In 2000, Rwanda’s national household survey recorded that 196 of every 1,000 children born died before their fifth birthday. Roughly one child in five. The country was six years out of a genocide that had killed hundreds of thousands of people and wrecked the institutions a health system depends on, including the health workforce itself.
Fifteen years later, the Rwanda Demographic and Health Survey of 2014-15 put the figure at 50 per 1,000.
That is one of the steepest sustained declines in child mortality recorded anywhere. It is also one of the most frequently cited, which makes it worth looking at carefully rather than admiringly. What did Rwanda actually do, what does the evidence support, and what should other systems be wary of borrowing?
The outcomes
The headline numbers come from successive rounds of the Rwanda Demographic and Health Survey (DHS), the nationally representative household surveys run with support from the international DHS Program:
- Under-five mortality: 196 per 1,000 live births in 2000, 50 per 1,000 in 2014-15
- Maternal mortality ratio: 1,071 per 100,000 live births in 2000, 210 per 100,000 in 2014-15
- Share of births in a health facility: 27% in 2000, 91% in 2014-15
These are survey estimates with confidence intervals, and maternal mortality in particular is hard to measure precisely. But the direction and scale of change are consistent across indicators and surveys, which is why they are taken seriously.
What changed: three reforms pushed together
No single intervention explains Rwanda’s trajectory. What stands out is that the government pursued several reforms at the same time, each addressing a different failure in the system.
Paying facilities for what they deliver
In 2005, the Government of Rwanda formally adopted performance-based financing as a health financing strategy. Instead of funding health centres purely on inputs, such as staff and supplies, part of their funding was linked to the volume and quality of specific services: deliveries in facilities, prenatal care, child preventive visits and others.
What makes Rwanda’s case unusually valuable is that the national rollout was evaluated rigorously. A study by Paulin Basinga and colleagues, published in The Lancet in 2011, took advantage of a phased rollout in which 166 facilities were randomly assigned at district level either to begin performance-based payments in 2006 or to continue with traditional input-based funding for about two years.
The results were mixed in an instructive way. Performance pay was associated with a 23% increase in institutional deliveries, a 56% increase in preventive care visits among children aged 23 months and younger, and a 132% increase among children aged 24 to 59 months, along with an improvement in measured prenatal care quality. The authors found the biggest effects on services that carried the highest payments and required the least effort from providers. Payment design mattered, not just the fact of paying for performance.
Lowering the price of showing up
Rwanda also expanded community-based health insurance, known as mutuelles de santé, into a national scheme. Households paid annual premiums into a pooled fund that covered services at health centres and district hospitals, with co-payments at the point of care.
The logic is straightforward. Paying health centres to deliver more services does little if families cannot afford to use them. Insurance tackled the demand side while performance-based financing tackled the supply side.
Bringing care to the village
The third element was a network of community health workers elected by their own villages, trained to provide basic services such as treating common childhood illnesses, following up pregnant women and referring cases to health centres. They became the link between households and the formal system, and a channel for the government to reach remote communities.
Why it worked, and what to be careful about
It is tempting to package this into a formula: pay for performance, insure the population, deploy community health workers. Many countries have tried versions of each, with far less dramatic results. So what was different?
- Coordination. The reforms reinforced each other. Community workers generated referrals, insurance made those referrals affordable, and facilities had a financial reason to handle them well.
- Government ownership. Donor money was substantial, but the strategy was set and driven by the ministry. Programmes were aligned to national priorities rather than run as parallel projects.
- Measurement. Rwanda repeatedly measured its own progress through national surveys and built its payment system on verified data. Measurement was not an afterthought; it was the currency of the financing model.
There are also reasons for caution. Rwanda is a small, densely populated country with a strong, highly centralised state. Some of what made implementation effective, such as top-down accountability and tight performance targets for local officials, has attracted criticism for limiting local discretion and dissent. What works in that political context may not transfer, and may not be desirable, elsewhere.
Performance-based financing is also not a proven universal solution. Evaluations in other countries have produced weaker or inconsistent results, and critics point to high administrative costs for verification. Rwanda’s randomised evaluation showed real effects, but it also showed that payment incentives shift what is paid for and may do less for what is not.
A timeline of how the pieces fitted together
It helps to see the reforms in sequence rather than as a list, because the order mattered.
In the years immediately after 1994, the priority was rebuilding the basic machinery: health centres, district hospitals and a workforce that had been killed, displaced or scattered. Performance-based financing started as smaller pilots supported by international partners before the government adopted it as national strategy in 2005. The randomised phased rollout that Basinga and colleagues evaluated began in 2006, which is why Rwanda has unusually strong evidence about its own reform: the government built an evaluation into the expansion rather than bolting one on afterwards.
