Estimating the Global Burden of Endemic Canine Rabies

Katie Hampson, Laurent Coudeville, Tiziana Lembo, Maganga Sambo, Alexia Kieffer, Michaël Attlan, J Barrat, Jesse D. Blanton, Deborah J. Briggs, Sarah Cleaveland, Peter Costa, Conrad M. Freuling, Elly Hiby, Lea Knopf, Fernando Leanes, François-Xavier Meslin, Artem Metlin, Mary Elizabeth Miranda, Thomas Müller, Louis H. Nel, Sergio Recuenco, Charles E. Rupprecht, Carolin L. Schumacher, Louise Taylor, Marco Antonio Natal Vigilato, Jakob Zinsstag, Jonathan Dushoff, on behalf of the Global Alliance for Rabies Control Partners for Rabies Prevention•View original

OverviewBalancedalloy voice

If you’ve ever wondered how a disease can be both nearly always fatal and somehow invisible, rabies is the case study. In many low-income settings, once a patient shows symptoms, death is almost certain. Yet the deaths often don’t show up in the books. Official counts miss them, health systems don’t see them, and what isn’t counted rarely gets funded. That invisibility isn’t just a statistical quirk. It’s a policy failure, a budgeting problem, and a tragedy in slow motion. So a group led by Katie Hampson set out to do something deceptively hard: count the uncounted. They wanted the first coherent global and country-level picture of dog-mediated rabies—how many people die, how many healthy years of life are lost, and what it costs across health, veterinary, and household budgets. Their stance was practical. Treat the missing data head-on. Synthesize what we do know. And use a model that can bridge the gaps without pretending the uncertainty isn’t there. The core idea goes like this: start with human bites, then work forward to deaths and disability using a chain of probabilities. If you like having the equation in your head, think of expected deaths as bite incidence times the chance that the bite was from a rabid dog, times the chance that the person did not get post-exposure prophylaxis, times the chance that an unprotected exposure leads to rabies. In plain words, that’s exposure, danger, access to care, then disease. Two of those links are anchored with simple constants. The team fixed the chance of developing rabies without treatment at 0.19. And in settings where dogs aren’t vaccinated, the maximum probability a bite is from a rabid animal was set at 0.74. Everything else flexes with real-world conditions: as dog vaccination pushes canine rabies down, the chance of a bite being from a rabid dog falls; as health systems make post-exposure prophylaxis more reachable and affordable, access increases. Knobel and colleagues, building on an earlier decision-tree framework that Sarah Cleaveland pioneered in Tanzania, took that structure global. They pulled in country-specific and cluster-level data on bite incidence, dog vaccination coverage, dog populations, livestock exposure, human and animal rabies surveillance, and how post-exposure prophylaxis is actually delivered. For the trickiest piece—who actually gets post-exposure prophylaxis—they didn’t guess. They inferred access from the ratio of reported deaths to reported post-exposure prophylaxis doses, combined with the chance that a bite is from a rabid dog and the probability of developing rabies without treatment, inside a generalized linear mixed-effects model that included the Human Development Index. That choice matters. Access tracks development tightly—the Human Development Index was a strong predictor with a p-value well below 0.001—which explains why access looks so different in rural sub-Saharan Africa than in urban Southeast Asia. Uncertainty wasn’t an afterthought; it was baked into the engine. Bite incidence and dog vaccination coverage were allowed to vary within triangular distributions, with minimums and maximums set by a method that pooled expert judgment. Access came with bootstrap resampling from the mixed model, and the chance of a bite being from a rabid dog and the probability of developing rabies without treatment were perturbed with permutation-based resampling from their underlying binomial data. Then they ran it all 1,000 times—Monte Carlo realizations—with the same quantile draw applied globally each round to keep country and cluster parameters in sync. On the output side, they converted deaths to disability-adjusted life years, or DALYs, using the Global Burden of Disease 2010 life table for years of life lost, and a small years-lived-with-disability component for adverse events from older nerve-tissue vaccines. They even tested an anxiety component—imagine the documented psychological burden in the weeks after a dangerous dog bite—that would add on the order of half a million DALYs. But they left it out of the headline totals because they couldn’t validate it everywhere. What drops out of that machinery is sobering. Globally, about 59,000 people die from dog-mediated rabies each year. The uncertainty bands are wide—roughly 25,000 to 159,000—but the central story doesn’t change. Almost all of those deaths are in Africa and Asia—about 36 percent in Africa and roughly 60 percent in Asia. The Americas, by contrast, account for less than a tenth of a percent—on the order of 182 deaths, with uncertainty ranging from about 84 up to a few hundred. India alone contributes about 35 percent of global deaths. That’s not because India has the worst dogs or people are less careful. It’s due to exposure incidence plus access to timely post-exposure prophylaxis—two levers that differ enormously by place. If you translate those deaths into healthy years of life lost, you get a second, equally stark picture. About 3.7 million DALYs a year globally, with a credible range of about 1.6 to 10.4 million. Virtually all of it—more than 99 percent—comes from years of life lost to early death. The small remainder, around 30,000 DALYs, reflects adverse events related to older nerve-tissue rabies vaccines that a minority of countries used historically. That mix tells you something fundamental. Rabies is not a disease of prolonged disability. It’s a disease that kills young and middle-aged people quickly, and that’s where the burden sits. Now turn that lens to money. Using