care.

On the acute and critical care side there is a separate grading system: under the Standards for Grading Hospitals' Emergency Medical Capability, established under the authority of the Emergency Medical Services Act, hospitals are rated by the types of emergency care they provide, their staffing and facilities, and their operational capacity into three levels of emergency responsibility hospital — advanced, intermediate, and general (verified 2026-07). The practical function of this grading is to let an ambulance decide on the scene "where should this patient go" — hospitals at the intermediate level or above must be able to handle acute stroke, acute coronary syndrome, high-risk pregnancy, and neonatal care and arrange appropriate transfers, while advanced-level hospitals must have the most complete emergency treatment capacity. So the basis for prehospital triage is not "which hospital is closest" but "which hospital can handle this patient."

Referral: Using Price to Buy Sequence

Taiwan's tiered care does not forbid you from going to a medical center; it makes you pay to cut the line. The current basic outpatient copayment (in force since 2017/4/15; the 2023/7/1 reform changed only outpatient-drug and emergency copayments and left this table unchanged; verified 2026-09):

Level	With referral	Without referral
Medical center	NT$170	NT$420
Regional hospital	NT$100	NT$240
District hospital	NT$50	NT$80
Primary-care clinic	NT$50	NT$50

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Dental and Chinese-medicine outpatient visits are NT$50 at every level; holders of a disability certificate pay NT$50 at every level)

What to read in this table is not the numbers but the slope: the higher the level, the heavier the penalty for going without a referral, while the primary-care clinic is NT$50 no matter what — the price differences are deliberately designed as a funnel that pushes patients toward primary care. In addition, follow-up visits within a certain period after a referral also enjoy the reduced copayment, to avoid patients being "bounced back to square one after being referred."

The other half of tiered care is making primary care strong. Since March 2003, NHI has piloted the Family Physician Integrated Care Program (verified 2026-07), in which several primary-care clinics join with a partner hospital to form a community healthcare group, jointly providing 24-hour consultation, disease prevention, chronic disease care, and referral assistance. It corresponds precisely to the core attributes of family medicine: accessibility (first contact), continuity, comprehensiveness, and coordination — without "someone who keeps caring for you," tiered care is nothing but a price hike.

⭐ Board-exam essentials

Chapter 4 Must-Knows

Article 88 of the Medical Care Act authorizes dividing medical care regions, establishing tiered care, and drawing up the healthcare network plan.
The country is divided into 6 primary medical regions (Taipei, Northern, Central, Southern, Kaohsiung-Pingtung, Eastern) / 17 secondary medical regions / 50 sub-regions.
Hospital emergency capability grading (under the Emergency Medical Services Act): advanced, intermediate, general; prehospital triage goes by "can handle," not "closest."
Basic outpatient copayment (since 2017/4/15; unchanged by the 2023 reform): medical center NT$170 with referral / NT$420 without; regional hospital 100/240; district hospital 50/80; clinic 50.
The design intent of the price gap = using price to buy sequence, pushing patients toward primary care; clinics charge NT$50 with or without a referral.
Family Physician Integrated Care Program (piloted 2003/3): primary-care clinics join with partner hospitals to form community healthcare groups.
Core attributes of family medicine: accessibility, continuity, comprehensiveness, coordination — the success or failure of tiered care depends on whether primary care is strong enough.

♪ Sing-along text — English rendering of the Mandarin song lyrics

A mother took her five-year-old, feverish for two days, straight to a medical center's pediatric clinic; she waited three hours, was seen for seven minutes, and was told it was an ordinary viral infection, and the basic outpatient copayment she paid was four hundred twenty NT dollars. Had she first gone to the clinic at the end of her lane, she would have paid fifty dollars, and if further tests were truly needed, the clinic would write a referral and she would pay only one hundred seventy dollars at the very same medical center. Same illness, same hospital, same doctor — the two-hundred-fifty-dollar difference in between buys not medical care but sequence. That is the whole logic of tiered care: it does not forbid you from going to a medical center; it makes you pay to cut the line.

Tiered care doesn't ban you from the big hospital — it just puts a visible price tag on cutting the line.

Before talking about tiers, you need a map. Medical resources naturally concentrate where population and income are dense — the inevitable result of the market and the geographic shape of health inequality — so Article eighty-eight of the Medical Care Act authorizes the central competent authority to divide the country into medical care regions, establish a tiered healthcare system, and draw up a healthcare network plan. The point of dividing is to cut an unanswerable question like whether the country has enough doctors into an auditable one like whether this sub-region has a community hospital. The country is divided into six primary medical regions — Taipei, Northern, Central, Southern, Kaohsiung-Pingtung, and Eastern — beneath which are seventeen secondary medical regions and fifty sub-regions, with the planning goal that every secondary region has a regional-level hospital and every sub-region has a community hospital.

On the acute and critical care side there is a separate grading system: under the Standards for Grading Hospitals' Emergency Medical Capability, established under the authority of the Emergency Medical Services Act, hospitals are rated by the types of emergency care they provide, their staffing and facilities, and their operational capacity as one of three kinds of emergency responsibility hospital — advanced, intermediate, or general. The practical function of this grading is to let an ambulance decide on the scene where the patient should go: hospitals at the intermediate level or above must be able to handle acute stroke, acute coronary syndrome, high-risk pregnancy, and neonatal care and arrange appropriate transfers, while advanced-level hospitals must have the most complete emergency treatment capacity. So the basis for prehospital triage is not which hospital is closest, but which hospital can handle this patient.

Tiered care doesn't ban you from the big hospital — it just puts a visible price tag on cutting the line.

The current basic outpatient copayment dates from the adjustment of April fifteenth, two thousand seventeen, and the reform of July two thousand twenty-three left it untouched; it works like this: at a medical center, one hundred seventy dollars with a referral and four hundred twenty without; at a regional hospital, one hundred with a referral and two hundred forty without; at a district hospital, fifty with a referral and eighty without; at a primary-care clinic, fifty across the board; dental and Chinese-medicine outpatient visits are fifty at every level, and holders of a disability certificate pay fifty at every level. What to read in this table is not the numbers but the slope: the higher the level, the heavier the penalty for going without a referral, while the primary-care clinic is fifty dollars no matter what — the price differences are deliberately designed as a funnel that pushes patients toward primary care. Follow-up visits within a certain period after a referral also enjoy the reduced copayment, to avoid patients being bounced back to square one after being referred. But the other half of tiered care is making primary care strong: since March two thousand three, NHI has piloted the Family Physician Integrated Care Program, in which several primary-care clinics join with a partner hospital to form a community healthcare group, jointly providing twenty-four-hour consultation, disease prevention, chronic disease care, and referral assistance. It corresponds precisely to the core attributes of family medicine: accessibility, continuity, comprehensiveness, and coordination. Without someone who keeps caring for you, tiered care is nothing but a price hike.

5. Translating Disability into Four Envelopes: Long-Term Care 2.0

🩺 At the bedside

An 82-year-old grandmother is left with right-sided hemiplegia after a stroke and is discharged home. Her daughter quits her job to care for her; three months later, the daughter herself starts losing sleep and hair, snaps at her mother, and then hides in the bathroom to cry. She applies for long-term care, and a care manager comes to the house to assess, scoring Grandma's daily-living functions item by item on a rating scale, and finally produces a number: Level 6. From that day on, the family has a fixed monthly allowance that can be exchanged for hours of a home care worker's time, and a yearly sum that lets the daughter catch her breath by sending her mother to a day-care center for a few days. Disability has been translated into a level, and the level has been translated into money.

Why National Health Insurance Can't Carry This

🔬 Mechanism, unpacked

Step one, National Health Insurance (NHI) covers medical care — diagnosis, treatment, acute-phase management — and its billing unit is "one medical act." Step two, but what Grandma needs is not treatment; it is someone to bathe her, turn her, and feed her every day. This is called long-term care, and its billing unit is "a stretch of time of daily-living support." Step three, the financial structures of the two are completely different: medical care is low in probability and high in cost per episode (well suited to insurance), whereas long-term care is high in probability and long in duration (closer to something financed by savings or taxes). Step four, so cramming long-term care into NHI would crush NHI at once; Taiwan chose to build a separate system with a separate funding source — LTC 2.0 runs on a tax-financed model, not an insurance model. Step five, this dividing line is the exam's favorite spot: acute medical care and post-acute care (PAC) are paid by NHI; home care services, day care, and respite services are paid by long-term care.

Long-Term Care 2.0 (LTC 2.0) has been rolled out since 2017 (ROC year 106), with a target population far broader than LTC 1.0 (newly adding people with dementia aged 50 and over, functionally disabled plains Indigenous people aged 55–64 — mountain Indigenous people were already covered under 1.0 — functionally disabled persons with disabilities aged 49 and under, and frail older adults aged 65 and over whose disability is limited to IADLs); LTC 3.0 (2026–2035), which took over from 2026, further admits people with dementia and disability at any age and patients enrolled in NHI's integrated post-acute care program. What truly changed how the public experiences the system was the new payment scheme launched in January 2018, popularly known as the "four envelopes" (四包錢) (verified as of July 2026):

"Envelope"	What it covers	Allowance at a glance	Copayment cap
① Care and professional services	Home care services, day care, family-based care; professional services such as home rehabilitation, home nutrition, and home pharmacy care	By disability Levels 2–8, roughly NT$10,000–36,000 per month	Up to 16%
② Transportation	Transport to medical appointments and rehabilitation	Applies to Level 4 and above; roughly NT$1,680–2,400 per month depending on distance	Up to 30%
③ Assistive devices and home accessibility modifications	Wheelchairs, air mattresses, stair-climbing machines; bathroom grab bars, non-slip flooring	Up to NT$40,000 (or NT$60,000 under the Group 2 option from July 2026) every 3 years	Up to 30%
④ Respite services	In-home respite, institutional respite, day-care-center respite, overnight temporary care at small-scale multifunctional centers, temporary care at neighborhood LTC stations	Up to about NT$48,510 per year	Up to 16%

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Contents and allowances of the four envelopes verified as of July 2026; actual amounts in any given year are subject to the competent authority's announcements)

The division-of-labor logic of the four envelopes deserves a full explanation: ① is service for the "person," ② and ③ support "mobility and environment," and ④ is for the "caregiver." The fourth envelope in particular — the beneficiary of respite services is not the disabled person but the family. This is the shift in Taiwan's long-term care system that is most likely to be tested and also most easily overlooked: it formally acknowledges that "the caregiver is also a person who needs care."

CMS Levels: The Ruler That Converts Function into Money

The degree of disability is determined by the care manager using an assessment scale and converted into an LTC need level (commonly called the CMS level), ranging from Level 1 to Level 8. The key test point: Level 1 means "does not meet LTC need" and receives no benefits; those who can actually apply for services are Levels 2 through 8, and the higher the level, the higher the allowance. And transportation applies only to Level 4 and above — this is the threshold most often turned into a question.

What LTC 2.0 does is translate the sentence "he can't stand up" into an integer that can be accounted for, audited, and funded.

The ABC Three-Tier Community-Based Integrated Care Model

Service delivery under LTC 2.0 follows the community-based integrated care service system (the ABC model), whose goal is aging in place:

Tier	Name	Role
A	Community Integrated Service Center (the "LTC flagship store")	Drafts care plans; links and dispatches Tier B and Tier C resources
B	Composite Service Center (the "LTC specialty store")	Actually delivers day care, home care services, professional services, and so on
C	Neighborhood LTC Station (the "LTC corner shop")	Provides nearby temporary care, communal meals, and courses to prevent and delay disability

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

A handles "planning and linking," B handles "delivering services," C handles "being close at hand" — the three are a division of labor, not a ranking of quality, and questions often flip this around on you.

Where the Money Comes From: Long-Term Care Is Not Insurance — It Is Tax

LTC 2.0 is funded through the Long-Term Care Services Development Fund, drawn mainly from four tax-and-levy sources (the Tobacco Health and Welfare Surcharge is a levy, not a tax; verified September 2026): the added revenue from changing the estate and gift tax from a single 10% rate to three progressive brackets of 10%/15%/20%, the added revenue from the tobacco tax increase, the Tobacco Health and Welfare Surcharge, and the consolidated income tax on housing and land (the balance remaining after deducting the central government's allocated share and the transfer to the housing fund), plus government budget appropriations, donations, and fund interest.

This matters enormously at the conceptual level: NHI is "premium-financed," LTC 2.0 is "tax-financed." The advantage of tax financing is that not everyone has to pay a contribution and it can be progressive by income; the disadvantage is that revenue fluctuates with the economy, and every year it has to compete with other government programs for budget — this is precisely the structural reason the stability of LTC funding has long been questioned, and the core of the essay question "LTC insurance model vs. tax-financed model."

🦦

PipGrandpa had a stroke and was admitted. The acute phase is over, but he still can't go home — so is the rehab that comes next paid by NHI or by long-term care?

🐻‍❄️

BarnabyYou have to memorize this boundary. Post-acute care (PAC) still belongs to NHI — it exists to concentrate training in the golden rehabilitation window and fight for functional recovery. Once his condition stabilizes and he returns to daily life in the community, home care services, day care, and respite services switch over to long-term care benefits. The criterion is not "is he in the hospital or not," but "is he still on the track of treatment and functional recovery." And while you're at it, lock this in: LTC Level 1 receives no benefits, benefits start at Level 2, and transportation requires Level 4 or above.

⭐ Board-exam essentials

Chapter 5 Must-Knows

NHI = medical care (low probability, high cost — suited to insurance); LTC = daily-living support (high probability, long duration — tax-financed).
LTC 2.0 has been rolled out since 2017; the "four envelopes" payment scheme took effect in January 2018; from 2026 it continues under LTC 3.0 (2026–2035), with the four-envelope framework retained.
The four envelopes: ① care and professional services (copayment cap 16%), ② transportation (30%, Level 4 and above only), ③ assistive devices and home accessibility (30%, NT$40,000 every 3 years, or NT$60,000 under the Group 2 option from July 2026), ④ respite services (16%).
The beneficiary of respite services is the "caregiver," not the disabled person.
LTC need level (CMS) 1–8; Level 1 receives no benefits; eligibility to apply starts at Level 2; transportation requires Level 4 or above.
ABC model: A Community Integrated Service Center (planning and linking) / B Composite Service Center (delivering services) / C Neighborhood LTC Station (nearby temporary care, communal meals, delaying disability); the goal is aging in place.
Funding = Long-Term Care Services Development Fund: added revenue from the three-bracket progressive estate and gift tax (10/15/20%) + added revenue from the tobacco tax increase + Tobacco Health and Welfare Surcharge + consolidated housing and land tax + government budget appropriations → tax-financed, not insurance-financed.
The dividing line: acute medical care and post-acute care (PAC) belong to NHI; home care services, day care, and respite belong to LTC.

♪ Sing-along text — English rendering of the Mandarin song lyrics

An eighty-two-year-old grandmother was left with right-sided hemiplegia after a stroke and discharged home. Her daughter quit her job to care for her, and three months later the daughter herself started losing sleep and hair, snapped at her mother, and then hid in the bathroom to cry. She applied for long-term care, and a care manager came to the house and scored Grandma's daily-living functions item by item, finally producing a number: Level Six. From that day on, the family had a fixed monthly allowance that could be exchanged for hours of a home care worker's time, and a yearly sum that let the daughter catch her breath. Disability had been translated into a level, and the level had been translated into money.

What long-term care does is translate the sentence "he can't stand up" into an integer that can be funded.

Why can't National Health Insurance carry this? Because National Health Insurance covers medical care — diagnosis, treatment, acute-phase management — and its billing unit is one medical act; but what Grandma needs is not treatment, it is someone to bathe her, turn her, and feed her every day. This is called long-term care, and its billing unit is a stretch of time of daily-living support. The financial structures of the two are completely different: medical care is low in probability and high in cost per episode, well suited to being spread out by insurance; long-term care is high in probability and long in duration, closer to something financed by savings or taxes. So cramming long-term care into National Health Insurance would crush it at once, and Taiwan chose to build a separate system with a separate funding source — Long-Term Care 2.0 runs on tax financing, not insurance.

Long-Term Care 2.0 has been rolled out since 2017, with a target population far broader than 1.0, newly adding people with dementia aged fifty and over, functionally disabled plains Indigenous people aged fifty-five to sixty-four, functionally disabled persons with disabilities aged forty-nine and under, and frail older adults aged sixty-five and over whose disability is limited to instrumental activities of daily living; Long-Term Care 3.0, which began in 2026, goes further and admits people with dementia and disability at any age, as well as patients in NHI's integrated post-acute care program. What truly changed how the public experiences the system were the four envelopes launched in January 2018. The first envelope is care and professional services, including home care services, day care, family-based care, and professional services such as home rehabilitation, nutrition, and pharmacy care; by disability Level Two through Level Eight it is roughly ten thousand to thirty-six thousand New Taiwan dollars a month, with a copayment of up to sixteen percent. The second envelope is transportation, available only from Level Four and above, roughly one thousand six hundred eighty to two thousand four hundred dollars a month depending on distance, with a copayment of up to thirty percent. The third envelope is assistive devices and home accessibility modifications, up to forty thousand dollars every three years — or, from July 2026, a Group 2 option of up to sixty thousand dollars for a narrower list of items — with a copayment of up to thirty percent. The fourth envelope is respite services, up to a little over forty-eight thousand dollars a year, with a copayment of up to sixteen percent. The division-of-labor logic of the four envelopes is this: the first is service for the person, the second and third support mobility and environment, and the fourth is for the caregiver — the beneficiary of respite services is not the disabled person but the family, and this is the most important and most often overlooked shift in Taiwan's long-term care system.

What long-term care does is translate the sentence "he can't stand up" into an integer that can be funded.

The degree of disability is determined by the care manager using an assessment scale and converted into an LTC need level from one to eight. Level One falls short of LTC need and receives no benefits; those who can actually apply for services are Levels Two through Eight, and the higher the level, the higher the allowance, while transportation applies only from Level Four and above. Service delivery follows the community-based integrated care service system, that is, the ABC model: A is the Community Integrated Service Center, responsible for drafting care plans and for linking and dispatching resources; B is the Composite Service Center, which actually delivers day care and home care services; C is the Neighborhood LTC Station, which provides nearby temporary care, communal meals, and courses to delay disability. The three are a division of labor, not a ranking of quality, and the goal is aging in place. Where does the money come from? The four tax-and-levy sources of the Long-Term Care Services Development Fund are the added revenue from changing the estate and gift tax to three progressive brackets, the added revenue from the tobacco tax increase, the Tobacco Health and Welfare Surcharge (strictly a levy, not a tax), and the consolidated housing and land tax remaining after deducting the centrally allocated share and the housing fund, plus government budget appropriations. So National Health Insurance is premium-financed while long-term care is tax-financed; the advantage of tax financing is that not everyone has to pay a contribution and it can be progressive by income, and the disadvantage is that revenue fluctuates with the economy and has to compete every year with other government programs for budget. Finally, lock in that dividing line: post-acute care still belongs to National Health Insurance, because it is concentrated training in the golden rehabilitation window to fight for functional recovery; only once the condition stabilizes and returns to the level of daily life in the community do home care services, day care, and respite services switch over to long-term care benefits. The criterion is not whether the patient is in the hospital, but whether they are still on the track of treatment and functional recovery.

6. The Red Line on the Water Gauge: NHI Finances, Premium Rates, and the Generational Ledger

🔬 Mechanism, unpacked

Step one, NHI is pay-as-you-go — this year's premiums pay this year's medical expenses; it is not a fund that saves money for your own future use. Step two, so whether its finances are healthy depends on one thing: the ratio of those who pay in to those who draw out. Step three, Taiwan's population structure is pushing this ratio in the worst possible direction — the main payers are young and middle-aged adults, the main users are older adults, and the share of older people keeps climbing while the working-age population keeps shrinking. Step four, this makes NHI's financial crisis fundamentally not a matter of "poor management" but an inevitability of population structure; any measure to raise revenue or cut spending only postpones it and does not solve it. Step five, so every discussion of NHI finances ultimately lands on the same political question: how much more must this generation pay so as not to leave the entire bill to the next?

The Reserve Fund: How to Read This Water Line

Article 78 of the National Health Insurance Act provides that the total NHI reserve shall be maintained at the equivalent of 1 to 3 months of insurance benefit expenditure, based on the most recent actuarial calculation (verified as of July 2026). This range is NHI's water gauge:

Below 1 month → under Article 26, the insurer must draw up a proposal to adjust the scope of insurance benefits for review by the National Health Insurance Committee and approval by the Executive Yuan (Article 78 itself only sets the 1-to-3-month principle); the premium rate is reviewed each year under Article 24 to balance income and expenditure.
Above 3 months → revenue is running high; the Act does not require a rate cut in that case, but the committee weighs it in the annual rate review (for example, the rate was cut from 4.91% to 4.69% in 2016 when finances were ample).

Premium-rate adjustments are not something the NHIA can make whenever it wishes: they must be reviewed and agreed on by the National Health Insurance Committee and reported to the Ministry of Health and Welfare for forwarding to the Executive Yuan for approval, and the statutory cap is 6% (verified as of July 2026). The current general premium rate is 5.17% and the supplementary premium rate is 2.11%, both in effect since January 1, 2021 and unchanged to this day (verified as of July 2026) — which means only about 0.83 percentage points of room remain below the statutory ceiling, and that is exactly why "should the statutory premium-rate cap be loosened" has become a policy focus in recent years.

In addition, NHI is subject to actuarial review: the law requires financial actuarial projections covering at least 25 years, reviewed at least once every 5 years (Article 25 verbatim: the insurer conducts an actuarial review at least once every five years, each projecting 25 years) — because NHI faces a population curve spanning decades, and judging its health by short-term income and expenditure will systematically mislead.

What Second-Generation NHI Changed, and What It Left Behind

🔬 Mechanism, unpacked

Step one, second-generation NHI (2G-NHI, implemented January 1, 2013) addressed three things: insufficient financing, insufficient participation, and lack of transparency. Step two, the answer on the financing side was the supplementary premium — it broadened the premium base but did not change the "wage-based" skeleton. Step three, the answer on the participation side was establishing the National Health Insurance Committee, placing the review and agreement of premium rates, the global budget, and the scope of benefits into a deliberative body jointly composed of representatives of payers, the medical community, experts, and relevant agencies, so that "the people who pay" hold formal seats. Step four, the answer on the transparency side was opening the Joint Drafting Meetings to the public together with information disclosure. Step five, but 2G-NHI left one core problem unresolved: the premium base is still mainly wages plus a few enumerated types of income and did not move to "total household income" — this is precisely the starting point for subsequent reform discussions (often called third-generation NHI), with issues including a total-household-income system, loosening the premium-rate cap, restructuring copayments, and whether to introduce a division of labor between private insurance and NHI (specific reform proposals ⚠️ pending verification).

⭐ Board-exam essentials

Chapter 6 Must-Knows

NHI is pay-as-you-go; the root of its financial problem is population structure (shrinking payers, expanding users), not simply a management problem.
The reserve should be maintained at the equivalent of 1–3 months of insurance benefit expenditure (National Health Insurance Act Article 78); below 1 month, the insurer must draft a benefit-scope adjustment proposal for the committee under Article 26.
Premium-rate adjustment procedure: review and agreement by the National Health Insurance Committee → MOHW → approval by the Executive Yuan; statutory cap 6%.
Current general premium rate 5.17%, supplementary premium rate 2.11% (since 2021/1/1).
Actuarial requirement: at least once every 5 years, each projection covering 25 years (Article 25).
The three main thrusts of 2G-NHI: broadening the premium base (supplementary premium), broadening participation (National Health Insurance Committee), transparency (opening the Joint Drafting Meetings).
2G-NHI's unfinished business = premium base still mainly wages, total household income not adopted; this is the core of subsequent reform discussions.

♪ Sing-along text — English rendering of the Mandarin song lyrics

National Health Insurance is pay-as-you-go: this year's premiums pay this year's medical expenses. It is not a fund that saves money for your own future use, so its financial health depends on one thing — the ratio of those who pay in to those who draw out. Taiwan's population structure is pushing this ratio in the worst possible direction: the main payers are young and middle-aged adults, the main users are older adults, and the share of older people keeps climbing while the working-age population keeps shrinking. This makes National Health Insurance's financial crisis fundamentally not a matter of poor management but an inevitability of population structure, and any measure to raise revenue or cut spending only postpones it rather than solving it.

National Health Insurance is not you saving for your future self; it is this generation paying the bill for the generation before.

Article seventy-eight of the National Health Insurance Act provides that the total reserve shall be maintained at the equivalent of one to three months of insurance benefit expenditure based on the most recent actuarial calculation, and this range is National Health Insurance's water gauge. Below one month, Article twenty-six requires the insurer to draw up a proposal to adjust the scope of insurance benefits and submit it to the National Health Insurance Committee for review; above three months means revenue is running high, and although the law does not require a rate cut, the committee weighs it in its annual rate review. Premium-rate adjustments are not something the National Health Insurance Administration can make whenever it wishes: they must be reviewed and agreed on by the National Health Insurance Committee and reported to the Ministry of Health and Welfare for forwarding to the Executive Yuan for approval, and the statutory cap is six percent. The current general premium rate of five point one seven percent and supplementary premium rate of two point one one percent both took effect on New Year's Day 2021 and remain in use today, which means less than one percentage point of room is left below the statutory ceiling — exactly why whether to loosen the premium-rate cap has become a policy focus. National Health Insurance is also subject to actuarial review: under Article twenty-five the insurer must carry out an actuarial review at least once every five years, each projecting twenty-five years, because it faces a population curve spanning decades, and judging its health by short-term income and expenditure will systematically mislead.

National Health Insurance is not you saving for your future self; it is this generation paying the bill for the generation before.

Second-generation National Health Insurance took effect on New Year's Day 2013, and it addressed three things: insufficient financing, insufficient participation, and lack of transparency. The answer on the financing side was the supplementary premium, which broadened the premium base without changing the wage-based skeleton. The answer on the participation side was establishing the National Health Insurance Committee, placing the review and agreement of premium rates, the global budget, and the scope of benefits into a deliberative body jointly composed of representatives of payers, the medical community, experts, and relevant agencies, so that the people who pay hold formal seats. The answer on the transparency side was opening the Joint Drafting Meetings to the public, together with information disclosure. But second-generation National Health Insurance left one core problem unresolved: the premium base is still mainly wages plus a few enumerated types of income and has not moved to total household income. This is precisely the starting point for subsequent reform discussions, with issues including a total-household-income system, loosening the premium-rate cap, restructuring copayments, and how labor should be divided between commercial insurance and National Health Insurance.

7. When NHI Began to Remember: From the IC Card to the Constitutional Court

How a Card Grew a Memory

In its early years, Taiwan's NHI used a paper card (replaced with a new one once its six boxes were used up). From January 1, 2004, the NHI IC card fully replaced the paper card (verified as of July 2026). The significance goes far beyond a change of material — the paper card was merely a voucher; the IC card is an interface: it allows every medical visit to be recorded, uploaded, and linked under the same identifier.

On top of this grew the NHI MediCloud System (including what is popularly called the PharmaCloud): during a visit, the physician can instantly pull up the patient's medications, laboratory and examination results, and images from across different hospitals and clinics. It tackles the most real disease in Taiwanese medicine: duplicate medications, duplicate tests, and drug interactions. And at the same time it is a form of clinical decision support — when the screen pops up "a drug with the same ingredient was prescribed at another facility three days ago," that is not an administrative audit; it is putting the information right at the moment the decision is made.

Dispute Review: What to Do When a Claim Is Cut or Denied

When insured persons, enrolling units, or contracted medical care institutions disagree with a determination by the NHIA (for example, premium calculation, determination of the insured amount, or payment and deduction of medical expense claims), they cannot go straight to court; they must go through NHI dispute review:

Dispute review (the MOHW's National Health Insurance Dispute Review Board) → if still dissatisfied, file an administrative appeal → if still dissatisfied, file administrative litigation

Review first, then appeal, then administrative litigation — this prerequisite procedure is a must-know sequence, and questions love to use "directly filing administrative litigation" as a distractor. Its institutional rationale: NHI disputes are highly technical and enormous in volume, so a filtering layer with medical expertise is needed to keep the administrative courts from being swamped by technical disputes.

That Database, and Its Constitutional Limits

🔬 Mechanism, unpacked

Step one, "secondary use" of the NHI database means taking personal NHI data originally collected to "administer the insurance," de-identifying it, and providing it to third parties for uses beyond the original purpose, such as research. Step two, its scientific value is extremely high — covering the entire population, with long follow-up and no selection bias — and Taiwan has produced a large body of world-class epidemiological research as a result. Step three, but the constitutional problem is: the data subjects never consented, and cannot opt out. Step four, Constitutional Court Judgment 111-Hsien-Pan-13 (August 12, 2022) held (verified as of July 2026): the current law lacks clear provisions on how NHI data may be used beyond its original purpose — its entities, purposes, requirements, scope, methods, and organizational and procedural oversight mechanisms — and therefore fails the principle of legal reservation, violating the constitutionally protected right to informational privacy; moreover, the lack of an independent oversight mechanism for personal data protection also risks being unconstitutional; and there is no "opt-out mechanism" allowing data subjects to request that the provision of their personal NHI data to third parties for secondary use be stopped. Step five, the judgment gave the authorities a deadline — as it turned out, when that deadline passed in August 2025 the NHIA first opened opt-out requests administratively, the Legislative Yuan then enacted the National Health Insurance Data Management Act (promulgated December 19, 2025; in force August 10, 2026), writing the opt-out right, an advisory council, review committees, and penalties into law, and the organic act for an independent Personal Data Protection Commission was still pending as of September 2026, with only a preparatory office in operation; the original order was to amend the National Health Insurance Act or other relevant laws within 3 years of the date of its announcement. Step six, so the true proposition of this judgment is not "is research good or bad," but — however great the public interest, it cannot waive three things: the legal authorization must be clear, the oversight must be independent, and the data subject must have an exit.

