Every health ministry has a drawer full of successful pilots. A district where maternal deaths fell. A province where enrollment in a new insurance scheme doubled in eighteen months. A slide deck with a before-and-after chart that made a donor’s quarter.
And then the national rollout arrives, and the chart flattens.
This is one of the most predictable patterns in global health, and one of the least discussed honestly. The WHO and World Bank’s 2023 monitoring report on universal health coverage found that roughly half the world’s population is still not fully covered by essential health services, and that progress on service coverage has largely stalled since 2015. That is not a shortage of pilots. It is a shortage of pilots that survived contact with a whole country.
So why does the thing that worked in one district stop working in two hundred?
The usual answers are money and political will. Both matter, but they are downstream. The deeper problem is that pilots are built to answer a question nobody at scale is asking, and the incentives around them reward the answer rather than the question.
A pilot is not a small version of a national program
The uncomfortable truth is that a pilot and a national program are different species that happen to share a name.
A pilot is chosen. Districts volunteer, or are selected because they have a capable health officer, a functioning hospital and a road that stays open in the rainy season. The team running it is small, motivated and usually paid from a separate budget line. Problems get escalated by phone to someone who can fix them the same week. A consultant visits every month.
A national program is inherited. It runs through every district, including the ones with no health officer, no data system and a hospital whose generator has been broken since the last election. It runs through the civil service as it actually exists, with its salary scales, its transfer rules and its procurement delays. Nobody is visiting monthly.
When the pilot “proves” something, what it often proves is that the intervention works under unusually good management. That is useful to know. It is not the same as proving the intervention works.
Why everyone prefers the pilot
It is worth being blunt about incentives, because nobody in this story is acting irrationally.
A funder needs results inside a grant cycle, usually three to five years. A pilot can deliver a clean before-and-after within that window; a national purchasing reform cannot. A minister needs something to open, announce and photograph. A consultancy is paid to design and evaluate, not to sit in a claims office for a decade. An evaluator is rewarded for a publishable effect size, and the surest way to get one is to run the intervention where it is most likely to work.
Each of these choices is defensible alone. Together they produce a global pipeline that is very good at generating proof of concept and structurally bad at generating proof of scale. The pilot becomes the product. Success is declared at exactly the moment the hard part begins.
This also explains why the same failure repeats across countries that share little else. It is not a local capacity problem. It is a market design problem in how development and domestic health money is allocated and how careers are made.
The money changes shape, not just size
Pilots tend to be funded like projects: a fixed grant, a start date and an end date. National coverage has to be funded like a promise: indefinitely, from domestic revenue, in competition with roads, schools and debt service.
That shift exposes three problems pilots never had to face.
First, per-person costs rarely stay flat. The easiest people to reach are reached first. Extending coverage to remote villages, informal workers and people who have never interacted with a formal health system costs more per person, not less.
Second, benefits get promised before the revenue exists. A pilot can offer a generous benefit package because the pilot population is small. Scaled nationally, the same package becomes a fiscal liability, so it gets quietly rationed through waiting lists, stock-outs and informal payments. Coverage exists on paper and not at the clinic.
Third, the payment system that worked for a few hundred facilities breaks at a few thousand. Claims pile up. Hospitals go months without reimbursement. They respond by turning away scheme members or asking them to pay “just for supplies.” Trust erodes, enrollment falls, and the scheme gets a reputation it never recovers from.
A back-of-envelope scale-up
Consider an illustrative scheme, with round numbers chosen to show the shape of the problem rather than to describe any real country.
The pilot covers 200,000 people in two well-run districts at a cost of $20 per person per year, so $4 million, mostly donor-funded. Staff time from the ministry and a technical assistance team is on a separate budget, so it does not appear in that $20. Say it adds another $5 per head. The true pilot cost is $25.
Now scale to a country of 30 million. The next ten million people are similar to the pilot population. The following ten million live further from facilities, work informally and have never been in a formal system; reaching and enrolling them might cost half as much again. The last ten million, in the places the pilot never touched, might cost double. Averaged, the national figure sits well above $25, not at $20, before a single extra drug is bought.
Then utilisation rises, because coverage is supposed to make people use care they were avoiding. A scheme that succeeds on its own terms gets more expensive. Multiply through and the programme that cost $4 million in the pilot needs a sum in the high hundreds of millions nationally, every year, from a finance ministry that was shown the $20 slide.
None of these numbers is exotic. The mistake is presenting the pilot’s average cost as the national marginal cost. The gap between those two figures is where most stalled schemes quietly die.
Political time and program time don’t match
A pilot usually fits inside one minister’s term. It has a launch, a result and a photograph. National coverage takes ten to twenty years of steady, unglamorous work: building a purchasing agency, training claims assessors, fixing drug supply, renegotiating provider payments.