a human-capital approach for mortality—a way of valuing the productive years lost when someone dies—plus explicit spending on care, veterinary responses, and livestock losses, Hampson and colleagues put the global economic bill at about 8.6 billion US dollars a year. The uncertainty is big—roughly 2.9 to 21.5 billion—but, again, the hierarchy holds. The single largest component is productivity lost to premature death, a bit over half the total. Direct medical spending on post-exposure prophylaxis is on the order of 1.7 billion dollars. The indirect costs of time—patients and caregivers traveling, waiting, returning for doses—add another roughly 1.3 billion. Livestock losses run to about 512 million. And here’s a counterintuitive piece: the veterinary sector’s direct costs, mainly dog vaccination, are tiny by comparison, less than 1.5 percent of the global bill, something like 130 million dollars. Across society, more than 70 percent of the burden lands outside hospitals and clinics, in the lost futures of people who die and the hours families spend seeking care; around a fifth sits with the medical sector; a thin slice accrues to veterinary budgets and communities through dog vaccination and livestock losses. Laboratory surveillance, the thing that tells us whether control programs are working, barely registers financially—around one-hundredth of a percent of costs. If you zoom in by region, the patterns tell a story of uneven investment. In the Americas, average per-person spending on dog vaccination is about eleven cents a year, and that alone makes up close to a fifth of the region’s rabies-related costs. In most other endemic, low-income settings, it’s essentially zero—less than two cents per person—and the share of costs borne by vaccination programs is correspondingly tiny. Unit prices swing wildly, too. A dose of dog vaccine can cost several dollars in parts of West Africa, about fifty cents in Chad, and as low as twenty to thirty cents in Tanzania. Human post-exposure prophylaxis doses range from roughly eleven dollars up to about one hundred fifty, depending on the vaccine and regimen. Those aren’t abstract numbers; they’re the thresholds at which a health center can afford to keep stock, and a family can afford to finish a course. The team also mapped practice patterns that shape effectiveness. Intradermal regimens—which use less vaccine for the same protection—are widely used in the Philippines, Sri Lanka, and Thailand, and to a lesser degree in India. Rabies immunoglobulin, the antibody preparation given for high-risk exposures, is used more substantially in Eastern Europe, North Africa, and those same Asian countries. And while many places have moved away from older nerve-tissue vaccines, some countries reported using them into the 2010s; Bangladesh discontinued them in 2011. Under the hood, a few modeling choices carry a lot of weight. Dog rabies incidence is treated as a function of dog vaccination coverage, so as coverage rises, the share of bites that are actually from rabid animals falls. That link is what makes dog vaccination show up as the single most powerful lever on human deaths. When detailed country data are missing, inputs are pooled into regional clusters with similar epidemiology, and endemic clusters are analyzed separately from places that are effectively rabies-free. When the dust settles, the model’s sensitivity analysis points to three drivers of mortality variation: bite incidence first, then access to post-exposure prophylaxis, then the baseline chance of developing rabies without treatment. In plain English, it’s how often people are bitten, whether they can get to care in time, and how biologically risky an untreated bite is. Put all of that together and the policy picture crystallizes. One lever is upstream: vaccinate dogs. By cutting canine rabies transmission, you drive down the probability that any given bite is dangerous—and deaths fall. This isn’t speculative. As Hampson and Knobel show across countries and years, higher dog vaccination coverage aligns with lower rabies incidence and lower human deaths. The other lever is downstream: make post-exposure prophylaxis easy to get and cheap to complete. Access, the probability that a bite victim receives prophylaxis, climbs with development and with program design. The World Health Organization, or WHO, recommended intradermal schedules can slash vaccine volumes by more than half compared to intramuscular regimens—often a sixty percent or greater saving—which stretches budgets and stocks without sacrificing protection. When access rises, the product of bite incidence times the chance that a bite is from a rabid dog times one minus access times the probability of developing rabies without treatment collapses toward zero, and lives are saved. There is, of course, a catch. The same things that make rabies invisible—weak surveillance, rural health deserts, fragmented veterinary systems—also make the model’s estimates uncertain. That’s why the team spent so much care on uncertainty analysis, and why they call out surveillance as a control tool, not just an accounting nicety. Underreporting and patchy bite data force the use of cluster averages. Variable supply chains for post-exposure prophylaxis and immunoglobulin introduce rural-urban disparities that matter on the ground but are hard to resolve in national statistics. Better surveillance won’t just tighten the credible intervals; it will show whether investment is working. And that’s the payoff. We’re not stuck waiting for a breakthrough. The demonstrated levers are already on the table. Mass dog vaccination is affordable relative to the losses we’re absorbing, yet woefully underfunded where the deaths happen. Post-exposure prophylaxis can be made more accessible and more efficient with regimens we already know how to deliver. And the price of seeing clearly—of strengthening surveillance—barely registers compared to the cost of staying blind. As Hampson and colleagues argue, if we want rabies to stop being an invisible killer, we have to do the unglamorous math, spend where the math points, and keep counting as we go.