🦦

PipBut the NHI database is already de-identified — doesn't that take care of personal data protection? How can it still be unconstitutional?

🐻‍❄️

BarnabyThat's exactly the most common misunderstanding. Of the three things the Constitutional Court singled out, not one can be solved by "de-identification": first, the law doesn't clearly specify who may use the data, for what, and within what scope — that's a legal reservation problem; second, there is no independent personal data protection oversight authority, which amounts to being both player and referee; third, data subjects have no opt-out mechanism — not even the right to say "don't use mine." De-identification is a technical safeguard; these three are institutional safeguards — however good the technology, it cannot replace the institution. The judgment required the law to be amended within three years (after the deadline, the Legislative Yuan passed the National Health Insurance Data Management Act, in force from August 10, 2026, which spells out the opt-out right); remember it as 111-Hsien-Pan-13, August 12, 2022.

⭐ Board-exam essentials

Chapter 7 Must-Knows

The NHI IC card fully replaced the paper card from 2004/1/1; its significance is "voucher → interface," enabling data to be linked across hospitals and clinics.
NHI MediCloud System (PharmaCloud): tackles duplicate medications, duplicate tests, and drug interactions; it is a form of clinical decision support.
Order of dispute remedies: NHI dispute review → administrative appeal → administrative litigation (review is a prerequisite procedure; you may not go straight to court).
Scope of dispute review: premiums (insured amount, premium calculation), benefits, payment and deduction of medical expense claims, etc.
Constitutional Court Judgment 111-Hsien-Pan-13 (2022/8/12): the entities, purposes, requirements, scope, methods, and oversight mechanisms for use of NHI data beyond its original purpose lack clear provisions, violating legal reservation and the right to informational privacy; no independent personal data oversight mechanism; no opt-out mechanism for data subjects; follow-up law = the National Health Insurance Data Management Act (promulgated 2025/12/19, in force 2026/8/10, opt-out right written in; the independent data-protection commission's organic act still pending as of 2026-09); amendment originally ordered within 3 years.
Core concept: de-identification is a technical safeguard; clear legal authorization, independent oversight, and the right to opt out are institutional safeguards — neither can substitute for the other.

♪ Sing-along text — English rendering of the Mandarin song lyrics

On August twelfth, 2022, the Constitutional Court announced a judgment. The petitioners were several ordinary citizens, backed by a decade of litigation by human rights groups, and their claim was: my medical records were taken for research; I never consented, and I have no way to make it stop. At the center of the constitutional review sat a database that researchers the world over dream of — the complete care trajectories of twenty-three million people over nearly thirty years, with almost no one lost to follow-up. What the court finally said was simple: this database may go on existing, but the law as it now stands is not enough to support it.

De-identification is technology; clear law, independent oversight, and a way for people to opt out — those three are institutions, and technology cannot replace institutions.

In its early years, Taiwan's National Health Insurance used paper cards, replaced with a new one once the six boxes were used up, and from New Year's Day 2004 the NHI integrated-circuit card fully replaced the paper card. The significance goes far beyond a change of material: the paper card was merely a voucher, while the chip card is an interface that allows every medical visit to be recorded, uploaded, and linked under the same identifier. On top of this grew the NHI MediCloud System, popularly known as PharmaCloud: during a visit, the physician can instantly pull up the patient's medications, laboratory and examination results, and images from across different hospitals and clinics. It tackles the most real disease in Taiwanese medicine — duplicate medications, duplicate tests, and drug interactions — and it is also a form of clinical decision support, because it puts the information right at the moment the decision is made.

If insured persons, enrolling units, or contracted medical care institutions disagree with a determination by the National Health Insurance Administration — for example, premium calculation, determination of the insured amount, or payment and deduction of medical expense claims — they cannot go straight to court. They must first go through NHI dispute review, file an administrative appeal only if dissatisfied with the review result, and file administrative litigation only if still dissatisfied after that. Review first, then appeal, then administrative litigation: this prerequisite procedure is a must-know sequence, and questions love to use directly filing administrative litigation as a distractor. Its institutional rationale is that NHI disputes are highly technical and enormous in volume, so a filtering layer with medical expertise is needed.

De-identification is technology; clear law, independent oversight, and a way for people to opt out — those three are institutions, and technology cannot replace institutions.

As for secondary use of the NHI database, it means taking personal NHI data originally collected to administer the insurance, de-identifying it, and providing it to third parties for uses beyond the original purpose, such as research. Its scientific value is extremely high — covering the entire population, with long follow-up and no selection bias — and Taiwan has produced a large body of world-class epidemiological research as a result. But the constitutional problem is that the data subjects never consented and cannot opt out. The Constitutional Court's Judgment Hsien-Pan number thirteen of ROC year one hundred eleven held that the current law lacks clear provisions on how NHI data may be used beyond its original purpose — its entities, purposes, requirements, scope, methods, and organizational and procedural oversight mechanisms — and therefore fails the principle of legal reservation and violates the constitutionally protected right to informational privacy; that it also lacks an independent oversight mechanism for personal data protection; and that it lacks an opt-out mechanism allowing data subjects to request that the provision of their personal NHI data to third parties for secondary use be stopped. The judgment ordered the relevant authorities to amend the National Health Insurance Act or other relevant laws within three years of the date of its announcement. After the deadline, the Legislative Yuan passed the National Health Insurance Data Management Act in December 2025, in force from August tenth, 2026, writing the opt-out right into law, while the organic act for an independent Personal Data Protection Commission is still awaiting passage. The true proposition of this judgment is not whether research is good, but that however great the public interest, it cannot waive three things: the legal authorization must be clear, the oversight must be independent, and the data subject must have an exit.

8. The Medical Regulatory System: One License, One Institution, One Lifesaving Line

🩺 At the bedside

At three in the morning, a resident just off a night shift is jolted awake by a LINE message. A senior colleague has sent a photo of someone's skin: "Take a look for me — it's my cousin's. Could you prescribe him something?" He stares at the screen for ten seconds, then replies: "Tell him to come to the ER." In those ten seconds, what flashed through his mind was not a differential diagnosis but Article 11 of the Physicians Act — a physician may not provide treatment, prescribe medication, or issue a diagnostic certificate without personally examining the patient. That same night, the light-box sign of the clinic next door read "Exclusive to this clinic, No. 1 in the nation, guaranteed results," and an inspector from the health bureau took a photo of it. These two events sit at the two ends of the same legal system.

Taiwan's medical regulations are not a pile of article numbers; they really answer only three questions: who may do it (Physicians Act), where it may be done (Medical Care Act), and who is responsible when there isn't time to do it (Emergency Medical Services Act). Put these three questions in the right order, and law questions turn from "memorizing article numbers" into "reasoning through logic."

The Physicians Act: Three Gates to a License

The Physicians Act governs the "person." It sets up three gates, and their order cannot be reversed.

The first gate is qualification: passing the examination and holding a physician certificate. The certificate is issued by the central government (the Ministry of Health and Welfare), only once in a lifetime, and signifies that "this person is qualified to be a physician."

The second gate is practice registration: a certificate alone does not let you see patients. A physician must apply for practice registration with the competent authority of the special municipality or county (city) where the practice is located and obtain a practice license before practicing, and practice is limited to one location (separate rules apply to reported support at other sites). The certificate is "who you are"; the license is "where you are practicing now" — exam questions love to swap these two.

The third gate is continuing education: the practice license must be renewed every 6 years, and renewal is conditional on completing continuing-education credits. Under Article 13 of the current Regulations Governing Practice Registration and Continuing Education of Medical Personnel (which replaced the old physician-specific regulations on July 1, 2013), physicians, Chinese medicine physicians, and dentists no longer follow the old rule of 180 points every 6 years, of which credits in "professional quality, professional ethics, and professional regulations" (categories two through four) had to reach 18 points; like other practitioner-level medical personnel, they now need categories two through four to total at least 12 points (anything above 24 counted as 24), including courses on infection control and gender issues, for a 6-year requirement of 120 points/12 points (holders of a valid specialist certificate need only document the 12 points at renewal; assistant-level personnel such as physical therapy assistants need 72/7 points; verified September 2026 against the version amended August 26, 2022).

🔬 Mechanism, unpacked

Step one, why must continuing education mandate a minimum number of "ethics and regulations" points? Step two, because professional knowledge updates itself — physicians naturally read new guidelines in order to treat patients — but ethics and regulations won't make today's clinic run any smoother, so without an outside push no one reads them. Step three, what the system's designers face is a classic incentive failure: learning with no immediate payoff for the individual but large external benefits for society. Step four, the solution is to turn it from "voluntary" into "a condition for keeping your license" — internalizing an external benefit as a personal cost. Step five, so "infection control" and "gender issues" are singled out by name not because they matter more than pharmacology, but because they are the least likely to be learned voluntarily.

The Physicians Act has two more substantive provisions that show up every year. The first is Article 21: a physician shall, for a patient in critical condition, immediately provide treatment or take necessary measures according to their professional ability and shall not delay without cause — this is the physician's personal duty to rescue, whereas Medical Care Act Article 60, "hospitals and clinics that encounter a patient in critical condition shall first provide appropriate emergency treatment," is an institutional duty; memorize the two separately. The second is the telemedicine exception in Article 11: in principle, a physician may not treat, prescribe, or issue a diagnostic certificate without personally examining the patient, but in mountain areas, offshore islands, or remote areas, or in special or urgent circumstances, a physician designated by the competent authority of the special municipality or county (city) may inquire about the patient's condition, examine, and prescribe by means of telecommunication, with health personnel delivering the medication to the patient. This narrow door later grew into the Regulations Governing Telemedicine Diagnosis and Treatment (⚠️ pending verification: dates of promulgation and amendment of the current version). Key trap: telemedicine relaxes the requirement of "being physically present," not "the duty to examine" itself — the physician must still perform an examination that meets the prevailing medical standard, and a wrong prescription still carries liability.

As for discipline, the test point is "who punishes, and with what." When a physician falls under the circumstances listed in Physicians Act Article 25 — serious or repeated negligence in practice, a crime committed by exploiting opportunities arising from practice with a final conviction, excessive medication or treatment that is not medically necessary, practice that violates medical ethics, and other improper conduct in practice — the medical association or the competent authority refers the physician for discipline, and the case is handled by the Physician Disciplinary Committee. Disciplinary measures from lightest to heaviest are: warning → order to complete a set number of additional hours of continuing education or clinical training → restriction of scope of practice, or suspension from practice for not less than one month and not more than one year → revocation of the practice license → revocation of the physician certificate. Note that "revocation of the practice license" and "revocation of the physician certificate" are penalties of very different weight: after the former you can still reapply; the latter is the end of a career.

The Medical Care Act: Institutions, Advertising, Medical Records, and That Line of Liability

The Medical Care Act governs "places and conduct."

Institution classification turns not on bed count but on whether a facility has wards admitting inpatients (Medical Care Act Article 12): one with such wards is a hospital, one offering outpatient care only is a clinic, and a clinic is limited to 9 beds or fewer (observation beds only; obstetric clinics may have up to 10 obstetric beds), so the common shorthand that facilities with 9 or fewer are clinics, and those with 10 or more are hospitals is only a rough mnemonic, not the legal definition; under the Standards for Establishment of Medical Care Institutions, hospitals are divided into general hospitals, chronic-care hospitals, psychiatric hospitals, Chinese medicine hospitals, dental hospitals, and others, while a teaching hospital is an accreditation status (Articles 94–95), not a facility type. In terms of ownership, besides individual practice there are medical corporations (medical foundations and medical associations) — the institutional basis for why Taiwanese hospitals may not be run as for-profit companies, and the legal source of the principle that "medical institutions may not engage in commercial promotion beyond medical advertising."

Medical advertising is a question tested every year. The core logic is a single sentence: medical care is not an ordinary commodity, so advertising works on a "positive list." Medical Care Act Article 85 restricts what may be published to: the name of the medical institution and its opening license number, address and telephone number, physicians' names and their education and experience, clinical departments and hours, opening, closure, suspension, resumption, and relocation, plus other approved matters. Non-medical institutions may not engage in medical advertising (Article 84). And the sentence in Article 87 — "advertising content that implies or alludes to medical services shall be deemed medical advertising" — exists specifically to catch the evasive trick of "not saying I'm a clinic, but writing as if I were one." Conversely, publications in academic journals, medical news or research reports, and patient health education that do not involve soliciting medical business are not deemed medical advertising. So the reason that "No. 1 in the nation, guaranteed results" sign is illegal is not that it brags, but that it is not on the positive list at all.

The medical record retention numbers must be memorized cold (verified as of July 2026, Medical Care Act Article 70): a medical institution must designate appropriate premises and personnel to keep its medical records, and retain them for at least 7 years; records of minors must be retained at least until 7 years after they reach adulthood; records of human trials must be retained permanently. The logic of the three tiers is "limitation period for claims → giving minors their own chance to assert their rights → the generational value of research data."

Consent forms are also frequently tested for which article they sit in: surgical consent and anesthesia consent forms (Article 63) and consent forms for invasive examinations or treatments (Article 64) require explanation to the patient or their legal representative, spouse, relatives, or related persons and obtaining a signed form; but when the patient lacks legal capacity or has limited legal capacity, or in emergencies, other procedures apply. A consent form cannot replace disclosure — a signature without adequate explanation is legally empty.

Advertising works on a positive list because the information gap in medicine is too wide: patients have no way to judge whether your bragging is true.

Medical Care Act Article 82, passed on third reading on December 29, 2017 and promulgated on January 24, 2018, is the single most important line in the whole Act. It frames the liability requirements for medical personnel as follows: for harm caused to a patient in the course of medical practice, they are liable for damages only where it results from intent, or from breach of the duty of care necessary in medicine together with exceeding the bounds of reasonable clinical professional discretion; criminal liability is addressed only for negligently causing a patient's death or injury, again only where there is a breach of that duty of care together with exceeding reasonable clinical discretion (intentional offenses fall outside this article and remain under the general Criminal Code). And breach of the duty of care and the scope of clinical professional discretion shall be judged by the objective circumstances of that medical field at that time and place — its medical customs, medical standards, medical facilities, working conditions, and urgency.

🔬 Mechanism, unpacked

Step one, why amend this article? Because under the old regime, the criminalization of medical disputes turned a physician's "probability of being sued" into a real cost of practice. Step two, once the probability of being sued becomes a cost, the rational response is defensive medicine — order one more CT, request one more consult, transfer high-risk patients elsewhere. Step three, the bill for defensive medicine is paid by everyone, and it makes the five major specialties (internal medicine, surgery, obstetrics and gynecology, pediatrics, emergency medicine) even harder to staff, ultimately harming patients' access to care. Step four, so legislators chose to write the liability requirements out clearly: liability depends not on how terrible the outcome was, but on the medical customs, standards, facilities, working conditions, and urgency at that time and place. Step five, the conclusion — this article is not a "physician immunity clause"; it is a clause fixing the point in time used as the standard of judgment, one that shuts hindsight out the door.

Medical Practice vs. Medical Auxiliary Acts: Where the Line Is Drawn

This is a high-frequency test point shared by nursing and medicine, and its logic is actually very clean. Medical practice (decisions about diagnosis, prescribing, surgery, anesthesia, and invasive procedures) must be carried out by the physician personally — this is the core protected by the offense of unlicensed practice (密醫罪) in Physicians Act Article 28. But if a physician had to do every single thing by hand, the healthcare system would be paralyzed, so the law opened an authorization pipeline: Article 24 of the Nursing Personnel Act lists four areas of nursing practice — nursing assessment of health problems, preventive health care nursing measures, nursing guidance and counseling, and medical auxiliary acts. The first three are nurses' independent practice; the fourth shall be performed under a physician's direction.

The medical auxiliary acts announced by the health authority include: assisting with invasive examinations; assisting with invasive treatments and procedures; assisting with surgery; assisting with childbirth; assisting with radiological examinations and treatment; assisting with chemotherapy; assisting with oxygen therapy (including inhalation therapy) and phototherapy; assisting with drug administration; assisting with psychological and behavioral therapies; monitoring and assessment of patients' vital signs; and other acts recognized by the central health authority, 11 items in all (Department of Health announcement Wei-Shu-Yi-Zi No. 0900017655, March 12, 2001; a separate ruling states that vaccination is a medical auxiliary act; verified September 2026). A true/false question needs only one question: does this task require "deciding," or "executing"? Requires deciding (whether to operate, what drug to prescribe, what the diagnosis is) = medical practice, which no one but a physician may perform; requires executing, and the physician has already given the order (give this injection, change this dressing) = a medical auxiliary act, which nurses may perform under direction.

The Emergency Medical Services Act: Turning "No Time" into a Protocol

The Emergency Medical Services Act deals with the stretch "before the patient reaches the hospital." Its skeleton has four blocks.

The first block is people: emergency medical technicians (EMTs) come in three levels — basic (EMT-1), intermediate (EMT-2), and advanced (EMTP) — and the emergency procedures each level may perform are set by the central competent authority and may not be exceeded.

The second block is institutions: hospitals' emergency care capability classification has advanced, intermediate, and general levels (commonly called the "responsibility hospital classification"). The point of the classification is not prestige but allocation of responsibility: where an ambulance should take a STEMI, a stroke, or major trauma is decided by this classification table. This is exactly the institutionalization driven by the 2005 case in which a girl with a brain injury was shuttled late at night from hospital to hospital, each citing "no beds, no specialist," until the delay in treatment killed her — in essence the same thing as the United States' 1986 Emergency Medical Treatment and Labor Act (EMTALA), which arose from hospitals "dumping" patients who couldn't pay onto public hospitals: in both countries, someone first died after being pushed from place to place, and only then was a law passed saying who is not allowed to push.

The third block is equipment: public places designated by the central competent authority shall be equipped with an automated external defibrillator (AED), and the public is encouraged to learn cardiopulmonary resuscitation.

The fourth block is immunity: a person other than emergency medical personnel who, in order to avert an imminent danger to another's life, uses emergency equipment or administers first aid is covered by the necessity-based exemptions from liability in the Civil Code and the Criminal Code — this is Taiwan's version of the Good Samaritan law. Its reason for existing is quite practical: if bystanders are afraid of being sued, the AED hanging on the wall is just decoration.

⭐ Board-exam essentials

Chapter 8 Must-Knows

Physician certificate (issued centrally, qualification) ≠ practice license (issued locally, practice registration, renewed every 6 years); practice limited to one location.
Continuing education: physicians 120 points per 6 years, of which the professional quality/ethics/regulations categories must total at least 12 points, counted up to 24 (the old pre-2013 rule was 180 and 18 points; must include infection control and gender issues); the same as other practitioner-level medical personnel, i.e. 120/12 points.
Physicians Act Article 21 = the physician's personal duty to rescue; Medical Care Act Article 60 = the institution's duty to provide emergency care.
Physicians Act Article 11: no treatment, prescribing, or diagnostic certificate without personal examination; exception = mountain areas, offshore islands, remote areas, or special urgent circumstances. Telemedicine does not waive the duty to examine.
Discipline from lightest to heaviest: warning → additional continuing education/clinical training → restriction of scope of practice or suspension for 1 month to 1 year → revocation of practice license → revocation of physician certificate.
Institution classification: wards admitting inpatients = hospital, outpatient only = clinic (clinics limited to 9 observation beds; a teaching hospital is an accreditation status, not a type); the juridical-person forms are medical foundations/medical associations.
Medical advertising uses a positive list; non-medical institutions may not engage in medical advertising; implying or alluding is deemed medical advertising; academic publications and health education without solicitation do not count.
Medical record retention: at least 7 years; minors until 7 years after adulthood; human-trial records kept permanently.
Medical Care Act Article 82: civil liability is limited to "intent" or "breach of the duty of care necessary in medicine together with exceeding reasonable clinical professional discretion," and criminal liability covers only negligent death or injury under the same two conditions (promulgated January 2018); both are judged by the medical customs, standards, facilities, working conditions, and urgency at that time and place.
Medical practice = "deciding," done personally by the physician; medical auxiliary acts = "executing," done by nurses under a physician's direction (11 announced items, e.g., assisting with invasive examinations and procedures, assisting with drug administration, monitoring vital signs; vaccination also counts). Nurses' independent practice consists of only three areas: assessment, preventive health care measures, guidance and counseling.
Emergency Medical Services Act: EMTs are EMT-1/EMT-2/EMTP; hospital emergency care capability is classified as advanced/intermediate/general; public places must have AEDs; rescue by people other than emergency personnel is covered by the necessity exemption (Good Samaritan clause).

♪ Sing-along text — English rendering of the Mandarin song lyrics

Taiwan's medical regulations look like a pile of article numbers, but they really answer only three questions: who may do it, where it may be done, and who is responsible when there isn't time to do it. The first question is answered by the Physicians Act, the second by the Medical Care Act, and the third by the Emergency Medical Services Act. Get the order right, and law questions turn from memorizing article numbers into reasoning through logic.

The Physicians Act governs people, and it sets up three gates. The first gate is qualification: pass the examination and hold a physician certificate, issued by the central government only once in a lifetime, which proves who you are. The second gate is practice registration: a certificate alone does not let you see patients; you must apply for practice registration with the competent authority of the city or county where you practice and obtain a practice license before you can practice, and practice is limited to one location. The certificate is who you are, the license is where you are practicing now, and exam questions love to swap the two. The third gate is continuing education: the practice license is renewed every six years, on condition that you complete your credits — physicians need one hundred twenty points every six years, of which the professional quality, professional ethics, and professional regulations categories must total at least twelve points and must include infection control and gender issues, the same as other practitioner-level medical personnel; the old physician rule of one hundred eighty and eighteen points was retired in 2013 with the new regulations.

The certificate says who you are, the license says where you are, and continuing education says whether you've kept up.

Why must continuing education mandate a minimum number of ethics and regulations points? Because professional knowledge updates itself — physicians naturally read new guidelines in order to treat patients — but ethics and regulations won't make today's clinic run any smoother, so without an outside push no one reads them. This is a classic incentive failure: learning with no immediate payoff for the individual but great benefit to society, and the solution is to make it a condition for keeping your license, internalizing an external benefit as a personal cost. Infection control and gender issues are singled out by name not because they matter more than pharmacology, but because they are the least likely to be learned voluntarily.

The Physicians Act has two more provisions tested every year. Physicians Act Article twenty-one says that a physician shall, for a patient in critical condition, immediately provide treatment or take necessary measures according to their professional ability and shall not delay without cause — this is the physician's personal duty. Medical Care Act Article sixty says that hospitals and clinics that encounter a patient in critical condition shall first provide appropriate emergency treatment — this is the institution's duty, and the two must be kept apart. Physicians Act Article eleven says that a physician may not treat, prescribe, or issue a diagnostic certificate without personally examining the patient, but in mountain areas, offshore islands, remote areas, or special urgent circumstances, a physician designated by the competent authority may inquire about the condition, examine, and prescribe by telecommunication, with health personnel then delivering the medication. This narrow door later grew into the Regulations Governing Telemedicine Diagnosis and Treatment. The most important trap is that telemedicine relaxes the requirement of being physically present, not the duty to examine itself, and a wrong prescription still carries liability. As for discipline, from lightest to heaviest: a warning; an order to complete additional continuing education or clinical training; restriction of scope of practice or suspension for one month to one year; revocation of the practice license; revocation of the physician certificate. The last two differ greatly in weight — after the former you can still reapply, while the latter is the end of a career.

The certificate says who you are, the license says where you are, and continuing education says whether you've kept up.

The Medical Care Act governs places and conduct. Institution classification turns not on bed count but on whether there are wards admitting inpatients: a facility with such wards is a hospital, one offering only outpatient care is a clinic, and a clinic may have at most nine observation beds, or up to ten obstetric beds for an obstetric clinic; hospitals are further divided under the establishment standards into general, chronic-care, psychiatric, Chinese medicine, and dental hospitals, among others, while a teaching hospital is an accreditation status rather than a separate type of facility; in terms of ownership there are medical foundations and medical associations, the institutional basis for why Taiwanese hospitals cannot be run as for-profit companies. Medical advertising works on a positive list, because medical care is not an ordinary commodity and patients have no way to judge whether your bragging is true; so the law allows publishing only the institution's name and opening license number, address and telephone, physicians' names, education, and experience, clinical departments and hours, opening, closure, suspension, resumption, and relocation, plus other approved matters. Non-medical institutions may not engage in medical advertising, and content that implies or alludes to medical services is deemed medical advertising — a sentence designed specifically to catch the evasive trick of writing exactly like a clinic while refusing to admit being one. Conversely, academic journal publications, medical news, and patient health education do not count as advertising as long as they do not solicit medical business. The medical record retention numbers must be memorized cold: at least seven years; for minors, at least until seven years after adulthood; human-trial records kept permanently. The logic of these three tiers is the limitation period for claims, giving minors their own chance to assert their rights, and the generational value of research data. Surgical and anesthesia consent forms and consent forms for invasive examinations and treatments all require explanation before signing, but a consent form cannot replace disclosure, and a signature without adequate explanation is legally empty.

Medical Care Act Article eighty-two, passed at the end of 2017 and promulgated in January 2018, is the single most important line in the whole Act. It says civil liability is limited to intent, or to breach of the duty of care necessary in medicine together with exceeding the bounds of reasonable clinical professional discretion, while criminal liability covers only negligently causing death or injury and likewise requires both a breach of that duty and exceeding reasonable discretion, and the benchmark for judgment is the objective circumstances of that medical field at that time and place — its medical customs, medical standards, medical facilities, working conditions, and urgency. Why amend it? Because the criminalization of medical disputes turned the probability of being sued into a real cost of practice, and once that probability becomes a cost, the rational response is defensive medicine — one more CT scan, one more consult, transferring high-risk patients elsewhere. That bill is paid by everyone, and it also makes internal medicine, surgery, obstetrics and gynecology, pediatrics, and emergency medicine even harder to staff, so in the end it is patients' access to care that gets hurt. So this article is not an immunity clause but a clause fixing the point in time used as the standard of judgment, one that shuts hindsight out the door.

There is also the line between medical practice and medical auxiliary acts. Medical practice means the things that require a decision — decisions about diagnosis, prescribing, surgery, anesthesia, and invasive procedures — and it must be done by the physician personally; this is the core protected by the offense of unlicensed practice. But if every task had to be done by a physician's own hands, the system would be paralyzed, so the law opened an authorization pipeline. The Nursing Personnel Act lists four areas of nursing practice: nursing assessment of health problems, preventive health care nursing measures, and nursing guidance and counseling — these three are independent practice — while the fourth, medical auxiliary acts, must be performed under a physician's direction; the 2001 Department of Health announcement lists eleven items, including assisting with invasive examinations, invasive treatments and procedures, surgery, childbirth, radiological examinations and treatment, chemotherapy, and oxygen and light therapy, assisting with drug administration, assisting with psychological and behavioral therapies, and monitoring and assessing vital signs, and vaccination has also been ruled a medical auxiliary act. A true/false question needs only one question: does this task require deciding or executing? What requires deciding belongs to the physician; what requires executing, once the physician has given the order, can be delegated. So a nurse giving vaccines in the community is performing a medical auxiliary act, not independent practice.

Finally, the Emergency Medical Services Act deals with the stretch before the patient reaches the hospital. On people: emergency medical technicians come in basic, intermediate, and advanced levels, and what each may do is set by the central government and may not be exceeded. On institutions: hospital emergency care capability is classified as advanced, intermediate, and general, and the classification is not prestige but allocation of responsibility — where an ambulance should take a heart attack, a stroke, or major trauma is decided by this table. This system was pushed through after a 2005 case in which a girl with a brain injury was shuttled late at night from hospital to hospital, each citing no beds and no specialist, until the delay killed her; in essence it is the same as the United States' Emergency Medical Treatment and Labor Act of 1986 — in both countries, someone first died after being pushed from place to place, and only then was a law passed saying who is not allowed to push. On equipment: designated public places shall be equipped with automated external defibrillators. On immunity: a person other than emergency medical personnel who uses emergency equipment or administers first aid to avert an imminent danger to another's life is covered by the necessity-based exemptions of the Civil Code and the Criminal Code — Taiwan's version of the Good Samaritan clause. Its reason for existing is quite practical: if bystanders are afraid of being sued, the defibrillator on the wall is just decoration.

The certificate says who you are, the license says where you are, and continuing education says whether you've kept up.

9. The Medical Accident Prevention and Dispute Resolution Act: Trading the Courtroom for a Table

Why the Old System Was Doomed to Fail

To understand the Medical Accident Prevention and Dispute Resolution Act (醫療事故預防及爭議處理法), which took effect in 2024 (hereafter the Medical Accident Act), you first have to see the three structural flaws of the old system.

