ICER, Dominance and the Cost-Effectiveness Plane: Reading the Thai Threshold of 160,000 Baht per QALY

อ่านฉบับภาษาไทย (Thai version)

Abstract

A district hospital committee must choose one of four ways to care for adults with a chronic condition. It will pay at most 160,000 THB, its threshold, for one extra year in full health. Usual care costs 100,000 THB per patient; a new preventive treatment, C, costs 250,000 THB and adds 1.00 such year. The analysts' report gives C three prices: 41,667, 112,500 and 150,000 THB per year. Only the last compares C with the option patients would otherwise get. The incremental cost-effectiveness ratio (ICER) is that extra cost divided by the extra health. Finding the right comparison means first removing any option that another beats on both cost and health, a rule called dominance, and any option that a mix of two others beats. Net monetary benefit, the health valued at the threshold minus the cost, reaches the same choice without a ratio: C leads by 10,000 THB per patient, and at 150,000 THB the two tie.

🔍 Click to enlarge Click anywhere to close

One committee, four options, three prices

A district hospital committee must choose one way to care for adults newly diagnosed with a chronic condition. Four options are on the table, and each patient can receive only one. Over a lifetime, usual care, A, costs 100,000 THB (Thai baht) per patient and gives the equivalent of 5.00 years in full health.

Adding nurse follow-up to usual care, option B, costs 160,000 THB and gives 5.20 years. A new preventive treatment, C, costs 250,000 THB and gives 6.00 years. A rival brand of that treatment, D, costs 270,000 THB and gives 5.90 years. All four are illustrative numbers for teaching.

The committee has also set a limit. It will pay at most 160,000 THB for one extra year in full health, the amount it judges such a year to be worth. This limit is called the threshold; the section on the Thai threshold near the end explains where the Thai figure comes from.

The analysts' report gives three prices for C. Dividing C's whole cost by its whole health gives 250,000 / 6.00 = 41,667 THB per year in full health. Against B, C costs 90,000 THB more for 0.80 more years, which is 112,500 THB per extra year. Against A, it costs 150,000 THB more for 1.00 more year, which is 150,000 THB per extra year.

The choice of price matters. At the committee's 160,000 THB all three say yes to C, but a committee with a limit of 140,000 THB would hear yes from the first two and no from the third. This article shows which price is the real one, which options to set aside before computing it, and how to check the answer without dividing at all.

Health economists count those years in a unit called the QALY. A QALY, a quality-adjusted life year, is a year of life weighted by its quality, from 1 for full health to 0 for death. Over 10 years, a state valued at 0.70 counts as 7.0 QALYs, so restoring full health gains 3.0; the full lesson is in QALY, DALY and the Utilities Behind Them: What Is Actually Being Measured. So A gives 5.00 QALYs per patient and C gives 6.00 QALYs.

Pay for what C adds

Start with the first price, 41,667 THB. It treats every year that patients on C live in full health as if C had bought it. Yet patients on usual care would have had 5.00 of those years anyway, for 100,000 THB. By the same division A looks even cheaper, at 20,000 THB per year, although it buys the least health.

So the committee should ask what C adds to what patients would otherwise receive. C costs 250,000 − 100,000 = 150,000 THB more than A and gives 6.00 − 5.00 = 1.00 more QALY. Each extra QALY therefore costs 150,000 / 1.00 = 150,000 THB, under the threshold of 160,000 THB.

This number has a name. The incremental cost-effectiveness ratio (ICER) is the extra cost of a treatment divided by the extra health it buys, compared with the next best option. In symbols, with $\Delta\text{Cost}$ the difference in cost and $\Delta\text{QALY}$ the difference in QALYs, each taken as the new option minus the option it is compared with:

$$\text{ICER}=\frac{\Delta\text{Cost}}{\Delta\text{QALY}}=\frac{250000-100000}{6.00-5.00}=150000$$

When an option costs more and gives more, a ratio under the threshold means the extra health is worth its extra cost. The hard part is the phrase "next best option": it means the next cheaper option still in the running once the beaten options are set aside. Comparing C with B gave 112,500 THB per QALY, and the sections "One option that beats another on both counts" and "Remove what a mix of two options beats" below decide whether B or A is that option.

