Global Health

The big picture

Start with a puzzle. A child in Sokoto, Nigeria and a child in Sheffield, England both get pneumonia this winter. Both have the same bacterium in the same lung. One will almost certainly live; one has perhaps a one in twenty chance of dying. Nothing about the biology explains that gap. Everything about oxygen supply, referral distance, antibiotic stock, household income, and whether anyone counted the death does. Global health is the field that takes that gap as its central object.

That sounds simple, and it is not, because the field arrived carrying baggage. It has an ancestor called tropical medicine that was built to keep colonial administrators and plantation labor alive. It has a second ancestor called international health that treated disease as something that crosses borders and therefore needs treaties. The current name, global health, is barely thirty years old, and part of what it was meant to signal was a break with both. Whether the break actually happened is one of the live arguments in the field, and you will meet it repeatedly in this course.

This lesson defines the field beside its two predecessors, tells the colonial-medicine history honestly because that history explains why today's institutions and funding flows look as they do, and introduces the fight over what the word global is supposed to mean, which turns out to be a fight about who counts as a subject of global health and who is only ever an object of it.

Key idea: Global health studies health problems and inequities that cross borders, and it is best understood as a deliberate attempt to move beyond two earlier fields, tropical medicine and international health, that it has not entirely escaped.

Three fields that sound alike

Students conflate these three constantly, and the conflation matters because each carries a different assumption about who has knowledge and who has need. Here is the cleanest way to separate them.

Field	Era of dominance	Central question	Direction of authority
Tropical medicine	1890s to 1940s	What diseases afflict people in hot climates, and how do we keep our people functional there?	Metropole to colony; expertise assumed to sit in London, Liverpool, Hamburg, Paris
International health	1940s to 1990s	How do sovereign states cooperate to stop disease crossing their borders, and how does the rich world assist the poor one?	State to state, and donor to recipient
Global health	1990s to present	What produces health inequity anywhere, and what transnational action reduces it?	Aspirationally multidirectional; in practice still heavily donor to recipient

Look at the last column. Tropical medicine organized disease by geography and, implicitly, by race, treating the tropics as a place with a pathology rather than as places where people happen to live. International health organized it by sovereignty, which is why its products are treaties, agencies, and bilateral aid. Global health claims to organize it by the problem itself, wherever it is, which is why its advocates insist a course like this should also cover overdose deaths in Ohio, not only child mortality in Chad.

The most cited definition came from a 2009 Lancet paper by Jeffrey Koplan and colleagues: an area for study, research, and practice that places a priority on improving health and achieving equity in health for all people worldwide, emphasizing transnational issues and collaboration across many disciplines. Notice the word equity doing the heavy lifting. Without it, global health collapses back into international health with a nicer name.

Key idea: Tropical medicine sorted disease by climate and empire, international health sorted it by sovereign states, and global health claims to sort it by health inequity itself, which is why equity is the load-bearing word in every serious definition.

The colonial ancestor, told straight

Tropical medicine was founded as an academic discipline in the last years of the nineteenth century, and it was founded on purpose, for reasons that were stated out loud at the time. The London School of Tropical Medicine opened in 1899 under Patrick Manson, who was simultaneously medical adviser to the Colonial Office. The Liverpool School of Tropical Medicine opened a year earlier in 1898, funded substantially by Alfred Lewis Jones, a shipping magnate whose fleet ran the West African trade. Joseph Chamberlain, the Colonial Secretary, argued explicitly that the study of tropical disease was a precondition for the development of the empire's tropical possessions. This is not a hostile reconstruction. It is the founders' own case for funding.

The science was real and often excellent. Ronald Ross demonstrated in 1897, working in Secunderabad in British India, that mosquitoes transmit the malaria parasite, and won the Nobel Prize for it in 1902. That discovery underlies every bed net and larval control program in Module 3. It is also worth noticing who else was in the room: Ross's assistant Kishori Mohan Bandyopadhyay did much of the painstaking dissection work and received a gold medal at Ross's request, and his name appears in almost no textbook. Multiply that pattern across a century and you have one of the field's structural problems in miniature.

What did colonial medicine do with its knowledge? Sometimes it saved lives at scale. Sometimes it did something else. The French and Belgian sleeping sickness campaigns in Central Africa from the 1910s onward assembled whole populations for compulsory examination and injection with arsenical drugs such as atoxyl that could and did cause blindness. In 1954, at Yokadouma in southeastern Cameroon, a mass preventive campaign with the drug Lomidine used contaminated equipment and produced hundreds of abscesses and dozens of deaths in people who had no disease at all. The historian Guillaume Lachenal's point in reconstructing it is not that colonial doctors were monsters, but that an apparatus built to manage populations rather than serve patients will eventually treat people as a throughput problem.

Key idea: Tropical medicine produced genuine and lasting science, including the discovery of mosquito transmission of malaria, inside an institutional project whose stated purpose was the economic and military viability of empire, and whose mass campaigns sometimes harmed the populations they enrolled.

How international health was assembled

Alongside the colonial schools ran a second, older track: states negotiating with each other about disease. It began with quarantine disputes. The first International Sanitary Conference met in Paris in 1851 to settle a practical argument, namely how long to hold ships suspected of carrying cholera or plague, since every extra day of quarantine cost shipping money and every day too few risked an epidemic. Fourteen more conferences followed, and out of them came the first permanent bodies: the Pan American Sanitary Bureau in 1902, which survives as the Pan American Health Organization and is the oldest international health agency in the world, the Office International d'Hygiene Publique in 1907, and the League of Nations Health Organisation in 1923.

The World Health Organization was created in 1948 as a specialized agency of the new United Nations, absorbing those predecessors. Its constitution opens with a claim that was radical then and is contested now: that the enjoyment of the highest attainable standard of health is a fundamental right of every human being. Its structure matters for everything that follows. It is an organization of member states, currently 194 of them, each with one vote in the World Health Assembly, with six unusually autonomous regional offices whose directors are elected by the countries of that region. It can set standards and declare emergencies, but it cannot compel a government to do anything.

The high-water mark of the international health era came in 1978 at Alma-Ata, in what is now Kazakhstan, where 134 countries signed a declaration committing to primary health care and to the slogan Health for All by the Year 2000. The declaration was expansive: it named community participation, intersectoral action, appropriate technology, and the New International Economic Order as health matters. Within a year it drew a famous counterproposal. Julia Walsh and Kenneth Warren argued in 1979 for selective primary health care, a short list of high-yield interventions that could actually be financed and delivered, later popularized by UNICEF as GOBI: growth monitoring, oral rehydration, breastfeeding, and immunization.

