The ratio's revenge.

Put them in the same units.

The reason we reached for a ratio in the first place is that cost and health are measured in different units — pounds and QALYs. You can't subtract £12,000 from 0.8 QALYs; the units don't match. Division sidestepped that by producing a rate (£ per QALY). But it cost us the signs.

There's another way to make the units match, and we already have the tool for it. The threshold, k, is an exchange rate between money and health: it says how many pounds one QALYs is worth (or, flipped, how much health one pound buys). With an exchange rate, you don't need a ratio — you can convert one quantity into the other and then simply subtract.

Two directions, two versions of net benefit:

• Convert money into health → net health benefit, measured in QALYs.
• Convert health into money → net monetary benefit, measured in pounds.

Same idea, same decision. Just pick the currency you want the answer in.

Net Health Benefit.

Start by converting money into health. The cost of a technology, ΔCost, would have produced some health elsewhere if left in the budget — exactly the displacement idea from the last lesson. At a threshold k, that displaced health is ΔCost ÷ k QALYs. So the net effect on population health is what you gain minus what you displace:

Measured in QALYs. If it's positive, the technology adds more health than it displaces — worth adopting. If it's negative, it destroys more than it creates.

And here's the thing: you've computed this before. Last lesson's displacement calculator showed QALYs gained minus QALYs displaced, and watched the net go negative when the technology cost too much per QALY. That net figure was the net health benefit. You built the concept before you had its name.

Net Monetary Benefit.

Now convert the other way — health into money. Value the QALYs gained at the threshold (ΔEffect × k), then subtract what they cost:

Measured in pounds. Positive means the health gained, priced at the threshold, is worth more than the money spent — worth adopting. Negative means it isn't.

NMB and NHB are the same statement in different currencies. In fact:

They cross zero at exactly the same point, so they always agree on the decision — a technology is either cost-effective under both or neither. The only difference is the unit the answer arrives in. Reporting to a clinical committee that thinks in health outcomes? Use NHB, in QALYs. Reporting to a finance director who thinks in budgets? Use NMB, in pounds. Same verdict, chosen language.

One point, three verdicts.

Drag the point — a technology versus standard care. Slide the threshold to change the exchange rate between money and health. Watch three numbers at once: the ICER, the NHB, and the NMB. Notice which ones stay honest as you move around the plane.

Drag into the dominant corner (cheaper and better): the ICER turns negative and useless, but NHB and NMB shoot clearly positive. Drag into the dominated corner: the ICER is negative again — same sign, opposite situation — while net benefit goes clearly, correctly negative. That's the whole point. The ICER's sign is a riddle; net benefit's sign is an answer. It means the same thing in all four quadrants.

Why the difference beats the ratio.

Three concrete wins, each one repairing a specific failure from the ICER lesson:

1. The sign always means the same thing. Positive net benefit = good deal, negative = bad deal, in every quadrant. No more "is this negative ICER dominant or dominated?" The subtraction keeps the direction the ratio threw away.
2. You can rank any number of options. Net benefit gives every option a single score on one scale, so you just sort them and take the top. ICERs can't be ranked that simply — with several mutually exclusive options they demand a fiddly step-by-step incremental analysis that's easy to get wrong (as the next screen shows).
3. It survives uncertainty. Real appraisals aren't one point — they're thousands of simulated points, because ΔCost and ΔEffect are uncertain. You can average net benefit across those simulations and get a meaningful expected value, because it's a simple linear expression. You cannot average ICERs — ratios blow up when the denominator nears zero. This is why the uncertainty methods in Module 9 are built on net benefit, not the ICER. Hold that thought; it's the reason net benefit quietly runs modern HTA.

Now you.

Ranking: where the ICER can't follow.

Here's where net benefit stops being elegant and starts being indispensable. Three mutually exclusive treatments — you can pick only one — each compared with standard care, at a £20,000 threshold:

Rank them by ICER and C looks best — the cheapest QALYs on the table at £12,000 each, with B the worst at £16,000. So the "lowest ICER" instinct picks C.

Now rank by net benefit. B wins outright at £6,000, then A, then C last. The instinct was backwards.

Why? All three buy QALYs below the threshold, so all three are cost-effective — the question isn't whether to treat but which delivers the most value overall. B produces the most health, and even at its higher price-per-QALY every one of those QALYs still clears the threshold, so it accumulates the most net benefit. C's QALYs are individually cheap, but there just aren't enough of them. A well-run incremental ICER analysis eventually reaches the same answer — B — but only after ruling out dominated and extended-dominated options in sequence, a procedure that trips up even experienced analysts. Net benefit gets there in one column of subtraction.

What does a positive NMB tell you?

Why this matters for HTA

Open a modern appraisal and you'll still see the ICER in the headline — but underneath, the machinery increasingly speaks net benefit. Knowing why changes how you read the whole document.

• A dossier that reports only an ICER is telling you half the story. The ICER gives cost per QALY but not scale. Two drugs with identical ICERs can have very different net benefits if one delivers far more health. When scale matters — and for budget-holders it always does — ask for the net benefit.
• Net benefit is the honest place to see the threshold's grip. Because NMB and NHB are computed at a chosen k, sliding that k shows exactly how fragile or robust a "cost-effective" verdict is. A drug with an NMB of +£200 is one threshold revision away from negative; one at +£40,000 is not. The ICER hides that sensitivity; net benefit displays it.
• It's the foundation everything uncertain is built on. Every probabilistic result you'll meet in Module 9 — expected net benefit, cost-effectiveness acceptability curves — exists because net benefit can be averaged and the ICER can't. Recognising net benefit now means those tools will look familiar, not foreign.

The ICER is the number that gets quoted. Net benefit is the number that gets used. Learn to read the second and you're reading what the committee is actually deciding on.

Net benefit, in one breath.

• Net benefit reformulates cost-effectiveness as a difference, not a ratio — using the threshold as an exchange rate to put cost and health in the same units.
• Two forms, one decision: NMB = (k × ΔEffect) − ΔCost (in £) and NHB = ΔEffect − (ΔCost / k) (in QALYs), with NMB = k × NHB. Positive = cost-effective.
• It fixes every flaw of the ICER: the sign always means the same thing, you can rank any number of options, and — uniquely — it can be averaged under uncertainty.
• The lowest ICER is not always the best option; the highest net benefit is.

Stop dividing, start subtracting. The same question — is the health worth the cost? — but asked in a way that never loses the answer.

That closes Module 7. You now have the whole toolkit: the plane to see the decision, the ICER to price it, the threshold to judge it, and net benefit to make it honest and rankable. But every one of these ran on two numbers we simply assumed — ΔEffect and ΔCost, measured over a patient's whole lifetime. Trials rarely run that long. So where do those lifetime numbers actually come from? That's Module 8: economic modelling — how we extrapolate beyond the data we have.