Official counts miss them, health systems don’t see them, and what isn’t counted rarely gets funded. That invisibility isn’t just a statistical quirk. It’s a policy failure, a budgeting problem, and a tragedy in slow motion.

So a group led by Katie Hampson set out to do something deceptively hard: count the uncounted. They wanted the first coherent global and country-level picture of dog-mediated rabies—how many people die, how many healthy years of life are lost, and what it costs across health, veterinary, and household budgets. Their stance was practical.

Treat the missing data head-on. Synthesize what we do know. And use a model that can bridge the gaps without pretending the uncertainty isn’t there.

The core idea goes like this: start with human bites, then work forward to deaths and disability using a chain of probabilities. If you like having the equation in your head, think of expected deaths as bite incidence times the chance that the bite was from a rabid dog, times the chance that the person did not get post-exposure prophylaxis, times the chance that an unprotected exposure leads to rabies. In plain words, that’s exposure, danger, access to care, then disease.

Two of those links are anchored with simple constants. The team fixed the chance of developing rabies without treatment at 0.19. And in settings where dogs aren’t vaccinated, the maximum probability a bite is from a rabid animal was set at 0.74.

Everything else flexes with real-world conditions: as dog vaccination pushes canine rabies down, the chance of a bite being from a rabid dog falls; as health systems make post-exposure prophylaxis more reachable and affordable, access increases.