Hold both sides of that argument. A program delivering four interventions leaves a country without a health system; a plan requiring the reorganization of the world economy is not something a health minister can execute next year. Nearly every debate in this course, about vertical programs, targets, and what donors will fund, descends from Alma-Ata versus GOBI.

Key idea: International health produced the treaty and agency architecture we still use, from PAHO in 1902 to WHO in 1948, and its central unresolved argument, comprehensive primary health care at Alma-Ata in 1978 versus selective high-yield interventions, still structures global health debate today.

Why the name changed in the 1990s

Three things happened at once. HIV made it obvious that a disease could not be managed as a national problem, and that the countries with the least money had the most disease. The World Bank's 1993 World Development Report, Investing in Health, introduced the disability-adjusted life year and made cost-effectiveness the organizing logic of health spending. And the end of the Cold War dissolved the framework in which health aid had been an instrument of bloc competition.

New actors poured in faster than anyone could coordinate them: the Gates Foundation from 1999, Gavi in 2000, the Global Fund in 2002, PEPFAR in 2003. You will map that landscape next lesson. The word global was chosen to signal three claims: that health problems are transnational, so climate, trade and supply chains are health topics; that non-state actors are central rather than peripheral; and that the field applies everywhere, so Baltimore and Bamako are both proper objects of study. That third claim is the one most often broken. Ask yourself how many global health departments in wealthy universities work on health inequity in their own city.

Key idea: Global health emerged in the 1990s from the collision of the HIV pandemic, the World Bank's cost-effectiveness turn, and the arrival of large private and multilateral funders, and the word global was meant to signal transnational causes, non-state actors, and relevance everywhere.

The definitional fight, which is really a fight about power

Here is a question worth sitting with before you read on. If a Ghanaian epidemiologist studies diabetes in Kumasi, is that global health, or is it just Ghanaian public health? And if an American epidemiologist studies diabetes in Kumasi, does the label change?

Most people's honest answer reveals the problem. In common usage, global health tends to mean health work done in low- and middle-income countries by people or institutions from high-income ones. The Nigerian physician and researcher Seye Abimbola calls this the foreign gaze: the field's default reader is assumed to be an outsider, so its writing explains local context to distant audiences and treats the local practitioner as a source of data rather than as the audience. Madhukar Pai and Abimbola put the question sharply in a 2020 Lancet comment, asking whether global health can survive its own decolonization, since a field defined by the outsider's position may not have a coherent identity once that position is given up.

There are serious counterarguments and you should not skip them. Transnational problems really do need transnational institutions, and dissolving the field into national public health would remove the only forum where a pandemic treaty or a pooled vaccine purchase can be negotiated. Decolonization talk can also become a branding exercise that changes conference panels while leaving the money exactly where it was. And many of the field's most effective institutions already sit in low- and middle-income countries, which suggests the sharper question is not what to call the field but who controls its budgets.

This course takes a position. Global health has a genuine record of saving lives, and its money, agenda-setting and prestige remain concentrated in a few high-income institutions in ways that demonstrably distort what gets studied. Both are true. A course that tells you only the first is a recruitment brochure; one that tells you only the second cannot explain why child mortality halved.

Key idea: In practice global health often means work done in poorer countries by institutions from richer ones, which is why the field's definition is contested; the honest position holds its real achievements and its real concentrations of power at the same time.

A worked example: one disease seen through three lenses

Take cholera, and watch how each framework generates a different research question, a different intervention, and a different measure of success.

Tropical medicine asks what the organism is and how it behaves in hot climates. It produces Vibrio cholerae, the biotypes, and the clinical picture of rice-water stools. Its intervention is a drug or a rehydration protocol, and its success measure is case fatality in the treated patient. International health asks how cholera moves between countries and what states owe each other. It produces the sanitary conferences, quarantine rules, notification requirements, and eventually the International Health Regulations you will study in Module 6. Its intervention is a legal obligation, and its success measure is whether an outbreak was reported and contained before it crossed a border.

Global health asks why cholera exists at all today, given that the answer, safe water and sanitation, has been known since John Snow in 1854. It produces analysis of municipal financing, of conflict-driven infrastructure collapse in Yemen and Haiti, of the oral cholera vaccine stockpile shortage that pushed WHO to a single-dose strategy in 2022 to stretch supply, and of the political economy that makes a water utility unbankable. Its intervention might be a vaccine, or it might be a debt renegotiation, and its success measure is whether the disease stops occurring in the poorest districts rather than merely being treated there.

None of the three is wrong. But only the third framework can explain why a disease with a nineteenth-century solution still kills, and only the third makes a water utility's balance sheet a legitimate object of health research.

Key idea: Applied to a single disease, the three frameworks generate different questions and different definitions of success, and only the global health frame treats the political and economic conditions producing the disease as part of the health problem itself.

Common misconceptions

• Global health means health in poor countries. By its own definition it means health inequity anywhere, including within wealthy countries; the narrower usage is a symptom of the field's imbalance, not its definition.
• Global health is just public health done abroad. The methods overlap heavily, but global health adds transnational determinants, multiple sovereignties, and a donor-recipient political economy that domestic public health does not face.
• Tropical medicine is simply an older name for the same thing. Tropical medicine classified disease by climate and served a colonial economic project; its institutions survive, but its organizing logic has been rejected.
• The colonial history is interesting background but not operationally relevant. It explains where the research institutions, journals, and funding relationships sit today, which shapes what gets studied now.
• Alma-Ata failed, so comprehensive primary health care was a bad idea. It was never funded at the scale it required; the selective alternative delivered measurable gains while leaving many countries without a functioning system, which is why the argument recurs.

Recap

• Global health is the study and practice of improving health and achieving health equity for all people worldwide, with an emphasis on transnational causes and multidisciplinary action.
• Tropical medicine, founded in Liverpool in 1898 and London in 1899, produced real science inside an explicitly colonial project.
• International health built the state-to-state architecture, from the 1851 sanitary conferences and PAHO in 1902 to WHO in 1948.
• Alma-Ata in 1978 committed to comprehensive primary health care; selective primary health care answered in 1979, and the argument continues.
• The name changed in the 1990s as HIV, the World Bank's 1993 report, and new funders such as Gavi, the Global Fund, and the Gates Foundation reshaped the field.
• The definition remains contested because in practice the field's money and agenda-setting stay concentrated in high-income institutions.

