Last week ended on a clean idea: the goal is value, the health outcomes that matter to a patient divided by the cost of achieving them. Easy to say, but here is the catch nobody enjoys.
A system pays for what it rewards, and it gets more of whatever it pays for. You cannot stand on a stage asking for value while your finance department is still paying, line by line, for activity. The payment model is not an administrative detail underneath the reform. It is the reform, and everything else is decoration.
So this piece is about the money mechanics: the handful of ways you can actually pay for care, what each one quietly encourages, and why the popular fixes mostly fail.
Four ways to pay, and what each one buys
Strip away the jargon and there are essentially four ways to pay a provider, which the OECD lays out as a spectrum. The further down you go, the more the reward shifts from "did you do things" to "did the patient end up well, affordably."
- Fee-for-service. Pay per consultation, scan, procedure. Simple, transparent, and the root of the problem: it pays for volume, so volume is what you get. Do the hip replacement, get paid. If it fails and you redo it, get paid again.
- Pay-for-performance. Fee-for-service, plus a bonus for hitting quality targets. The provider carries almost no real risk; the bonus just sits on top.
- Bundled, or episode, payments. One price for the whole episode of a condition, from operation to recovery. Now the redo comes out of your margin, so suddenly you care about getting it right the first time.
- Population-based payments. A provider group is accountable for the outcomes of a whole population against a budget, and keeps a share of what it saves if quality holds up. This is the shared-savings logic behind Kinzigtal from Part 2.
Most European systems live almost entirely at the top of that list, then bolt on small experiments lower down and call it transformation.
Why the easy fix usually fails
When systems decide to "pay for quality," they almost always reach for pay-for-performance, because it is the least disruptive. It is also the one with the thinnest evidence.
Despite being adopted in more than a dozen countries, there is remarkably little proof that paying providers bonuses for hitting targets has improved care. The clearest case is Britain's Quality and Outcomes Framework, the largest primary-care scheme of its kind in the world, which at its peak tied more than a quarter of a GP's income to performance measures. After two decades, the improvements have been modest and their cost-effectiveness is genuinely doubtful; by one careful estimate the programme sits well outside the range the NHS would normally consider value for money.
Worse, paying for metrics changes behaviour in ways you did not intend:
- It invites gaming, hitting the number without improving the underlying care.
- It causes tunnel vision, as attention drifts toward what is measured and away from everything that is not.
- It can crowd out the intrinsic motivation that made good professionals good in the first place, and there is evidence it eroded continuity of care, the very thing this series argues matters most.
The lesson echoes Part 2 and Part 5. Bolting a bonus onto fee-for-service does not change the logic of fee-for-service. It just adds a target to chase.
Where bundles work, and where they break
Move one step down and bundled payments have a better track record, with a sharp limit.
The most studied example is the American mandatory bundle for hip and knee replacements. It produced real but modest savings, on the order of a percent or so per episode, with no measurable loss of quality. A success, of a kind. But look at how the savings happened: mostly by cutting back on post-operative care in rehabilitation and nursing facilities. And much of what the system saved was handed straight back to providers as bonuses, so the net gain to the payer was slim. Only about half the hospitals managed to save anything at all.
More importantly, notice why it worked where it did. A joint replacement is the ideal candidate for a bundle: a discrete, predictable procedure on a mostly healthy patient, with a clear start, a clear end, and an easy outcome to measure.
The patient this whole series is about, the multimorbid one whose care has no clean start or end, is exactly the patient a bundle cannot hold.
And there is a warning in the data. Push the incentive to cut too hard and you do not save money, you defer it. Some bundled models showed more readmissions and worse outcomes for the frailest patients. An incentive to spend less is also an incentive to do less, and those are not always the same thing.
The model that actually fits continuous care
Which leaves the bottom of the spectrum, and it is no accident that it is where Kinzigtal lived.
Population-based, shared-savings payment is the only model whose logic matches a resolved case across a whole person rather than a single procedure. You make a provider group accountable for the outcomes of a defined population, give them a budget, and let them keep a share of what they save, but only if measured outcomes hold. Get a patient genuinely well and keep them out of the expensive ward, and everyone wins.
But this model carries the two failure modes you must design against from day one:
- Skimping. If you are paid to spend less, you might simply under-treat. The guard is the one from Part 5: gate every euro of shared savings on real, patient-reported outcomes. No outcomes, no bonus.
- Skimming. If you are paid per healthy population, you might quietly avoid the sick. The guard is the one from Part 4: allocate by clinical need, with the public system deciding who is in, so the complex patients cannot be filtered out.
Pay for the population and the outcome, measure the outcome honestly, and refuse to let anyone choose their patients. That combination is hard to assemble and rare to find, which is precisely why it works when it exists.
The uncomfortable conclusion
There is no magic payment model. Each one moves a single lever and carries a single, predictable failure. Fee-for-service buys volume. Pay-for-performance buys gaming. Bundles buy efficient procedures and struggle with complex lives. Population payment buys aligned incentives and tempts you toward skimping and skimming.
So payment reform is necessary and nowhere near sufficient. It only produces value when it is welded to the two things the earlier pieces described: honest measurement of outcomes that matter, and a clear owner accountable for the case. Change the cheque without changing those, and you have just bought a more sophisticated way to optimise the wrong thing.
If you are going to take on the outcomes of a whole population, though, the first question is brutally practical: which population, and where does the cost actually sit? It turns out to sit in a startlingly small group of people.
Next week: the economics of the few.
Sources
- OECD (2016), Better Ways to Pay for Health Care, OECD Health Policy Studies, OECD Publishing, Paris. doi.org/10.1787/9789264258211-en. On the payment-method spectrum and shared-savings design.
- National Academies of Sciences, Engineering, and Medicine (2021), Strengthening Primary Care Delivery through Payment Reform; and Mendelson A. et al. and related reviews on pay-for-performance. On the weak evidence base for P4P.
- On the UK Quality and Outcomes Framework: "Modelling the cost-effectiveness of pay-for-performance in primary care in the UK" (2018), and "Is the evidence on the effectiveness of pay for performance schemes in healthcare changing? A meta-regression analysis" (2021). On modest, costly, and inconsistent effects, gaming and crowding-out.
- On bundled payments: "Evaluation of Economic and Clinical Outcomes Under CMS Mandatory Bundled Payments for Joint Replacements" (2019), and the Healthcare Value Hub evidence summary on the Comprehensive Care for Joint Replacement (CJR) model. On modest, post-acute-driven savings and uneven results.
- Hildebrandt H. et al. (2010), "Gesundes Kinzigtal Integrated Care," International Journal of Integrated Care. On the population-based shared-savings model.
Part 6 of the series. Part 5 asked what we are buying. Part 7 covers the economics of the few: where cost and value concentrate.
Published by CW1. Malm is CW1's continuous-care programme for patients: malm.care. Nortb is CW1's operations partner for clinics and practices: nortb.com.