Other units for the bottom of the ratio

Read first on this site:

• How a trial's two risks give a risk difference

The bottom of the ratio can be any measure of health, and the analysis is named after it. In the trial behind C, 8 of 66 treated patients and 13 of 60 controls had a complication, risks of 12.12% and 21.67%. Complications therefore fell by about 9.55 per 100 patients treated during the trial's follow-up. Per 100 patients the extra cost is 150,000 × 100 = 15,000,000 THB, so each prevented complication costs about 15,000,000 / 9.55 = 1,570,681 THB: a cost-effectiveness analysis (CEA), counted in a clinical unit.

Counted in QALYs, as this article does for C, it is a cost-utility analysis (CUA), the only form that can be held against a threshold set per QALY. Counted in money, it is a cost-benefit analysis (CBA): the health must first be given a baht value, and the verdict is the money gained per baht spent, the return on investment (ROI). C's 150,000 THB per QALY is a price, not a return: it says what one QALY costs, not how many baht come back for each baht spent.

Comparing costs alone, a cost-minimisation analysis (CMA), is enough only when the options give the same health. C costs 250,000 THB against D's 270,000 THB, but D also gives 0.10 QALY less, so a comparison of costs alone would leave that difference out; a small trial with no significant difference has not shown equal health either.

# Four options: lifetime cost (THB) and QALYs per patient. Writes: none (stdout only).
cost <- c(A = 100000, B = 160000, C = 250000, D = 270000)
qaly <- c(A = 5.00, B = 5.20, C = 6.00, D = 5.90)
f <- function(x, d = 0) formatC(x, format = "f", digits = d, big.mark = ",")
cat("QALY recap: 10 years at 0.70 =", f(10 * 0.70, 1), " full health gains", f(10 * 0.30, 1), "\n")
cat("Total cost / total QALYs, A and C:", f(cost["A"] / qaly["A"]), f(cost["C"] / qaly["C"]), "\n")
dc <- cost["C"] - cost[c("A", "B")]
dq <- qaly["C"] - qaly[c("A", "B")]
cat("C vs A and C vs B: THB", f(dc), " QALYs", f(dq, 2), " THB per QALY", f(dc / dq), "\n")
risk <- c(treated = 8 / 66, control = 13 / 60)
cat("Risk %:", f(100 * risk, 2), " fewer complications per 100:", f(100 * diff(risk), 2), "\n")
cat("Extra cost for 100 patients on C instead of A, THB:", f(100 * dc["A"]), "\n")
cat("THB per complication prevented:", f(100 * dc["A"] / round(100 * diff(risk), 2)), "\n")

QALY recap: 10 years at 0.70 = 7.0  full health gains 3.0
Total cost / total QALYs, A and C: 20,000 41,667
C vs A and C vs B: THB 150,000 90,000  QALYs 1.00 0.80  THB per QALY 150,000 112,500
Risk %: 12.12 21.67  fewer complications per 100: 9.55
Extra cost for 100 patients on C instead of A, THB: 15,000,000
THB per complication prevented: 1,570,681

Read the line, not the sign

A ratio hides which way each difference points. Take the rival brand, D, against C. D costs 270,000 − 250,000 = 20,000 THB more and gives 5.90 − 6.00 = −0.10 QALY, so its ratio is 20,000 / −0.10 = −200,000 THB per QALY. Swap the roles and C against D saves 20,000 THB and gains 0.10 QALY, a ratio of −20,000 / 0.10 = −200,000 THB per QALY as well.

The same −200,000 describes the worst kind of switch and the best kind. A negative ratio cannot be read until you know which difference is negative.

A chart fixes this. Put the option you switch from at the centre, draw the extra health across, to the right for more, and the extra cost up, upward for more. Every switch then lands in one of four quarters: more health for more cost, more health for less, less health for more, or less health for less. This chart is the cost-effectiveness plane, and its quarters are called quadrants.

Now draw a straight line through the centre that rises 160,000 THB for each QALY to the right: the threshold drawn as a slope. A switch below the line buys health for less than 160,000 THB per QALY, or saves more than 160,000 THB for each QALY it gives up. So below the line is worth it in every quadrant, and above it is not.

One option that beats another on both counts

Read the chart switch by switch. C instead of A sits at 1.00 QALY and 150,000 THB, just under the line: yes. A instead of C, the same switch backwards, sits at −1.00 QALY and −150,000 THB.

That switch saves only 150,000 THB for a lost QALY worth 160,000 THB, so it lands above the line: no. C instead of D lands below the line, and D instead of C lands above it.