The first flaw was time. Under the old system, a patient could only file a civil suit, lodge a criminal complaint, or ask the health authority to send the case to the Medical Review Committee (醫事審議委員會) for expert assessment. Expert assessment plus the successive levels of appeal routinely took three to five years; by the time a result came out, the patient's harm was long past repair, and the physician's career had already been dragged down.

The second flaw was information. Litigation is an adversarial process, and each side's best strategy is to hide unfavorable information. So the truth does not surface in litigation — litigation was not designed to find causes; it was designed to allocate liability.

The third flaw was learning. A hospital knows a particular operation went wrong, yet dares not write it down, dares not report it, dares not perform a root cause analysis, because anything written down can be subpoenaed. The result is that the same error happens again and again in the same hospital, and every time it is treated as "an individual physician's negligence."

🔬 Mechanism, unpacked

Step one, put these three flaws side by side and you will find they share a common root: the pursuit of accountability and the learning of causes are tied into one and the same procedure. Step two, as long as "telling the truth" raises the probability of "being found liable," rational actors will stop telling the truth. Step three, aviation, nuclear power, and the railways solved this long ago — de-identified reporting + protection for those who report + reported data inadmissible as evidence for assigning blame — so that the learning system and the accountability system run on separate tracks. Step four, this is exactly what the Medical Accident Act does: it did not abolish litigation; instead, it attached a pipeline of "explanation and care → mediation → reporting and learning" in front of litigation, and used limits on the admissibility of evidence to cut this pipeline off from the courtroom. Step five, so the core keyword of the Medical Accident Act is not "compensation" — it is "blame-free error reduction."

Three Principles, Three Mechanisms

The Medical Accident Prevention and Dispute Resolution Act was promulgated on June 22, 2022 (ROC year 111) and, by order of the Executive Yuan, took effect on January 1, 2024 (ROC year 113) (verified as of July 2026). It explicitly declares three principles: immediate care after medical accidents, mediation first for medical disputes, and prevention of medical accidents to improve quality. Each principle is paired with a mechanism.

Mechanism one: the explanation-and-care team. A medical institution shall form a medical accident care team, which, within 5 working days starting from the day after a medical accident occurs, must explain and communicate with the patient, the family, or their representative, and provide assistance and care services; however, hospitals with 99 beds or fewer, and clinics, may designate professional staff or commission a professional institution or organization to do this (verified as of July 2026). The most crucial sentence comes next: expressions of regret, apologies, or similar statements made during the care process may not be used as evidence in litigation or as a basis for a judgment. This one sentence fishes "I'm sorry" out of legal risk and returns it to where it belonged all along — between one human being and another.

Mechanism two: mediation first. In a medical dispute, a party that has not first gone through mediation may not directly file a civil suit; criminal cases involving medical disputes that are under prosecutorial investigation or on trial in court shall be referred to the medical dispute mediation committee of the competent local authority for mediation. Mediation is handled by medical dispute mediation committees set up by the competent authorities of each special municipality and county (city); their members include experts in medicine and law, as well as other persons with professional knowledge or impartial persons of good public standing, and a committee may request a medical dispute assessment (a neutral professional opinion provided by an institution designated by the central authority) as a basis for mediation (⚠️ pending verification: the current number of days allowed for mediation and the number of extensions permitted).

Litigation was not designed to find causes; it was designed to allocate liability — so if you want to find causes, you have to build a separate road.

Mechanism three: reporting, root cause analysis, and prevention. Medical institutions above a certain size shall establish internal mechanisms for managing risk events and handling patient safety events, and for major medical accidents shall conduct a root cause analysis (RCA), propose an improvement plan, and report to the competent authority. The spirit of RCA is systems thinking: it does not ask "who did it wrong?" but "why did this system allow this to happen?" Its methodology comes from quality management — the Swiss cheese model tells us that accidents rarely arise from a single lapse; rather, the holes in multiple layers of defense happen to line up. So the target of improvement is the holes (labeling, workflow, staffing, alert design), not the person who happened to be standing at the mouth of the hole. And the legal condition that holds all of this up is, once again, the same sentence: reported and analyzed data may not be used as evidence in judicial cases or for litigation purposes.

The Childbirth Accident Emergency Relief Act: Taiwan's Pilot Experiment

The Medical Accident Act did not spring from nowhere. It had a forerunner that, counting from its 2012 (ROC year 101) pilot program, had built up more than a decade of experience by the time the Medical Accident Act took effect — the Childbirth Accident Emergency Relief Act (生產事故救濟條例), promulgated on December 30, 2015 (ROC year 104) and in effect from June 30, 2016 (ROC year 105) (verified as of July 2026).

The legislative spirit of this Act was written as four phrases: "no blame, no fault, timely relief, error correction and improvement." It classifies "the risks women face in giving birth" as a social risk that the state should bear — because childbearing creates positive externalities for society, and the incidence of childbirth accidents can never be driven to zero. By design, whenever the mother, fetus, or newborn dies or suffers serious injury as a result of childbirth, relief can be applied for without first proving that the medical personnel were at fault; at the same time, expressions of regret, apologies, or similar statements made by medical personnel during explanation, communication, assistance, or care services may not be used as evidence in litigation or as a basis for a judgment. In exchange, cases that receive relief must cooperate with analysis of the accident's causes and improvement.

🔬 Mechanism, unpacked

Step one, why is no-fault compensation workable? Because it changes the question from "did this physician do something wrong?" to "who should bear this harm?" Step two, once fault no longer has to be proven, the enormous cost of proof disappears, and payment speed shifts from "years" to "months." Step three, because no individual has to be blamed, physicians have no motive to conceal anything, and only then can cause analysis get real data. Step four, so no-fault compensation and error-reduction learning are preconditions for each other — neither works without the other; this is the same logic by which New Zealand's ACC (Accident Compensation Corporation) has, since 1974, replaced personal-injury litigation with comprehensive no-fault compensation. Step five, the price: no-fault compensation does not address the sense of justice. It gives money and gives answers, but it does not give "a person found guilty." That is why it must be paired with care and explanation; otherwise the family's emotions have nowhere to land.

⭐ Board-exam essentials

Chapter 9 Must-Knows

Medical Accident Prevention and Dispute Resolution Act: promulgated June 22, 2022 (ROC 111), in effect January 1, 2024 (ROC 113). Three principles = immediate care, mediation first, prevention to improve quality.
Explanation-and-care team: explain and communicate within 5 working days starting from the day after the accident; hospitals with 99 beds or fewer, and clinics, may delegate this to professional staff/institutions.
Apologies are not evidence: expressions of regret, apologies, or similar statements during the care process may not be used as evidence in litigation or as a basis for a judgment.
Mediation first: without mediation, no directly filing a civil suit; criminal cases shall be referred to mediation; handled by the county/city medical dispute mediation committee, which may request a medical dispute assessment.
Reporting and RCA: major accidents must be reported and undergo root cause analysis; reported and analyzed data may not be used as judicial evidence. The spirit is systems thinking (the Swiss cheese model) — don't ask who was wrong; ask why the system allowed it.
Childbirth Accident Emergency Relief Act: promulgated ROC 104 (2015), in effect June 30, ROC 105 (2016); four principles = no blame, no fault, timely relief, error correction and improvement; the forerunner of the Medical Accident Act.
International comparison: New Zealand's ACC no-fault compensation (since 1974) replaces personal-injury litigation; the U.S. EMTALA (1986) deals with access, not compensation — don't mix them up.
Trap: the Medical Accident Act does not exempt physicians from liability and does not ban litigation; what it changes is the order of procedures and the admissibility of evidence.

♪ Sing-along text — English rendering of the Mandarin song lyrics

Start by seeing why the old system was doomed to fail. A patient bled after surgery, went back to the operating room, and died three days later; in the corridor, the family grabbed the resident and asked what had actually happened, and the standard answer was silence, because the legal department would tell you that any "I'm sorry" could become evidence in court. So the family was left with only two paths, sue or let it go, and the physician with only one, keep quiet — both sides lose, and they keep losing for five years.

The old system had three structural flaws. The first was time: a civil suit, a criminal complaint, or an expert assessment by the Medical Review Committee, plus the successive levels of appeal, routinely took three to five years, and by the time a result came out, the harm was long past repair and the physician's career had been dragged down too. The second was information: litigation is an adversarial process, and each side's best strategy is to hide unfavorable information, so the truth does not surface in litigation, because litigation was not designed to find causes; it was designed to allocate liability. The third was learning: a hospital knows a particular operation went wrong, yet dares not write it down, dares not report it, dares not perform a root cause analysis, because whatever is written down can be subpoenaed; the result is that the same error happens again and again in the same hospital, and every time it is treated as an individual physician's negligence.

Accountability must be pursued and causes must be learned, but the two cannot travel the same road — put them on one road, and no one dares tell the truth.

Put the three flaws side by side and you find the common root: the pursuit of accountability and the learning of causes are tied into one and the same procedure. As long as telling the truth raises the probability of being found liable, rational people will stop telling the truth. Aviation, nuclear power, and the railways solved this long ago — de-identified reporting, protection for those who report, and reported data inadmissible as evidence for assigning blame — so that the learning system and the accountability system run on separate tracks. That is exactly what the Medical Accident Act does: it did not abolish litigation, but attached a pipeline of explanation and care, mediation, and reporting and learning in front of litigation, and used limits on the admissibility of evidence to cut that pipeline off from the courtroom. So the core keyword of the Medical Accident Act is not compensation; it is blame-free error reduction.

The Medical Accident Prevention and Dispute Resolution Act was promulgated on June twenty-second of ROC year one hundred eleven, that is, 2022, and by order of the Executive Yuan took effect on January first of ROC year one hundred thirteen, that is, 2024. It declares three principles — immediate care after medical accidents, mediation first for medical disputes, and prevention of medical accidents to improve quality — and each principle is paired with a mechanism. The first mechanism is the explanation-and-care team: a medical institution must form a care team that, within five working days starting from the day after the accident, explains and communicates with the patient and family and provides assistance and care, while hospitals with ninety-nine beds or fewer and clinics may designate professional staff or commission a professional institution or organization to do it. The most crucial sentence comes next: expressions of regret, apologies, or similar statements made during the care process may not be used as evidence in litigation or as a basis for a judgment — this sentence fishes "I'm sorry" out of legal risk and returns it to the space between one human being and another.

Accountability must be pursued and causes must be learned, but the two cannot travel the same road — put them on one road, and no one dares tell the truth.

The second mechanism is mediation first: a party to a medical dispute who has not first gone through mediation may not directly file a civil suit, and criminal cases involving medical disputes under investigation or on trial shall be referred to the medical dispute mediation committee of the local competent authority; the committee is made up of experts in medicine and law along with impartial persons, and it may request a neutral medical dispute assessment as a basis. Note that this does not strip anyone of the right to sue — if mediation fails, you can still sue, and the limitation period does not run during mediation; what really changes is who speaks first: under the old system the first words came from a lawyer, and under the new system they come from the hospital. The third mechanism is reporting and root cause analysis: medical institutions above a certain size must establish mechanisms for managing risk events and handling patient safety events, and for major accidents must conduct a root cause analysis, propose an improvement plan, and report it. The spirit of root cause analysis is systems thinking — it does not ask who did it wrong, but why this system allowed it to happen. The Swiss cheese model tells us that accidents rarely arise from a single lapse; rather, the holes in multiple layers of defense happen to line up, so the target of improvement is the holes — labeling, workflow, staffing, and alert design — not the person who happened to be standing at the mouth of the hole. And what holds all of this up is, once again, the same sentence: reported and analyzed data may not be used as judicial evidence.

The Medical Accident Act did not spring from nowhere; it had a forerunner that, counting from its 2012 pilot program, had more than a decade of experience by the time the Medical Accident Act took effect, the Childbirth Accident Emergency Relief Act, promulgated on December thirtieth of ROC year one hundred four, that is, 2015, and in effect from June thirtieth of ROC year one hundred five, that is, 2016, whose legislative spirit is no blame, no fault, timely relief, and error correction and improvement. It classifies the risks women face in giving birth as a social risk the state should bear, because childbearing creates positive externalities for society, and the incidence of childbirth accidents can never be driven to zero. Whenever the mother, fetus, or newborn dies or is seriously injured as a result of childbirth, relief can be applied for without first proving fault; medical personnel's apologies are likewise inadmissible as evidence in litigation; and in exchange, cases must cooperate with cause analysis and improvement. Why is no-fault compensation workable? Because it changes the question from whether this physician did something wrong to who should bear this harm; once fault no longer has to be proven, the enormous cost of proof disappears, and payment speed shifts from years to months; because no individual has to be blamed, physicians have no motive to conceal anything, and only then can cause analysis get real data. So no-fault compensation and error-reduction learning are preconditions for each other, and neither works without the other — the same logic by which New Zealand's Accident Compensation Corporation has, since 1974, replaced personal-injury litigation with comprehensive no-fault compensation. Its price is that it does not address the sense of justice: it gives money and gives answers, but it does not give a person found guilty, so it must be paired with care and explanation, or the family's emotions have nowhere to land.

Accountability must be pursued and causes must be learned, but the two cannot travel the same road — put them on one road, and no one dares tell the truth.

10. The Patient-Rights Trilogy: From Withholding, to Deciding for Yourself, to Handing Over Your Body

Patient autonomy in Taiwan grew in three stages, and each stage pushed one step further. To understand how the three relate, you need only ask three questions: Who does it apply to? Who decides? What can be refused?

Part One: The Hospice Palliative Care Act (2000) — Making "Not Doing" Legal

The Hospice Palliative Care Act was promulgated on June 7, 2000 (ROC year 89); it is Taiwan's first law to recognize that "not resuscitating is also a medical decision."

It applies to only one group: terminal patients — people with a serious injury or illness who, as diagnosed by physicians, are considered incurable, and for whom there is medical evidence that progression to death in the near term is unavoidable. The determination requires diagnosis by two physicians, who must hold the relevant specialist qualifications.

Only two things can be refused: cardiopulmonary resuscitation (CPR) and life-sustaining treatment (interventions that maintain vital signs but have no curative effect, such as ventilators, vasopressors, and dialysis). Artificial nutrition and hydration are not among them — this is exactly where that daughter's request got stuck.

Who decides: the patient's own letter of intent takes priority; when the patient is comatose or unable to express their wishes clearly, the closest relative issues a letter of consent. The order of closest relatives is: spouse → adult children and grandchildren → parents → siblings → grandparents → great-grandparents, great-grandchildren, or collateral blood relatives within the third degree → first-degree lineal relatives by marriage. The patient's wishes may be annotated on the National Health Insurance card (NHI IC card).

🔬 Mechanism, unpacked

Step one, why is "withdrawing" harder than "withholding"? Step two, because intuitively, never putting him on the ventilator feels like "letting the disease take him," while taking the ventilator off feels like "I took him away." Step three, but in ethics the two are the same thing — withholding and withdrawing are morally equivalent, because the legitimacy of the decision comes from "this treatment no longer benefits this patient," not from whether the machine happens to be connected right now. Step four, if withdrawal is treated as more serious than withholding, it produces a terrible side effect: physicians will be afraid to start treatment for fear that "they won't be able to stop it," and people who might have been saved lose the chance to try. Step five, so the law later spelled out the procedure for withdrawal (since the 2013 amendment, no medical ethics committee review is required; for a terminal patient with no letter of intent and no closest relative, it is done after a hospice palliative care consultation, on a physician's order issued in the patient's best interest in place of a consent form), so that "daring to start" and "daring to stop" can both hold true at the same time.

Part Two: The Patient Right to Autonomy Act (Promulgated 2016 / In Effect 2019) — Handing the Steering Wheel Back to the Patient

The Patient Right to Autonomy Act was promulgated on January 6, 2016 (ROC year 105) and took effect on January 6, 2019 (ROC year 108); it is Asia's first dedicated law that puts the patient at the center (verified as of July 2026).

Its first breakthrough concerns the order of disclosure. Past practice in Taiwan was "tell the family first, and let the family decide whether the patient should know." Articles 4 and 5 of the Act specify that the patient has the right to know their condition, the medical options, and the possible efficacy, risks, and prognosis of each option, and has the right to choose and decide among the medical options the physician offers; the physician shall inform the patient personally (and, unless the patient expressly objects, may also inform the patient's related persons), and the family may not obstruct this. The patient comes first in line for information — this is the spiritual starting point of the entire Act.

The second breakthrough is planning in advance. A person with full legal capacity (aged 18 or older; since 2023 the Civil Code sets both adulthood and the minimum marriage age at 18, so married minors no longer exist) may, through advance care planning (ACP), sign an advance decision (AD) and have it annotated on the NHI card. ACP is a formal, documented consultation; its participants must include the declarant and at least one relative within the second degree of kinship (with separate provisions for exceptions), and if a health care agent (HCA) has been designated, that agent should also take part; the medical institution's team providing the consultation must include at least one physician, one nurse, and one psychologist or social worker.

The third breakthrough is expansion to five clinical conditions (verified as of July 2026):

#	Clinical condition	Key determination points
1	Terminal illness	Incurable; death in the near term unavoidable
2	Irreversible coma	Brain damage causing loss of consciousness, no response to any stimulus, eyes persistently closed; observation for 3–6 months
3	Permanent vegetative state	Loss of consciousness, but the eyes may open spontaneously or reflex responses are present; unable to care for oneself, learn, or work; observation for 3–6 months
4	Very severe dementia	Persistent impairment of consciousness; unable to care for oneself, learn, or work — e.g., urinary and fecal incontinence, failure to recognize family and friends, loss of judgment
5	Other conditions announced by the central competent authority	Unbearable suffering, an incurable disease, and no other suitable solution under the medical standards of the time

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Determination procedure: each condition must be confirmed by two physicians with the relevant specialist qualifications, and confirmed through at least two consultations with the palliative care team (verified as of July 2026).

The fourth breakthrough is that the scope of what can be refused grows: in addition to life-sustaining treatment (LST), it also includes artificial nutrition and hydration (ANH) — exactly the step the Hospice Act cannot take and the Patient Right to Autonomy Act can.

Comparison	Hospice Palliative Care Act (2000)	Patient Right to Autonomy Act (in effect 2019)
Who it applies to	Terminal patients only	Five clinical conditions
Who decides	The patient's letter of intent; if comatose, the closest relative's letter of consent	The patient's own advance decision (AD); the family may not override it
Prior procedure	Signing the letter of intent is enough	ACP consultation required before signing the AD
What can be refused	CPR, life-sustaining treatment	Life-sustaining treatment + artificial nutrition and hydration
Agent	Health care agent (may express wishes on the patient's behalf)	Health care agent, with more clearly defined powers
Annotation	NHI IC card	NHI IC card

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Part Three: The Human Organ Transplant Act and Brain Death Determination — One Last Act of Autonomy

When autonomy extends beyond death, we enter the territory of the Human Organ Transplant Act. Every provision here exists to block two risks at once: declaring death prematurely, and the buying and selling of organs.

How death is determined. A physician may remove organs from a cadaver only after death has been determined by two or more of the patient's treating physicians; where death is determined by brain death, the Brain Death Determination Guidelines (腦死判定準則) issued by the central competent authority must be followed.

The logic of brain death determination is a strict chain of exclusion (verified as of July 2026):

🔬 Mechanism, unpacked

Step one, first confirm the prerequisites — the patient is in a deep coma, must depend on a mechanical ventilator to breathe, and the cause of the coma has been established. Step two, exclude every reversible cause of coma: hypothermia, drug intoxication or residual sedatives, and metabolic or endocrine disturbances. Step three, wait out an adequate observation period — 12 hours for primary brain injury with a clear cause; with suspected drug intoxication, wait until the drug's half-life has passed, then observe another 12 hours; when the type of drug is unknown, observe for at least 72 hours. Step four, perform the brainstem reflex tests: pupillary light reflex, corneal reflex, oculocephalic reflex (doll's eyes), oculovestibular reflex (cold caloric irrigation), cough reflex, and gag reflex — all must be absent. Step five, perform the apnea test: first give high-concentration oxygen, then disconnect the ventilator and let the PaCO₂ rise to 60 mmHg or above; only if there is still no spontaneous respiratory effort is the test positive. Step six, all of the above must be completed in two determinations, 4 hours apart, and the determining physicians must hold the specified specialist qualifications and training and may not be members of the organ transplant team — this conflict-of-interest rule is the moral fulcrum of the entire procedure.

Where organs come from. Taiwan uses explicit consent (opt-in): the deceased gave written or testamentary consent while alive, or the closest relative consents in writing; the wish to donate can be annotated on the NHI card. This contrasts with Spain's presumed consent (opt-out: no stated objection is treated as consent) combined with its system of dedicated transplant coordinators — Spain has long led the world in donation rates, and what it relies on is not just opt-out but the coordinators present in every hospital, who turn "asking the family" into a standard procedure.

Four locks on living donation. ① The donor must be an adult with mental capacity (the statutory threshold is actually age 20 — higher than the Civil Code's age of majority of 18; the only age exception is that people 18 or older may donate part of the liver to relatives within the fifth degree); ② recipients are limited to blood relatives within the fifth degree of kinship or a spouse; ③ a spouse must have been married to the donor for two years or more or have had a child with the donor (to prevent marrying for an organ); ④ the donation must be reviewed and approved by a medical ethics committee, and the donor must act voluntarily and without payment.

Fairness in allocation. Organ allocation follows the Regulations on the Allocation and Management of Human Organ Transplantation (人體器官移植分配及管理辦法) and rests mainly on medical suitability (blood type, tissue matching, body size, waiting time, disease severity); among the relative factors, one deserves special mention: if a transplant candidate's spouse or a blood relative within the third degree was once a deceased organ donor, the candidate receives allocation priority (verified as of July 2026). The meaning of this rule is not reward but reciprocity — it creates a visible link between "giving an organ away" and "getting one when you are the one waiting."

Every step of brain death determination is not there to prove "he is dead" — it is there to prove "we did not give up too early."

⭐ Board-exam essentials

Chapter 10 Must-Knows

Hospice Palliative Care Act (2000): applies only to terminal patients (diagnosed by two physicians); may refuse CPR and life-sustaining treatment, not including artificial nutrition; if the patient is comatose, the closest relative's letter of consent (order: spouse → adult children and grandchildren → parents → siblings → grandparents → …).
Patient Right to Autonomy Act: promulgated ROC 105, in effect ROC 108 (2019), Asia's first dedicated patient-centered law. The first in line to be informed is the patient personally.
AD requirements: aged 18 or older with full legal capacity → go through ACP (must include at least one relative within the second degree; the team includes a physician + nurse + psychologist or social worker) → sign the AD → annotate it on the NHI card.
Five clinical conditions: terminal illness, irreversible coma, permanent vegetative state, very severe dementia, and announced conditions of unbearable suffering with no solution; confirmation by two specialist physicians + two palliative care team consultations.
The Patient Right to Autonomy Act allows refusing life-sustaining treatment + artificial nutrition and hydration (the step the Hospice Act cannot take).
Patient Right to Autonomy Act ≠ euthanasia: what it permits is refusal/withdrawal, not actively causing death; Taiwan has not legalized euthanasia.
Brain death determination: prerequisites (deep coma, ventilator dependence, established cause) + exclusion of reversible factors → observe 12 hours (drug intoxication: past the half-life, then another 12 hours; unknown drug: at least 72 hours) → all brainstem reflexes absent → apnea test (PaCO₂ ≥ 60 mmHg) → two determinations, 4 hours apart; determining physicians may not be members of the transplant team.
Source of organs: Taiwan uses explicit consent (opt-in), which can be annotated on the NHI card; Spain uses presumed consent (opt-out) + transplant coordinators.
Living donation: the donor must be an adult (the statute says age 20 or older; from 18, only part of the liver may be donated, to relatives within the fifth degree); recipients limited to blood relatives within the fifth degree or a spouse; a spouse must be married 2 years or more or have a child together; buying and selling prohibited.
Allocation priority: if a spouse or a blood relative within the third degree was once a deceased donor, the candidate has allocation priority.

♪ Sing-along text — English rendering of the Mandarin song lyrics

Patient rights in Taiwan grew in three stages, each pushing one step further, and to understand them you need only ask three questions: who does it apply to, who decides, and what can be refused.

The first part is the Hospice Palliative Care Act, promulgated in ROC year eighty-nine, that is, 2000 — Taiwan's first law to recognize that not resuscitating is also a medical decision. It applies to only one group, terminal patients: people with a serious injury or illness who have been diagnosed by physicians as incurable, with medical evidence, and for whom death in the near term is unavoidable — the diagnosis must be made by two physicians, and they must hold the relevant specialist qualifications. Only two things can be refused: cardiopulmonary resuscitation and life-sustaining treatment, meaning interventions that maintain vital signs but have no curative effect, such as ventilators, vasopressors, and dialysis. Artificial nutrition and hydration are not included, so a family's request to stop nasogastric tube feeding has no basis under this law. As for who decides, the patient's own letter of intent comes first; when the patient is comatose or cannot express their wishes clearly, the closest relative issues a letter of consent, in the order of spouse, adult children and grandchildren, parents, siblings, grandparents, then great-grandparents, great-grandchildren, or collateral blood relatives within the third degree, and then first-degree lineal relatives by marriage; the patient's wishes can be annotated on the National Health Insurance card.

Withholding and withdrawing are morally the same thing, because the reason comes from the treatment no longer helping, not from whether the machine is connected right now.

Why does withdrawing feel harder than withholding? Because intuitively, never putting him on the ventilator feels like letting the disease take him, while taking the ventilator off feels like I took him away. But ethically the two are equivalent; the legitimacy comes from the fact that this treatment no longer benefits this patient. If withdrawal is treated as more serious than withholding, it produces a terrible side effect: physicians become afraid to start for fear they won't be able to stop, and people who might have been saved lose the chance to try. So the law later spelled out the procedure for withdrawal — when a terminal patient has no letter of intent and no closest relative, a physician may, after a hospice palliative care consultation, issue an order in the patient's best interest in place of a consent form, and since the 2013 amendment no medical ethics committee review is required — precisely so that daring to start and daring to stop can both hold true at the same time.

The second part is the Patient Right to Autonomy Act, promulgated in ROC year one hundred five and in effect from ROC year one hundred eight, that is, 2016 and 2019 — Asia's first dedicated law that puts the patient at the center. Its first breakthrough concerns the order of disclosure: past practice was to tell the family first and let the family decide whether the patient should know, but this Act specifies that the patient has the right to know their condition, the medical options, and each option's efficacy, risks, and prognosis, and has the right to choose and decide among those options; the physician shall inform the patient personally, the family may not obstruct this, and the patient comes first in line for information. The second breakthrough is planning in advance: a person aged eighteen or older with full legal capacity, after advance care planning, signs an advance decision and has it annotated on the NHI card. The consultation is a formal, documented meeting that must include the declarant and at least one relative within the second degree; if a health care agent has been designated, the agent must take part as well; and the team providing it must include at least one physician, one nurse, and one psychologist or social worker. The third breakthrough is expansion to five clinical conditions — terminal illness, irreversible coma, permanent vegetative state, very severe dementia, and other centrally announced conditions involving unbearable suffering, an incurable disease, and no suitable solution — of which coma and vegetative state require three to six months of observation, and all five require confirmation by two physicians with the relevant specialty plus at least two consultations with the palliative care team. The fourth breakthrough is a wider scope of what can be refused: besides life-sustaining treatment, it also includes artificial nutrition and hydration — exactly the step the Hospice Act cannot take and the Patient Right to Autonomy Act can.

Here is a trap someone falls into every year: the Patient Right to Autonomy Act does not equal euthanasia. What it permits is refusal and withdrawal, letting the disease run its natural course; euthanasia is the active administration of a lethal intervention, and Taiwan has not legalized it.

Withholding and withdrawing are morally the same thing, because the reason comes from the treatment no longer helping, not from whether the machine is connected right now.

The third part is the Human Organ Transplant Act, where every provision blocks two risks at once: declaring death prematurely and the buying and selling of organs. A physician may remove organs from a cadaver only after two or more treating physicians have determined death, and where death is determined by brain death, the Brain Death Determination Guidelines must be followed. Brain death determination is a strict chain of exclusion. First confirm the prerequisites: the patient is in a deep coma, must depend on a mechanical ventilator, and the cause of the coma has been established. Next exclude all reversible causes — hypothermia, drug intoxication or residual sedatives, and metabolic and endocrine disturbances. Then wait out the observation period: twelve hours for primary brain injury with a clear cause; when drug intoxication is suspected, wait past the drug's half-life and then observe another twelve hours; and when the type of drug is unknown, observe for at least seventy-two hours. Then test the brainstem reflexes — pupillary light, corneal, oculocephalic, oculovestibular cold caloric, cough, and gag — all must be absent. Finally, the apnea test: first give high-concentration oxygen, then disconnect the ventilator and let the partial pressure of carbon dioxide rise to sixty millimeters of mercury or above, and only if there is still no spontaneous breathing is the test positive. All of this must be done twice, four hours apart, and the determining physicians must hold the specified specialist qualifications and training and may not be members of the organ transplant team — this conflict-of-interest rule is the moral fulcrum of the whole procedure.