Ministers rotate. Priorities rotate with them. The official who championed the scheme moves on, and the successor has every incentive to launch something new with their name on it rather than finish someone else’s program. Pilots get stranded, not because they failed, but because they were orphaned.
There is a second-order effect here that gets less attention. Every orphaned pilot teaches frontline staff a lesson: new programmes come and go, so do not reorganise your work around them. By the fourth or fifth initiative, district managers have learned to comply on paper and wait. That learned scepticism is itself a cost, and it makes the next genuinely good reform harder to land.
The countries that got through the gap
The countries often cited as UHC successes did not simply scale a pilot. They changed the system around it.
Thailand is the classic example. When it launched its Universal Coverage Scheme in the early 2000s, it did not bolt a new program onto a weak system. It had spent the previous decades building district hospitals and a rural health workforce, including mandatory rural service for new doctors. When the financing reform arrived, there were facilities and staff for the money to flow to. It also created a dedicated purchasing body, the National Health Security Office, and paid facilities mainly through capitation for outpatient care, which kept costs predictable.
Rwanda took a different route. It expanded community-based health insurance nationally while simultaneously investing in community health workers and performance-based financing for facilities. Coverage and service quality were pushed together, not in sequence.
Neither model is copy-paste. But both share a feature most stalled programs lack: they treated the rollout itself as the main event, not as an afterthought to a successful pilot.
The strongest case for pilots
The counterargument deserves a fair hearing. Pilots exist for good reasons. They let a country find out, cheaply, that an idea is bad before spending a national budget on it. They build a cadre of people who understand the model. They generate political proof that something is possible, which is often what unlocks domestic money in the first place.
That case partly holds. A pilot that fails is valuable, and a pilot that shows an intervention cannot work even under ideal conditions saves a great deal of money. The problem is asymmetry: pilots are informative when they fail and much less informative when they succeed. A positive result under hand-picked conditions tells you the ceiling, not the floor. Most ministries treat it as the floor.
So the answer is not fewer pilots. It is pilots designed to be informative when they succeed, which means designing them to meet the conditions the national system will actually face.
What a pilot should actually be designed to test
If pilots keep producing results that do not scale, the fix is partly in how pilots are designed. A few questions that should be asked before a pilot starts, not after:
- Is the pilot site typical? If it was chosen because it is well run, the results tell you about good management, not about the intervention.
- Is it running through the real system? Parallel staff, parallel budgets and parallel data systems make pilots look good and make scale-up impossible.
- What does it cost at the margin? Not the average cost in the pilot, but the cost of reaching the next, harder-to-reach person.
- Who owns it when the grant ends? If the answer is “the ministry, eventually,” the answer is nobody.
- What would failure look like at scale? Pilots should test the weak points of national rollout: claims processing, drug supply, provider payment delays. Not only the clinical intervention.
What each actor can change
For ministries of health, the practical shift is to put the scale-up owner in the room from day one. That usually means the purchasing agency, the finance ministry’s budget office and the civil service commission, not only the programme team. If they cannot see how the pilot maps onto their systems, it will not scale.
For finance ministries, it means asking for marginal cost curves rather than average costs, and for a named line in the medium-term budget framework before a pilot is expanded. A programme without a line is a programme on borrowed time.
For donors and funders, the hardest change is timelines. Funding the dull middle, claims systems, drug supply chains, provider payment reform, over ten years rather than three is less photogenic. It is also where the returns are. Funders can also require that pilots run through government systems and staff even when that makes early results look worse.
For hospital leaders, the lesson is about cash flow. Facilities that join a scheme are extending credit to the government. Knowing how long reimbursement takes, and negotiating on that before signing up, protects both the hospital and the patients it would otherwise turn away.
For evaluators and journals, it means valuing implementation research, including reports of rollouts that stalled, as highly as effect sizes from well-controlled pilots. We learn more from the district where it broke than from the district where it shone.
What to watch
Over the next year or two, the useful signals are unglamorous. Watch whether countries expanding insurance schemes publish claims processing times and reimbursement delays, not only enrolment figures. Watch how much of external health funding moves through national systems rather than parallel ones, especially as several major donors have cut or reorganised aid budgets in 2025, which will force some pilots to find domestic owners faster than planned. And watch whether the benefit packages being promised match the budget lines being passed. Where they diverge, expect coverage on paper and queues in practice.
The real question
Universal health coverage does not stall because countries lack evidence that it can work. It stalls because the evidence usually comes from conditions that will never exist nationally, and because the hard parts of the job, financing, purchasing, payment systems and the civil service, are rarely what pilots are built to test.
The ministries making progress are the ones that stopped asking “did the pilot work?” and started asking “what will break when this reaches the district nobody wants to visit?”
That is a less exciting question. It is also the only one that gets you to coverage.
What’s the biggest gap you’ve seen between a health pilot and its national rollout? We’d like to hear it.