Sources

1. Koplan, J. P., Bond, T. C., Merson, M. H., Reddy, K. S., Rodriguez, M. H., Sewankambo, N. K., and Wasserheit, J. N. (2009). Towards a common definition of global health. The Lancet, 373(9679), 1993-1995. doi.org/10.1016/S0140-6736(09)60332-9
2. Abimbola, S., and Pai, M. (2020). Will global health survive its decolonisation? The Lancet, 396(10263), 1627-1628. doi.org/10.1016/S0140-6736(20)32417-X
3. World Health Organization. (n.d.). History of WHO. who.int
4. Encyclopaedia Britannica. (n.d.). Sir Ronald Ross. britannica.com
5. Wikipedia contributors. (n.d.). Tropical medicine. en.wikipedia.org

The big picture

Picture a health post in rural Malawi. One nurse, two health surveillance assistants, a solar fridge, a paper register. Now count the institutions whose decisions determine what happens in that building this month. The antiretrovirals arrived through a Global Fund grant. The vaccines came through Gavi, co-financed by the Malawian treasury. The fridge was donated by a bilateral donor. The nurse's salary comes from the Ministry of Health, whose budget is shaped by an International Monetary Fund fiscal ceiling. The bed nets were procured by a nongovernmental organization under a subcontract. The malaria treatment guideline was written by WHO. The data the nurse enters at the end of the day will be aggregated into a district report, a national report, and eventually an estimate published by an institute in Seattle.

That is not a caricature. It explains two things at once: why global health can move enormous resources quickly, and why a health minister in a low-income country may spend more time managing donors than managing health. This lesson maps the actors, follows the money, and then asks the question that matters most: who actually decides what the world works on? You will need the map later. When you read in Module 3 that HIV treatment reached tens of millions, the answer to how involves the Global Fund and PEPFAR; when you read in Module 4 that mental health receives almost nothing, the answer to why involves the same institutions.

Key idea: Global health is not run by any single institution; it is a crowded field of funders, norm-setters, and implementers whose overlapping decisions determine what a frontline health worker actually has to work with.

The cast, in one table

Actor	Founded	What it actually does	What it cannot do
World Health Organization	1948	Sets norms and guidelines, classifies disease, declares emergencies, convenes member states	Compel any government; fund programs at scale
World Bank	1944	Lends and grants for health systems, produces influential economic analysis, hosts the Pandemic Fund	Set clinical or technical norms
Global Fund	2002	Channels grants for HIV, TB and malaria through country-led applications	Implement anything itself; fund outside its three diseases and system strengthening
Gavi	2000	Pools demand and subsidizes vaccines for lower-income countries, with country co-financing	Deliver vaccines; operate outside eligible countries
Gates Foundation	2000	Funds research, product development, and advocacy; shapes agendas through grant choices	Be voted out; be held accountable to any electorate
Bilateral agencies	Various	Fund and often direct programs in partner countries, sometimes at very large scale	Outlast their own domestic politics
National health ministries	Various	Employ the workforce, run the facilities, set national policy, pay most of the bill	Escape donor conditionality or fiscal ceilings easily

Read the last column carefully. Every actor in this field is powerful in one dimension and helpless in another, and most of the dysfunction you will study comes from the gaps between those columns rather than from anyone's bad intentions.

Key idea: Each global health actor holds a narrow kind of power, normative, financial, or operational, and the system's characteristic failures happen in the spaces where no actor holds the power that a situation requires.

WHO: authority without money

WHO is the only body in this field with genuine global legitimacy, because it belongs to its 194 member states and each has one vote. It is also, for its mandate, remarkably poor. Its approved programme budget for the 2024-2025 biennium was roughly 6.8 billion US dollars for two years, covering everything from cancer guidelines to Ebola response, which is less than many single large hospital systems spend annually.

The structure of that money matters more than the amount. WHO has two income streams. Assessed contributions are the dues each member state owes, calculated from population and wealth; they are flexible and predictable, and WHO can spend them as it judges best. Voluntary contributions come from states, foundations, and other agencies, and they are overwhelmingly earmarked for specific programs chosen by the donor. For decades assessed contributions were frozen in nominal terms while voluntary funding grew, so that by the 2020s roughly four fifths of WHO's income arrived tied to somebody else's priorities.

Think about what that does to an organization. If a donor will pay for polio but not for hypertension, WHO's polio staff grow and its hypertension staff do not, regardless of what the World Health Assembly voted for. Member states recognized the problem and in 2022 agreed to raise assessed contributions to 50 percent of the base budget by the 2030-2031 biennium, a genuine structural reform that will take a decade to bite.

Then the ground moved. In January 2025 the United States, historically WHO's largest single funder at close to a fifth of total income, announced its withdrawal. WHO responded by cutting its proposed 2026-2027 base budget substantially, to around 4.2 billion dollars, reducing senior management and merging departments, while the 2025 World Health Assembly approved a further 20 percent rise in assessed contributions. At the same time the dismantling of the United States Agency for International Development removed the largest bilateral health donor from many countries at short notice. The full consequences were still being counted when this course was written in 2026, so treat any specific figure you read about the aftermath as provisional and check the current source.

Key idea: WHO holds normative authority but very little discretionary money, because most of its income is earmarked voluntary contributions, and the 2025 withdrawal of its largest funder made that dependence visible in the sharpest possible way.

The World Bank and the arrival of economics

The World Bank became a health actor almost by accident and then became a decisive one. Its 1993 World Development Report, Investing in Health, argued that health spending should be judged by how much health it buys per dollar, introduced the disability-adjusted life year as the unit of account, and proposed an essential package of cost-effective services. Within a few years the Bank was lending more for health than WHO was spending, and the vocabulary of cost-effectiveness had become the common language of the field. You will work with that vocabulary yourself in Module 2.

