Why do health systems spend so little on prevention?

Where the money in a European health system actually goes, what fee-for-service and case-based payment do to prevention, and what that leaves for one person.

A budget is the honest document. I trained as an engineer and I read budgets the way other people read mission statements, because a budget can't posture. If you want to know what an institution is for, don't open its strategy. Open its accounts.

The accounts of European health systems say treatment. Prevention gets a sliver, and the usual reading of that sliver is wrong: it isn't evidence that anybody inside doubts prevention works. Most people I've met in a health system believe in it, often more seriously than the people selling supplements do. The sliver tells you what the building can see, fund, code and defend in front of a finance committee. That's an accounting problem, and accounting problems don't get fixed by finding better people.

So the claim here is narrow. These systems find and repair illness that has already announced itself, and they do it well. What hasn't arrived yet sits outside their field of view. A person who stays well generates no billable event, and institutions do the things they can put on a bill. That changes what it makes sense to walk in and ask for.

This is an educational and strategic perspective, not personal medical advice. The views are the author's own and not statements by Atlas Cove Lda.

Three percent, and what that number is really counting

Across OECD countries, the share of health expenditure booked as preventive care has hovered near 3 percent for years. It rose to something like 6 percent while the pandemic ran and then slid back to roughly its previous level (OECD, 2023). The rest, better than nine euros in every ten, goes to conditions that have already declared themselves.

Worth knowing what that figure is and isn't. It's a spending classification, counting what national accounts label preventive: immunisation programmes, screening, health promotion, a slice of public health administration. It misses prevention that happens incidentally inside a consultation billed as something else, and the very large amount that happens nowhere near a health budget, in housing, in road design, in what a food industry may sell. So 3 percent understates preventive effort rather than describing it exactly. The gap it points at is wide enough that the imprecision rescues nothing.

A number like that tells you what the machinery can notice, pay for and enter in a ledger. It tells you little about what the people inside it think. Strategy documents describe what an organisation would like to be true about itself. Accounts describe what it did.

Fee-for-service and DRG, two models with one blind spot

Fee-for-service pays a provider per item of service delivered, so revenue tracks billable volume. Respond rationally to that signal and you do more of whatever the coding system recognises. That isn't corruption, it's arithmetic.

Case-based payment was built to stop precisely that. A Diagnosis-Related Group system sorts every hospital admission into one of several hundred classes, defined by the principal diagnosis, the procedures carried out and how complicated the patient turned out to be, then pays a fixed tariff for the class instead of for each thing done inside it (Busse et al., 2011). Portugal, where I live and work, runs on them. So do Germany, France and Spain. Inside the hospital it works: a fixed price per case makes an admission something to handle briskly rather than lavishly, which was the entire reason for introducing DRGs.

Neither model produces a unit of revenue for the case that never arrived. One pays for events, the other for classified events. An illness prevented is not an event at all, so it registers in neither.

Porter and Teisberg laid out the general form of this two decades ago, arguing that these systems end up competing on service volume, and on pushing cost onto one another, instead of on the result a patient carries away from an entire episode of care (Porter and Teisberg, 2006). Their book is about American healthcare, which a reader should hold in mind. What transfers is the observation underneath, about what a unit of payment can refer to at all, and every payer in Europe still has to choose one.

The fifteen minute appointment answers its own question correctly

This reaches down into the individual consultation too, which is where it stops looking like economics and starts feeling like something you personally got wrong.

You arrive with a question about direction. Where is this heading over the next ten years, given what you can see today? The consultation has one output to produce: is there something diagnosable here, and if so, what follows. A handful of markers drifting the same way across a few years, every one still inside its reference interval, is a real signal to anybody thinking in decades and a complete non-event for that particular decision. The room did the job it was built to do. Lisa wrote the patient's-eye version of that moment in her piece on labs that come back fine.

Berwick made the general point in 2002, in a commentary on the Institute of Medicine's quality work. A health system's output follows from its design and from whatever that design elects to measure and reward, which is why telling staff to try harder inside an unchanged structure has never produced anything different (Berwick, 2002). Two things about that citation: it's over twenty years old, and it's a commentary rather than a trial. I use it because the argument is structural and has aged well, not because anybody measured it.

Procedures and coded cases get counted precisely. Illness that never happened gets counted badly, because one of those has an identifier attached and the other has none.

The shortfall appears even in the work these systems fully mean to do. The most cited attempt at measuring it looked at adults in the United States and found the recommended standard of care reached them about 55 percent of the time, with the authors blaming systems that fail to implement knowledge they already hold rather than clinicians who lacked it (McGlynn et al., 2003). I want to be straight about that figure, because it's American data doing work inside a European argument. It came from telephone interviews and medical record review across twelve US metropolitan areas, under a financing arrangement resembling nothing in Europe, and it dates from 2003. A European replication would produce different numbers, better ones in several places. The 55 was never the transferable part. What transfers is that a system delivers inconsistently whatever it isn't structured to measure.