Knobel and colleagues, building on an earlier decision-tree framework that Sarah Cleaveland pioneered in Tanzania, took that structure global. They pulled in country-specific and cluster-level data on bite incidence, dog vaccination coverage, dog populations, livestock exposure, human and animal rabies surveillance, and how post-exposure prophylaxis is actually delivered. For the trickiest piece—who actually gets post-exposure prophylaxis—they didn’t guess.

They inferred access from the ratio of reported deaths to reported post-exposure prophylaxis doses, combined with the chance that a bite is from a rabid dog and the probability of developing rabies without treatment, inside a generalized linear mixed-effects model that included the Human Development Index. That choice matters. Access tracks development tightly—the Human Development Index was a strong predictor with a p-value well below 0.001—which explains why access looks so different in rural sub-Saharan Africa than in urban Southeast Asia.

Uncertainty wasn’t an afterthought; it was baked into the engine. Bite incidence and dog vaccination coverage were allowed to vary within triangular distributions, with minimums and maximums set by a method that pooled expert judgment. Access came with bootstrap resampling from the mixed model, and the chance of a bite being from a rabid dog and the probability of developing rabies without treatment were perturbed with permutation-based resampling from their underlying binomial data.

Then they ran it all 1,000 times—Monte Carlo realizations—with the same quantile draw applied globally each round to keep country and cluster parameters in sync. On the output side, they converted deaths to disability-adjusted life years, or DALYs, using the Global Burden of Disease 2010 life table for years of life lost, and a small years-lived-with-disability component for adverse events from older nerve-tissue vaccines. They even tested an anxiety component—imagine the documented psychological burden in the weeks after a dangerous dog bite—that would add on the order of half a million DALYs.

But they left it out of the headline totals because they couldn’t validate it everywhere.

What drops out of that machinery is sobering. Globally, about 59,000 people die from dog-mediated rabies each year. The uncertainty bands are wide—roughly 25,000 to 159,000—but the central story doesn’t change.

Almost all of those deaths are in Africa and Asia—about 36 percent in Africa and roughly 60 percent in Asia. The Americas, by contrast, account for less than a tenth of a percent—on the order of 182 deaths, with uncertainty ranging from about 84 up to a few hundred. India alone contributes about 35 percent of global deaths.

That’s not because India has the worst dogs or people are less careful. It’s due to exposure incidence plus access to timely post-exposure prophylaxis—two levers that differ enormously by place.

If you translate those deaths into healthy years of life lost, you get a second, equally stark picture. About 3.7 million DALYs a year globally, with a credible range of about 1.6 to 10.4 million. Virtually all of it—more than 99 percent—comes from years of life lost to early death.

The small remainder, around 30,000 DALYs, reflects adverse events related to older nerve-tissue rabies vaccines that a minority of countries used historically. That mix tells you something fundamental. Rabies is not a disease of prolonged disability.

It’s a disease that kills young and middle-aged people quickly, and that’s where the burden sits.

Now turn that lens to money. Using a human-capital approach for mortality—a way of valuing the productive years lost when someone dies—plus explicit spending on care, veterinary responses, and livestock losses, Hampson and colleagues put the global economic bill at about 8.6 billion US dollars a year. The uncertainty is big—roughly 2.9 to 21.5 billion—but, again, the hierarchy holds.

The single largest component is productivity lost to premature death, a bit over half the total. Direct medical spending on post-exposure prophylaxis is on the order of 1.7 billion dollars. The indirect costs of time—patients and caregivers traveling, waiting, returning for doses—add another roughly 1.3 billion.

Livestock losses run to about 512 million. And here’s a counterintuitive piece: the veterinary sector’s direct costs, mainly dog vaccination, are tiny by comparison, less than 1.5 percent of the global bill, something like 130 million dollars. Across society, more than 70 percent of the burden lands outside hospitals and clinics, in the lost futures of people who die and the hours families spend seeking care; around a fifth sits with the medical sector; a thin slice accrues to veterinary budgets and communities through dog vaccination and livestock losses.

Laboratory surveillance, the thing that tells us whether control programs are working, barely registers financially—around one-hundredth of a percent of costs.