The Bank's influence cuts both ways. It brought analytic rigor and made it possible to compare an intervention in Ghana with one in Nepal on a common scale. It also arrived during structural adjustment, when Bank and IMF lending conditions pushed governments to cut public payrolls and introduce user fees at health facilities. The evidence that user fees suppress care-seeking among the poor is strong, and most countries that introduced them have since removed them. That episode explains why the phrase Washington consensus still draws a cold response in health ministries, and why arguments about fiscal ceilings on health worker hiring remain live.

Key idea: The World Bank's 1993 report made cost-effectiveness the field's common language and the Bank a major health financier, and its structural adjustment era, including user fees that measurably suppressed care among the poor, explains much of the enduring suspicion of donor economic advice.

The vertical funds and the bilateral giants

The institutions created after 2000 share a design philosophy: pick a problem, raise money against it, buy measurable results. The Global Fund, created in 2002, does not implement anything. It receives applications developed by country coordinating mechanisms that must include government, civil society, and people affected by the diseases, disburses against performance, and reports results. By its own 2024 accounting it has disbursed well over 60 billion dollars and contributed to saving 65 million lives, a modeled figure rather than a body count, and one worth treating as an estimate with wide uncertainty.

Gavi, launched in 2000, solves a different problem: vaccine markets. Poor countries were too small and too uncertain a market for manufacturers to price for, so Gavi pools demand across dozens of countries, negotiates long-term prices, and subsidizes doses on a sliding scale while requiring countries to co-finance an increasing share until they graduate. It reports having helped immunize more than a billion children since 2000. Its 2026-2030 replenishment sought around 11.9 billion dollars and raised roughly 9 billion at its 2025 pledging summit, with the United States declining to pledge.

PEPFAR, the United States President's Emergency Plan for AIDS Relief, launched in 2003, is the largest bilateral commitment to a single disease in history, with over 110 billion dollars spent and more than 25 million lives credited to it. It is also the clearest case study in both the strength and the fragility of the vertical model: it moved faster than any multilateral body could have, and it left several national HIV programs dependent on annual appropriations by a foreign legislature, a dependence that became acute in 2025.

Key idea: The vertical funds created after 2000 delivered enormous measurable gains by concentrating money on named diseases with performance-based disbursement, and the same concentration left national programs exposed when the funding politics changed.

Philanthropy and the accountability question

The Gates Foundation is the largest private funder in global health and has been for two decades, frequently ranking among WHO's top two or three contributors. Its money has done things no government would have done: it took the financial risk on vaccine development for diseases with no commercial market, funded the Global Fund and Gavi into existence, and sustained polio eradication when state interest wavered. In 2025 the foundation announced it would spend down its endowment and close by the end of 2045, committing to distribute roughly 200 billion dollars over the intervening two decades.

The critique is not that the money is unwelcome but that it is unaccountable. A foundation's board answers to no electorate and can be neither voted out nor petitioned. When one private actor's grantmaking can shift what the world's researchers study, the concern is agenda capture: an emphasis on measurable technological fixes over slower structural work on health systems and social determinants. Defenders answer that public funders have their own distortions and shorter political cycles, and that someone had to take the risk on a malaria vaccine. Both positions are serious, and the useful habit is to ask, of any global health priority, who paid for it to become a priority.

Key idea: Private philanthropy funded work no government would risk, and it exercises agenda-setting power that no electorate can check, which is why the question of who paid for a priority to become a priority is a standard analytic move in this field.

Who actually pays: the number that surprises people

Here is the correction most newcomers need. Development assistance for health is a minority of health spending almost everywhere. IHME's Financing Global Health series tracks it rising from roughly 8 billion dollars a year around 1990 to about 40 billion by the late 2010s, spiking above 70 billion in 2021 as COVID money surged, then falling back. Set that against total global health spending, which runs in the trillions. Aid is on the order of one percent of what the world spends on health.

Concentration is what makes it matter. In low-income countries as a group, external assistance funds roughly a quarter to a third of health spending, and in a handful of countries more than half. So the same dollar is simultaneously trivial in global terms and decisive in the specific places where it lands. Meanwhile, African Union states committed in the 2001 Abuja Declaration to spend 15 percent of government budgets on health, and only a small number have ever reached it. Domestic revenue, not aid, is the long-run answer, which is why tax capacity has become a health policy subject.

Now compare burden with money. Noncommunicable diseases cause roughly three quarters of deaths worldwide, and they have consistently attracted under 2 percent of development assistance for health. HIV, which causes a small fraction of deaths, has for years drawn the single largest share. This is not a scandal so much as a fact with an explanation: aid follows advocacy, emergency framing, and the possibility of a discrete measurable product, and NCDs offer none of those easily. But hold the mismatch in mind, because it is the single most important thing to know about how priorities get set.

Key idea: Health aid is roughly one percent of global health spending yet up to a third or more of health spending in low-income countries, and its allocation tracks advocacy and measurability rather than burden, which is why NCDs receive under 2 percent of it while causing about three quarters of deaths.

The standard critiques, at full strength

Four criticisms recur, each with real evidence behind it. Verticalization: disease-specific programs build parallel supply chains, information systems, and salary scales a ministry cannot sustain, and they pull experienced staff out of general services into better-paid donor projects, an internal brain drain that hollows out the system the program depends on. Fragmentation: a ministry may report to dozens of donors with incompatible formats and timetables, with senior officials spending large parts of the year on donor missions. Volatility: aid arrives and departs on foreign political cycles, as 2025 demonstrated. Dependency: Angus Deaton argues that large external flows weaken the link between a government and its own citizens, since a ministry funded from abroad answers to funders rather than voters.

Now the other side, just as plainly. Antiretroviral therapy reached over 30 million people, which did not happen on domestic revenue alone. Child deaths fell from roughly 12.8 million a year in 1990 to under 5 million, with vaccination, bed nets, and oral rehydration doing much of that work and external financing paying for a large part of it. Studies of health aid specifically tend to find measurable mortality effects, which is not reliably true of aid in general. The honest position is that health aid has a stronger empirical record than aid as a category and that the structural critiques of its delivery are also correct. Holding both is not fence-sitting; it is the accurate description.

Key idea: Verticalization, fragmentation, volatility, and dependency are documented harms of the aid architecture, and health aid also has a measurable mortality record that general development aid lacks, so both claims must be held together.