Four appointments, and I blamed the wrong thing

For two winters and the springs that followed them, whole days went to severe migraine, flu-like aching and a fatigue that sleeping through didn't dent. I went in, the bloods came back unremarkable, somebody said it all looked fine, and I went home with nothing. Four times. Possibly five, I've lost count. Somewhere in there I filed the whole business under lack of interest.

I had that wrong. I'd assigned blame before I bothered to check anything.

Each of those visits answered its own question, and answered it correctly: nothing diagnosable present. The question I actually carried in was about regulation across systems, and it belonged to no single organ, so I kept addressing it to the only institution in my life with no channel for it. My own version does carry a name, as it happens. Cold urticaria, driven by mast cells, and behind it a family history of autoimmune dysregulation. A name is worth having. It still isn't a plan. Regulation is a subject of its own, and I've written about the budget a nervous system runs on.

What eventually shifted things took months and would have shown up as nothing on a chart. I stopped drinking. Bedtime and waking became fixed hours I no longer argued with. I pulled so much intensity out of my training that it stopped really being training. Recovery returned first, performance a long way behind it, an order I hadn't anticipated and didn't enjoy. Nobody prescribes that in a fifteen minute slot, and not because the slot is short. It's that none of those are things anybody can do to you once and be finished.

Bismarck, Beveridge, and why switching model doesn't help

The obvious objection turns up around here. Europe runs several quite different models, so surely at least one handles this better.

Two definitions, since the terms get used loosely. A Bismarck system finances care through compulsory social insurance contributions from employers and employees, administered by a plurality of sickness funds sitting apart from the state. A Beveridge system pays for care from general tax revenue, with the state as the main purchaser and usually the owner of the hospitals too. Germany, where I'm from, is Bismarck. Portugal, where I live now, is Beveridge. Plenty of countries sit somewhere in the middle, with their own mixtures of co-payment and waiting list.

Inside them the two feel nothing alike, and I've been a patient in each. The prices differ, the waiting differs, and getting to a specialist is a different exercise entirely. Nobody who has queued in both Berlin and Lisbon would tell you the experiences are interchangeable. In neither country has anyone ever asked me where my health was heading over the coming decade.

What they share is the direction the money runs. In both, and in the hybrids, the euro moves once an existing illness gets diagnosed and treated (OECD, 2016). Prevention consists of actions, and actions are in principle purchasable, so the obstacle isn't philosophical. The obstacle is the trigger. A payment fires when something diagnostic or therapeutic happens to an illness that already exists. Work whose whole success consists in an illness never arriving has nothing to fire on.

A hospital paid per case manages its case mix and its lengths of stay. A provider running inside a fixed global budget has to hold demand within that ceiling, and prevention pushes demand down years later while spending money now, which pays back nothing inside the period that provider is judged on. A mixed system inherits both logics and lands exactly where the other two land. Changing country is no escape from that, and when the OECD last reviewed the ways care gets paid for, rewarding prevention was still filed under unsolved design problems (OECD, 2016).

You can't invoice an absence

None of this requires bad faith, and that's worth stating flatly, because an incentive structure, viewed from inside it, looks a great deal like malice when it's only economics doing what economics does. There's a separate argument about who profits when health becomes a product. Mine is duller, and it needs nobody to be greedy.

Treatment is easy to define, which is exactly why it's easy to pay for. Every procedure has its code and every diagnosis its tariff, an admission leaves behind recorded activity and the revenue that goes with it, and a payer can audit the lot afterwards. Prevention fails all of those tests. Nobody has ever sent an invoice for a heart attack that failed to occur, and there's no identifier for ten years of function that simply held.

A second economic property sits underneath, and it's the part that holds my attention. Medical care is a credence good, a term from economics for a class of goods where the buyer cannot assess quality even after purchase, and therefore has to take the seller's word for what was required at all. A car repair is the standard example: you mostly cannot tell whether the part needed replacing. That asymmetry explains a great deal about why medicine ended up licensed and bound by codes of conduct.

Prevention is the extreme case of the type. Its output, when it works, is a thing that didn't happen, and a thing that didn't happen cannot be inspected, compared, or credited to anyone. Markets price the permanently invisible badly. No finance minister finds it easy to defend a line item whose payoff shows up in a later electoral cycle, in another institution's accounts, under a different name.

That gap matters more than it first appears, because medical care was never the largest determinant of health anyway. Work on the determinants of population health has attributed something like a tenth of premature deaths to gaps in medical care itself, with behaviour, social circumstance, genetics and environment carrying the remainder (McGinnis, Williams-Russo and Knickman, 2002). Marmot reached a compatible place from the other end, documenting how living and working conditions set the health of a population in ways clinical care afterwards cannot undo (Marmot, 2010).

Both are old, and worth flagging as such. The McGinnis figure is modelled rather than measured and has been argued over since 2002, and Marmot's review appeared in 2010 describing England at one moment. I'd back the direction of both and hold the numbers loosely. If either turned out materially off, this argument moves rather than collapses, since it rests on the payment trigger.