D's corner has a name. D costs more than C and gives less health, so no threshold can ever make D the better buy. Dominance is the rule that removes an option when another costs less and gives at least as much health. C dominates D, and D is set aside before any ratio is computed.

Comparing every option with usual care alone hides this. Against A, D costs 170,000 THB more for 0.90 more QALY, or 188,889 THB per QALY, so it would pass any threshold above 188,889 THB. Yet C gives more health for less money.

Remove what a mix of two options beats

B is harder to spot. Against A, nurse follow-up costs 160,000 − 100,000 = 60,000 THB more for 5.20 − 5.00 = 0.20 more QALY: 300,000 THB per QALY. Going on from B to C costs 250,000 − 160,000 = 90,000 THB more for 0.80 more QALY: 112,500 THB per QALY. The second step is cheaper per QALY than the first, which is a warning sign.

Suppose the committee gave C to 20% of patients and A to the other 80%. The mix would cost 100,000 + 0.20 × 150,000 = 130,000 THB per patient on average and give 5.00 + 0.20 × 1.00 = 5.20 QALYs, the same health as B. B buys that health for 160,000 THB, so it costs 30,000 THB more than a mix of its two neighbours. Nobody needs to run such a mix; it only shows that B buys its health at a poor price per QALY.

Extended dominance and the efficiency frontier

This is extended dominance: an option is removed when a mix of the options on either side of it gives the same health for less [3,4]. B is extendedly dominated by A and C. On a chart of cost against QALYs, B sits above the straight line from A to C, and anything above that line loses to a point on it.

With D and B gone, the options left, A and C, form the efficiency frontier: the line of options that give the most health for each amount spent. The right comparison for C is its neighbour on that line, A, so the real price is 150,000 THB per QALY, not 112,500 THB. The 112,500 THB came from comparing C with an option that should already have been removed.

The rule in short: rank the options by cost, remove the dominated ones, then remove any middle option whose ratio is higher than the next one, until the ratios rise from step to step. Here one ratio is left, 150,000 THB per QALY for C against A.

For the same steps on a second case with more options, read:

• A heart team choosing one of four follow-up plans

# Who beats whom: each option against A, D against C, a mix of A and C against B. Writes: none.
cost <- c(A = 100000, B = 160000, C = 250000, D = 270000)
qaly <- c(A = 5.00, B = 5.20, C = 6.00, D = 5.90)
f <- function(x, d = 0) formatC(x, format = "f", digits = d, big.mark = ",")
dc <- cost[-1] - cost["A"]
dq <- qaly[-1] - qaly["A"]
cat("Against A, B C D: extra cost THB", f(dc), " extra QALYs", f(dq, 2), "\n")
cat("Against A, B C D: THB per extra QALY", f(dc / dq), "\n")
dd <- c(cost["D"] - cost["C"], qaly["D"] - qaly["C"])
cat("D minus C: THB", f(dd[1]), " QALYs", f(dd[2], 2), " ratio", f(dd[1] / dd[2]), "\n")
share_c <- dq["B"] / dq["C"]
mix <- cost["A"] + share_c * dc["C"]
cat("Mix,", f(share_c, 2), "on C: THB", f(mix), " B costs more by THB", f(cost["B"] - mix), "\n")

Against A, B C D: extra cost THB 60,000 150,000 170,000  extra QALYs 0.20 1.00 0.90
Against A, B C D: THB per extra QALY 300,000 150,000 188,889
D minus C: THB 20,000  QALYs -0.10  ratio -200,000
Mix, 0.20 on C: THB 130,000  B costs more by THB 30,000

One score in money: net monetary benefit

Ratios need care: the right option to compare with, the right quadrant, and no dividing by zero when two options give the same health. There is a way to reach the answer without dividing. Turn each option's health into money at the threshold, then subtract what the option costs.

At 160,000 THB per QALY, C's 6.00 QALYs are worth 960,000 THB; minus its cost of 250,000 THB, C is left with 710,000 THB. A's 5.00 QALYs are worth 800,000 THB, which leaves 700,000 THB after its cost of 100,000 THB. B is left with 832,000 − 160,000 = 672,000 THB and D with 944,000 − 270,000 = 674,000 THB. The committee picks the highest score: C.