As for where organs come from, Taiwan uses explicit consent: written or testamentary consent given by the deceased while alive, or written consent from the closest relative, and the wish can be annotated on the NHI card. Spain uses presumed consent — no stated objection is treated as consent — plus the transplant coordinators present in every hospital, who turn asking the family into a standard procedure; that is the real reason its donation rate has long been the highest in the world. Living donation has four locks. The donor must be at least twenty with mental capacity, stricter than the civil age of majority of eighteen, and the only age exception is that people eighteen or older may donate part of the liver to relatives within the fifth degree; recipients are limited to blood relatives within the fifth degree or a spouse; a spouse must have been married to the donor for two years or more or have had a child with them, to prevent marrying for an organ; and the donation must be reviewed by a medical ethics committee and be voluntary and unpaid, with buying and selling prohibited. Allocation is based mainly on medical suitability — blood type, tissue matching, body size, waiting time, and disease severity — and among the relative factors, one is worth remembering: a transplant candidate whose spouse or blood relative within the third degree once donated organs after death receives allocation priority. That is not a reward but reciprocity; it creates a visible link between giving an organ away and getting one when you are the one waiting.

Withholding and withdrawing are morally the same thing, because the reason comes from the treatment no longer helping, not from whether the machine is connected right now.

11. The Communicable Disease Control Act: Five Drawers, Two Clocks, and Taiwan's Own Diseases

The Five-Drawer Classification Is Really a Classification by "Response Intensity"

Article 3 of the Communicable Disease Control Act divides notifiable communicable diseases into five categories. Many people assume the basis of classification is "severity," but it isn't — the basis of classification is "how much force the state needs to respond with."

Category	Defining logic	Representative diseases	Reporting deadline
Category I	Requires compulsory isolation and treatment; extremely grave hazard	Smallpox, plague, severe acute respiratory syndrome (SARS), rabies	Within 24 hours
Category II	Requires active surveillance and possible isolation; high risk of local outbreaks	Dengue fever, measles, rubella, diphtheria, typhoid fever, paratyphoid fever, shigellosis, amoebic dysentery, cholera, malaria, hantavirus syndrome, anthrax, enterohemorrhagic E. coli infection, acute viral hepatitis A, multidrug-resistant tuberculosis, acute flaccid paralysis	Within 24 hours
Category III	Requires long-term surveillance and health education; mostly vaccine-preventable or chronic communicable diseases	Tuberculosis, pertussis, Japanese encephalitis, tetanus, enterovirus infection with severe complications, human immunodeficiency virus infection, syphilis, gonorrhea, Hansen's disease, acute viral hepatitis B/C/D/E, mumps, Legionnaires' disease, congenital rubella syndrome (note: invasive pneumococcal disease and influenza with severe complications are Category IV — do not list them here)	Within one week
Category IV	Known communicable diseases outside the first three categories that the central authority deems to require epidemic surveillance or control	Scrub typhus, leptospirosis, Q fever, melioidosis, botulism, Creutzfeldt–Jakob disease, Lyme disease, endemic typhus, severe fever with thrombocytopenia syndrome, invasive pneumococcal disease, influenza with severe complications, COVID-19 (moved from Category V to Category IV on May 1, 2023, ROC year 112; renamed COVID-19 with severe complications, 新冠併發重症, in September 2024)	Per central-authority announcement, varying by disease (botulism, melioidosis, etc.: 24 hours; listeriosis: 72 hours; Creutzfeldt–Jakob disease: one month; the rest mostly within one week)
Category V	Emerging communicable diseases or syndromes outside the first four categories that may affect public health	Novel influenza A, Ebola virus disease, Lassa fever, Marburg hemorrhagic fever, Middle East respiratory syndrome coronavirus infection, yellow fever, Rift Valley fever	Within 24 hours

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(Verified as of July 2026: Categories I, II, and V are reported within 24 hours; Category III within one week; Category IV per central-authority announcement, varying by disease: mostly within one week, but 24 hours for botulism, melioidosis, and others, 72 hours for listeriosis, and one month for Creutzfeldt–Jakob disease.)

🔬 Mechanism, unpacked

Step one, why must "emerging" Category V diseases be reported as fast as Category I — even faster than tuberculosis in Category III? Step two, because reporting deadlines are designed not around how frightening a disease is right now, but around how costly it is not to know. Step three, tuberculosis spreads slowly, we know its natural history, and there are standard regimens — learning of a case three days late will not change the nation's disease-control deployment; but with a never-before-seen pathogen, every day of delay lets exponential growth run one more generation, and we don't even know yet which masks to wear or how many days to isolate. Step four, so Category V shares the "24-hour" clock with Category I, on the grounds that uncertainty itself is a risk. Step five, this also explains why COVID-19 was "downgraded" from Category V to Category IV — not because it became milder, but because it became known, and the state no longer needed to respond with the force reserved for the unknown.

Three Weapons: Isolation, Quarantine, Requisition

Once a notifiable communicable disease is established, the law gives the state three interventions of differing intensity, and each one draws a line between individual liberty and population health.

Isolation and quarantine are not synonyms — this pair of definitions is tested every year: isolation targets people who are already sick or confirmed; quarantine targets people who have been exposed and may be infected but have not yet developed disease. So "home quarantine" is for arrivals from abroad or contacts, while "isolation treatment" is for confirmed cases. Patients with Category I diseases shall be placed in compulsory isolation treatment; patients with Category II and III diseases may be placed in compulsory isolation treatment when necessary. For people who have had contact or are suspected of being infected, the competent authority may impose quarantine, isolation, or other necessary measures.

Requisition is the heaviest weapon. Article 54 of the Communicable Disease Control Act provides that while the Central Epidemic Command Center is in operation, the central competent authority may requisition or deploy privately owned land, structures, buildings, epidemic-prevention equipment, facilities, drugs, medical devices, pollution-treatment facilities, means of transport, and other necessary supplies, and shall provide appropriate compensation. Requisition needs an explicit legal provision because it directly intrudes on the constitutionally protected right to property — emergency powers must have a legal basis, and they must come with compensation and a sunset.

The legal basis of the Central Epidemic Command Center (CECC) is Article 17: when the central competent authority deems it necessary to coordinate all kinds of resources and equipment and to integrate personnel from the relevant agencies, it may, with the approval of the Executive Yuan, establish the center; the commander may direct local competent authorities to carry out control work and may enlist relevant personnel to assist.

Two more articles are easily overlooked. Article 10: competent authorities and medical personnel may not disclose information sufficient to identify a communicable disease patient — disease control and privacy are not an either-or; the law requires both to hold at once. Article 50: the competent authority may restrict or prohibit the moving of the bodies of those who died of communicable diseases, and when necessary may order encoffining and cremation within 24 hours — the most painful controversy during SARS and COVID-19, and a classic question type in which "public health necessity" collides head-on with "freedom of religious and funeral practice."

Taiwan's Own Diseases: Eleven Local Priorities to Cover One by One

Communicable disease questions on the national exam rarely test "Ebola in Africa"; they test the diseases you will actually encounter in Taiwan in your lifetime. For each one below: mechanism first, then the test points.

Dengue fever｜Category II｜reported within 24 hours. The vectors are the yellow fever mosquito (*Aedes aegypti*, found mainly south of Chiayi) and the Asian tiger mosquito (*Aedes albopictus*, found island-wide); both bite during the day and breed in standing water in artificial containers — so the mainstay of control is not spraying insecticide but source reduction (inspect, empty, clean, scrub). Dengue virus has four serotypes, DENV-1 to DENV-4, and infection with one confers lifelong immunity only to that serotype. The real killer is secondary infection with a different serotype: antibodies from the old serotype bind the new virus but cannot neutralize it, and instead let the virus hitch a ride into monocytes via Fc receptors — this is antibody-dependent enhancement (ADE), which leads to massive cytokine release and increased vascular permeability → dengue hemorrhagic fever/dengue shock syndrome. The single most important clinical sentence: the danger comes at the moment the fever "breaks" — days 3–7 after fever onset, and 24–48 hours after defervescence, are the peak of plasma leakage. Warning signs include abdominal pain or tenderness, persistent vomiting, clinical fluid accumulation, mucosal bleeding, lethargy or restlessness, liver enlargement > 2 cm, and a rising hematocrit with a rapid drop in platelets. Treatment is supportive care and fluid replacement; aspirin and nonsteroidal anti-inflammatory drugs (NSAIDs) are contraindicated (they worsen bleeding and platelet dysfunction), and the only acceptable antipyretic is acetaminophen.

Japanese encephalitis｜Category III｜reported within one week. This question has only one trap, but someone steps in it every year: the vector is not *Aedes* but *Culex tritaeniorhynchus*, *Culex annulus*, and *Culex fuscocephala*, which are active at dusk and dawn and breed in rice paddies and large bodies of standing water. The ecological chain: pigs are the amplifying host — the virus replicates massively in pigs and is carried to humans by mosquito bites; humans are dead-end hosts, whose viral load is too low to pass the virus back to mosquitoes, so there is no human-to-human transmission, which is also why the areas around pig farms are high-risk zones. The epidemic season runs from May to October, peaking in June and July. Most infections are asymptomatic, with only about 1 in 300 developing disease, but once disease develops the encephalitis is severe, the case-fatality rate is high, and survivors are often left with motor and cognitive sequelae. Vaccine: Taiwan's current routine childhood schedule has switched to a cell-culture-derived live attenuated chimeric virus vaccine, with dose 1 given at 15 months of age and dose 2 given 12 months later (at 27 months).

Scrub typhus｜Category IV. The pathogen is *Orientia tsutsugamushi* (family Rickettsiaceae), and the vector is the larva of the trombiculid mite (the chigger) — note that only the larval stage bites humans, and it feeds on tissue fluid, not blood. The classic triad is eschar + high fever + lymphadenopathy, often with headache and rash. The eschar is the black crust at the larval bite site, most often found in skin folds along the edges of clothing — the axillae, groin, and waistline — and failing to find one is often simply because the patient was never undressed for the exam. In Taiwan, the offshore islands of Kinmen, Penghu, and Matsu, along with eastern Taiwan, are high-incidence areas. The drug of choice is doxycycline. Why are β-lactams completely ineffective? Because rickettsiae are obligate intracellular bacteria, and their cell walls lack the typical peptidoglycan cross-linking target — the drug can't get in, and there's nothing for it to hit, so you must use a tetracycline that can enter cells.

Hantavirus syndrome｜Category II｜reported within 24 hours. The reservoir is rodents (rats and mice); humans are infected not by being bitten but by inhaling aerosols of dried rodent urine and feces — and that one fact dictates the entire control strategy: never dry-sweep when cleaning up after a rodent infestation; first spray with diluted bleach to wet and disinfect, let it stand, and then wipe clean, or else every puff of dust kicked up by the broom becomes a source of infection. The Old World type (such as the Seoul virus commonly seen in Taiwan) causes hemorrhagic fever with renal syndrome (HFRS) — the triad of fever, bleeding tendency, and acute kidney injury; the New World type causes hantavirus pulmonary syndrome (HPS), dominated by abrupt noncardiogenic pulmonary edema and shock, with an extremely high case-fatality rate.

Rabies｜Category I｜reported within 24 hours. The pathogen is *Lyssavirus* (family Rhabdoviridae). Its mechanism is the most elegant, and the most cruel, pathway in all of infectious disease: the virus enters through the bite wound in saliva → replicates in local muscle → binds acetylcholine receptors to enter nerve terminals → ascends along peripheral nerves by retrograde axonal transport to the spinal cord and brain → after massive replication in the central nervous system, spreads anterogradely to the salivary glands, turning the next bite into transmission. The incubation period is long, 1 to 3 months, and the closer the bite is to the brain, the shorter the incubation — a direct inference from the distance of retrograde transport. Once hydrophobia, aerophobia, pharyngeal muscle spasms, and autonomic dysfunction appear, it is almost 100% fatal. The pathological hallmark is the eosinophilic inclusion in the neuronal cytoplasm, the Negri body. Taiwan had no indigenous human case after 1959 and no animal case after 1961, and was regarded as rabies-free for half a century, but in 2013 rabies virus was detected in wild ferret-badgers (*Melogale moschata*), and Taiwan became a rabies-affected area again — an event that teaches a public health concept: "eradication" refers only to people and domestic animals no longer falling ill; it does not mean the pathogen has vanished from the wildlife ecosystem. The order of post-exposure prophylaxis (PEP) must not be wrong: ① immediately wash the wound with soap and copious clean water for at least 15 minutes (physical removal plus disruption of the lipid envelope — this step yields the greatest benefit); ② do not suture the wound, or delay suturing; ③ for those never vaccinated, give human rabies immune globulin (HRIG) infiltrated around the wound + five doses of vaccine (days 0, 3, 7, 14, 28); ④ those previously fully vaccinated need only 2 booster doses, with no HRIG (because a memory response already exists, and HRIG would actually interfere with active immunity).

Tuberculosis｜Category III｜reported within one week (multidrug-resistant tuberculosis belongs to Category II). Taiwan is still an intermediate-prevalence country. The core of TB control is not new drugs but getting people to finish their drugs — because TB treatment takes six months, and drugs left unfinished breed resistant bacteria. So Taiwan runs the DOTS program (Directly Observed Treatment, Short-course): care workers deliver the pills into the patient's hand, watch them go into the mouth, and leave only after they are swallowed. The standard initial regimen is 2 months of HRZE (isoniazid + rifampin + pyrazinamide + ethambutol) + 4 months of HR. Contact tracing uses testing for latent tuberculosis infection (LTBI) — the interferon-gamma release assay (IGRA) is preferred over the tuberculin skin test because it is not affected by BCG vaccination (which matters especially in Taiwan, where the whole population receives BCG).

Human immunodeficiency virus (HIV) infection｜Category III. Besides the Communicable Disease Control Act, there is a dedicated law, the HIV Infection Control and Patient Rights Protection Act (人類免疫缺乏病毒傳染防治及感染者權益保障條例). The test points focus on rights rather than virology: anonymous screening must be offered; discrimination is prohibited (no refusal in schooling, medical care, employment, residential care, or housing); people living with HIV shall inform medical personnel of their infection when seeking care; anyone who knowingly conceals their infection and engages in risky sexual behavior or shares needles with others bears criminal liability; pregnant women should undergo prenatal HIV screening to enable prevention of mother-to-child (vertical) transmission (antenatal antiretrovirals + cesarean delivery + prophylactic drugs for the newborn + no breastfeeding). Three abbreviations to remember among the prevention tools: PrEP (pre-exposure prophylaxis), PEP (post-exposure prophylaxis, started within the golden 72 hours and continued for 28 days), and U=U (Undetectable = Untransmittable: once the viral load is undetectable, HIV is not sexually transmissible) — this last one is at once a medical fact and anti-stigma policy language.

Enterovirus infection with severe complications｜Category III. Enteroviruses come in dozens of serotypes, but enterovirus 71 (EV71) is the most neuroinvasive. Clinically, first tell its two faces apart: hand, foot, and mouth disease (vesicular rash on the palms, soles, mouth, buttocks, and knees) and herpangina (vesicles and ulcers on the soft palate and posterior pharyngeal wall, with high fever). What you must memorize cold are the warning signs of severe disease — lethargy, altered consciousness, poor activity, limb weakness, myoclonic jerks (sudden whole-body muscle contractions during sleep), persistent vomiting, and tachypnea or tachycardia. Why are these signs so crucial? Because EV71's path to severe disease is brainstem encephalitis → sympathetic storm → neurogenic pulmonary edema → cardiopulmonary failure, and from the first sign to shock may take only a few hours — the time window is extremely narrow. One technical test point in control: alcohol does not work against enteroviruses, because they are non-enveloped viruses, and alcohol acts mainly by destroying lipid envelopes; what works is chlorine bleach (500 ppm for general surfaces, 1000 ppm where contaminated by patient secretions) and thorough handwashing.

Measles｜Category II｜reported within 24 hours. Its transmissibility tops all common communicable diseases, with a basic reproduction number (R₀) of about 12–18, and it spreads by the airborne route — the virus can stay suspended in an enclosed space for up to 2 hours, so patients need negative-pressure isolation and caregivers need N95 respirators. The clinical picture comes in three beats: a prodrome of the 3 Cs (cough, coryza, conjunctivitis) + fever; then, 1–2 days before the rash, Koplik spots — pinpoint white spots on the buccal mucosa and the only pathognomonic sign; then the red rash starts at the hairline behind the ears and spreads downward to the face, trunk, and limbs, leaving desquamation and hyperpigmentation as it fades. The infectious period runs from 4 days before to 4 days after rash onset, which is the basis for identifying contacts. Among complications, pneumonia is the most common cause of death, along with otitis media, encephalitis, and subacute sclerosing panencephalitis (SSPE), which appears years later. Post-exposure management: susceptible individuals receive catch-up MMR within 72 hours of exposure, or immune globulin within 6 days (for infants, pregnant women, and the immunocompromised).

Pertussis｜Category III. The pathogen is *Bordetella pertussis*. The course has three stages: the catarrhal stage (like a cold, the most contagious, and also when antibiotics work best) → the paroxysmal stage (a string of explosive coughs followed by a forceful inspiration that produces the whoop, often coughing until vomiting, lasting several weeks) → the convalescent stage. The most endangered are infants under 6 months: they often do not produce the classic whoop and instead present with apnea or cyanosis, the main cause of pertussis deaths in infants. Treatment and post-exposure prophylaxis for contacts use macrolide antibiotics (azithromycin, etc.). And the most elegant prevention strategy is Tdap vaccination for pregnant women at weeks 28 to 36 of every pregnancy — so that the mother's IgG is actively transported across the placenta to the fetus, covering the immunity gap before the newborn receives the first dose of the pentavalent vaccine at 2 months. This is more direct and effective than the cocooning strategy of vaccinating only family members.

Invasive pneumococcal disease (IPD)｜Category IV｜reported within one week. The key to the definition is "invasive": *Streptococcus pneumoniae* isolated from normally sterile sites such as blood or cerebrospinal fluid. High-risk groups are children under 2, adults 65 and older, and people with asplenia or splenic dysfunction, sickle cell anemia, cochlear implants, cerebrospinal fluid leaks, or immunosuppression — why is asplenia so deadly? Because the spleen is the main organ for clearing encapsulated bacteria; without a spleen, you lose the first line of defense against capsular polysaccharides.

🔬 Mechanism, unpacked

Step one, why do young children get the 13-valent conjugate vaccine (PCV13) rather than the 23-valent polysaccharide vaccine (PPV23)? Step two, pure polysaccharide antigens are T-cell independent: they stimulate B cells directly, but without going through T helper cells there is no class switching and no memory B cells. Step three, and in infants and toddlers under 2, marginal-zone B cells are still immature, so they respond very poorly to such antigens. Step four, a conjugate vaccine covalently links the polysaccharide to a carrier protein, so antigen-presenting cells can load carrier-protein peptides onto MHC II and present them to T cells → it becomes a T-cell–dependent response → producing high-affinity IgG and memory B cells. Step five, so "children get the 13-valent vaccine; older adults and high-risk groups may add the 23-valent to broaden serotype coverage (current: since 2026, Taiwan's publicly funded adult program gives a single dose of PCV20 or PCV21 instead of adding PPV23)" is not a pairing to memorize — it is an inevitability dictated by immunological structure.

⭐ Board-exam essentials

Chapter 11 Must-Knows

The logic of the five categories = the intensity of the state's response; reporting deadlines: Categories I, II, and V within 24 hours, Category III within one week, Category IV per announcement, varying by disease (mostly one week; some 24 hours; Creutzfeldt–Jakob disease one month). COVID-19 has been moved from Category V to Category IV.
Isolation = people already sick; quarantine = exposed but not yet sick. Category I shall be compulsorily isolated; Categories II and III may, when necessary, be compulsorily isolated.
Requisition (Article 54) must take place while the command center is in operation, with appropriate compensation; the legal basis of the Central Epidemic Command Center is Article 17, and it is established with the approval of the Executive Yuan.
Dengue: *Aedes aegypti*/*Aedes albopictus*, daytime biting, standing water in artificial containers; secondary infection with a different serotype → ADE → hemorrhagic fever/shock; 24–48 hours after defervescence is the danger period; no NSAIDs/aspirin.
Japanese encephalitis: *Culex tritaeniorhynchus* and other *Culex* mosquitoes (not *Aedes*); pigs are amplifying hosts, humans are dead-end hosts, no human-to-human transmission; about 1 in 300 develop disease; vaccine dose 1 at 15 months, dose 2 twelve months later.
Scrub typhus: *Orientia tsutsugamushi*, bite of chigger larvae, eschar in clothing folds; drug of choice doxycycline, β-lactams ineffective (obligate intracellular).
Hantavirus: inhalation of aerosolized rodent urine and feces; never dry-sweep during cleanup — wet and disinfect first; HFRS (fever + bleeding + renal failure) / HPS (pulmonary edema).
Rabies: Category I; retrograde axonal transport, incubation 1–3 months and shorter the closer the bite is to the brain; Negri body; ferret-badgers in 2013 made Taiwan an affected area again. PEP = wash with soap for 15 minutes → don't suture → HRIG infiltrated around the wound + vaccine on days 0/3/7/14/28; previously fully vaccinated: only 2 boosters, no HRIG.
Tuberculosis: DOTS — into the hand, into the mouth, leave after it's swallowed; 2 months HRZE + 4 months HR; LTBI via IGRA (unaffected by BCG); MDR-TB is Category II.
HIV: the dedicated law guarantees anonymous screening and non-discrimination; people with HIV must disclose when seeking care; prenatal screening of pregnant women + prevention of vertical transmission (no breastfeeding); PrEP / PEP (within 72 hours, for 28 days) / U=U.
Enterovirus: EV71 is the most neuroinvasive; severe-disease warning signs = lethargy, altered consciousness, poor activity, limb weakness, myoclonic jerks, persistent vomiting, tachypnea or tachycardia; the path is brainstem encephalitis → neurogenic pulmonary edema → cardiopulmonary failure; alcohol is ineffective (non-enveloped) — use chlorine bleach.
Measles: R₀ 12–18, airborne, negative pressure required; 3 Cs + Koplik spots (the only pathognomonic sign); rash spreads downward from the hairline behind the ears; infectious from 4 days before to 4 days after rash onset; post-exposure MMR within 72 hours or immune globulin within 6 days; among complications, pneumonia is the most common cause of death, with SSPE as a late complication.
Pertussis: the catarrhal stage is the most contagious; infants may present with apnea alone; treat with macrolides; Tdap at weeks 28–36 of every pregnancy (better than cocooning).
IPD: isolated from a sterile site; asplenia / under age 2 / 65 and older at high risk; conjugate (PCV13) → T-cell dependent → memory, hence used in young children; pure polysaccharide (PPV23) → T-cell independent, poor response under age 2.

♪ Sing-along text — English rendering of the Mandarin song lyrics

The Communicable Disease Control Act divides notifiable communicable diseases into five categories. Many people assume the basis is severity, but it isn't — the basis is how much force the state needs to respond with. Category I covers diseases that require compulsory isolation and treatment and pose an extremely grave hazard, such as smallpox, plague, severe acute respiratory syndrome, and rabies. Category II covers diseases with a high risk of local outbreaks that require active surveillance and possible isolation, such as dengue fever, measles, rubella, typhoid fever, shigellosis, cholera, malaria, hantavirus syndrome, anthrax, and multidrug-resistant tuberculosis. Category III covers diseases that require long-term surveillance and health education, mostly vaccine-preventable or chronic, such as tuberculosis, pertussis, Japanese encephalitis, tetanus, enterovirus infection with severe complications, human immunodeficiency virus infection, syphilis, and gonorrhea. Category IV covers known communicable diseases outside the first three categories that the central authority considers to need surveillance or control, such as scrub typhus, leptospirosis, melioidosis, Creutzfeldt–Jakob disease, invasive pneumococcal disease, and influenza with severe complications, and COVID-19 was later moved from Category V into Category IV as well. Category V covers emerging diseases, such as novel influenza A, Ebola, and Middle East respiratory syndrome. As for reporting deadlines, Categories I, II, and V are all within twenty-four hours, Category III is within one week, and Category IV follows central announcements and varies by disease — mostly within one week, but twenty-four hours for botulism, melioidosis, and others, and one month for Creutzfeldt–Jakob disease.

A reporting deadline measures not how frightening a disease is now, but how costly it is not to know.

Why must emerging Category V diseases be reported as fast as Category I, even faster than tuberculosis in Category III? Because tuberculosis spreads slowly, we know its natural history, and there are standard regimens, so learning of it three days late won't change the national deployment; but with a never-before-seen pathogen, every day of delay lets exponential growth run one more generation, and we don't even know yet which masks to wear or how many days to isolate — so uncertainty itself is a risk. That also explains why COVID-19 was downgraded from Category V to Category IV: not because it became milder, but because it became known, and the state no longer needed to respond with the force reserved for the unknown.

The law provides three weapons. For the first, you must tell two words apart: isolation targets people who are already sick or confirmed, while quarantine targets people who have been exposed and may be infected but have not yet developed disease — so home quarantine is for arrivals and contacts, and isolation treatment is for confirmed cases; Category I patients shall be placed in compulsory isolation treatment, and Category II and III patients may be when necessary. The second is requisition: while the command center is in operation, the central authority may requisition or deploy land, buildings, epidemic-prevention equipment and facilities, drugs, medical devices, and means of transport, and shall provide appropriate compensation; it requires an explicit legal provision because it directly intrudes on property rights, and emergency powers must have a legal basis, compensation, and a sunset. The third is really organizational: the Central Epidemic Command Center can be established only with the approval of the Executive Yuan. Two more provisions are easily overlooked: competent authorities and medical personnel may not disclose information sufficient to identify a patient, because disease control and privacy are not an either-or; and the competent authority may restrict the moving of the bodies of those who died of communicable diseases and, when necessary, order encoffining and cremation within twenty-four hours — a classic question type in which public health necessity collides head-on with freedom of religious and funeral practice.

Next come Taiwan's own diseases. Dengue fever is transmitted by Aedes aegypti and Aedes albopictus, which bite during the day and breed in standing water in artificial containers, so the mainstay of control is source reduction rather than spraying. The virus has four serotypes, and infection with one gives lifelong immunity only to that type; the real killer is secondary infection with a different serotype, in which antibodies from the old type bind the new virus but cannot neutralize it, instead letting the virus hitch a ride through receptors into monocytes — this is called antibody-dependent enhancement, and it leads to massive cytokine release and increased vascular permeability, progressing to hemorrhagic fever or shock syndrome. The most important clinical sentence is that the danger arrives the moment the fever breaks: days three to seven after fever onset, and twenty-four to forty-eight hours after defervescence, are the peak of plasma leakage, and warning signs include abdominal pain or tenderness, persistent vomiting, fluid accumulation, mucosal bleeding, lethargy or restlessness, liver enlargement of more than two centimeters, and a rising hematocrit with a rapid drop in platelets. Treatment is supportive fluid replacement; aspirin and nonsteroidal anti-inflammatory drugs are contraindicated, and the only acceptable antipyretic is acetaminophen.

Japanese encephalitis has only one trap, but someone steps in it every year: the vector is not Aedes but Culex tritaeniorhynchus and other Culex mosquitoes, which are active at dusk and dawn and breed in rice paddies and large bodies of standing water. The ecological chain is that pigs are the amplifying host — the virus replicates massively in pigs and is then carried to humans by mosquitoes — while humans are dead-end hosts, with no human-to-human transmission, which is why the areas around pig farms are high-risk. The season runs from May to October, peaking in June and July; most infections are asymptomatic, with about one in three hundred developing disease, but once disease develops, the fatality rate is high and the sequelae severe. The current vaccine is a cell-culture live attenuated chimeric virus vaccine, with the first dose at fifteen months of age and the second dose twelve months later.

Scrub typhus is caused by Orientia tsutsugamushi, and the vector is the larva of the trombiculid mite, the chigger; only the larval stage bites humans, and it feeds on tissue fluid. The hallmark is eschar plus high fever plus lymphadenopathy, and the eschar most often lies in skin folds along the edges of clothing, such as the axillae, groin, and waistline — failing to find it is usually just because the patient was never undressed for the exam. In Taiwan, the offshore islands and the east are high-incidence areas. The drug of choice is doxycycline, and beta-lactams are completely ineffective, because rickettsiae are obligate intracellular organisms and lack the typical peptidoglycan target: the drug can't get in, and there's nothing for it to hit.

The reservoir of hantavirus is the rodent; humans are infected not by being bitten but by inhaling aerosols of dried rodent urine and feces, so a rodent-infested area must never be dry-swept — spray it wet with diluted bleach first, let it stand, then wipe. The Old World type causes hemorrhagic fever with renal syndrome — fever and bleeding plus acute kidney injury; the New World type causes hantavirus pulmonary syndrome, dominated by abrupt noncardiogenic pulmonary edema and shock.

A reporting deadline measures not how frightening a disease is now, but how costly it is not to know.

Rabies is Category I. The virus enters through the bite in saliva, replicates in local muscle, binds acetylcholine receptors to enter nerve terminals, ascends by retrograde axonal transport along peripheral nerves to the spinal cord and brain, and after massive replication spreads anterogradely to the salivary glands, turning the next bite into transmission. The incubation period is one to three months, and the closer the bite is to the brain, the shorter it is — a direct inference from transport distance. Once hydrophobia, aerophobia, pharyngeal muscle spasms, and autonomic dysfunction appear, it is almost one hundred percent fatal, and the pathological hallmark is the inclusion in the neuronal cytoplasm called the Negri body. Taiwan had no indigenous human case after 1959 and no animal case after 1961, and was regarded as rabies-free for half a century, but in 2013 the virus was detected in wild ferret-badgers and Taiwan became an affected area again — teaching us that eradication refers only to people and domestic animals no longer falling ill and does not mean the pathogen has vanished from the wildlife ecosystem. The order of post-exposure prophylaxis must not be wrong: first wash with soap and copious clean water for at least fifteen minutes, the step with the greatest benefit; do not suture the wound, or delay suturing; for those never vaccinated, infiltrate immune globulin around the wound and add five doses of vaccine, on days zero, three, seven, fourteen, and twenty-eight; those previously fully vaccinated get only two boosters and no immune globulin, because outside antibodies would interfere with their own memory response.