Hold that against how the money is distributed and the shape is hard to miss. Nearly all the spending, and nearly all the attention, goes to the narrow slice of health that medical care directly determines. The far larger slice, where almost all the room for prevention sits, falls outside what gets funded or counted. Individual choices carry enormous weight there. Nobody is paid to be present while they're being made.

Four questions to aim at the system

I'd hate this to become an argument against clinicians. They understand prevention perfectly well as a rule, they want to help, and they work inside a structure none of them designed and none of them can move alone. The reason to look at the structure is smaller than blame. See it clearly and you stop putting one particular request to an institution that has no way of answering it.

These four locate where a given system stands. Point them at the system, never at whoever is in the room with you.

• Is prevented illness counted here at all, or does the counting stop at illness that got treated?

• If I stay well for a decade and use fewer services, does anything come back to me, or to whoever would have treated me, or does the saving simply vanish?

• When somebody measures something in me, can they say why that marker and not a different one?

• Does prevention have a budget line of its own here, or is it bolted onto the money for treating people?

Inside most European systems the honest answers land in the same direction, and I didn't put a single one of these to anybody until my thirties. That's not embarrassing, it's just how it goes. The way a thing is financed stays invisible from inside until somebody goes looking.

What this changes for one person

Repairing this at system level is slow collective work. It would mean redefining success, rebuilding the measurement around that definition, and moving resources across whole institutions and more than one budget cycle. I'd like it to happen. I don't build plans on the assumption that it will.

An individual has a narrower move available, and it's the more useful one. Stop asking a treatment system to behave like a prevention system. Use it well, and with real gratitude, for acute and diagnosable problems, which it handles with genuine skill. Then build your prevention layer on purpose, elsewhere, rather than waiting for an institution to hand you one. That's the whole of how we build the week.

Sources

1. OECD (2023). Health at a Glance 2023: OECD Indicators. OECD Publishing, Paris. DOI: 10.1787/7a7afb35-en

2. Busse, R., Geissler, A., Quentin, W., and Wiley, M. (Eds.) (2011). Diagnosis-Related Groups in Europe: Moving Towards Transparency, Efficiency and Quality in Hospitals. European Observatory on Health Systems and Policies / Open University Press. ISBN 9780335245574

3. Porter, M. E., and Teisberg, E. O. (2006). Redefining Health Care: Creating Value-Based Competition on Results. Harvard Business School Press, Boston. ISBN 9781591397786

4. Berwick, D. M. (2002). A user's manual for the IOM's 'Quality Chasm' report. Health Affairs, 21(3), 80-90. DOI: 10.1377/hlthaff.21.3.80

5. McGlynn, E. A., et al. (2003). The quality of health care delivered to adults in the United States. New England Journal of Medicine, 348(26), 2635-2645. DOI: 10.1056/NEJMsa022615

6. McGinnis, J. M., Williams-Russo, P., and Knickman, J. R. (2002). The case for more active policy attention to health promotion. Health Affairs, 21(2), 78-93. DOI: 10.1377/hlthaff.21.2.78

7. Marmot, M. (2010). Fair Society, Healthy Lives: The Marmot Review. Institute of Health Equity, UCL.

8. OECD (2016). Better Ways to Pay for Health Care. OECD Health Policy Studies, OECD Publishing, Paris. DOI: 10.1787/9789264258211-en

Common questions

Does 3 percent mean health systems don't believe in prevention?

No, and that reading has it backwards. Nearly everyone inside these systems believes in prevention, and plenty of them know the evidence better than the people selling it.

The figure describes which activities the accounting machinery can recognise. An organisation does what it can define, code, pay for and audit. Belief doesn't survive contact with a payment system that has nowhere to put it.

Which system is better for prevention, the German or the Portuguese one?

Neither, in the sense that matters here, and I've been a patient in both. They differ enormously in cost, in waiting and in how you reach a specialist.

They don't differ in the direction the money runs. Payment attaches itself to diagnosed and treated illness in both. Choosing between them on this criterion is a wasted decision.

Does going private fix the problem?

It changes who sends the bill, which isn't nothing. It doesn't change the underlying property. Private provision still needs a billable event, and any business selling prevention hits the same difficulty proving that it worked.

That applies to us as much as to anyone. What I'd look for is whether the reasoning gets explained and the measurements get repeated, rather than whether the setting looks expensive.

What should I actually ask my doctor?

Ask the acute question, because that's what the appointment is built to answer. Don't walk in with a ten year question and leave irritated that nobody engaged with it. I did that four or five times and blamed the wrong people.

The direction question needs somewhere else to be asked, with time, repeated measurements and a chain of reasoning you're allowed to see.

Where Atlas Cove fits

The gap described here is why my sister and I started Atlas Cove. The week sits in the long ordinary stretch of a life where no institution has any reason to be present. It starts with somebody's own measurements and ends in decisions that come with their reasoning attached, and it replaces nothing, because medical care stays the right answer to genuine illness. I'm confident the gap is real. Whether seven days is the correct size of response to it, I'm considerably less sure, and I'd rather say so.