If you zoom in by region, the patterns tell a story of uneven investment. In the Americas, average per-person spending on dog vaccination is about eleven cents a year, and that alone makes up close to a fifth of the region’s rabies-related costs. In most other endemic, low-income settings, it’s essentially zero—less than two cents per person—and the share of costs borne by vaccination programs is correspondingly tiny.

Unit prices swing wildly, too. A dose of dog vaccine can cost several dollars in parts of West Africa, about fifty cents in Chad, and as low as twenty to thirty cents in Tanzania. Human post-exposure prophylaxis doses range from roughly eleven dollars up to about one hundred fifty, depending on the vaccine and regimen.

Those aren’t abstract numbers; they’re the thresholds at which a health center can afford to keep stock, and a family can afford to finish a course.

The team also mapped practice patterns that shape effectiveness. Intradermal regimens—which use less vaccine for the same protection—are widely used in the Philippines, Sri Lanka, and Thailand, and to a lesser degree in India. Rabies immunoglobulin, the antibody preparation given for high-risk exposures, is used more substantially in Eastern Europe, North Africa, and those same Asian countries.

And while many places have moved away from older nerve-tissue vaccines, some countries reported using them into the 2010s; Bangladesh discontinued them in 2011.

Under the hood, a few modeling choices carry a lot of weight. Dog rabies incidence is treated as a function of dog vaccination coverage, so as coverage rises, the share of bites that are actually from rabid animals falls. That link is what makes dog vaccination show up as the single most powerful lever on human deaths.

When detailed country data are missing, inputs are pooled into regional clusters with similar epidemiology, and endemic clusters are analyzed separately from places that are effectively rabies-free. When the dust settles, the model’s sensitivity analysis points to three drivers of mortality variation: bite incidence first, then access to post-exposure prophylaxis, then the baseline chance of developing rabies without treatment. In plain English, it’s how often people are bitten, whether they can get to care in time, and how biologically risky an untreated bite is.

Put all of that together and the policy picture crystallizes. One lever is upstream: vaccinate dogs. By cutting canine rabies transmission, you drive down the probability that any given bite is dangerous—and deaths fall.

This isn’t speculative. As Hampson and Knobel show across countries and years, higher dog vaccination coverage aligns with lower rabies incidence and lower human deaths. The other lever is downstream: make post-exposure prophylaxis easy to get and cheap to complete.

Access, the probability that a bite victim receives prophylaxis, climbs with development and with program design. The World Health Organization, or WHO, recommended intradermal schedules can slash vaccine volumes by more than half compared to intramuscular regimens—often a sixty percent or greater saving—which stretches budgets and stocks without sacrificing protection. When access rises, the product of bite incidence times the chance that a bite is from a rabid dog times one minus access times the probability of developing rabies without treatment collapses toward zero, and lives are saved.

There is, of course, a catch. The same things that make rabies invisible—weak surveillance, rural health deserts, fragmented veterinary systems—also make the model’s estimates uncertain. That’s why the team spent so much care on uncertainty analysis, and why they call out surveillance as a control tool, not just an accounting nicety.

Underreporting and patchy bite data force the use of cluster averages. Variable supply chains for post-exposure prophylaxis and immunoglobulin introduce rural-urban disparities that matter on the ground but are hard to resolve in national statistics. Better surveillance won’t just tighten the credible intervals; it will show whether investment is working.

And that’s the payoff. We’re not stuck waiting for a breakthrough. The demonstrated levers are already on the table.

Mass dog vaccination is affordable relative to the losses we’re absorbing, yet woefully underfunded where the deaths happen. Post-exposure prophylaxis can be made more accessible and more efficient with regimens we already know how to deliver. And the price of seeing clearly—of strengthening surveillance—barely registers compared to the cost of staying blind.

As Hampson and colleagues argue, if we want rabies to stop being an invisible killer, we have to do the unglamorous math, spend where the math points, and keep counting as we go.

Related lectures