The core of tuberculosis control is not new drugs but getting people to finish their drugs, because treatment lasts six months and drugs left unfinished breed resistant bacteria, so Taiwan runs directly observed treatment, in which care workers deliver the pills into the hand, watch them go into the mouth, and leave only after they are swallowed. The standard initial regimen is four drugs for two months, then two drugs for four months. Contact tracing for latent tuberculosis infection uses the interferon-gamma release assay, because it is not affected by BCG vaccination, which matters especially in Taiwan, where everyone receives BCG. Multidrug-resistant tuberculosis is listed in Category II.

HIV and AIDS have their own dedicated law, and the test points focus on rights: anonymous screening is offered; discrimination in schooling, medical care, employment, residential care, and housing is prohibited; people living with HIV must disclose their infection when seeking care; those who knowingly conceal their infection and engage in risky sexual behavior or share needles bear criminal liability; and pregnant women should undergo prenatal screening to enable prevention of mother-to-child transmission, including antenatal antiretroviral drugs, cesarean delivery, prophylactic drugs for the newborn, and no breastfeeding. Also remember three abbreviations: pre-exposure prophylaxis; post-exposure prophylaxis, which must start within seventy-two hours and continue for twenty-eight days; and undetectable equals untransmittable.

Among enteroviruses, the most neuroinvasive is type seventy-one, and clinically you first tell apart its two faces, hand, foot, and mouth disease and herpangina. What you must memorize cold are the warning signs of severe disease — lethargy, altered consciousness, poor activity, limb weakness, myoclonic jerks, persistent vomiting, and tachypnea or tachycardia — because the path to severe disease runs from brainstem encephalitis to sympathetic storm to neurogenic pulmonary edema to cardiopulmonary failure, and from the first sign to shock may take only a few hours. There is also a technical test point in control: alcohol does not work against enteroviruses because they have no lipid envelope; what works is chlorine bleach and thorough handwashing.

Measles tops all common communicable diseases in transmissibility, with a basic reproduction number of about twelve to eighteen, and it is airborne — the virus can stay suspended in an enclosed space for two hours — so patients require negative-pressure isolation, and caregivers must wear N95 respirators. The prodrome is cough, coryza, and conjunctivitis plus fever; one or two days before the rash, Koplik spots appear, pinpoint white spots on the buccal mucosa and the only pathognomonic sign; then the red rash spreads downward from the hairline behind the ears to the face, trunk, and limbs, leaving desquamation and hyperpigmentation as it fades. The infectious period runs from four days before to four days after rash onset, which is the basis for identifying contacts. Among complications, pneumonia is the most common cause of death, and there is also the late complication of subacute sclerosing panencephalitis. Post-exposure management is catch-up vaccination for susceptible individuals within seventy-two hours, or immune globulin within six days.

Pertussis has three stages. The catarrhal stage looks like a cold but is the most contagious, and it is also when antibiotics work best; the paroxysmal stage is a string of explosive coughs followed by a forceful inspiration that produces the whoop, often coughing until vomiting, and it can last for weeks; and only then comes the convalescent stage. The most endangered are infants under six months, who often don't produce the classic whoop and instead present with apnea or cyanosis, the main cause of infant pertussis deaths. Treatment and prophylaxis use macrolide antibiotics. The most elegant strategy is for pregnant women to receive pertussis vaccine at weeks twenty-eight to thirty-six of every pregnancy, so the mother's antibodies are transported across the placenta to the fetus, covering the immunity gap before the newborn's first dose at two months — more direct and effective than the cocooning strategy of vaccinating only family members.

The key to the definition of invasive pneumococcal disease is the word invasive: the bacterium must be isolated from a sterile site such as blood or cerebrospinal fluid. High-risk groups are children under two and adults sixty-five and older, as well as people with asplenia or splenic dysfunction, sickle cell anemia, cochlear implants, cerebrospinal fluid leaks, or immunosuppression. Why is asplenia so deadly? Because the spleen is the main organ for clearing encapsulated bacteria. As for why young children get the thirteen-valent conjugate vaccine rather than the twenty-three-valent polysaccharide vaccine, the answer lies in immunological structure: pure polysaccharides are T-cell–independent antigens, and without going through T helper cells there is no class switching and no memory B cells, while in children under two the marginal-zone B cells are still immature and respond very poorly; a conjugate vaccine covalently links the polysaccharide to a carrier protein, so antigen-presenting cells can load the protein's peptides onto major histocompatibility complex class II molecules and present them to T cells, turning it into a T-cell–dependent response that produces high-affinity antibodies and memory B cells. So children getting the conjugate vaccine, older adults and high-risk groups once adding the polysaccharide vaccine, and Taiwan's adult public program switching in 2026 to a single dose of a higher-valent conjugate vaccine are not pairings to memorize — they follow inevitably from immunological structure.

A reporting deadline measures not how frightening a disease is now, but how costly it is not to know.

12. Immunization: Inside the Price of Every Vaccine Dose Lies a Compensation Reserve

🩺 At the bedside

In July 1984, Taiwan's public health centers began giving newborns a vaccine that the rest of the world was still watching warily from the sidelines. At the time, the hepatitis B surface antigen carrier rate among Taiwanese adults was staggeringly high; cirrhosis and hepatocellular carcinoma were a "national disease" in every sense of the word, and physicians had long known that the biggest route of transmission was not shared needles but the mother handing the virus to her child at the moment of birth. Thirteen years later, a National Taiwan University team published a result in *The New England Journal of Medicine*: the incidence of hepatocellular carcinoma among children in the vaccinated cohorts had fallen significantly. It was the first time in the history of medicine that a vaccine had been shown to prevent cancer.

Taiwan's Pride: That One Shot in 1984

Taiwan was the first country in the world to launch a nationwide neonatal hepatitis B vaccination program (it began in July 1984 with newborns of hepatitis B surface antigen (HBsAg)-positive mothers and was extended to all newborns in July 1986). This matters not merely because "we were first," but because it accomplished, all at once, three things that are extremely hard to achieve simultaneously in public health: targeting the highest-flow route, mother-to-child transmission; intervening within 24 hours of birth with vaccine plus HBIG; and turning the blocking of carriage into the prevention of liver cancer decades later.

🔬 Mechanism, unpacked

Step one, to break a chain of transmission, you must first find the segment carrying the greatest flow. In Taiwan, hepatitis B spreads mainly by mother-to-child vertical transmission; e-antigen-positive mothers transmit to their newborns with very high probability, and the younger the age at infection, the higher the proportion who become chronic carriers (about 90% of infections acquired in the neonatal period become chronic carriage, whereas most infections acquired in adulthood clear on their own). Step two, so the best moment to intervene is not adolescence, not adulthood, but within 24 hours after birth — the single most critical judgment of the entire program. Step three, the intervention has to be "double-barreled": active immunization (the vaccine) switches on the infant's own antibody production, while passive immunization (hepatitis B immune globulin, HBIG) immediately neutralizes any virus already encountered, and the two are given together at different sites. Step four, because chronic hepatitis B is the main causal agent of hepatocellular carcinoma, blocking the carrier state is equivalent to blocking cancer decades later — a causal chain spanning thirty years. Step five, so "a drop in childhood liver cancer" was not a side effect of the vaccine but the primary endpoint anticipated from the design stage — first confirmed in children 13 years on (1997), though verifying it into adulthood took about thirty years.

The program's results: the hepatitis B surface antigen (HBsAg) carrier rate in children fell from about 10% before vaccination to below 1%, and the incidence of hepatocellular carcinoma in children dropped significantly (⚠️ pending verification: the exact carrier-rate percentages and the incidence figures in the original publication; expansion to all newborns in July 1986 is verified). This passage deserves a place in the self-identity of every Taiwanese medical student: Taiwan is not a follower in public health policy — on this one, we were the ones who wrote the exam.

The Current Routine Childhood Immunization Schedule: A Living Document That Keeps Being Rewritten

Below is the main framework of Taiwan CDC's current childhood immunization schedule (per the Taiwan CDC schedule in effect since January 2025, verified 2026-09; HPV is the HPA's school-based program).

Age at vaccination	Vaccine
Within 24 hours of birth	Hepatitis B immune globulin (HBIG) (for infants of HBsAg-positive mothers, regardless of e-antigen status since July 2019), hepatitis B vaccine, dose 1
1 month	Hepatitis B vaccine, dose 2
2 months	*Diphtheria-tetanus-acellular pertussis, Haemophilus influenzae* type b, and inactivated polio pentavalent vaccine (DTaP-Hib-IPV), dose 1; 13-valent pneumococcal conjugate vaccine (PCV13), dose 1**
4 months	Pentavalent, dose 2; PCV13, dose 2
5 months	Bacille Calmette-Guérin (BCG), 1 dose
6 months	Hepatitis B vaccine, dose 3; pentavalent, dose 3; influenza vaccine, dose 1
12 months	Measles-mumps-rubella vaccine (MMR), dose 1; varicella vaccine, 1 dose; PCV13 booster; hepatitis A vaccine moved to 18 months (since 2025)
15 months	Japanese encephalitis vaccine (live attenuated chimeric virus vaccine), dose 1
18 months	Pentavalent, dose 4; hepatitis A vaccine, dose 1 (moved here from 12 months in 2025)
27 months	Japanese encephalitis vaccine, dose 2 (12 months after dose 1); hepatitis A vaccine, dose 2
Age 5 until elementary school entry	Diphtheria-tetanus-acellular pertussis and inactivated polio tetravalent vaccine (DTaP-IPV), 1 dose; MMR, dose 2
Elementary to senior high students	Influenza vaccine (school-based mass vaccination); publicly funded human papillomavirus (HPV) vaccine for junior-high students (girls since 2018, boys added September 2025)

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

The three most frequently tested "whys":

Why must hepatitis B dose 1 be given within 24 hours? Because it is simultaneously routine immunization and post-exposure prophylaxis — the newborn has already been exposed in the birth canal, and one day late may be too late.

Why was BCG pushed back to 5 months? Taiwan originally vaccinated within a few days after birth, but because the reporting rate of BCG osteitis as an adverse reaction ran high, the schedule was moved, after evaluation, to 5 months of age (with completion recommended between 5 and 8 months) (in effect since January 1, 2016). This is the best teaching example of the complete loop "safety signal → policy re-evaluation → national schedule revision": the immunization schedule is not carved in stone by a medical textbook; it is a living document that surveillance data can rewrite.

Why was the Japanese encephalitis vaccine switched? The old regimen used an inactivated vaccine manufactured from mouse-brain tissue, which required multiple doses and raised more concerns about reaction rates; the current regimen uses a cell-culture-derived live attenuated chimeric virus vaccine, with the schedule simplified to 2 doses. The switch in vaccines was a switch in manufacturing technology, not a change in the virus.

Where the Line Between Publicly Funded and Self-Paid Vaccines Is Drawn

This is the most commonly misunderstood question. Students often assume that "publicly funded = important disease, self-paid = unimportant disease" — that is wrong. The real criterion is the product of four factors:

Criterion	Content	Example
Disease burden	Incidence × severity × cost of sequelae	Measles is extremely contagious with serious complications → must be publicly funded
Cost-effectiveness	Cost per case averted / per QALY	Comparing a vaccine's unit price with the cost of treating the disease
Herd-immunity externality	Whether vaccination protects "those who are not vaccinated"	Measles and polio have strong externalities → a strong case for public funding
Safety and supply stability	Adverse-reaction rate, stability of production capacity and price	Affects whether it enters the national routine schedule

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Procedurally, the immunization panel of the Ministry of Health and Welfare's Communicable Disease Control Advisory Committee (Advisory Committee on Immunization Practices, ACIP) provides expert recommendations, and the competent authority then decides, based on budget and supply, whether to add a vaccine to the routine schedule.

🦦

PipVaccines are so great — why not just vaccinate everyone and make every vaccine publicly funded? "Universal vaccination is always in the public interest," right?

🐻‍❄️

BarnabyThat sentence is precisely the standard wrong answer choice on the exam. Vaccine policy is never about "does it work" but about "risk–benefit and resource allocation." Three reasons: first, every vaccine has adverse reactions, and when a disease has nearly vanished, the risks may outweigh the benefits (that is exactly why smallpox vaccination was stopped); second, a public budget is exclusive — one more vaccine means one less of something else, so you have to look at cost-effectiveness; third, some diseases have no herd-immunity externality (tetanus, for example, does not spread from person to person), so the justification for public funding has to be found elsewhere. So the criteria are always those four: disease burden, cost-effectiveness, herd externality, safety and supply.

The Vaccine Injury Compensation Fund: The Reserve Hidden in the Price of a Vaccine

The deepest lesson of vaccine policy lies not in how well we vaccinate, but in what we do when vaccination goes wrong.

Legal basis: Article 30 of the *Communicable Disease Control Act* authorizes the establishment of the Vaccine Injury Compensation Fund, with a subsidiary regulation, the Regulations Governing the Collection and Review of the Vaccine Injury Compensation Fund.

Funding source (verified 2026-07): the levy is paid by vaccine manufacturers or importers according to "the number of vaccine doses that pass inspection," on a per-dose basis, with a different per-dose levy for each vaccine (e.g., NT$22 per dose of COVID-19 vaccine, NT$2 per dose of BCG); the total fund may be no less than NT$150 million and no more than NT$400 million — when it falls below the floor, the competent authority adjusts the levy rate.

Review mechanism (verified 2026-07): the Ministry of Health and Welfare has set up the Vaccine Injury Compensation Review Panel (Vaccine Injury Compensation Program, VICP), composed of medical experts in infectious diseases, immunology, neurology, pathology, and other fields, together with legal experts and impartial members of the public; review outcomes are classified by causal relationship into three categories — "related," "cannot be determined," and "unrelated" — and the first two receive compensation payments.

🔬 Mechanism, unpacked

Step one, why handle vaccine injury with a "fund" rather than "litigation"? Step two, because litigation requires the plaintiff to prove negligence and causation, yet vaccine adverse reactions are mostly known, low-probability events that cannot be entirely avoided — no one is at fault — and if you handle them through tort law, the result is inevitably that the injured get nothing. Step three, worse still, if manufacturers exit the market because of litigation risk, the whole population is left with no vaccine at all (the wave of pertussis-vaccine lawsuits in the United States in the 1980s, which sharply shrank the number of vaccine manufacturers, is the empirical proof of where this road leads). Step four, so the solution is to shift compensation from fault-based liability to risk pooling: vaccination is undertaken for the public good of herd immunity, so the losses of the very few who are harmed should be borne jointly by all who benefit, and the most efficient collection point is each and every dose of vaccine. Step five, and so "from day one, a small slice of every vaccine's price was never meant for the drug company — it is skimmed off in advance, waiting to compensate the one unlucky person" — this is the ethical principle of fairness and reciprocity translated into an accounting line item.

The most elegant international comparison is the United States: the 1986 National Childhood Vaccine Injury Act levies (via a companion tax law enacted in 1987) an excise tax, since 1997 a flat US$0.75 per dose for each disease a vaccine prevents, paid into the Vaccine Injury Compensation Trust Fund; the injured file with the "Vaccine Court" (a special proceeding of the Court of Federal Claims), where their case is matched against the qualifying conditions of the Vaccine Injury Table; those who meet them are presumed to have causation and are paid directly. Taiwan wrote it into its Communicable Disease Control Act; the United States wrote it into its tax code and a special statute — yet the conclusion is the same sentence: the price of a vaccine ought, from the start, to include a compensation reserve.

From day one, a small slice of every vaccine's price was never meant for the drug company — it is skimmed off in advance, waiting to compensate the one unlucky person.

⭐ Board-exam essentials

Chapter 12 Must-Knows

In 1984 Taiwan became the first country in the world to launch a nationwide neonatal hepatitis B vaccination program; the main route of transmission is mother-to-child vertical transmission, and the younger the age at infection, the higher the chronic carrier rate, so dose 1 must be given within 24 hours of birth; infants of HBsAg-positive mothers (regardless of e-antigen status since July 2019) also receive HBIG (active + passive immunization, at different sites). Results: childhood HBsAg carrier rate fell below 1% and childhood hepatocellular carcinoma incidence declined — the first vaccine in history proven to prevent cancer.
Key ages in the schedule: within 24 hours, HepB dose 1; 2 months, pentavalent + PCV13; 5 months, BCG; 6 months, first influenza dose; 12 months, MMR + varicella + PCV13 booster; 15 months, JE dose 1; 18 months, pentavalent dose 4 + HepA dose 1; 27 months, JE dose 2 + HepA dose 2; before elementary school, DTaP-IPV + MMR dose 2; publicly funded HPV for junior-high students (boys included since September 2025).
The reason BCG was postponed to 5 months was reports of BCG osteitis — the immunization schedule is a living document that safety data can rewrite.
The Japanese encephalitis vaccine switched from mouse-brain-derived inactivated to cell-culture live attenuated chimeric vaccine (schedule simplified to 2 doses).
Publicly funded vs. self-paid criterion = disease burden × cost-effectiveness × herd-immunity externality × safety and supply; recommendations come from ACIP. "Universal vaccination is always in the public interest" is a false statement.
Vaccine Injury Compensation Fund: legal basis is Article 30 of the *Communicable Disease Control Act*; funded by levies paid by vaccine manufacturers/importers per dose passing inspection; total fund NT$150 million–NT$400 million; decisions are made by the VICP review panel (medical + legal experts + impartial members of the public), with outcomes classified as related / cannot be determined / unrelated, and the first two are compensated.
The underlying spirit: not fault-based liability, but the socialization of risk — herd immunity is a public good, so injury should not be borne by the individual alone.
International comparison: the US 1986 National Childhood Vaccine Injury Act; a companion 1987 law levies an excise tax, since 1997 US$0.75 per dose per disease prevented, paid into a trust fund, handled through the Vaccine Court and the Vaccine Injury Table.

♪ Sing-along text — English rendering of the Mandarin song lyrics

In July 1984, Taiwan's public health centers began giving newborns a vaccine that the rest of the world was still watching from the sidelines. Back then, the hepatitis B carrier rate among Taiwanese adults was staggeringly high, cirrhosis and hepatocellular carcinoma were a national disease in every sense, and physicians had long known that the biggest route of transmission was not shared needles but the mother handing the virus to her child at the moment of birth. Thirteen years later, a National Taiwan University team published a result: the incidence of hepatocellular carcinoma among children in the vaccinated cohorts had fallen significantly. It was the first time in the history of medicine that a vaccine had been shown to prevent cancer. Taiwan was the first country in the world to launch a nationwide neonatal hepatitis B vaccination program.

To break a chain of transmission, find the segment with the greatest flow, then strike at the earliest possible moment.

Why this matters is not only that we were first, but that it accomplished at once three things that are very hard to achieve together: targeting the highest-flow route, mother-to-child transmission; intervening within twenty-four hours of birth with vaccine plus immune globulin; and turning the blocking of carriage into the prevention of liver cancer decades later. To break a chain of transmission, you first have to find the segment with the greatest flow. In Taiwan, hepatitis B spreads mainly by mother-to-child vertical transmission; e-antigen-positive mothers pass it to their newborns with very high probability, and the younger the age at infection, the higher the proportion who become chronic carriers — about nine in ten infections acquired as a newborn become chronic, while most infections acquired in adulthood clear on their own. So the best time to intervene is not adolescence, nor adulthood, but within twenty-four hours after birth, and that is the most critical judgment of the whole program. The intervention has to be double-barreled: the vaccine provides active immunity, while immune globulin provides passive immunity that immediately neutralizes any virus already encountered, and the two are given at different sites. Because chronic hepatitis B is the main causal agent of hepatocellular carcinoma, blocking the carrier state is equivalent to blocking cancer decades later — a causal chain spanning thirty years. So the drop in liver cancer was not a lucky windfall; it was the primary endpoint anticipated at the design stage, first confirmed in children thirteen years on, though verifying it into adulthood took about thirty years.

The current routine childhood immunization schedule can be memorized by age. Within twenty-four hours of birth comes hepatitis B vaccine dose one, with immune globulin given at the same time to infants whose mothers are positive for hepatitis B surface antigen; at one month, hepatitis B dose two; at two months, pentavalent dose one and 13-valent pneumococcal conjugate dose one; at four months, the second doses of those two; at five months, BCG; at six months, hepatitis B dose three and pentavalent dose three, and influenza vaccination begins; at twelve months, measles-mumps-rubella dose one, varicella vaccine, and the pneumococcal booster; at fifteen months, Japanese encephalitis dose one; at eighteen months, pentavalent dose four and hepatitis A dose one; at twenty-seven months, Japanese encephalitis dose two and hepatitis A dose two; from age five until entering elementary school, the tetravalent vaccine and measles-mumps-rubella dose two; and junior-high students receive publicly funded human papillomavirus vaccine, with boys included since September 2025.

There are three whys to think through. Why must hepatitis B dose one come within twenty-four hours? Because it is both routine immunization and post-exposure prophylaxis — the newborn has already been exposed in the birth canal, and a day late may be too late. Why was BCG pushed back to five months? Because the reporting rate of BCG osteitis as an adverse reaction ran high, and after evaluation the schedule was moved later. This is the best teaching example of the complete loop of safety signal, policy re-evaluation, and national schedule revision: the immunization schedule is not carved in stone by a textbook, but a living document that surveillance data can rewrite. Why was the Japanese encephalitis vaccine switched? Because the old regimen used an inactivated vaccine made from mouse-brain tissue, and it has now been replaced by a cell-culture live attenuated chimeric virus vaccine, simplified to two doses — what changed was the manufacturing technology, not the virus.

To break a chain of transmission, find the segment with the greatest flow, then strike at the earliest possible moment.

How is the line between publicly funded and self-paid drawn? People often assume that publicly funded means an important disease and self-paid means an unimportant one, and that is wrong. The real criterion is four factors: disease burden, meaning incidence times severity times the cost of sequelae; cost-effectiveness, meaning how much it costs to avert one case or to gain one quality-adjusted life year; herd-immunity externality, meaning whether my getting vaccinated protects those who are not; and finally safety and supply stability. In decision-making, the immunization panel provides expert recommendations, and the competent authority then decides based on budget and supply. So the sentence "universal vaccination is always in the public interest" is the standard wrong answer, because every vaccine has adverse reactions, and when a disease has nearly vanished the risks may outweigh the benefits — that is exactly why smallpox vaccination was stopped; a public budget is exclusive, so one more vaccine means one less of something else; and some diseases have no herd-immunity externality at all — tetanus does not spread from person to person — so the justification for public funding has to be found elsewhere.

The deepest lesson of vaccine policy lies not in how well we vaccinate, but in what we do when vaccination goes wrong. Article thirty of the Communicable Disease Control Act authorizes the establishment of the Vaccine Injury Compensation Fund. It is financed by vaccine manufacturers or importers paying a levy per dose according to the number of doses that pass inspection, with a different per-dose amount for each vaccine; the fund may be no less than one hundred fifty million and no more than four hundred million New Taiwan dollars, and when it falls below the floor the competent authority adjusts the rate. Review is handled by the Vaccine Injury Compensation Review Panel set up by the Ministry of Health and Welfare, whose members include medical experts in infectious diseases, immunology, neurology, and pathology, as well as legal experts and impartial members of the public; outcomes are classified by causal relationship as related, cannot be determined, or unrelated, and the first two are compensated.

Why a fund rather than litigation? Because litigation requires the plaintiff to prove negligence and causation, yet vaccine adverse reactions are mostly known, low-probability events that cannot be entirely avoided, and no one is at fault; handle them through tort law, and the injured inevitably get nothing. Worse still, if manufacturers leave the market because of litigation risk, the whole population has no vaccine to receive — the wave of pertussis-vaccine lawsuits in the United States in the 1980s drove manufacturers out in large numbers, the empirical proof of where this road leads. So the solution is to shift compensation from fault-based liability to risk pooling: vaccination serves the public good of herd immunity, so the losses of the very few who are harmed should be borne jointly by all who benefit, and the most efficient collection point is every single dose of vaccine. The American version is the National Childhood Vaccine Injury Act of 1986, funded by a vaccine excise tax enacted in 1987 that, since 1997, charges seventy-five cents per dose for each disease a vaccine prevents and flows into a trust fund; the injured go through the Vaccine Court and are matched against the Vaccine Injury Table, and those who meet the conditions are presumed to have causation and are paid directly. Taiwan wrote it into its Communicable Disease Control Act, the United States into its tax code and a special statute, yet the conclusion is the same sentence: the price of a vaccine ought, from the start, to include a compensation reserve.

To break a chain of transmission, find the segment with the greatest flow, then strike at the earliest possible moment.

13. The Prevention of Rare Diseases and Orphan Drug Act: The World's Fifth, and the One Most Like a Complete System of Care

A Dedicated Law That Writes "Prevention and Control" and "Drugs" into a Single Statute

The Rare Disease and Orphan Drug Act was promulgated on February 9, 2000 (ROC year 89) (verified 2026-07; official English title: Prevention of Rare Diseases and Orphan Drug Act), making Taiwan the fifth country in the world to enact a dedicated law for rare diseases — after the United States, Japan, Australia, and the European Union (strictly, the fifth jurisdiction, since the EU is not a country) (verified 2026-07).

But what truly makes this law special is not the ranking of "fifth" but its scope. In most countries, rare-disease legislation is essentially an orphan-drug law — the US *Orphan Drug Act* of 1983 addressed a purely market problem: too few patients for research and development costs to be recouped, so it rewrote the drug companies' financial equation with market exclusivity, tax credits for R&D expenses, fee waivers, and clinical-trial grants. Its subject is how drugs get made.

The Chinese name of Taiwan's law contains two nouns: "prevention and control" and "drugs."

🔬 Mechanism, unpacked

Step one, what is the real predicament of rare-disease patients? It is not only that "there is no drug." Step two, look earlier in the timeline — many rare diseases are inherited metabolic disorders, and if they are not found before symptoms appear, even the most effective drug arrives too late (once the brain damage of PKU has occurred, it is irreversible). So newborn screening, together with genetic counseling and prenatal testing, must be brought in. Step three, look later in the timeline — even once a drug has been made, orphan drugs carry extremely high unit prices, and without reimbursement it is as if there were no drug at all. So National Health Insurance reimbursement, special-access applications, and supply guarantees must be brought in. Step four, look sideways — patients and their families face the pressures of caregiving, schooling, employment, and reproductive choices, so reporting and registry, medical care, supply of nutritional products, and reproductive support must be brought in. Step five, only by connecting these four segments do you get a patient's real life. Taiwan's legislators chose to write "find them," "keep it from happening again," "have drugs available," and "be able to afford them" into a single law — this is its fundamental difference from a pure orphan-drug law.

Institutional framework: the central competent authority establishes the Review Committee on Rare Diseases and Orphan Drugs (Article 4), which is responsible for the recognition and announcement of rare diseases, the review and approval of orphan drugs, the review of special-access applications, and related duties. Recognition of a rare disease centers on prevalence (the current standard is 1 in 10,000 or lower, with heritability and difficulty of diagnosis and treatment also considered; ⚠️ pending verification: the numerical prevalence threshold in the current official announcement).

The three institutional tools for orphan drugs:

1. Special approval for manufacture or import — a drug that has not yet obtained marketing authorization can, after review, be imported on a special-access basis for specific patients, filling the gap where "the drug exists but is not marketed in Taiwan."

2. Market exclusivity — approved orphan drugs enjoy market protection for a set number of years (⚠️ pending verification: the current term).

3. National Health Insurance reimbursement — rare-disease drugs and special nutritional foods for rare diseases are reimbursed by National Health Insurance, with special treatment in the reimbursement rules.

Newborn Screening: The Technology and Ethics Behind the 21 Conditions

Current status (2019 to June 2026; verified 2026-09): the Health Promotion Administration has, since October 1, 2019 (ROC year 108), fully expanded newborn screening for congenital metabolic disorders to 21 conditions; a heel-prick blood sample is collected once the infant is 48 hours old (2 full days); the subsidy is NT$200 per case for ordinary newborns, and NT$550 per case for infants from low-income households and for those born at medical institutions or midwifery clinics in areas lacking medical resources for genetic health services.