Community-based insurance grew over the same period from local schemes into a national system. The design then changed in an important way. From 2011, premiums were stratified using Ubudehe, Rwanda’s community-based classification of households by socioeconomic status. The poorest households had their premiums fully subsidised by the government, while better-off households paid more. A flat premium had been a real barrier for the poorest families; stratification was an attempt to make the insurance progressive rather than simply universal on paper.
The DHS rounds of 2005, 2010 and 2014-15 then tracked what happened as these reforms matured together. The steepest improvements came after the reforms were running at scale, which is consistent with the story that they worked, though, as we discuss below, it does not prove it on its own.
The counterfactual problem
The honest difficulty with any national success story is that we cannot see the Rwanda that did not reform. Many things changed at once between 2000 and 2015.
Rwanda was recovering from catastrophe, and some improvement would have happened simply because the extreme conditions of the 1990s ended. The economy grew for much of this period, raising household incomes, improving nutrition and making it easier to pay for care. International aid for health rose substantially across Africa during the 2000s, funding bed nets, vaccines and HIV treatment in many countries at once. Global progress on child survival was fast during this era in general.
So the right claim is not that Rwanda’s three reforms caused a three-quarter fall in child mortality. It is that Rwanda’s decline was unusually steep relative to countries facing similar conditions, that it coincided with a deliberate set of reforms, and that at least one of those reforms, performance-based financing, was shown by a randomised design to change the services it paid for. That is a strong case. It is not a controlled experiment on the whole system.
This matters because the most common mistake in borrowing from Rwanda is to treat its headline outcomes as the measured effect of a specific policy. The Basinga study measured changes in service use and quality over about two years. It did not measure the fall in child deaths. The link from more facility deliveries and more preventive visits to fewer deaths is plausible and well supported in general, but it runs through a chain of assumptions.
Why copies have disappointed
Performance-based financing spread widely across low- and middle-income countries in the decade after Rwanda’s evaluation, often with support from the World Bank. Results elsewhere have been more mixed, and it is worth asking why.
One reason is that the payment was often doing a different job. In Rwanda, it sat inside a system that had insurance, community outreach and strong government management. In many other settings, it was introduced as a largely standalone donor project. Paying facilities to deliver more services has limited effect if patients cannot afford to come, if drugs are out of stock, or if staff are not present to respond to the incentive.
A second reason is verification cost. Paying for performance requires someone to check that the reported services happened and met a quality standard. That checking is expensive and can crowd out the very management capacity it is meant to strengthen.
A third is the design lesson Basinga’s team found in Rwanda itself: incentives shift effort towards services that are well paid and easy to deliver. Programmes that did not think carefully about which services to pay for, and at what rate, could end up rewarding the simple and neglecting the difficult.
Second-order effects worth weighing
Several consequences of Rwanda’s model deserve attention from anyone considering a similar path.
Strong performance contracts for local officials helped drive implementation, but they also create pressure to report good numbers. When data is the currency of funding, the reliability of that data becomes a governance question, not just a technical one. Independent surveys such as the DHS are an important check precisely because they sit outside the payment system.
Community-based insurance improved access, but premiums and co-payments still represent a real cost for poor families just above the subsidy line. Stratification depends on classifying households accurately, and misclassification can push families into a premium band they cannot afford.
Community health workers took on a large and growing share of frontline work. Programmes that rely on community workers have to answer questions about their pay, training, supervision and workload, or the model can become a way to deliver care on the cheap.
What to watch
The next test for Rwanda’s model is not child mortality alone. As infectious causes of death decline, the burden shifts towards newborn complications and non-communicable diseases, which need more sophisticated facility care and longer-term management. Whether a system designed around counting and paying for specific services can adapt to chronic, complex care is the open question. For readers elsewhere, the more useful thing to watch is how Rwanda funds its system as donor support changes, because financing sustainability is where many of the countries that copied parts of the model have struggled.
The lesson for other systems
The most transferable lesson is probably not any single tool. It is that Rwanda treated demand, supply and outreach as one problem. Many health systems fix one part of the chain, say by building clinics, and then wonder why outcomes do not follow, because people cannot afford to use them, or the staff have no reason to deliver more, or nobody reaches the families who need care most.
Over to you
Have you worked with performance-based financing, community-based insurance or community health worker programmes? Did they reinforce each other in your setting, or pull in different directions? Tell us what you have seen in the comments, and whether you think the Rwandan model travels.