Common misconceptions

• WHO runs global health. WHO sets norms and declares emergencies but controls a small budget, most of it earmarked by donors, and cannot compel any government to act.
• Foreign aid pays for health care in poor countries. Domestic government revenue and household out-of-pocket payments dominate almost everywhere; aid is roughly a quarter to a third of health spending in low-income countries and about one percent globally.
• Money follows the biggest health problems. It follows advocacy, emergency framing, and measurable products; NCDs cause about three quarters of deaths and receive under 2 percent of health aid.
• Vertical disease programs are simply bad. They produced the largest measured mortality gains of the past twenty-five years while also building parallel systems that weakened general services.
• Private foundations are just another donor. Their scale, permanence, and freedom from electoral accountability give them agenda-setting power that state donors do not have in the same way.

Recap

• Global health is governed by no one; it is a crowded field of norm-setters, financiers, and implementers with mismatched powers.
• WHO has 194 member states and global legitimacy but a two-year budget near 6.8 billion dollars, roughly four fifths of it earmarked by donors.
• The World Bank's 1993 report made cost-effectiveness the field's language; its structural adjustment era left lasting distrust.
• The Global Fund, Gavi, and PEPFAR concentrate money on named problems with measurable results, at the cost of parallel systems and political exposure.
• The Gates Foundation shapes agendas at a scale no electorate can check, and announced in 2025 that it will spend down and close by 2045.
• Health aid is about one percent of global health spending, but up to a third or more of it in low-income countries, and its allocation tracks advocacy rather than burden.

Sources

1. World Health Organization. (n.d.). How WHO is funded. who.int
2. The Global Fund to Fight AIDS, Tuberculosis and Malaria. (n.d.). Results and impact. theglobalfund.org
3. Gavi, the Vaccine Alliance. (n.d.). About our alliance. gavi.org
4. Institute for Health Metrics and Evaluation. (n.d.). Financing Global Health. healthdata.org
5. U.S. Department of State. (n.d.). PEPFAR: The U.S. President's Emergency Plan for AIDS Relief. state.gov
6. Bill and Melinda Gates Foundation. (n.d.). About the foundation. gatesfoundation.org

The big picture

Start with an uncomfortable fact. A majority of the deaths that happen in the world this year will never be registered anywhere. No death certificate, no cause recorded, no entry in any national system. WHO has estimated for years that fewer than half of all deaths worldwide are registered with a cause of death, and in much of sub-Saharan Africa and South Asia the share is far lower. In some countries the civil registration system captures under a tenth of deaths.

Sit with what that means. Every number you will read in this course, every headline about child mortality falling or maternal deaths rising, rests on a data foundation that in many places consists of household surveys, sample systems, and statistical models rather than counted events. That does not make the numbers useless. It makes them estimates with uncertainty, produced by identifiable institutions using stated methods, and it makes reading them carefully a professional skill rather than a formality.

This lesson teaches three headline measures, life expectancy, under-5 mortality, and maternal mortality, with the arithmetic worked out so you can compute them yourself. It then shows where the underlying data comes from and how to interrogate a figure before you repeat it. The methods of study design and causal inference sit in Atlas's Epidemiology and Public Health (PUBH 310); here we are concerned with the population indicators themselves and with their honest interpretation.

Key idea: Most of the world's deaths go unregistered, so the headline statistics of global health are usually modeled estimates rather than counts, and reading them well means knowing who produced them, from what data, and with how much uncertainty.

Life expectancy: the most misread number in the field

Life expectancy at birth is not a forecast of how long a baby born today will live. It is the average number of years a hypothetical cohort would live if it experienced, at every age, the death rates observed in a given population in a given year. Demographers call this a period life table, and the cohort it describes is synthetic. If mortality improves over the next seventy years, and it usually does, actual babies will outlive the figure.

Here is the second and larger misreading. When you hear that life expectancy in a poor country is 53, or that it was around 35 in medieval Europe, most people picture a population of people dying in middle age. That is not what those numbers describe. Life expectancy is a mean, and means are dragged hard by deaths at very young ages.

Work it yourself. Imagine a village of ten people. Four die in their first year of life; six live to age 70. The mean age at death is (4 times 1, plus 6 times 70) divided by 10, which is (4 plus 420) divided by 10, or 42.4 years. Life expectancy in that village is 42, and yet every single person who survives infancy reaches 70. Now suppose you prevent three of those four infant deaths, and those three children also live to 70. The mean becomes (1 plus 9 times 70) divided by 10, or 63.1 years. Life expectancy jumped 21 years and not one adult lived a day longer. This is why child survival dominates the history of life expectancy gains, and why an improving life expectancy figure usually tells you about babies rather than about the elderly.

The current global picture, with dates attached because these move: global life expectancy at birth reached about 73 years by 2019, fell roughly 1.8 years during 2020 and 2021 as COVID-19 erased close to a decade of gains, and has been recovering since. The spread across countries is enormous. Japan, Switzerland, Spain and a handful of others sit in the mid-80s; Chad, Nigeria, Lesotho and the Central African Republic sit in the low- to mid-50s. A gap of roughly thirty years in the expected span of a human life is the single starkest number in global health.

One refinement is worth knowing. Healthy life expectancy, or HALE, estimates the years lived in good health rather than merely alive, by subtracting time lived with disability. Globally it runs roughly ten years below life expectancy, which is a compact way of saying that the last decade of the average life is lived in some degree of ill health. Extending life expectancy without extending HALE simply lengthens that decade.

Key idea: Life expectancy at birth is a synthetic period measure heavily driven by deaths in early childhood, so a low figure usually signals high infant mortality rather than a population dying in middle age, and gains in it are mostly gains in child survival.

Under-5 mortality: the field's favorite indicator

Under-5 mortality is the workhorse indicator of global health, and for good reason. It is sensitive to nutrition, infection, water, birth care, and health system access all at once, so it moves when almost anything important changes. Get the definitions right, because three related measures are constantly confused.

Indicator	What it counts	Denominator	Typical global figure (2023)
Neonatal mortality rate	Deaths in the first 28 days of life	Per 1,000 live births	About 17
Infant mortality rate	Deaths before the first birthday	Per 1,000 live births	About 27
Under-5 mortality rate	Probability of dying before the fifth birthday	Per 1,000 live births	About 37
Maternal mortality ratio	Maternal deaths during pregnancy or within 42 days of its end	Per 100,000 live births	About 197

Note that the under-5 rate is technically a probability derived from a life table rather than a simple division, but for a single year in a single district the simple calculation is close enough to build intuition. Try it. A district records 8,000 live births in a year and 240 deaths of children under five. The rate is 240 divided by 8,000, times 1,000, which is 30 per 1,000 live births. A neighboring district with the same 8,000 births records 24 such deaths, giving 3 per 1,000. Same country, tenfold difference, and that kind of internal gap is extremely common.