From July 1, 2026, the HPA fully subsidizes the NT$750 test fee and has added spinal muscular atrophy (SMA), making 22 publicly funded conditions. The original 21 fall into several broad categories (per the HPA's official list, verified 2026-09): endocrine disorders (congenital hypothyroidism, CHT; congenital adrenal hyperplasia, CAH), amino acid disorders (phenylketonuria, PKU; homocystinuria; maple syrup urine disease, MSUD; citrullinemia types I and II), organic acidemias (methylmalonic acidemia, propionic acidemia, isovaleric acidemia, glutaric acidemia type I, 3-hydroxy-3-methylglutaric aciduria [HMG-CoA lyase deficiency], holocarboxylase synthetase deficiency), fatty acid oxidation disorders (medium-chain acyl-CoA dehydrogenase deficiency, MCAD; very-long-chain acyl-CoA dehydrogenase deficiency, VLCAD; primary carnitine deficiency; carnitine palmitoyltransferase I and II deficiencies, CPT I/II; early-onset glutaric acidemia type II, GA II), carbohydrate metabolism disorders (galactosemia), red-cell enzyme deficiency (glucose-6-phosphate dehydrogenase deficiency, also known as favism, G6PD deficiency). By contrast, the often-listed lysosomal storage disorders (Pompe disease, Fabry disease) and immunodeficiency (severe combined immunodeficiency, SCID) are not among the 21 publicly funded conditions; they are self-paid add-on screens.

🔬 Mechanism, unpacked

Step one, why could the screening panel leap from five conditions in the early years all the way to 21? The intuitive answer is "the government became more generous," but that is wrong. Step two, the real reason is that technology changed the cost structure: in the early days, each disease required its own dedicated biochemical assay, so every added condition added another layer of cost. Step three, once tandem mass spectrometry (MS/MS) arrived, the same drop of blood, in a single run on the instrument, could simultaneously quantify dozens of acylcarnitines and amino acids — in other words, the marginal cost of screening for one more disease approaches zero. Step four, when the marginal cost approaches zero, the decision shifts from "is this condition worth screening for?" to "is there anything we can do about this condition?" Step five, so the real bottleneck for adding conditions is no longer the assay technology but that one line in the Wilson & Jungner screening principles: "Is there an accepted treatment?" — if nothing can be done once it is detected, all screening brings is anxiety and stigma for the family, which is why not every disease that can be measured gets included.

Newborn screening is the textbook example of secondary prevention, and the most elegant real-world landing point of the "three levels, five stages" framework of prevention: the disease already exists (the gene is there), but clinical symptoms have not yet appeared, and the goal of intervention is to change the natural history before irreversible damage occurs. PKU is the classic case: early dietary control with a low-phenylalanine special formula can completely prevent intellectual disability; start three months late, and nothing can be recovered.

🦦

PipSurely with screening, the more you test the better — every extra condition saves one more child, and the machine is already running anyway.

🐻‍❄️

BarnabyThat is exactly the classic mistake. Screening must first pass the Wilson & Jungner test: the disease must be important, there must be a recognizable early asymptomatic stage, there must be an acceptable and effective treatment, the tool must be safe and reliable, and the cost-effectiveness must be reasonable. For a disease that can be detected but not treated, screening only manufactures anxiety and stigma. And one more common error: a positive screen does not equal a diagnosis. Newborn screening is designed to be highly sensitive and less specific, so there must always be recall for repeat testing and confirmatory diagnostic testing — because at this stage, better to recall too many than to miss a single one.

The 21 conditions are not the product of generosity but of marginal cost falling to nearly zero; the real threshold was never the machine — it is "what can you do once you've found it."

⭐ Board-exam essentials

Chapter 13 Must-Knows

The *Prevention of Rare Diseases and Orphan Drug Act* was promulgated on February 9, 2000 (ROC year 89); Taiwan was the fifth jurisdiction in the world to enact a dedicated rare-disease law (after the US, Japan, Australia, and the EU).
What makes it special: it covers both "prevention and control" (newborn screening, genetic counseling, prenatal testing and genetic diagnosis, reporting and registry, reproductive support) and "drugs" (supply, special approval for manufacture or import, market exclusivity, National Health Insurance reimbursement); the US 1983 *Orphan Drug Act* deals only with drug-market incentives (market exclusivity, tax credits, fee waivers).
Governing body: the Review Committee on Rare Diseases and Orphan Drugs; recognition of a rare disease uses prevalence as its core criterion.
Newborn screening: fully expanded to 21 conditions from October 2019 (ROC year 108); 22 with SMA and a fully subsidized test fee from July 2026; heel-prick blood collected at 48 hours of age; the subsidy is NT$200 per case in general before July 2026, and NT$550 for low-income households and areas lacking medical resources.
The 21 publicly funded conditions (Pompe disease, Fabry disease, and SCID are self-paid add-ons, not among them) include: CHT, CAH, PKU, homocystinuria, MSUD, citrullinemia types I/II, galactosemia, G6PD deficiency (favism), MCAD, VLCAD, primary carnitine deficiency, CPT I/II, early-onset GA II, the organic acidemias (including HMG-CoA lyase and holocarboxylase synthetase deficiencies), and others.
The technical reason the panel could expand = tandem mass spectrometry (MS/MS) drives the marginal cost of each added condition toward zero; the real threshold is Wilson & Jungner's "Is there an accepted treatment?"
Newborn screening is secondary prevention; a positive screen ≠ a diagnosis — recall for repeat testing and confirmatory diagnostic testing is required (the design prioritizes high sensitivity).

♪ Sing-along text — English rendering of the Mandarin song lyrics

On the third day of life, a nurse holds a baby girl who has just turned forty-eight hours old, disinfects her heel and gives it a gentle prick, letting a few drops of blood fall into the circles on the filter paper. Two weeks later the phone rings: the tandem mass spectrometer has found abnormally elevated phenylalanine in her blood. At that moment the child's life splits onto two tracks. Without that call, within a few months she would begin to show developmental delay, seizures, and irreversible intellectual disability; with that call, she drinks a special formula from then on and grows up to be an ordinary child. The difference between those two lives is a drop of blood, a machine, and a law.

The Prevention of Rare Diseases and Orphan Drug Act was promulgated on February 9, 2000, making Taiwan the fifth jurisdiction in the world to enact a dedicated law for rare diseases, after the United States, Japan, Australia, and the European Union. But what truly makes this law special is not the ranking of fifth — it is its scope. In most countries, rare-disease legislation is at heart an orphan-drug law. The American Orphan Drug Act of 1983 dealt with a purely market problem: too few patients for research costs to be recouped, so it rewrote the drug companies' financial equation with market exclusivity, research tax credits, fee waivers, and trial grants — its subject is how drugs get made. The name of Taiwan's law holds two nouns: prevention and control, and drugs.

Find them, keep it from happening again, have drugs to use, be able to pay — only when the four segments join do you get a patient's real life.

Why write it this way? Because the real predicament of rare-disease patients is not only that there is no drug. Look earlier in the timeline: many rare diseases are inherited metabolic disorders, and if they are not found before symptoms appear, even the most effective drug arrives too late — once the brain damage of phenylketonuria has occurred, it is irreversible — so newborn screening, genetic counseling, and prenatal testing must be brought in. Look later: even once a drug has been made, orphan drugs carry extremely high unit prices, and without reimbursement it is as if there were no drug, so National Health Insurance reimbursement, special-access applications, and supply guarantees must be brought in. Look sideways: patients and families face the pressures of caregiving, schooling, employment, and reproductive choices, so reporting and registry, medical care, supply of nutritional products, and reproductive support must be brought in. Only by joining these four segments do you get a patient's real life. Institutionally, the central government established the Review Committee on Rare Diseases and Orphan Drugs, responsible for recognizing and announcing rare diseases, reviewing and approving orphan drugs, and reviewing special-access applications, with prevalence as the core criterion for recognizing a rare disease. Orphan drugs have three tools: a drug not yet granted marketing authorization can receive special approval for manufacture or import for use by specific patients; approved orphan drugs enjoy market exclusivity for a set number of years; and rare-disease drugs and special nutritional foods for rare diseases are reimbursed by National Health Insurance.

As for newborn screening today, the Health Promotion Administration fully expanded it to twenty-one conditions on October 1, 2019; heel-prick blood is collected once the baby is forty-eight hours old; ordinary newborns are subsidized two hundred New Taiwan dollars per case, and infants from low-income households or born in areas lacking medical resources for genetic health services are subsidized five hundred fifty dollars per case. The twenty-one conditions cover endocrine disorders such as congenital hypothyroidism and congenital adrenal hyperplasia; amino acid disorders such as phenylketonuria, homocystinuria, maple syrup urine disease, and citrullinemia types one and two; organic acidemias such as methylmalonic acidemia, propionic acidemia, isovaleric acidemia, glutaric acidemia type one, HMG-CoA lyase deficiency, and holocarboxylase synthetase deficiency; fatty acid oxidation disorders such as medium-chain and very-long-chain acyl-CoA dehydrogenase deficiency, primary carnitine deficiency, carnitine palmitoyltransferase deficiency types one and two, and early-onset glutaric acidemia type two; galactosemia among the carbohydrate disorders; and glucose-6-phosphate dehydrogenase deficiency, better known as favism, among the red-cell enzyme deficiencies. Pompe disease, Fabry disease, and severe combined immunodeficiency, often listed here by mistake, are not among the twenty-one publicly funded conditions but self-paid add-on screens. From July 2026, public screening also added spinal muscular atrophy, making twenty-two conditions, with the test fee fully subsidized.

Find them, keep it from happening again, have drugs to use, be able to pay — only when the four segments join do you get a patient's real life.

Why could the panel leap from five conditions in the early years all the way to twenty-one? The intuitive answer is that the government became more generous, but that is wrong. The real reason is that technology changed the cost structure. In the early days, each disease required its own dedicated biochemical assay, so every added condition added another cost; once tandem mass spectrometry arrived, the same drop of blood in a single run could simultaneously quantify dozens of acylcarnitines and amino acids, which means the marginal cost of screening for one more disease approaches zero. When the marginal cost approaches zero, the decision shifts from whether this condition is worth screening for to whether there is anything we can do about it. So the real bottleneck for adding conditions is no longer the assay technology but that one line in the screening principles: is there an accepted treatment? If nothing can be done once it is detected, all screening brings is anxiety and stigma for the family, which is why not every disease that can be measured gets included.

Newborn screening is the textbook example of secondary prevention: the disease already exists, but clinical symptoms have not yet appeared, and the goal of intervention is to change the natural history before irreversible damage occurs. Phenylketonuria is the classic case: early dietary control with a low-phenylalanine special formula can completely prevent intellectual disability, and three months late, nothing can be recovered. Finally, two concepts that are easy to get wrong. Screening is not a case of the more the better; it must first clear the hurdles of importance, a recognizable early asymptomatic stage, an acceptable and effective treatment, a safe and reliable tool, and reasonable cost-effectiveness. And a positive screen does not equal a diagnosis: newborn screening is designed with high sensitivity as the priority, so recall for repeat testing and confirmatory diagnostic testing is always required, because at this stage it is better to recall too many than to miss a single one.

Find them, keep it from happening again, have drugs to use, be able to pay — only when the four segments join do you get a patient's real life.

14. Occupational Safety and Health and Workers' Compensation: Writing the Job into the First Line of the Chart

The Occupational Safety and Health Act: From "Protecting Laborers" to "Protecting All Workers"

The Occupational Safety and Health Act was amended and renamed in 2013 (ROC year 102) from the former *Labor Safety and Health Act*, and its biggest change was in scope of application: from "employed laborers" to "workers" in all industries — including the self-employed and other persons who perform labor under the direction or supervision of the person in charge of the workplace. The significance of this step is its acknowledgment that hazards do not pass you by just because your employment arrangement is different.

The health management the law requires of employers has two tiers:

General health examinations — frequency varies by age: once every 5 years under age 40, once every 3 years from age 40 to under 65, and once a year from age 65 on.

Special health examinations — for those engaged in operations with special health hazards (noise, dust, ionizing radiation, lead, tetraalkyl lead, organic solvents, specified chemical substances, high temperature, abnormal air pressure, yellow phosphorus, bipyridyl or paraquat, etc.; heavy physical labor is not on this list), at least once a year, with graded management under the *Labor Health Protection Rules*:

Level	Examination findings	Required action
Level 1 management	All normal, or partly abnormal but judged by the physician on overall assessment to be not abnormal	General health management
Level 2 management	Partly or wholly abnormal, judged by the physician on overall assessment to be abnormal but unrelated to work	Provide health guidance
Level 3 management	Abnormal, and work-relatedness cannot be determined	An occupational medicine specialist must be consulted for evaluation
Level 4 management	Abnormal, and work-related	Hazard control and related management measures must be implemented immediately

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

🔬 Mechanism, unpacked

Step one, why is the keyword in the grading not "how sick" but "whether it is related to work"? Step two, because the purpose of the special health examination is not to diagnose disease — diagnosing disease is the business of clinical medicine. Step three, its purpose is to feed back into the work environment: if a person's worsening hearing is work-related, what needs fixing is not his ears but the noise of that machine. Step four, so Level 3 (uncertain relationship) is sent up to an occupational medicine specialist for clarification, while Level 4 (confirmed relationship) must trigger hazard control — change the engineering controls, change the protective equipment, change the work method, or even reassign the worker away from the operation. Step five, one sentence to sum it up: the special health examination is "the human-body version of environmental monitoring" — what it measures is not the patient, but the workplace.

How an Occupational Disease Gets "Recognized"

The mistake clinicians make most often is treating "occupational disease" as a diagnosis. It is not — it is a judgment of causal attribution, and it must satisfy five conditions at once:

1. A definite disease exists (objective clinical, imaging, and laboratory evidence);

2. A definite occupational exposure exists (what was done, for how long, at what concentration);

3. There is a plausible temporal sequence between exposure and disease (exposure first, onset later, with a latency consistent with the known natural history);

4. Epidemiologic or toxicologic literature supports that the exposure can cause the disease (ideally with a dose–response relationship);

5. Other, non-occupational causes are reasonably excluded.

These five are in fact an operationalized, individual-case version of the Hill criteria for causation. And the fact that the latency of pneumoconiosis can run 10 to 30 years shows exactly why condition 3 says "latency is consistent" rather than "still employed at onset" — pneumoconiosis that appears only after retirement is still an occupational disease.

From Labor Insurance Add-On to Standalone Insurance: The Labor Occupational Accident Insurance and Protection Act

The *Labor Occupational Accident Insurance and Protection Act* was promulgated on April 30, 2021 (ROC year 110) and took effect on May 1, 2022 (ROC year 111) (verified 2026-07), splitting occupational accident insurance out of Labor Insurance to become its own insurance under its own act, and consolidating the former *Act for Protecting Workers of Occupational Accidents*. Four key changes:

Expanded coverage: workers employed by registered business entities, regardless of the size of the entity (including those with 4 or fewer employees), are compulsorily insured, with coverage taking effect on the first day of employment.
Higher benefits: the upper and lower limits of insured salary were raised, and the levels of injury and sickness, disability, death, and other benefits were increased.
Occupational injury and disease diagnosis and treatment network: a medical network for diagnosing and treating occupational injuries and diseases, along with occupational injury and disease prevention centers, was established, so that "who recognizes occupational disease" has institutions to take it on.
Rehabilitation system: medical, social, and vocational rehabilitation were integrated to address the truly long-term question of "can they go back to work once the injury heals?"

🔬 Mechanism, unpacked

Step one, why split occupational accident insurance out of Labor Insurance? Step two, mainly not because of pricing: old Labor Insurance was a combined scheme in which occupational benefits could not be raised on their own, and workers at firms with fewer than five employees were not compulsorily covered; as for pricing, old Labor Insurance already charged occupational premiums by industry-specific rates, with experience rating for employers of 70 or more, so it is simply not true that high-risk industries never feel the price of their own risk. Step three, the new act keeps and sharpens this pricing: on top of industry-specific accident rates, employers of 50 or more pay premiums adjusted by industry-specific experience rating (based on three years of claims plus safety performance) — within the same industry, companies with more accidents pay more. Step four, this turns "safety" from a moral exhortation into a cost that hits the company's bottom line, giving employers a financial reason to invest in protective equipment. Step five, it follows the same line of thinking as US black lung compensation: the *Federal Coal Mine Health and Safety Act* of 1969 created the benefits program, and the *Black Lung Benefits Revenue Act* of 1977 taxed coal to fund the Black Lung Disability Trust Fund — forcing the cost of disease to be internalized back into the industry itself.

Overwork: Epidemiology's Dose–Response Written into Regulation

In Taiwan, "overwork" (karoshi) is not just an adjective; it has an administrative recognition document you can look up: the Ministry of Labor's Reference Guidelines for the Recognition of Work-Related Cerebrovascular and Cardiac Diseases (Excluding Those Caused by Trauma). It breaks workload down into three time scales:

Dimension assessed	Observation period	Content
Abnormal events	From the time of onset back to the previous day	Mental strain, physical load, or abrupt changes in the work environment
Short-term excessive workload	About 1 week before onset	Exceptionally heavy workload
Long-term excessive workload	About 6 months before onset	Accumulated fatigue quantified by overtime hours

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Quantitative thresholds for long-term excessive workload (verified 2026-07):

More than 100 hours of overtime in the 1 month before onset, or an average of more than 80 hours of overtime per month over the 2 to 6 months before onset → very strong association with onset;
An average of 45 to 80 hours of overtime per month over the 1 to 6 months before onset → the longer the overtime, the stronger the association;
An average of less than 45 hours of overtime per month → weak association.

Target diseases include cerebrovascular conditions (cerebral hemorrhage, subarachnoid hemorrhage, cerebral infarction, hypertensive encephalopathy) and cardiac conditions (myocardial infarction, acute heart failure, aortic dissection, angina pectoris, severe arrhythmia, sudden cardiac death).

This guideline's teaching value far exceeds its administrative force: it is one of the clearest examples (noise exposure limits are another case of dose–response written into rules) of the Hill criterion of "dose–response relationship" being written by the state into a threshold that pays out benefits. Keep its legal rank in mind as well — it is an administrative guideline, not a law, and its function is to standardize recognition criteria and reduce case-by-case inconsistency.

Six Occupational Diseases You Must Know in Taiwan

Noise-induced hearing loss. Long-term exposure at 85 dBA or above. Mechanism: as acoustic energy travels through the cochlea, the outer hair cells of the basal turn bear the greatest mechanical stress and are damaged first, and the basal turn is precisely the region responsible for high frequencies — so the pure-tone audiogram first shows a dip at 4000 Hz (the 4k notch), and only with further deterioration does the loss erode the speech frequencies (500–2000 Hz); this is also why the stage in which the patient "can hear people talking but can't make out the words" is the easiest to overlook. Its hallmarks are sensorineural, bilateral and symmetric, and irreversible. On the regulatory side: an 8-hour time-weighted average sound level (TWA) of 85 dBA requires a hearing conservation program; 90 dBA is the statutory permissible exposure limit, with a 5-dB exchange rate — for every 5-dB increase, the permissible exposure time is halved (90 dBA for eight hours, 95 dBA for four hours, 100 dBA for two hours).

Pneumoconiosis.

Type	Exposure settings	Imaging features	Associated risks
Silicosis	Tunneling, sandblasting, quarrying, ceramics, quartz countertop fabrication	Upper-lobe nodules, eggshell calcification of lymph nodes, progressive massive fibrosis	Markedly increased tuberculosis risk, lung cancer
Asbestosis	Shipbuilding, insulation materials, brake linings, demolition of old building materials	Lower-lobe interstitial fibrosis, pleural plaques	Malignant mesothelioma, lung cancer
Coal workers' pneumoconiosis (CWP)	Coal mining	Coal macules, progressive massive fibrosis (PMF)	Chronic obstructive pulmonary disease

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Asbestos comes with one must-know interaction: asbestos and smoking have a multiplicative effect on "lung cancer", but asbestos-induced malignant mesothelioma shows no clear multiplicative relationship with smoking — mesothelioma is almost asbestos's exclusive signature.

Lead poisoning. Sources of exposure: lead-acid batteries, lead soldering, scrap-metal recycling, and lead-containing paints and glazes. The mechanism is an elegant chain of enzyme inhibition: lead inhibits δ-aminolevulinic acid dehydratase (ALAD) and ferrochelatase → heme synthesis is blocked → microcytic anemia appears, residual ribosomes in red cells form basophilic stippling, and at the same time urinary δ-aminolevulinic acid (δ-ALA) and blood zinc protoporphyrin (ZPP) rise. Neurologically it is predominantly motor, the classic sign being wrist drop from radial nerve palsy (with sensation relatively spared); in the gut, lead colic and the gingival lead line (Burton's line); in the kidney, proximal tubular dysfunction and saturnine gout (reduced uric acid excretion). Monitoring uses blood lead. Management: removing the exposure is always the first step, and severe cases are treated with chelating agents (CaNa₂EDTA, succimer/DMSA, BAL).

Hand-arm vibration syndrome (HAVS). Exposure to localized vibration from rock drills, chainsaws, pneumatic tools, and the like. Mechanism: repeated vibration causes digital artery endothelial injury and an exaggerated sympathetic vasoconstrictor response, plus peripheral sensory neuropathy → secondary Raynaud's phenomenon (commonly called vibration white finger): on exposure to cold, the fingertips go through the triphasic sequence of pallor → cyanosis → rubor, accompanied by numbness and difficulty with fine movements. Key differential: primary Raynaud's disease is more common in young women and is bilateral and symmetric; vibration white finger has a clear history of vibrating-tool exposure and often appears first in the dominant hand.

Occupational skin disease. It is among the most frequently reported occupational diseases worldwide, and the key point is that the two types have entirely different mechanisms:

	Irritant contact dermatitis (ICD)	Allergic contact dermatitis (ACD)
Mechanism	Non-immunologic direct chemical/physical injury	Type IV (delayed-type) hypersensitivity, T-cell mediated
Prior sensitization needed?	No — can occur on first contact	Yes — flares only on re-exposure after sensitization
Timing	Minutes to hours after contact	24–72 hours after re-exposure
Dose relationship	Clearly dose-related	Trace amounts can trigger it
Share of cases	About 80%, more common	About 20%
Diagnosis	History and exposure assessment	Patch test
Common causes	Wet work, detergents, cutting oils, solvents	Chromates (cement, leather), nickel, rubber accelerators, epoxy resins

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

Organic-solvent exposure and causal recognition: the RCA Taoyuan plant. The Radio Corporation of America (RCA) set up a plant in Taoyuan in 1970 and pulled out in 1992; for years it used organic solvents such as trichloroethylene and tetrachloroethylene, contaminating the groundwater at the site. After the contamination came to light in 1994, the employees' self-help association launched a collective lawsuit that dragged on for more than twenty years, part of which was made final by a Supreme Court judgment in 2018. The teaching value of this case is enormous: when the company has already been dissolved, the plant demolished, and individual exposure records long since lost, how do you prove "it was them"? The answer — the evidence the court ultimately accepted was population-level epidemiological inference, not any one person's medical record. That single sentence turns three abstract exam points — "cohort studies can establish temporality," "the Hill criteria for causation," and "the allocation of the burden of proof" — into a real courtroom battle.

The special health examination measures not the patient but the workplace — what Level 4 management has to fix is not that person's ears, but that machine.

⭐ Board-exam essentials

Chapter 14 Must-Knows

The Occupational Safety and Health Act (renamed in ROC year 102 [2013] from the Labor Safety and Health Act) extends its scope to "workers" in all industries (including the self-employed).
General health exam: under 40 every 5 years, 40–64 every 3 years, 65 and over every year; special health exam: for operations with special health hazards, once a year.
Graded management of special health exams: Level 1 = not abnormal; Level 2 = abnormal but unrelated to work; Level 3 = abnormal but work-relatedness uncertain → refer to an occupational medicine specialist; Level 4 = abnormal and work-related → immediate hazard control.
Five criteria for recognizing occupational disease: evidence of disease, evidence of exposure, plausible temporal sequence, support from the literature and a dose–response relationship, exclusion of non-occupational causes. Pneumoconiosis latency is 10–30 years; onset after retirement still counts as occupational disease.
Labor Occupational Accident Insurance and Protection Act: promulgated in ROC year 110, in force May 1, 2022 (ROC year 111); occupational accident insurance becomes a standalone insurance under its own act; compulsory coverage from the first day of employment regardless of employer size; establishes the occupational injury and disease diagnosis and treatment network and triple rehabilitation — medical / social / vocational.
Overwork recognition (guideline on work-related cerebrovascular and cardiac diseases): > 100 hours of overtime in the 1 month before onset, or a monthly average > 80 hours over the preceding 2–6 months → very strong association; monthly average 45–80 hours, the longer the stronger; < 45 hours, weak. It is an administrative guideline, not a law.
Noise-induced hearing loss: basal-turn outer hair cells damaged first → 4000 Hz notch; sensorineural, bilateral, irreversible; 85 dBA triggers a hearing conservation program, 90 dBA is the permissible limit, 5-dB exchange rate.
Silicosis = upper lobes + eggshell calcification + increased TB risk; asbestosis = lower lobes + pleural plaques + mesothelioma; asbestos × smoking is multiplicative for lung cancer, not for mesothelioma.
Lead poisoning: inhibits ALAD and ferrochelatase → microcytic anemia + basophilic stippling + elevated ZPP/δ-ALA; wrist drop (predominantly motor), Burton's line, lead colic, saturnine gout; monitor with blood lead; remove the exposure first, then chelate.
HAVS: localized vibration → endothelial injury and sympathetic over-reaction → vibration white finger (pallor → cyanosis → rubor); the distinction from primary Raynaud's disease lies in the exposure history.
Occupational skin disease: ICD (non-immunologic, can occur on first contact, dose-related, about 80%) vs ACD (type IV delayed hypersensitivity, requires sensitization, 24–72 hours, patch test).
RCA Taoyuan plant (1970–1992, trichloroethylene/tetrachloroethylene): with exposure records lost, the court adopted population-level epidemiological inference — a real-life case of temporality, the Hill criteria, and the burden of proof.

♪ Sing-along text — English rendering of the Mandarin song lyrics

A fifty-four-year-old man came to the chest clinic with progressive shortness of breath. His chest X-ray showed multiple small nodules in the upper lobes, and several mediastinal lymph nodes showed eggshell calcification. The physician asked the one question no one else had asked — what did you used to do for work — and the answer was tunnel construction, twenty-two years, retired two years ago. Dust from twenty-two years ago, a clinic visit two years on. The greatest diagnostic barrier in occupational disease is not that the test can't be done; it is that no one asks that one question.

The Occupational Safety and Health Act was renamed from the Labor Safety and Health Act in 2013, and its biggest change was its scope of application, expanding from employed laborers to workers in all industries, including the self-employed and others who perform labor under the direction or supervision of the person in charge of a workplace — because hazards do not pass you by just because your employment arrangement is different. The health management the law requires has two tiers. General health examinations vary in frequency by age: once every five years under forty, once every three years from forty to under sixty-five, and once a year from sixty-five on. Special health examinations are for those engaged in operations with special health hazards, at least once a year, with graded management. Level one is all normal, or partly abnormal but judged not abnormal on overall assessment; level two is abnormal but unrelated to work; level three is abnormal with work-relatedness that cannot be determined, and an occupational medicine specialist must evaluate it; level four is abnormal and work-related, and hazard control must be implemented immediately.

The special health exam measures not the patient but the workplace; what level four must fix is not the ears, but that machine.

Why is the keyword in the grading not how sick someone is, but whether it is related to work? Because the purpose of the special health examination is not to diagnose disease — diagnosing disease is the business of clinical medicine — but to feed back into the work environment. If a person's worsening hearing is work-related, what needs fixing is not his ears but the noise of that machine. So level three is sent up to a specialist for clarification, and level four must trigger hazard control: change the engineering controls, change the protective equipment, change the work method, or even reassign the worker away from the operation.

Occupational disease is not a diagnosis; it is a judgment of causal attribution, and it must satisfy five conditions at once: a definite disease exists; a definite occupational exposure exists; the temporal sequence between exposure and disease is plausible, with a latency consistent with the known natural history; epidemiologic or toxicologic literature offers support, especially a dose–response relationship; and other non-occupational causes are reasonably excluded. These five are in fact an individual-case, operational version of Hill's criteria for causation. And the fact that pneumoconiosis can have a latency of ten to thirty years shows exactly why the third condition says the latency must be consistent, rather than that the person must still be employed at onset — pneumoconiosis that appears only after retirement is still an occupational disease.

The big institutional event is the Labor Occupational Accident Insurance and Protection Act, promulgated in 2021 and in force from May 1, 2022, which split occupational accident insurance out of Labor Insurance into its own insurance under its own act, and consolidated the former Act for Protecting Workers of Occupational Accidents. There are four changes. Expanded coverage: workers employed by registered business entities, regardless of size, are insured from their first day on the job. Higher benefits: the upper and lower limits of insured salary and the level of each benefit were raised. A medical network for diagnosing and treating occupational injuries and diseases, so that the question of who recognizes occupational disease has institutions to take it on. And the integration of medical, social, and vocational rehabilitation, to address the truly long-term question of whether people can return to work once the injury heals. Why split occupational accident insurance out? Mainly because the old scheme was bundled into the combined Labor Insurance statute, so occupational benefits could not be raised on their own, and small firms with fewer than five employees were not compulsorily covered. As for pricing, old Labor Insurance already set occupational premiums by industry and applied experience rating to employers of seventy or more; the new act keeps this, lowers the threshold to fifty, and adds grading by safety performance, so within an industry those with more accidents pay more, and safety turns from a moral exhortation into a cost that hits the bottom line, giving employers a financial reason to invest in protective equipment. This follows the same line of thinking as US black lung compensation: the Federal Coal Mine Health and Safety Act of 1969 created the benefits program, and the Black Lung Benefits Revenue Act of 1977 taxed coal to fund the Black Lung Disability Trust Fund — forcing the cost of disease to be internalized back into the industry itself.