This score is called net monetary benefit (NMB): it values health in money at the threshold and subtracts the cost [1]. With $\lambda$ for the threshold:

$$\text{NMB}=\lambda\times\text{QALY}-\text{Cost},\quad 160000\times 6.00-250000=710000$$

Comparing two options is then a subtraction. C's 710,000 THB minus A's 700,000 THB is 10,000 THB, the same as 160,000 × 1.00 − 150,000: the health C adds, valued in money, minus the extra money it costs. This difference is the incremental net monetary benefit (INMB), $\text{INMB}=\lambda\times\Delta\text{QALY}-\Delta\text{Cost}$, and a positive value means the extra cost is worth paying.

It also works in every quadrant. D's 674,000 THB is lower than C's 710,000 THB, so switching from C to D loses, and no negative ratio has to be read. It even works when two options give exactly the same health, where a ratio would divide by zero.

Where the choice flips

The threshold is a judgement, so the committee should see how its choice moves with it. At 150,000 THB per QALY, A's 5.00 QALYs are worth 750,000 THB and C's 6.00 are worth 900,000 THB, so after their costs both are left with 650,000 THB. At 140,000 THB per QALY, A's QALYs are worth 700,000 THB and C's 840,000 THB, so A leads with 600,000 THB against C's 590,000 THB. B and D stay lower at both thresholds (B 620,000 and 568,000 THB; D 615,000 and 556,000 THB).

So C wins above 150,000 THB per QALY and A wins below it. The switching point is exactly the ratio of C against A: the ratio is the threshold at which the extra cost and the money value of the extra health balance.

# Net monetary benefit per patient: QALYs valued at the threshold, minus cost. Writes: none.
cost <- c(A = 100000, B = 160000, C = 250000, D = 270000)
qaly <- c(A = 5.00, B = 5.20, C = 6.00, D = 5.90)
f <- function(x) formatC(x, format = "f", digits = 0, big.mark = ",")
for (threshold in c(160000, 150000, 140000)) {
  cat("Threshold", f(threshold), "THB per QALY\n")
  cat("  QALYs valued in THB, A B C D:", f(threshold * qaly), "\n")
  cat("  NMB THB, A B C D:", f(threshold * qaly - cost), "\n")
}
inmb <- 160000 * (qaly["C"] - qaly["A"]) - (cost["C"] - cost["A"])
cat("INMB of C vs A at 160,000, THB:", f(inmb), "\n")
cat("A and C tie at, THB per QALY:", f((cost["C"] - cost["A"]) / (qaly["C"] - qaly["A"])), "\n")

Threshold 160,000 THB per QALY
  QALYs valued in THB, A B C D: 800,000 832,000 960,000 944,000
  NMB THB, A B C D: 700,000 672,000 710,000 674,000
Threshold 150,000 THB per QALY
  QALYs valued in THB, A B C D: 750,000 780,000 900,000 885,000
  NMB THB, A B C D: 650,000 620,000 650,000 615,000
Threshold 140,000 THB per QALY
  QALYs valued in THB, A B C D: 700,000 728,000 840,000 826,000
  NMB THB, A B C D: 600,000 568,000 590,000 556,000
INMB of C vs A at 160,000, THB: 10,000
A and C tie at, THB per QALY: 150,000

What 160,000 THB per QALY means

Thailand has used 160,000 THB per QALY since 2013 as the threshold for decisions like this one (the second edition of its health technology assessment guidelines is described in [5]). It is a demand-side figure: an estimate of what society is willing to pay for one more QALY. Treat it as a stated assumption of the analysis, not a fact of nature.

A supply-side threshold asks a different question: how much health the same money would buy if it were spent on other services instead. The health given up there is the opportunity cost of funding C. This article gives no Thai supply-side estimate, so 160,000 THB is read only as a willingness to pay. A supply-side figure, if one were estimated, would be the number to compare against, and a lower one could remove C's narrow margin.

Rules that set the threshold at one to three times gross domestic product (GDP) per person are easy to compute. They have no evidence base for what health a budget actually displaces. A clear analysis states its threshold and shows how the choice changes at other values, as the switching point at 150,000 THB per QALY did above.

C's lead is narrow: 10,000 THB per patient in net monetary benefit, and a ratio only 10,000 THB under the threshold. Every input behind it, from the trial's effect to the price of C, is uncertain, so a small change could reverse the choice. A sound report therefore shows how often each option comes out best when the inputs vary [2].