The global trend is one of the genuine triumphs in this course, and it should be stated plainly. Around 12.8 million children died before their fifth birthday in 1990. By 2023 the UN Inter-agency Group for Child Mortality Estimation put the figure at roughly 4.8 million, with the rate falling from about 93 to about 37 per 1,000 live births. That is a decline of about 60 percent in the rate during a period when the number of births rose. Vaccination, insecticide-treated nets, oral rehydration, better nutrition, and rising incomes all contributed; you will meet each of them later in the course.

Two features of the remaining deaths matter for policy. First, the burden has shifted toward the newborn period: roughly 2.3 million of those 4.8 million deaths occurred in the first 28 days of life, which means the intervention target has moved from the village to the labor ward. Second, the deaths are geographically concentrated, with about half occurring in sub-Saharan Africa; a child born there is on the order of fifteen times more likely to die before turning five than a child born in western Europe.

Key idea: Under-5 mortality per 1,000 live births fell from roughly 93 in 1990 to about 37 in 2023, one of the field's clearest successes, and the remaining deaths are increasingly newborn deaths concentrated in sub-Saharan Africa.

Maternal mortality: rare, hard to count, and politically loaded

A maternal death is defined as the death of a woman while pregnant or within 42 days of the end of pregnancy, from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes. That definition does a lot of work. A woman who dies of eclampsia is a maternal death; a woman who dies in a car crash at seven months pregnant is not.

The maternal mortality ratio is maternal deaths per 100,000 live births. Note the denominator carefully, because it is a different scale from child mortality and mixing them up produces absurdities. Note also the distinction from the maternal mortality rate, which uses women of reproductive age as its denominator and is used far less often. When someone says MMR in global health, they almost always mean the ratio.

Work an example. A country records 1,000,000 live births in a year and 5,000 maternal deaths. The MMR is 5,000 divided by 1,000,000, times 100,000, which is 500 per 100,000 live births. For comparison, several high-income countries report figures in the low single digits. A ratio of 500 versus 3 is not a difference of degree; it is a difference in whether a woman giving birth faces a meaningful risk of dying.

The global figures, with their vintage attached: UN estimates put maternal deaths at about 287,000 in 2020, an MMR of roughly 223, and the round of estimates released in 2025 put 2023 at roughly 260,000 deaths and an MMR near 197. Sub-Saharan Africa accounts for around 70 percent of the world total. The starkest way to express the gap is lifetime risk, the cumulative probability that a 15-year-old girl will eventually die of a maternal cause: roughly 1 in 40 in sub-Saharan Africa against roughly 1 in 16,000 in Australia and New Zealand.

Why is this so hard to measure? Because maternal deaths are statistically rare. In a district with 5,000 births a year and an MMR of 400, you expect twenty maternal deaths. Random variation alone will swing that number substantially year to year, so a district cannot reliably detect improvement from its own counts. Add misclassification, since a death from postpartum hemorrhage may be recorded simply as hemorrhage, and underreporting of deaths outside facilities, and you understand why the field uses indirect methods: the sisterhood method, which asks adult respondents about the survival of their sisters; verbal autopsy, which reconstructs a probable cause from a structured interview with relatives; and confidential enquiry, which reviews each death in detail. All of these feed statistical models rather than a register.

Key idea: Maternal mortality is measured as deaths per 100,000 live births, it is concentrated in sub-Saharan Africa where lifetime risk approaches 1 in 40, and because maternal deaths are rare and often misclassified the published figures are modeled estimates built from surveys and indirect methods.

Where the numbers actually come from

Four sources feed almost everything you will read. Civil registration and vital statistics systems record births and deaths continuously and are the gold standard, but they function well in only about half the world's countries. Household surveys, above all the Demographic and Health Surveys and UNICEF's Multiple Indicator Cluster Surveys, interview women about their full birth histories and are the backbone of child mortality estimation in low-income settings. Sample registration systems, such as India's, continuously monitor a representative sample of the population rather than everyone. Censuses provide periodic anchors.

Those inputs then go into modeling groups. The UN Inter-agency Group for Child Mortality Estimation, run jointly by UNICEF, WHO, the World Bank and UN DESA, produces the official child mortality series. The Maternal Mortality Estimation Inter-agency Group does the same for maternal deaths. The Institute for Health Metrics and Evaluation in Seattle produces a separate and independent set of estimates. These groups do not always agree, and the disagreements are not trivial: IHME and the UN agencies have at times differed by tens of thousands of maternal deaths for the same year, because they weight sources differently and make different assumptions where data is missing.

That is not a scandal; it is what estimation under uncertainty looks like. But it has a practical consequence. When a news story says maternal deaths rose or child deaths fell, ask which series it is quoting, whether the change exceeds the uncertainty interval, and whether the whole historical series was revised at the same time. Retrospective revision is routine: a new survey can change the estimate for a year twenty years in the past, which means two reports published three years apart may disagree about 1990.

Key idea: Global health statistics flow from civil registration, household surveys, sample registration, and censuses into competing modeling groups, and different groups produce genuinely different numbers for the same year, so the source and vintage of a figure are part of the figure.

A checklist for reading any global health number

Adopt these six questions as a habit, and you will avoid most of the errors that appear in print. What exactly is the numerator, and does it match the definition I think it does? What is the denominator, per 1,000 live births, per 100,000 live births, or per 100,000 population? What year does it refer to, which is often several years before publication? Who produced it, and is it observed or modeled? What is the uncertainty interval, and does the reported change exceed it? Has the series been revised, so that comparing today's number to a figure you remember from an older report is comparing two different models?

Apply the checklist to a real sentence: maternal deaths fell by 40 percent since 2000. The numerator is modeled maternal deaths, the denominator is live births, the endpoint is probably three years before the publication date, the producer is likely the UN inter-agency group, the uncertainty interval on the percentage change is wide, and the entire series was recalculated when the estimates were released. All of that can be true while the sentence remains substantially correct. Precision is not cynicism.