The special health exam measures not the patient but the workplace; what level four must fix is not the ears, but that machine.

Overwork in Taiwan is not just an adjective; there is a recognition reference guideline you can look up, and it breaks workload into three time scales. Abnormal events cover the period from onset back to the previous day; short-term excessive workload covers about one week before onset; long-term excessive workload covers about six months before onset and is quantified by overtime hours. More than one hundred hours of overtime in the month before onset, or an average of more than eighty hours per month over the preceding two to six months, means a very strong association; with an average of forty-five to eighty hours per month, the longer the overtime, the stronger the association; below forty-five hours, the association is weak. Target diseases include cerebral hemorrhage, subarachnoid hemorrhage, cerebral infarction, and hypertensive encephalopathy, as well as myocardial infarction, acute heart failure, aortic dissection, angina pectoris, severe arrhythmia, and sudden cardiac death. The teaching value of this document far exceeds its administrative force: it is one of the clearest examples of the dose–response criterion being written by the state into a threshold that pays out benefits — but remember that its rank is that of an administrative guideline, not a law.

Next come the occupational diseases you must know in Taiwan. Noise-induced hearing loss comes from long-term exposure at eighty-five decibels or above. The mechanism is that the outer hair cells of the cochlea's basal turn bear the greatest mechanical stress and are damaged first, and the basal turn handles high frequencies, so the pure-tone audiogram first shows a dip at four thousand hertz, and only with further deterioration does the loss erode the speech frequencies — which is also why the stage in which the patient can hear but cannot hear clearly is the easiest to overlook. It is sensorineural, bilateral and symmetric, and irreversible. By regulation, an eight-hour time-weighted average sound level of eighty-five decibels requires a hearing conservation program, ninety decibels is the permissible exposure limit, and a five-decibel exchange rate applies: for every five-decibel increase, the permissible time is halved.

There are three types of pneumoconiosis to tell apart. Silicosis comes from tunneling, sandblasting, quarrying, ceramics, and quartz countertop fabrication; on imaging it shows upper-lobe nodules and eggshell-calcified lymph nodes, and its most important complication is a markedly increased risk of tuberculosis. Asbestosis comes from shipbuilding, insulation materials, brake linings, and the demolition of old building materials; on imaging it shows lower-lobe interstitial fibrosis and pleural plaques, and its complications are malignant mesothelioma and lung cancer. Coal workers' pneumoconiosis comes from coal mining and can progress to progressive massive fibrosis. Asbestos has one must-know interaction: asbestos and smoking have a multiplicative effect on lung cancer, but asbestos-induced malignant mesothelioma shows no clear multiplicative relationship with smoking — mesothelioma is almost asbestos's exclusive signature.

Lead poisoning comes from lead-acid batteries, lead soldering, scrap-metal recycling, and lead-containing paints and glazes. Lead inhibits aminolevulinic acid dehydratase and ferrochelatase, blocking heme synthesis, so microcytic anemia and basophilic stippling in the red cells appear, and urinary aminolevulinic acid and blood zinc protoporphyrin rise. Neurologically it is predominantly motor, the classic sign being wrist drop from radial nerve palsy, with sensation relatively spared; in the gut there are lead colic and the gingival lead line; in the kidney, proximal tubular dysfunction and saturnine gout. Monitoring uses blood lead, and the first step in management is always removing the exposure, with chelating agents used only in severe cases.

Hand-arm vibration syndrome comes from the localized vibration of rock drills, chainsaws, and pneumatic tools. Repeated vibration causes digital artery endothelial injury and an exaggerated sympathetic vasoconstrictor response, which, together with peripheral sensory neuropathy, produces secondary Raynaud's phenomenon, known as vibration white finger: on exposure to cold, the fingertips turn pale, then blue, then red in sequence. The key to distinguishing it from primary Raynaud's disease is the exposure history — the primary form is more common in young women and is bilateral and symmetric, while vibration white finger has a clear history of tool exposure and often starts in the dominant hand.

For occupational skin disease, remember that the two types have entirely different mechanisms. Irritant contact dermatitis is non-immunologic direct chemical or physical injury; it needs no prior sensitization, can occur on first contact, is clearly dose-related, and accounts for about eighty percent of cases, commonly arising from wet work, detergents, cutting oils, and solvents. Allergic contact dermatitis is type IV delayed-type hypersensitivity mediated by T cells; it requires prior sensitization, flares twenty-four to seventy-two hours after re-exposure, can be triggered by trace amounts, and is diagnosed by patch testing, with common sensitizers being the chromates in cement and leather, nickel, rubber accelerators, and epoxy resins.

Finally, the case of the Radio Corporation of America's Taoyuan plant. The company moved in in 1970 and pulled out in 1992; for years it used organic solvents such as trichloroethylene and tetrachloroethylene and contaminated the groundwater at the site. After the contamination came to light in 1994, the employees' self-help association fought a legal battle for more than twenty years, part of which was made final by a Supreme Court judgment in 2018. Its teaching value lies in a single question: when the company has already been dissolved, the plant demolished, and individual exposure records long since lost, how do you prove it was them? The answer is that the evidence the court ultimately accepted was population-level epidemiological inference, not any one person's medical record. That single sentence turns three abstract exam points — that cohort studies can establish temporality, the Hill criteria for causation, and the allocation of the burden of proof — into a real courtroom battle.

The special health exam measures not the patient but the workplace; what level four must fix is not the ears, but that machine.

15. Disability, Ethnicity, and Geography: The Prescription for Health Inequality Is Not Written in the Exam Room

ICF: Moving "Disability" from the Person to the Space Between Person and Environment

Since July 11, 2012 (ROC year 101), Taiwan's disability determination and needs assessment has used the new system based on the ICF (International Classification of Functioning, Disability and Health) (verified as of July 2026). This is not swapping one form for another; it is swapping one worldview for another.

🔬 Mechanism, unpacked

Step one, the old system's worldview came from the WHO's early ICIDH model: impairment → disability → handicap, a one-way causal line — the body breaks, so the person cannot do things, so social participation is restricted. Step two, this line locates "disability" entirely in the individual's body, so the solution naturally runs in only one direction: treat or correct this person. Step three, but those two spinal-cord-injury patients prove the line wrong — their bodies are in exactly the same condition, yet their social participation is worlds apart, and the difference comes from the environment. Step four, so the ICF turns the model into an interactive one: health condition ↔ body functions and structures ↔ activity ↔ participation, and this whole set is modulated at the same time by environmental factors (buildings, transport, assistive devices, attitudes, policies, services) and personal factors (age, education, coping style). Step five, the conclusion is powerful: disability is not an attribute; it is a state of "mismatch between a person and the environment." So intervention can be placed at any node — treating the body, training activity capacity, modifying the environment, reforming the system — and interventions on the environmental side are often the cheapest and the most effective.

Three concrete changes to the system:

Item	Old system (before 2012)	New ICF system
Classification	16 categories of disease/disability	Eight major systems of body function and structure
Assessment	Assessed by a physician alone	Assessed by a professional team of medical staff, social workers, and others, with an added evaluation of activity/participation and environmental factors
Output	Disability handbook	Disability certificate + needs assessment (linked to services)

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

The eight systems (Article 5 of the People with Disabilities Rights Protection Act, verified as of July 2026): (1) structure of the nervous system and mental and psychological functions; (2) eye, ear, and related structures, sensory functions, and pain; (3) structures and functions involved in voice and speech; (4) structures and functions of the circulatory, hematopoietic, immune, and respiratory systems; (5) structures and functions related to the digestive, metabolic, and endocrine systems; (6) structures and functions related to the urinary and reproductive systems; (7) movement-related structures and functions of the nerves, muscles, and bones; (8) skin and related structures and functions.

The single most important word is "certificate" replacing "handbook," followed by "needs assessment" — because under the ICF logic, the purpose of assessment is not to hand out a grade label but to determine what services this person needs (assistive devices, personal assistants, home accessibility modifications, transport, employment services). This act also mandates quota employment: under Article 38, government agencies, public schools, and state-owned enterprises with 34 or more employees must fill at least 3% of their workforce with people with disabilities, and private schools, organizations, and private enterprises with 67 or more employees at least 1%, and no fewer than one person.

The Indigenous Peoples Health Act: Can a Dedicated Law Close a Six-Year Life-Expectancy Gap?

The Indigenous Peoples Health Act was promulgated on June 21, 2023 (ROC year 112) (verified as of July 2026). It is Taiwan's first health law dedicated to a specific population group.

It does several very concrete things in governance terms: the central competent authority must set up a dedicated unit for Indigenous health affairs and allocate an adequate budget; it must regularly survey Indigenous health status and needs and build an Indigenous health database; it must train Indigenous healthcare professionals; it must research and promote traditional Indigenous medical and healthcare knowledge; and it must convene an Indigenous Peoples health policy council, in which members with Indigenous status may not be fewer than one-half (verified as of July 2026). This last provision is the soul of the whole act — it grants not just resources but seats at the decision-making table.

Why is a dedicated law needed? Because the numbers speak for themselves: the life-expectancy gap between Indigenous peoples and the national average narrowed from 8.17 years in 2017 to 6.19 years in 2022 (verified as of July 2026). Narrowing by two years is good news, but there are still six years to go.

🔬 Mechanism, unpacked

Step one, first block a dangerous misconception: these six years are not caused by "ethnic constitution." Step two, lay out the Indigenous disease profile — accidental injury, chronic liver disease and cirrhosis, alcohol use disorder, tuberculosis, oral cancer, diabetes and metabolic syndrome, hyperuricemia and gout — what these diseases share is not genes but the fact that they are all strongly associated with modifiable social conditions. Step three, these social conditions are the social determinants of health (SDOH): employment structure (a high share of high-risk manual labor), education and health literacy, the accessibility and availability of healthcare, the availability and cultural context of tobacco, alcohol, and betel quid, and historically accumulated resource allocation. Step four, so intervention cannot sit only in the exam room — what the exam room can do lies in the second and third of the "three stages and five levels" of prevention, whereas this six-year gap is generated mainly before the first level. Step five, this is exactly the role of a dedicated law: it moves the point of intervention from "the physician's prescription pad" to the four places that can truly change the denominator — budget, seats, data, and workforce training.

IDS: Health Insurance Arrived, but the Doctors Still Didn't Come

The most awkward fact after Taiwan's National Health Insurance (NHI) launched in 1995: enrollment exceeded 99%, yet mountain and offshore-island areas had no doctors. The insurance card solved "can you afford it" but did nothing about "is anyone coming" — this is the split between accessibility and availability.

The National Health Insurance Administration therefore launched the "Medical Benefit Improvement Plan for Mountainous and Offshore-Island Areas" (山地離島地區醫療給付效益提昇計畫; Integrated Delivery System, IDS) (rolled out nationwide in November 1999; its official name is prefixed with National Health Insurance). The approach: the healthcare of an entire mountain township or offshore island is contracted out to a single hospital on a bundled, capitation-like basis, providing fixed-site clinics, mobile outreach care, 24-hour consultation, and emergency transfer.

🔬 Mechanism, unpacked

Step one, why use bundled payment rather than fee-for-service? Step two, because under fee-for-service (FFS), medical services generate revenue only when "someone comes in to be seen," while mountain townships have sparse populations and high transport costs — the revenue from each clinic session cannot cover its costs, so a rational provider simply will not go. Step three, once payment switches to contracting a whole township for one lump-sum budget, the contracting hospital's revenue is decoupled from "how many sessions it runs" and tied instead to "keeping this population well." Step four, so the incentives flip — prevention, health education, chronic-disease control, and fewer emergency transfers all become in the contractor's interest, because costs stay low only if patients don't deteriorate. Step five, this is the core principle of payment systems: the larger the settlement unit, the stronger the incentive to economize and prevent (fee-for-service < per diem < per-case payment (DRG) < capitation). IDS is the closest real-world experiment in capitation within Taiwan's NHI.

The legal counterpart is that narrow gate: Article 11 of the Physicians Act allows that, in mountainous, offshore-island, or remote areas, or in special or urgent circumstances, a physician designated by the competent authority may examine patients and prescribe by telecommunication, with health workers then delivering the medication. Why does this exception specifically name "mountainous and offshore-island areas"? The answer is IDS's very reason for existing — the law's exception exists to patch a market failure.

Blackfoot Disease: Taiwan's Chapter in the International Textbooks

The truth was one of environmental epidemiology's most classic victories. The local shallow-well water was brackish and bitter, so residents dug deep wells to obtain drinkable water — and that deep-well water contained high concentrations of inorganic arsenic. Long-term chronic arsenic exposure caused peripheral arterial occlusive disease (blackfoot disease proper), while also bringing a whole set of arsenic-related malignancies: skin cancer (Bowen's disease, i.e., squamous cell carcinoma in situ, plus invasive squamous cell and basal cell carcinoma; arsenical keratosis is the precancerous lesion), bladder cancer (especially transitional cell carcinoma), lung cancer, and liver cancer, along with raindrop pigmentation of the skin (raindrop-like hyperpigmentation and depigmentation).

And the most important ending of this story: blackfoot disease was not cured by a drug; it was cured by tap water. Once tap-water pipelines reached those townships, new cases disappeared.

The cure for blackfoot disease was not any drug but tap water — the cleanest proof ever that "environmental intervention beats individual treatment."

Taiwan's blackfoot-disease cohort data were later written into the evidence base for international arsenic risk assessment and drinking-water standards (the World Health Organization's guideline value for arsenic in drinking water is 10 μg/L). This is another mark Taiwanese public health has left on the world — we paid the price in villages and bought the world a standard.

🦦

PipIndigenous people have more chronic disease — isn't that because of their constitution or genes?

🐻‍❄️

BarnabyStep on that intuition and you're wrong — dangerously wrong. Look at the disease profile — accidental injury, chronic liver disease, alcohol use disorder, tuberculosis, oral cancer, metabolic syndrome — what they share isn't genes; it's that they're all strongly associated with modifiable social conditions. That's called the social determinants of health: employment structure, health literacy, the accessibility and availability of healthcare, the availability of substances, and historical resource allocation. So the prescription isn't in the exam room; it's in budget, seats, data, and workforce — which is exactly why the Indigenous Peoples Health Act requires that members with Indigenous status make up no fewer than one-half of the policy council.

⭐ Board-exam essentials

Chapter 15 Must-Knows

The new ICF system took effect on July 11, 2012 (ROC year 101): classification changed from 16 categories to eight major systems; assessment by a medical + social-work team with an added evaluation of activity/participation and environmental factors; disability handbook → disability certificate + needs assessment.
The core idea of the ICF: from the ICIDH's one-way linear model (impairment → disability → handicap) to an interactive model — disability results from a mismatch between person and environment, and environmental factors can either worsen or eliminate disability.
Order of the eight systems: nervous & mental/psychological → eye/ear, senses & pain → voice & speech → circulatory, hematopoietic, immune & respiratory → digestive, metabolic & endocrine → urinary & reproductive → neuromusculoskeletal movement → skin.
The Indigenous Peoples Health Act was promulgated on June 21, 2023 (ROC year 112), Taiwan's first health law for a specific population group; key points = dedicated unit, adequate budget, regular surveys and a database, training Indigenous healthcare professionals, traditional medical and healthcare knowledge, and members with Indigenous status must make up no fewer than one-half of the policy council.
Indigenous vs national-average life-expectancy gap: 8.17 years in 2017 → 6.19 years in 2022. The cause is SDOH, not ethnic constitution.
IDS (Medical Benefit Improvement Plan for Mountainous and Offshore-Island Areas): a bundled/capitation-like contract giving one hospital the care of a whole township; principle = the larger the settlement unit, the stronger the incentive to prevent and economize (FFS < per diem < DRG < capitation).
Accessibility ≠ availability: NHI solved "can you afford it"; IDS solves "is anyone coming." The telemedicine exception in Article 11 of the Physicians Act exists precisely to patch a market failure.
Blackfoot disease: chronic exposure to inorganic arsenic in deep-well water on the Chiayi–Tainan coast → peripheral arterial occlusive gangrene + skin cancer (Bowen's disease [SCC in situ], SCC, BCC; arsenical keratosis is precancerous), bladder cancer, lung cancer, liver cancer + raindrop pigmentation changes; the cure was tap water (environmental intervention), not drugs; WHO drinking-water arsenic guideline value 10 μg/L.

♪ Sing-along text — English rendering of the Mandarin song lyrics

Two patients carry exactly the same diagnosis: complete spinal cord injury at the sixth thoracic segment. The first lives in Taipei, where his neighborhood has elevators, the MRT stations have accessible facilities, and his company has given him an adjusted desk, so he commutes to work on his own every day; the second lives in the mountains of Nantou, with seven steps at his front door, the nearest bus stop two kilometers away, and no rehabilitation resources in his village, and he hasn't left his house in two years. Same spinal cord, same segment, two completely different degrees of disability — and the difference doesn't lie in them; it lies around them.

Since July 11, 2012, Taiwan's disability determination and needs assessment has used the new system based on the International Classification of Functioning, Disability and Health. This is not swapping one form for another; it is swapping one worldview for another. The old worldview came from an early World Health Organization model in which impairment causes disability and disability causes handicap — a one-way causal line that located disability entirely in the individual's body, so the solution naturally ran in only one direction: treat or correct this person. But those two spinal-cord-injury patients prove the line wrong: their bodies are in exactly the same condition, yet their social participation is worlds apart, and the difference comes from the environment. So the new model became interactive: the health condition, body functions and structures, activity, and participation all interact with one another, and the whole set is modulated at the same time by environmental factors and personal factors, with environmental factors including buildings, transport, assistive devices, attitudes, policies, and services. The conclusion is powerful: disability is not an attribute; it is a state of mismatch between a person and the environment, so intervention can be placed at any node, and interventions on the environmental side are often the cheapest and the most effective.

Disability isn't something that grows on a person; it's the gap where a person and the environment fail to meet.

There were three changes to the system. Classification changed from sixteen categories of disease and disability to eight major systems of body function and structure: in order, the structure of the nervous system and mental and psychological functions; the eye, ear, and related structures, sensory functions, and pain; structures and functions involved in voice and speech; the circulatory, hematopoietic, immune, and respiratory systems; the digestive, metabolic, and endocrine systems; the urinary and reproductive systems; the movement-related structures of the nerves, muscles, and bones; and the skin and related structures. Assessment changed from a physician assessing alone to a professional team of medical staff, social workers, and others, with an added evaluation of activity, participation, and environmental factors. The output changed from a disability handbook to a disability certificate plus a needs assessment. Most important of all, the certificate replaces the handbook and is followed by a needs assessment, because under the new logic the purpose of assessment is not to hand out a grade label but to determine what services this person needs: assistive devices, personal assistants, home accessibility modifications, transport, and employment services. This act also sets employment quotas: public-sector bodies with thirty-four or more employees must employ at least three percent people with disabilities, and private employers with sixty-seven or more at least one percent, and no fewer than one person.

The Indigenous Peoples Health Act was promulgated on June 21, 2023, as Taiwan's first health law dedicated to a specific population group. It requires the central government to set up a dedicated unit for Indigenous health affairs and allocate an adequate budget, to regularly survey health status and needs and build an Indigenous health database, to train Indigenous healthcare professionals, to research and promote traditional Indigenous medical and healthcare knowledge, and to convene an Indigenous Peoples health policy council in which members with Indigenous status may not be fewer than one-half. That last provision is the soul of the whole act: it grants not just resources but seats at the decision-making table.

Why a dedicated law? Because the numbers speak for themselves: the life-expectancy gap between Indigenous peoples and the national average narrowed from 8.17 years in 2017 to 6.19 years in 2022. Narrowing by two years is good news, but there are still six years to go. These six years are not caused by ethnic constitution. Lay out the Indigenous disease profile — accidental injury, chronic liver disease and cirrhosis, alcohol use disorder, tuberculosis, oral cancer, diabetes and metabolic syndrome, hyperuricemia and gout — and what these diseases share is not genes but that they are all strongly associated with modifiable social conditions. These conditions are the social determinants of health: employment structure, education and health literacy, the accessibility and availability of healthcare, the availability and cultural context of tobacco, alcohol, and betel quid, and historically accumulated resource allocation. So intervention cannot sit only in the exam room: what the exam room can do lies in the second and third of the three stages and five levels of prevention, while this six-year gap is generated mainly before the first level. That is exactly the role of a dedicated law: it moves the point of intervention from the physician's prescription pad to the four places that can truly change the denominator — budget, seats, data, and workforce training.

Disability isn't something that grows on a person; it's the gap where a person and the environment fail to meet.

The most awkward fact after National Health Insurance launched: enrollment exceeded ninety-nine percent, yet mountain and offshore-island areas had no doctors. The insurance card solved whether you could afford care but not whether anyone would come — this is the split between accessibility and availability. The National Health Insurance Administration therefore launched the Medical Benefit Improvement Plan for Mountainous and Offshore-Island Areas, contracting out the healthcare of an entire mountain township or offshore island to a single hospital on a bundled, capitation-like basis, providing fixed-site clinics, mobile outreach care, twenty-four-hour consultation, and emergency transfer. Why bundled payment rather than fee-for-service? Because under fee-for-service, services generate revenue only when someone comes in to be seen, and mountain townships have sparse populations and high transport costs, so the revenue from each clinic session cannot cover its costs and a rational provider won't go; once payment switches to contracting a whole township for one lump-sum budget, the contracting hospital's revenue is decoupled from how many sessions it runs and tied instead to keeping this population well, so prevention, health education, chronic-disease control, and fewer emergency transfers all become in the contractor's interest, because costs stay low only if patients don't deteriorate. This is the core principle of payment systems: the larger the settlement unit, the stronger the incentive to economize and prevent — from fee-for-service to per diem, to per-case payment, and on to capitation, the incentive grows stronger at every step. The legal counterpart is the narrow gate of Article 11 of the Physicians Act: in mountainous, offshore-island, or remote areas, or in special or urgent circumstances, a designated physician may examine and prescribe by telecommunication, with health workers then delivering the medication. Why does this exception specifically name mountains and offshore islands? The answer is this plan's very reason for existing: the law's exception exists to patch a market failure.

Finally, blackfoot disease. In the 1950s and 60s, along the Chiayi–Tainan coast — Beimen, Xuejia, Budai, Yizhu — a strange disease kept appearing: patients' toes first turned cold and painful with intermittent claudication, then blackened into gangrene and finally fell off on their own. Doctors could find no embolism, no diabetes, no vasculitis, until someone noticed that the villages where it struck had all dug deep wells just a few years earlier. The truth was one of environmental epidemiology's most classic victories: the local shallow-well water was brackish and bitter, so residents dug deep wells to get drinkable water, and the deep-well water contained high concentrations of inorganic arsenic. Long-term chronic arsenic exposure caused peripheral arterial occlusive disease — blackfoot disease proper — while also bringing a whole set of arsenic-related malignancies: skin cancers such as Bowen's disease (squamous cell carcinoma in situ) and invasive squamous and basal cell carcinoma, with arsenical keratosis as their precancerous forerunner, plus bladder cancer, lung cancer, and liver cancer, with raindrop-like hyperpigmentation and depigmentation appearing on the skin. And the most important ending of this story is that blackfoot disease was not cured by a drug; it was cured by tap water — once tap-water pipelines reached those townships, new cases disappeared. Taiwan's blackfoot-disease cohort data were later written into the evidence base for international arsenic risk assessment and drinking-water standards. We paid the price in villages and bought the world a standard.

Disability isn't something that grows on a person; it's the gap where a person and the environment fail to meet.

16. Tobacco, Alcohol, Betel Quid, and the Four Cancer Screenings: Using Tobacco Money to Screen for the Cancers Tobacco Causes

The 2023 Tobacco Hazards Prevention Act: Seven Key Points and Their Logic

The Tobacco Hazards Prevention Act was amended and promulgated on February 15, 2023 (ROC year 112) and took effect on March 22, 2023 (verified as of July 2026). Each of the amendment's seven key points corresponds to a policy logic:

Amendment point	Content	Underlying policy logic
Total ban on "tobacco-like products" (e-cigarettes)	Manufacture, import, sale, supply, display, advertising, and use all banned	The net new-user effect of nicotine delivery devices — e-cigarettes mainly attract adolescents who have never smoked, creating a new gateway to addiction
"Designated tobacco products" (including heated tobacco) must pass a health risk assessment review	Without passing review, manufacture, import, and sale are all illegal	Reversed burden of proof: it is not the government that must prove the product harmful, but the industry that must prove it may be marketed
Minimum smoking age raised from 18 to 20	Tobacco may not be supplied to anyone under 20	Delaying the age of initiation = lowering the lifetime addiction rate (the adolescent brain is the most sensitive to nicotine)
Warning-graphic area raised from 35% to 50%	In force from March 22, 2024 (ROC year 113) (a buffer period for the industry)	The package is the last advertising space — turn it into a health-education space
Expanded smoke-free places	Complete smoking bans (including outdoors) extended from senior high schools and below to all schools including universities, plus kindergartens, infant care centers, and home-based childcare settings; bars and nightclubs smoke-free indoors except sealed smoking rooms (indoor workplaces of three or more: smoke-free since 2009)	Protects groups who cannot choose their own environment
Heavier penalties	Substantially raised maximum fines across the board	Raises the expected cost of violations
Statutory definition of tobacco-like products	Electronic or non-electronic delivery systems that use non-tobacco materials, or alter the physical properties of tobacco, to simulate the use of tobacco products	Closes the loophole of "rename it and it's no longer tobacco"

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

(The seven points above verified as of July 2026: total ban on e-cigarettes, health risk assessment review for heated tobacco, minimum smoking age 20, 50% warning graphics in force from March 22, 2024.)

🔬 Mechanism, unpacked

Step one, why is the e-cigarette "totally banned" while heated tobacco gets a "review system"? This question appears every year, and many people assume it is a policy inconsistency. Step two, the key is the product definition: heated tobacco contains tobacco and is therefore a "tobacco product" as defined by the Tobacco Hazards Prevention Act; the law cannot apply different market-entry principles to products within the same tobacco-product category, so it can only add the gate of "must pass a health risk assessment before marketing." Step three, e-cigarettes contain no tobacco and fall under the newly defined "tobacco-like products," so the law can bar them from the market outright. Step four, and the substantive policy reason for banning them is the net effect: if e-cigarettes were used only by existing smokers as a substitute, they might reduce harm; but the evidence shows they mainly attract adolescents who have never smoked, in effect opening a new gateway to addiction. Step five, so memorize this line firmly: "heated tobacco that has not passed review is illegal in Taiwan" — a review system does not mean open access; until it passes review, it can no more be sold than an e-cigarette.

The WHO's six MPOWER measures are the skeleton of every tobacco question: Monitor (monitor tobacco use and prevention policies), Protect (protect people from secondhand smoke), Offer (offer help to quit), Warn (warn about the dangers of tobacco), Enforce (enforce bans on tobacco advertising, promotion, and sponsorship), Raise (raise taxes and prices on tobacco). The evidence shows that the single most cost-effective measure is raising tobacco prices (price elasticity is especially large among adolescents).

And Taiwan has a particularly elegant financing design: the health and welfare surcharge on tobacco products (菸品健康福利捐) — a dedicated levy on every pack, with earmarked uses that feed the NHI safety reserve, cancer prevention and control, cancer screening, and more. This ties "user pays" to "earmarking": using tobacco money to screen for the cancers tobacco causes.

Smoking-cessation treatment has three main drug classes: nicotine replacement therapy (NRT: patches/gum/inhalers), varenicline (an α4β2 nicotinic acetylcholine receptor "partial agonist" — partial agonism means it both relieves withdrawal symptoms and occupies the receptor so that smoking again brings no reward), and bupropion (a dopamine and norepinephrine reuptake inhibitor). Taiwan's second-generation smoking-cessation treatment program (since March 2012) is funded by the HPA from the tobacco health and welfare surcharge rather than paid by NHI; drugs are priced as under NHI, with a copayment of at most NT$200 per prescription.

Alcohol: From 0.15 to 0.05, and That Flushing Face

The legal thresholds must be memorized in two tiers: administrative penalties (under the Act Governing the Punishment of Violation of Road Traffic Regulations) begin at a breath alcohol concentration of 0.15 mg/L or a blood alcohol concentration of 0.03%; criminal liability (Article 185-3 of the Criminal Code, the offense of driving while unable to drive safely) has a threshold of breath 0.25 mg/L or blood 0.05% or above (figures per Article 114 of the Road Traffic Safety Rules and Article 185-3 of the Criminal Code; heavier penalties apply when death or serious injury results or on a repeat offense within ten years; verified 2026-09).

ALDH2 deficiency is the single most important pharmacogenomic fact in East Asian medicine. Alcohol is metabolized in two steps: ethanol → (ADH) → acetaldehyde → (ALDH2) → acetate. Roughly 30% to 50% of East Asians carry the ALDH2*2 variant, which sharply reduces the enzyme activity of the second step.

🔬 Mechanism, unpacked

Step one, low ALDH2 activity → acetaldehyde accumulates. Step two, acetaldehyde is a potent vasodilator and sympathetic stimulant, producing flushing, palpitations, headache, and nausea — the so-called "Asian flush" (alcohol flushing). Step three, the key is not the discomfort but that acetaldehyde is an IARC Group 1 carcinogen. Step four, so in people who carry ALDH2*2 yet keep drinking, the mucosa of the esophagus and upper aerodigestive tract is steeped for years in high concentrations of acetaldehyde → a sharply increased risk of esophageal squamous cell carcinoma. Step five, the conclusion is counterintuitive but vital: "flushing when you drink" is not low tolerance — it is impaired clearance of a carcinogen; these people should not train their tolerance; they should not drink at all.