Probabilistic sensitivity analysis (PSA) reruns the model many times with every uncertain input drawn from its distribution, so the result is a share of runs rather than one number. Suppose C's INMB over A varies from run to run along a bell-shaped (normal) curve with mean 10,000 THB and standard deviation 40,000 THB. C is ahead in every run where the INMB is above zero. Zero lies 10,000 / 40,000 = 0.25 standard deviations below the mean, and a normal curve puts 0.5987 of runs above that point.

So about 599 of 1,000 runs favour C and 401 favour A, a share of 0.60 for C; the full lesson is in Probabilistic Sensitivity Analysis: From Parameter Distributions to CEAC.

A report like this also follows the reporting checklist CHEERS 2022 (Consolidated Health Economic Evaluation Reporting Standards) [6].

Value for money is also not the same as affordable or fair. Funding C for every new patient in the district changes the yearly budget, and the health gained may fall unevenly across groups; Cost-Effective Is Not Affordable, and Efficient Is Not Fair: Budget Impact and Equity takes up both questions.

# How sure is C's lead? Its extra net benefit over A, spread across 1,000 runs. Writes: none.
inmb <- c(mean = 10000, sd = 40000)
runs <- 1000
f <- function(x, d = 0) formatC(x, format = "f", digits = d, big.mark = ",")
cat("INMB of C over A, mean and standard deviation, THB:", f(inmb), "\n")
z <- inmb["mean"] / inmb["sd"]
cat("Zero lies this many standard deviations below the mean:", f(z, 2), "\n")
share_c <- pnorm(z)
cat("Share of a normal curve above zero:", f(share_c, 4), "\n")
cat("Runs out of", f(runs), "favouring C and A:", f(runs * share_c), f(runs * (1 - share_c)), "\n")
cat("Share of runs favouring C:", f(share_c, 2), "\n")

INMB of C over A, mean and standard deviation, THB: 10,000 40,000
Zero lies this many standard deviations below the mean: 0.25
Share of a normal curve above zero: 0.5987
Runs out of 1,000 favouring C and A: 599 401
Share of runs favouring C: 0.60

The series, in learning order

1. Health Economics Series Guide: Read the Evidence Before the Costs
2. Incidence, Prevalence and the Right Denominator: The Epidemiology Beneath Cost-of-Illness and Markov Models
3. Design, Model and Estimand: What Each Study Design Can Actually Estimate
4. Risk Ratio, Odds Ratio, Risk Difference and NNT: One 2×2 Table, Four Answers
5. The Hazard Ratio Is Not a Risk Ratio: Proportional Hazards, Restricted Mean Survival Time and Model Inputs
6. Chance, Bias and Confounding: Why an Association Is Not Yet a Treatment Effect
7. Beyond the Evidence Pyramid: GRADE Certainty and Choosing Effectiveness Inputs for Economic Models
8. From Sensitivity to Cost per Case Detected: Predictive Values, Prevalence and the Economics of Screening
9. Whose Cost, in Which Year? Perspective, Opportunity Cost, Inflation and Discounting
10. QALY, DALY and the Utilities Behind Them: What Is Actually Being Measured
11. ICER, Dominance and the Cost-Effectiveness Plane: Reading the Thai Threshold of 160,000 Baht per QALY (this article)
12. Decision Trees, Markov Cohorts and Probabilistic Sensitivity Analysis: How a Model Turns Evidence into a Decision
13. Cost-Effective Is Not Affordable, and Efficient Is Not Fair: Budget Impact and Equity
14. Parameter Sourcing, Reference Cost Lists and Utility Catalogues When Local Data Are Sparse
15. Early Health Technology Assessment, Headroom Pricing and Benefit-Package Decisions
16. Communicating Health Economic Evidence: From Model Outputs to Policy Briefs

Related in the wiki: Whose Cost, in Which Year? Perspective, Opportunity Cost, Inflation and Discounting · QALY, DALY and the Utilities Behind Them: What Is Actually Being Measured · Decision Trees, Markov Cohorts and Probabilistic Sensitivity Analysis: How a Model Turns Evidence into a Decision · Cost-Effective Is Not Affordable, and Efficient Is Not Fair: Budget Impact and Equity · The Role of Health Economic Evaluation in National Drug Selection and Decision-Making · Decision Trees and Markov Models in Health Economic Evaluation: Principles, Construction, and Applications · ICER and Net Monetary Benefit Across Strategies: Find the Relevant Frontier · Probabilistic Sensitivity Analysis: From Parameter Distributions to CEAC