Key idea: A disciplined reader asks six questions of every figure, about numerator, denominator, year, producer, uncertainty, and revision, and doing so sharpens rather than undermines the real findings.

Common misconceptions

• Life expectancy of 45 means people die around 45. It is a mean dragged down by infant deaths; in such populations most people who survive childhood live into their sixties or beyond.
• Life expectancy predicts how long a baby born today will live. It is a synthetic period measure using this year's death rates at every age, so improving mortality means real cohorts usually outlive it.
• Under-5 mortality and infant mortality are the same thing. Infant mortality covers the first year; under-5 mortality covers the first five, and neonatal mortality covers the first 28 days, which is now nearly half the total.
• The maternal mortality ratio is per 1,000 births. It is per 100,000 live births, because maternal death is far rarer than child death.
• Published global figures are counts. Most are modeled estimates with uncertainty intervals, produced by named groups whose numbers sometimes differ substantially from each other.

Recap

• Fewer than half the world's deaths are registered with a cause, so most global figures are estimates rather than counts.
• Life expectancy at birth is a period life-table measure dominated by early-childhood deaths; global life expectancy reached about 73 years in 2019 and fell about 1.8 years in 2020-2021.
• Under-5 mortality fell from roughly 93 to about 37 per 1,000 live births between 1990 and 2023, with deaths concentrated in the newborn period and in sub-Saharan Africa.
• The maternal mortality ratio is deaths per 100,000 live births; global estimates were about 287,000 deaths in 2020 and about 260,000 in 2023.
• Maternal deaths are rare and often misclassified, so measurement relies on surveys, the sisterhood method, verbal autopsy, and modeling.
• Different estimation groups produce different numbers for the same year, and series are routinely revised backward.

Sources

1. UN Inter-agency Group for Child Mortality Estimation. (n.d.). Child mortality estimates. childmortality.org
2. UNICEF. (n.d.). Under-five mortality. data.unicef.org
3. World Health Organization. (n.d.). Maternal mortality (fact sheet). who.int
4. World Health Organization. (n.d.). Global Health Observatory data. who.int
5. Wikipedia contributors. (n.d.). Life expectancy. en.wikipedia.org

The big picture

A health minister has one budget and two proposals on the desk. The first prevents 200 child deaths from malaria. The second treats 5,000 adults with moderate depression for a year. Which buys more health? You cannot answer with death counts, because the second proposal prevents no deaths at all. You cannot answer with case counts, because a case of malaria and a case of depression are not the same kind of thing. You need a common currency.

The disability-adjusted life year, or DALY, is that currency, and by the end of this lesson you will be able to compute both sides of that comparison yourself. The DALY was invented for the World Bank's 1993 World Development Report and has since become the unit in which the world's health is described. Its sibling, the quality-adjusted life year or QALY, does similar work from the other direction and dominates health technology assessment in high-income countries.

Then comes the harder half of the lesson. Putting depression and child death on one scale requires value judgments, and the DALY contains several of them. Some were made explicitly and later reversed. Others are still contested. A serious global health practitioner uses these numbers constantly and knows exactly what is buried in them, which is the combination this lesson is aiming for.

Key idea: DALYs and QALYs exist because comparing a fatal disease with a disabling one requires a single unit, and building that unit necessarily embeds value judgments that must be understood rather than ignored.

Building the DALY from its two halves

One DALY is one lost year of healthy life. The formula is simple: DALYs equal YLL plus YLD, years of life lost plus years lived with disability.

Years of life lost measures fatal loss. For each death you take the standard remaining life expectancy at the age of death, drawn from a single reference life table used for every country and both sexes, and add it up. The GBD reference table sets life expectancy at birth at roughly 88 years, based on the lowest observed mortality at each age anywhere in the world. Using one table for everyone is deliberate: it means a death at 40 in Malawi counts exactly as much as a death at 40 in Norway. Earlier versions of the method used sex-specific tables, which counted men's deaths as smaller losses; that was dropped.

Years lived with disability measures non-fatal loss. For each health state you multiply the number of people in it by a disability weight between 0, meaning full health, and 1, meaning a loss equivalent to death, and by the fraction of a year spent in that state. Equivalently, for short conditions, incidence times average duration times weight.

Now work the minister's problem. Proposal one: 200 children die of malaria at an average age of 2. Standard remaining life expectancy at age 2 is about 86 years, so each death costs 86 YLL and the total is 200 times 86, which is 17,200 YLL. Suppose the same program also prevents 60,000 uncomplicated malaria episodes, each lasting about 14 days, which is 0.038 years, with a disability weight of about 0.051 for a moderate acute infectious episode. That gives 60,000 times 0.038 times 0.051, which is about 116 YLD. Total burden averted: roughly 17,316 DALYs, of which 99.3 percent is years of life lost. When children die, the fatal term swamps everything else.

Proposal two: 5,000 adults with moderate depression, disability weight about 0.396, treated for a full year with complete relief. That is 5,000 times 1 times 0.396, which is 1,980 YLD and zero YLL. Divide by 86 and you get the equivalent of about 23 averted child deaths.

So the DALY says the malaria program buys roughly nine times as much health. Notice what just happened. The comparison is now decidable, which is progress. It is decidable only because we accepted that 1,980 person-years of moderate depression is worth about 23 child lives, which is a moral claim wearing a lab coat. Both observations are correct at once, and that tension runs through everything that follows.

Key idea: DALYs equal years of life lost plus years lived with disability, with YLL computed against a single universal reference life table and YLD as prevalence times a disability weight, and the resulting comparability is bought with an explicit exchange rate between illness and death.

Where disability weights come from

The obvious question is who decided that moderate depression is 0.396. In the earliest rounds, small expert panels did. That was criticized hard, and rightly: physicians and health economists are not a representative sample of humanity, and they systematically misjudge what living with a condition is like.

The 2010 and 2013 rounds replaced expert panels with large population surveys. Tens of thousands of respondents across several countries, plus an open web survey, were shown pairs of health states described in plain lay language and asked which person they judged healthier. Paired comparisons of that kind can be converted into a single scale without ever asking anyone to name a number, which avoids a lot of known bias. The resulting weights, such as roughly 0.187 for blindness, 0.396 for a moderate depressive episode, and 0.658 for a severe one, are population judgments about health states rather than expert judgments about people.