Clinical tools: the AUDIT / AUDIT-C / CAGE screening questionnaires. For acute management, remember two things: delirium tremens from alcohol withdrawal usually appears 48–96 hours after the last drink, and the first-line treatment is a benzodiazepine; and give thiamine before glucose — because glucose metabolism consumes thiamine, giving sugar first may precipitate Wernicke encephalopathy.

Betel Quid and Oral Cancer: That Patch on the Cancer Map of Taiwanese Men

The most important sentence comes first: the areca nut in betel quid — even without betel inflorescence or tobacco added — has itself been classified by the International Agency for Research on Cancer (IARC) as a Group 1 carcinogen. This is a trap people fall into every year — "I only chew it with white lime, no inflorescence" does not bring the risk down to zero. Mechanistically, arecoline and the nitrosated derivatives it forms in the mouth are genotoxic and cytotoxic; combined with the mechanical friction of the fibers and the alkaline irritation from slaked lime, they together produce chronic mucosal injury.

The three sibling precancerous lesions must be kept straight:

Lesion	Appearance	Features
Leukoplakia	A white patch that cannot be scraped off	Most common; moderate malignant transformation rate
Erythroplakia	A red, velvety patch	Less common, but the highest malignant transformation rate
Oral submucous fibrosis (OSF)	Pale, stiff mucosa with palpable fibrous bands	Most specifically linked to betel quid; hallmark is restricted mouth opening (trismus)

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

And the most important thing about risk is interaction: when tobacco, alcohol, and betel quid are used together, oral-cancer risk is "multiplicative," not "additive" — this is the structural reason oral-cancer incidence among Taiwanese men ranks among the highest in the world, and also why, in the four national cancer screenings, the target population for oral cancer is defined as "betel quid chewers (including former chewers) or smokers." There is another very practical reason oral cancer is worth screening for: it can be seen and felt, and the screening tools are just a flashlight and a pair of gloves — the cost is very low, and the cure rate when caught early is high.

Cancer Screening: Who, How Often, and Three Must-Know Biases

Publicly funded cancer screening provided by the Health Promotion Administration (HPA) was substantially expanded from 2025 (ROC year 114) (verified as of July 2026):

Cancer	Target population	Frequency	Tool
Cervical cancer	Women aged 25–29	Once every 3 years	Pap smear
	Women aged 30 and over	Once a year (at least once every 3 years recommended)	Pap smear
	Women aged 35, 45, and 65	Once at each age	Human papillomavirus (HPV) testing (added in 2025)
Breast cancer	Women aged 40–74	Once every 2 years	Mammography
Colorectal cancer	Aged 45–74; aged 40–44 with a family history	Once every 2 years	Fecal immunochemical test (FIT)
Oral cancer	Aged 30 and over who chew betel quid (including former chewers) or smoke; Indigenous people aged 18 and over who chew betel quid	Once every 2 years	Oral mucosal examination
Lung cancer	Heavy smokers: aged 50–74, smoking history ≥ 20 pack-years, currently smoking or quit less than 15 years ago	Once every 2 years	Low-dose computed tomography (LDCT)
	Family history of lung cancer: women aged 40–74, men aged 45–74, with a first-degree relative (parent, child, or sibling) diagnosed with lung cancer	Once every 2 years	LDCT

Swipe or scroll sideways to compare every column; keyboard: focus the table and use arrow keys.

One common misstatement on the timeline needs correcting: publicly funded LDCT lung-cancer screening was launched by the HPA as the "Lung Cancer Early Detection Program" on July 1, 2022 (ROC year 111); from 2025 (ROC year 114) the service was expanded (the age criterion for those with a family history lowered by 5 years, colorectal screening lowered to age 45, breast screening extended to ages 40–44 and 70–74, and cervical screening extended to ages 25–29 with HPV testing added) (verified as of July 2026).

🔬 Mechanism, unpacked

Step one, why does "higher five-year survival in the screened group" not prove that screening works? Step two, because of lead-time bias: screening merely moves the date of diagnosis earlier; if treatment does not change the disease course, the patient's date of death stays the same, but the stretch "from diagnosis to death" is lengthened — survival time grows longer, yet life is not extended by a single day. Step three, there is also length-time bias: slow-growing tumors stay in the "detectable by screening" phase longer, so screening inherently tends to catch less aggressive tumors, and the screened group's prognosis naturally looks better. Step four, push length-time bias to its extreme and you get overdiagnosis — diagnosing a tumor that "would never have caused symptoms or death in this lifetime," then giving the patient real surgery and real anxiety. Step five, so to judge whether screening truly works, the only reliable evidence is a fall in disease-specific mortality in randomized trials — which is exactly why lung-cancer LDCT has NLST and NELSON behind it while many other screenings cannot stand up. Step six, LDCT's biggest controversy is also overdiagnosis (indolent lung adenocarcinomas and large numbers of false-positive nodules), so it is offered only to high-risk groups, not as a population-wide health-checkup item.

The one thing the tobacco health and welfare surcharge gets right: using tobacco money to screen for the cancers tobacco causes.

⭐ Board-exam essentials

Chapter 16 Must-Knows

Tobacco Hazards Prevention Act: amended and promulgated February 15, 2023 (ROC year 112), in force March 22. Seven key points: total ban on tobacco-like products (e-cigarettes), designated tobacco products (including heated tobacco) must pass a health risk assessment review, minimum smoking age raised to 20, warning graphics 35% → 50% (in force March 22, 2024), expanded smoke-free places (universities, kindergartens, infant care centers, and home-based childcare, including outdoors; bars and nightclubs indoors), heavier penalties, and a statutory definition of tobacco-like products.
The difference between the e-cigarette total ban and the heated-tobacco review system lies in whether the product contains tobacco (tobacco-like product vs tobacco product); heated tobacco that has not passed review is illegal in Taiwan.
MPOWER: Monitor/Protect/Offer/Warn/Enforce/Raise; the single most cost-effective measure is raising tobacco prices. Health and welfare surcharge on tobacco products = user pays + earmarking, feeding the NHI safety reserve and cancer prevention and screening.
Cessation drugs: NRT, varenicline (α4β2 partial agonist), bupropion.
Drunk driving: administrative penalty 0.15 mg/L (blood 0.03%); criminal offense of driving while unable to drive safely 0.25 mg/L (blood 0.05%).
ALDH2 deficiency (about 30–50% of East Asians) → acetaldehyde accumulation → flushing and palpitations; acetaldehyde is a Group 1 carcinogen → sharply raised risk of esophageal squamous cell carcinoma. "Flushing when you drink" is a clearance defect, not low tolerance.
Withdrawal: delirium tremens usually 48–96 hours after the last drink, first-line benzodiazepine; thiamine before glucose (to prevent Wernicke).
The areca nut itself (without betel inflorescence or tobacco) is an IARC Group 1 carcinogen; arecoline and its nitrosated derivatives are genotoxic. Precancerous lesions: leukoplakia (most common), erythroplakia (highest malignant transformation rate), oral submucous fibrosis (most specific to betel quid; hallmark is restricted mouth opening). Tobacco + alcohol + betel quid act multiplicatively.
Target populations for the four cancer screenings + lung cancer: cervical (ages 25–29 every 3 years; 30 and over annually; HPV test added at ages 35/45/65), breast (ages 40–74, every 2 years), colorectal (ages 45–74; ages 40–44 with a family history, every 2 years, FIT), oral (aged 30+ who chew betel quid or smoke; Indigenous people aged 18+ who chew betel quid, every 2 years), lung LDCT (ages 50–74 with ≥20 pack-years and currently smoking or quit less than 15 years ago; family history: women 40–74, men 45–74, every 2 years).
Timeline: lung-cancer LDCT launched July 1, 2022 (ROC year 111); 2025 (ROC year 114) was an expansion (not the launch year).
The three big screening biases: lead-time bias (earlier diagnosis falsely lengthens survival time), length-time bias (tends to catch slow-growing tumors), and overdiagnosis. The only reliable evidence of benefit is a fall in disease-specific mortality in randomized trials, not five-year survival.

♪ Sing-along text — English rendering of the Mandarin song lyrics

In a dental clinic, a forty-six-year-old taxi driver opens his mouth wide, yet it opens only two fingers' width; his buccal mucosa feels like dried leather, pale and stiff, with palpable cord-like fibrous bands. He has chewed for twenty years and thought it was a jaw-joint problem, but it isn't: it is oral submucous fibrosis, a precancerous lesion — and the areca nut itself is a Group 1 carcinogen, whether or not betel inflorescence is added.

The Tobacco Hazards Prevention Act was amended and promulgated on February 15, 2023, and took effect on March 22, and each of its seven key points corresponds to a policy logic. The total ban on tobacco-like products, meaning e-cigarettes, prohibits manufacture, import, sale, supply, display, advertising, and use, because they mainly attract adolescents who have never smoked, creating a new gateway to addiction. Designated tobacco products, including heated tobacco, must pass a health risk assessment review before they can be manufactured, imported, or sold — a reversed burden of proof: it is not the government that must prove the product harmful, but the industry that must prove it may be marketed. The minimum smoking age was raised from eighteen to twenty, because delaying the age of initiation lowers the lifetime addiction rate — the adolescent brain is the most sensitive to nicotine. The warning-graphic area was raised from thirty-five percent to fifty percent, in force from March 22, 2024, because the package is the last advertising space, and this turns it into a health-education space. Smoke-free places were expanded: universities, kindergartens, infant care centers, and home-based childcare settings became completely smoke-free, including outdoors, and bars and nightclubs smoke-free indoors, protecting groups who cannot choose their own environment. Add heavier penalties and a statutory definition of tobacco-like products, sealing off the loophole of "rename it and it's no longer tobacco."

Using tobacco money to screen for the cancers tobacco causes — earmarking at its most elegant.

Why a total ban on e-cigarettes but a review system for heated tobacco? Many people assume it's a policy inconsistency, but the key is the product definition. Heated tobacco contains tobacco and is a tobacco product under the law, and the law cannot apply different market-entry principles to products in the same tobacco-product category, so it can only add the gate of a pre-market health risk assessment; e-cigarettes contain no tobacco and fall under the newly defined tobacco-like products, so the law can bar them from the market outright. And remember firmly: a review system does not mean open access — until it passes review, heated tobacco is just as illegal in Taiwan. The World Health Organization's six measures are the skeleton of every tobacco question: monitoring, protecting people from secondhand smoke, offering help to quit, warning about the dangers of tobacco, banning advertising, promotion, and sponsorship, and raising tobacco taxes and prices — and of these, the single most cost-effective measure is raising tobacco prices, with price elasticity especially large among adolescents. Taiwan also has an elegant financing design: the health and welfare surcharge on tobacco products is a dedicated levy on every pack, earmarked to feed the NHI safety reserve and cancer prevention and screening — in effect, using tobacco money to screen for the cancers tobacco causes. Cessation drugs fall into three classes: nicotine replacement therapy; varenicline, a partial agonist that both relieves withdrawal and occupies the receptor so that smoking again brings no reward; and bupropion, which acts on dopamine and norepinephrine reuptake.

The legal thresholds for alcohol come in two tiers: administrative penalties start at a breath concentration of 0.15 milligrams per liter or a blood concentration of 0.03 percent, while the criminal offense of driving while unable to drive safely has a threshold of 0.25 milligrams on the breath or 0.05 percent in the blood or above. Medically, the most important point is aldehyde dehydrogenase 2 deficiency. Alcohol is metabolized in two steps: ethanol is converted to acetaldehyde by alcohol dehydrogenase, and acetaldehyde is converted to acetate by aldehyde dehydrogenase; roughly thirty to fifty percent of East Asians carry a variant that sharply reduces the enzyme activity of the second step, so acetaldehyde accumulates, causing flushing, palpitations, headache, and nausea. The key is not the discomfort but that acetaldehyde is recognized by the International Agency for Research on Cancer as a Group 1 carcinogen, so in people who carry the variant yet keep drinking, the mucosa of the esophagus and upper aerodigestive tract is steeped for years in high concentrations of acetaldehyde, and the risk of esophageal squamous cell carcinoma rises sharply. The conclusion is counterintuitive yet vital: flushing when you drink is not low tolerance; it is impaired clearance of a carcinogen — these people should not train their tolerance; they should not drink at all. Clinically we screen with questionnaires, and for acute management remember two things: delirium tremens from alcohol withdrawal usually appears forty-eight to ninety-six hours after the last drink, with benzodiazepines as the first choice; and give thiamine before glucose, because glucose metabolism consumes thiamine, and getting the order wrong may precipitate Wernicke encephalopathy.

Using tobacco money to screen for the cancers tobacco causes — earmarking at its most elegant.

On betel quid, the first sentence must come first: the areca nut itself, even without betel inflorescence or tobacco, has already been classified as a Group 1 carcinogen. Mechanistically, arecoline and the nitrosated derivatives it forms in the mouth are genotoxic and cytotoxic, and together with the mechanical friction of the fibers and the alkaline irritation of the lime, they produce chronic mucosal injury. The precancerous lesions are three siblings: leukoplakia is a white patch that cannot be scraped off and is the most common; erythroplakia is red and velvety, less common, but with the highest malignant transformation rate; and oral submucous fibrosis is the most specifically linked to betel quid, marked by pale, stiff mucosa and restricted mouth opening. The most important thing about risk is interaction: when tobacco, alcohol, and betel quid are used together, oral-cancer risk multiplies rather than adds — the structural reason oral-cancer incidence among Taiwanese men ranks among the highest in the world, and the reason the screening target is defined as betel quid chewers, including former chewers, or smokers. There is another very practical reason oral cancer is worth screening for: it can be seen and felt, the tools are just a flashlight and a pair of gloves, the cost is very low, and the cure rate when caught early is high.

Publicly funded cancer screening was substantially expanded from 2025. Cervical cancer: women aged twenty-five to twenty-nine get a Pap smear once every three years; from age thirty, a Pap smear is subsidized every year, with at least one every three years recommended; and at ages thirty-five, forty-five, and sixty-five, a human papillomavirus test is added once each. Breast cancer: women aged forty to seventy-four get a mammogram once every two years. Colorectal cancer: people aged forty-five to seventy-four, plus those aged forty to forty-four with a family history, get a fecal immunochemical test once every two years. Oral cancer: people aged thirty and over who chew betel quid, including former chewers, or who smoke, plus Indigenous people aged eighteen and over who chew betel quid, get an oral mucosal examination once every two years. Lung cancer uses low-dose computed tomography: for heavy smokers, it is ages fifty to seventy-four with a smoking history of twenty pack-years or more who still smoke or quit less than fifteen years ago; for those with a family history of lung cancer, it is women aged forty to seventy-four and men aged forty-five to seventy-four with a first-degree relative diagnosed with lung cancer — both once every two years. One common misstatement on the timeline needs correcting: publicly funded low-dose CT lung-cancer screening launched on July 1, 2022, and 2025 marked an expansion, not the launch.

Finally, the three must-know screening biases. Why does higher five-year survival in the screened group not prove that screening works? Because of lead-time bias: screening only moves the date of diagnosis earlier, and if treatment does not change the disease course, the date of death stays the same, but the stretch from diagnosis to death gets longer — survival time grows, yet life is not extended by a single day. There is also length-time bias: slow-growing tumors stay in the detectable phase longer, so screening inherently tends to catch less aggressive tumors, and the screened group's prognosis naturally looks better. Push length-time bias to its extreme and you get overdiagnosis: diagnosing a tumor that would never have caused symptoms or death in this lifetime, then giving the patient real surgery and real anxiety. So to judge whether screening truly works, the only reliable evidence is a fall in disease-specific mortality in randomized trials, not five-year survival. The biggest controversy around low-dose CT is precisely overdiagnosis and large numbers of false-positive nodules, which is why it is offered only to high-risk groups and is not a population-wide health-checkup item.

Using tobacco money to screen for the cancers tobacco causes — earmarking at its most elegant.

17. The Mental Health Act: Handing the Pen That Takes Away Liberty to a Judge

The Three Pillars of the 2022 Overhaul

The Mental Health Act was amended and promulgated on December 14, 2022 (ROC year 111), its full text substantially expanded, and it took effect on December 14, 2024 (ROC year 113) (the Chapter 5 provisions on compulsory admission took effect separately on August 1, 2026; verified September 2026). This amendment rests on three pillars.

Pillar one: compulsory admission is now decided by court ruling. Under the old system, the central competent authority set up a "Review Committee for Compulsory Admission and Compulsory Community Treatment of Mental Illness" to conduct an administrative review; under the new system, the designated psychiatric institution petitions the court for a ruling, and the court uses a lay-assessor system (參審制) — a collegiate panel composed of a judge, a lay assessor who is a psychiatric specialist, and a representative of a patient-rights advocacy group, deciding by majority opinion (verified as of July 2026). Court-ordered compulsory admission may not exceed 60 days; if extension is necessary, a petition for a ruling to extend must be filed with the court 14 days before the period expires, only one extension petition is allowed, and the extension may not exceed 60 days (verified as of July 2026).

🔬 Mechanism, unpacked

Step one, why must it be the court? Step two, because compulsory admission is, in essence, a deprivation of personal liberty, and the principle of judicial reservation under Article 8 of the Constitution requires that any deprivation of personal liberty, in principle, be decided by a court through due legal process. Step three, under the old system an administrative agency conducted a paper review, which amounted to placing "the petitioning side" and "the deciding side" within the same administrative system — procedurally lacking a neutral third party. Step four, but simply handing it to judges has problems too — judges don't know psychiatry and might become a rubber stamp or be overly conservative. Step five, so Taiwan chose the lay-assessor system: the judge handles legal requirements and procedure, the psychiatrist lay assessor handles medical judgment, and the patient-rights group representative brings the patient's perspective — three kinds of knowledge checking and balancing one another within the same collegiate panel. Step six, the safeguards are a cap on duration (60 days) and a cap on extensions (one) — because protection without a time limit is imprisonment.

Pillar two: procedural safeguards for patients' rights. Public-interest groups promoting patients' rights that are recognized by the central competent authority may conduct case monitoring of compulsory admission, compulsory community treatment, and emergency placement; when they find something improper, they must notify the competent authority to take corrective measures, and, out of concern for safeguarding the severe patient's autonomy, equality, and interests, they may petition the court for a ruling to terminate compulsory admission. A severe patient or their protector may also petition the court for a ruling terminating emergency placement, compulsory admission, or compulsory community treatment, and petitions and appeals in such matters are exempt from court fees (verified as of July 2026) — the court-fee exemption looks very technical, but it determines whether this right is actually affordable to use.

Pillar three: a community support system. The new act writes community support and reasonable accommodation into law, strengthens the rollout of community mental health centers, and establishes cross-agency collaboration among health, social welfare, police, fire services, education, and labor authorities together with case management mechanisms, while emphasizing support for patients and their protectors (family members). This pillar answers the most practical question: if the community has nothing, compulsory admission becomes the only tool; to use compulsion less, you must first build the community.

The Definitions Tested Every Single Year

A "severe patient" is not a "psychiatric patient." This is the highest-frequency trap. Under the Mental Health Act, a severe patient (嚴重病人), under current Article 3 as amended in 2022, is a patient who exhibits a mental state detached from reality, rendering them unable to manage their own affairs, as diagnosed and determined by a specialist physician; the 2007 wording was a patient who exhibits bizarre thoughts and strange behavior detached from reality, rendering them unable to manage their own affairs, as diagnosed and determined by a specialist physician. Only a severe patient who poses a risk of harming themselves or others and has been diagnosed by a specialist as needing full-time inpatient treatment but refuses enters the compulsory procedure. In other words: a diagnosis is not a reason for compulsion; "dangerousness + treatment necessity + refusal" is.

Emergency placement is a short-term measure preceding compulsory admission: when a severe patient poses a risk of harming themselves or others, a designated psychiatric institution may carry out emergency placement and conduct a compulsory assessment; emergency placement has a statutory maximum duration (Article 60: placement lasts 7 days, and the compulsory assessment must be completed within 3 days from the day after placement begins), and only if the assessment finds compulsory admission necessary is a petition filed with the court for a ruling.

Compulsory community treatment is the middle option of "no admission needed, but treatment must be ensured": it may be imposed on a severe patient who refuses community treatment, and its components include medication, testing of drug blood levels, screening for alcohol and other substance use, and other measures that can prevent deterioration or improve the patient's ability to adapt to daily life. Its reason for existing is the principle of proportionality — if a less intrusive means can achieve the goal, the most intrusive means should not be used.

🦦

PipHe has schizophrenia and won't take his meds, and his family is at the end of their rope — so let's just send him for compulsory admission?

🐻‍❄️

BarnabyStop. A diagnosis by itself is never a reason for compulsion. Three requirements must all be met together: he must be a legally defined "severe patient" (a mental state detached from reality that leaves him unable to manage his own affairs, as diagnosed and determined by a specialist), he must pose a risk of harming himself or others, and he must have been diagnosed as needing full-time admission yet refuse it. Every one is required. And under the new system, the final call belongs to the court, not the hospital — the court uses a lay-assessor system, the period may not exceed 60 days, and extension is limited to once. Short of compulsory admission, there's also compulsory community treatment, a less intrusive option — that's the principle of proportionality.

An Implementation Timeline That Must Be Stated Honestly

Although the parent act took effect on December 14, 2024 (ROC year 113), the part that transfers compulsory admission to court rulings requires judicial infrastructure — court organization, detailed adjudication rules, and the selection of lay assessors — so its effective date was set separately by the Executive Yuan together with the Judicial Yuan (Article 91), and it took formal effect on August 1, 2026 (ROC year 115) (per the MOHW announcement of July 30, 2026; verified September 2026). This in itself is excellent teaching material on institutions: between a right written into law and a right you can actually use lie the supporting arrangements.

Finally, one more law on the same axis: the Suicide Prevention Act establishes mechanisms for reporting suicidal behavior and follow-up care, and regulates media reporting — avoiding detailed descriptions of suicide methods and locations, and avoiding sensationalism (to prevent the Werther effect, i.e., copycat suicide), and encouraging the simultaneous provision of help-seeking resources and stories of recovery (to promote the Papageno effect) (the Suicide Prevention Act was promulgated on June 19, 2019, and took effect on the date of promulgation).

Protection without a time limit is imprisonment; without community support, compulsory admission becomes the only tool.

⭐ Board-exam essentials

Chapter 17 Must-Knows

Mental Health Act: amended and promulgated December 14, 2022 (ROC year 111), in force December 14, 2024 (ROC year 113); the part transferring compulsory admission to court rulings had a separate effective date because of the judicial infrastructure it needs and has been in force since August 1 (August 2026, ROC year 115).
The court uses a lay-assessor system: judge + lay assessor who is a psychiatric specialist + patient-rights advocacy group representative, deciding by majority opinion. Constitutional basis = judicial reservation for personal liberty.
Time limits: court-ordered compulsory admission may not exceed 60 days; extension must be petitioned 14 days before expiry, only once, and may not exceed 60 days.
Patient-rights groups may monitor cases of compulsory admission, compulsory community treatment, and emergency placement, and may petition the court for termination; severe patients or their protectors may also petition for termination, and petitions and appeals are exempt from court fees.
Definition of "severe patient": bizarre thoughts and strange behavior detached from reality, rendering the patient unable to manage their own affairs (2007 wording; current Article 3 says a mental state detached from reality), as diagnosed and determined by a specialist physician. Diagnosis ≠ grounds for compulsion; requirements for compulsion = severe patient + risk of harm to self or others + need for full-time admission yet refusal.
Emergency placement is a short-term measure (7 days) before compulsory admission, accompanied by a compulsory assessment; compulsory community treatment is a less intrusive middle option (medication, blood-level testing, alcohol/drug screening, etc.) embodying the principle of proportionality.
Community support and reasonable accommodation written into law, strengthening community mental health centers and cross-agency collaboration — using compulsion less presupposes that the community has something to offer.
Suicide Prevention Act: reporting and follow-up care; media guidelines to avoid the Werther effect and promote the Papageno effect.

♪ Sing-along text — English rendering of the Mandarin song lyrics

In the emergency department's safe room, a thirty-year-old man has been brought in by his family together with the police. He hasn't slept in three days, is convinced his neighbor has planted listening devices in the walls, and this afternoon he took a hammer to the neighbor's door. The psychiatrist's assessment is that he needs admission, but he refuses. Before August 2026, what would have happened next was assessment by two specialist physicians, referral to the central review committee for a paper review, and, if approved, compulsory admission — a person's liberty decided by a single official document. And now, that pen is being handed to another pair of hands.

The Mental Health Act was amended and promulgated on December 14, 2022, and took effect on December 14, 2024, and this amendment rests on three pillars. The first is that compulsory admission is now decided by court ruling. Under the old system, a review committee set up by the central competent authority conducted an administrative review; under the new system, the designated psychiatric institution petitions the court for a ruling, and the court uses a lay-assessor system, with a collegiate panel of a judge, a lay assessor who is a psychiatric specialist, and a representative of a patient-rights advocacy group, deciding by majority opinion. Court-ordered compulsory admission may not exceed sixty days; when extension is needed, a petition must be filed with the court fourteen days before the period expires, only one extension petition is allowed, and the extension also may not exceed sixty days.

Protection without a time limit is imprisonment; without community support, compulsory admission becomes the only tool.

Why must it be the court? Because compulsory admission is, in essence, a deprivation of personal liberty, and the principle of judicial reservation in Article 8 of the Constitution requires that any deprivation of personal liberty, in principle, be decided by a court through due legal process. Under the old system an administrative agency conducted a paper review, which placed the petitioning side and the deciding side within the same administrative system, procedurally lacking a neutral third party. But simply handing it to judges has problems too: judges don't know psychiatry and might become a rubber stamp or be overly conservative. So Taiwan chose the lay-assessor system: the judge handles legal requirements and procedure, the psychiatrist lay assessor handles medical judgment, and the patient-rights representative brings the patient's perspective — three kinds of knowledge checking and balancing one another within the same collegiate panel. And the safeguards are a sixty-day cap on duration and a one-time cap on extensions, because protection without a time limit is imprisonment.

The second pillar is procedural safeguards for patients' rights. Public-interest groups promoting patients' rights that are recognized by the central authority may monitor cases of compulsory admission, compulsory community treatment, and emergency placement, notify the competent authority to make improvements when they find something improper, and, out of concern for safeguarding the severe patient's autonomy, equality, and interests, petition the court for a ruling to terminate compulsory admission. The severe patient or their protector may also petition the court to terminate emergency placement, compulsory admission, or compulsory community treatment, and petitions and appeals in such matters are exempt from court fees. The court-fee exemption looks very technical, but it determines whether this right is actually affordable to use.

The third pillar is a community support system. The new act writes community support and reasonable accommodation into law, strengthens the rollout of community mental health centers, establishes cross-agency collaboration and case management across health, social welfare, police, fire services, education, and labor authorities, and emphasizes support for patients and their protectors. This pillar answers the most practical question: if the community has nothing, compulsory admission becomes the only tool, and to use compulsion less, you must first build the community.

Protection without a time limit is imprisonment; without community support, compulsory admission becomes the only tool.

Next come the definitions tested every single year. A severe patient is not the same as a psychiatric patient — this is the highest-frequency trap. Under the law, a severe patient is a patient who exhibits a mental state detached from reality, leaving them unable to manage their own affairs, as diagnosed and determined by a specialist physician. Only when a severe patient poses a risk of harming themselves or others, and has been diagnosed by a specialist as needing full-time inpatient treatment but refuses, does the compulsory procedure begin. In other words, a diagnosis by itself is never a reason for compulsion; dangerousness plus treatment necessity plus refusal — only all three together make one. Emergency placement is a short-term measure before compulsory admission: when a severe patient poses a risk of harming themselves or others, a designated psychiatric institution may carry out emergency placement and conduct a compulsory assessment; placement lasts seven days, with the compulsory assessment completed within three, and only if the assessment finds compulsory admission necessary is a petition filed with the court for a ruling. Compulsory community treatment is the middle option for when admission isn't needed but treatment must be ensured; it is imposed on severe patients who refuse community treatment, and it includes medication, testing of drug blood levels, screening for alcohol and other substance use, and other measures that can prevent deterioration or improve the ability to adapt to daily life. Its reason for existing is the principle of proportionality: if a less intrusive means can achieve the goal, the most intrusive means should not be used.

Finally, one timeline must be stated honestly. Although the parent act took effect on December 14, 2024, the part transferring compulsory admission to court rulings needs judicial infrastructure — court organization, detailed adjudication rules, and the selection of lay assessors — so its effective date was set separately, and it took formal effect on August 1, 2026. This in itself is excellent teaching material on institutions: between a right written into law and a right you can actually use lie the supporting arrangements. On the same axis there is also the Suicide Prevention Act, which establishes mechanisms for reporting suicidal behavior and follow-up care, and regulates media reporting: avoid detailed descriptions of suicide methods and locations and avoid sensationalism, to prevent the Werther effect of copycat suicide, and encourage the simultaneous provision of help-seeking resources and stories of recovery, to promote the Papageno effect.

Protection without a time limit is imprisonment; without community support, compulsory admission becomes the only tool.