One misreading has to be stamped out. A disability weight is not a valuation of a person or of a life. It is an estimate of how much health is lost in a state relative to full health. Disability rights scholars have argued, seriously and persistently, that in practice the distinction erodes: if a year of life with blindness counts as 0.813 of a year, then a program extending the lives of blind people generates fewer DALYs averted than one extending the lives of sighted people, and in a cost-effectiveness league table that is discrimination by arithmetic. The counterargument is that refusing to measure non-fatal loss at all would render every chronic and mental illness invisible, which harms the same groups more. There is no clean resolution, and you should know that there is not.

Key idea: Modern disability weights come from large lay population surveys using paired comparisons rather than expert panels, and although a weight is formally a property of a health state rather than of a person, cost-effectiveness rankings built from weights can still disadvantage people living with disability.

The value choices that were quietly removed

The original 1993 DALY contained two features that no longer exist, and knowing them tells you a great deal about how the field's values shifted.

Age weighting counted a year of life in young adulthood as worth more than a year in infancy or old age, on the reasoning that young adults support dependents and produce economically. It was a defensible instrumental argument and it was also, put plainly, a statement that some people's years matter more. It was dropped in the GBD 2010 round.

Time discounting reduced the value of health gains occurring in the future, typically at 3 percent a year, by analogy with financial discounting. Discounting has a real justification in economics and a perverse effect in health: it makes preventing a child's death in forty years worth far less than preventing an adult's death today, which systematically disfavors prevention, vaccination, and anything to do with climate. It too was dropped from the standard GBD calculation, though many cost-effectiveness analyses still apply it, so always check.

Key idea: Age weighting and 3 percent time discounting were built into the original DALY and removed in GBD 2010, and their removal shows that the unit's parameters are policy choices open to revision rather than facts of nature.

QALYs and the arithmetic of cost-effectiveness

The QALY runs the same logic in reverse. A DALY counts health lost, so lower is better; a QALY counts health gained, so higher is better. One QALY is one year in full health. Utilities run from 0 for death to 1 for perfect health, and they are typically elicited from patients or the public using instruments such as the EQ-5D questionnaire, the time trade-off, or the standard gamble.

	DALY	QALY
Direction	Health lost; fewer is better	Health gained; more is better
Typical use	Describing burden across populations and countries	Appraising a specific technology or treatment
Scale anchor	Disability weight: 0 is full health, 1 is death	Utility: 1 is full health, 0 is death
Where the values come from	Lay population surveys of health-state descriptions	Preference elicitation from patients or the public

Work a QALY example. A new treatment costs 5,000 dollars per patient. Without it, a patient lives 2 years at utility 0.5, which is 1.0 QALYs. With it, the patient lives 5 years at utility 0.7, which is 3.5 QALYs. The gain is 2.5 QALYs, so the incremental cost-effectiveness ratio is 5,000 divided by 2.5, or 2,000 dollars per QALY gained. In England, NICE has generally treated interventions under about 20,000 to 30,000 pounds per QALY as good value, so this one would be an easy approval there.

Now a warning that matters in practice. For two decades a rule of thumb circulated that an intervention costing less than one times a country's GDP per capita per DALY averted was highly cost-effective, and up to three times was cost-effective. Those thresholds came from a WHO commission and were never intended as decision rules. They have been publicly disowned by WHO-affiliated authors, because a threshold set at national income tells you nothing about what a health budget can actually afford and would endorse spending far more than most low-income countries have. Current practice tries to derive thresholds from what a specific health system actually displaces when it spends a marginal dollar. If you see the GDP-multiple rule used in a paper, treat it as a warning sign about the paper.

Key idea: QALYs measure health gained and drive technology appraisal through cost per QALY, DALYs measure health lost and drive burden description, and the old rule that an intervention is cost-effective if it costs under one to three times GDP per capita per DALY averted has been disowned and should not be used.

What the Global Burden of Disease study actually is

The GBD began as a study commissioned for the 1993 World Development Report, led by Christopher Murray and Alan Lopez. Before it, the honest answer to what people die of worldwide was that nobody knew; national reporting covered a minority of deaths and there was no consistent way to add them up. The GBD imposed one requirement that changed everything: every death must be assigned to exactly one cause, and the causes must sum to the total. That forced the field to confront the deaths nobody was counting.

Today the study is run from the Institute for Health Metrics and Evaluation in Seattle with several thousand collaborators worldwide. The GBD 2021 round, published in The Lancet in 2024, estimated incidence, prevalence, deaths, YLLs, YLDs and DALYs for 371 diseases and injuries and 88 risk factors, for 204 countries and territories, by age and sex, for every year since 1990. It is available through a free public visualization tool, and you should spend an hour in it at some point in this course.

Its findings restructured global health priorities. The GBD is the reason the field knows that noncommunicable diseases dominate the global burden, that mental and substance use disorders account for an enormous share of non-fatal loss that death counts had made invisible, that road injury is a leading killer of young people, and that a handful of risk factors, high blood pressure, tobacco, air pollution, and dietary risks, sit behind a very large fraction of total loss.

Key idea: The Global Burden of Disease study forced every death to be assigned to one cause and every non-fatal condition to be quantified, and in doing so it made noncommunicable disease, mental illness, and injury visible as global priorities for the first time.

The critiques, taken seriously

Four lines of criticism deserve your attention, and none of them is frivolous.

The first is data sparsity. GBD publishes an estimate for every country and every year, including countries with almost no primary data, by borrowing strength from covariates, regional neighbors, and historical patterns. The uncertainty intervals around those estimates are wide, and they are almost always stripped out when the number reaches a press release or a slide. A point estimate for a country with no vital registration is a model output, and treating it as a measurement is a category error.

The second is opacity and ownership. The modeling machinery is genuinely complex, which makes independent replication hard even though IHME publishes code and tools. More sharply, national statistical offices sometimes find that an institute in Seattle has published figures for their country that differ substantially from their own official statistics, and that the international press quotes Seattle. WHO and IHME have disagreed publicly about specific estimates. When national institutions have less authority over their own numbers than a foreign institute does, that is a governance problem as much as a technical one.

The third is philosophical, developed most rigorously by Sudhir Anand and Kara Hanson in 1997. Their argument is that the DALY presents contestable value judgments as technical parameters, and that aggregating health loss into a single total is blind to distribution: a DALY averted counts the same whether it is averted among the richest or the poorest, so a purely DALY-maximizing allocation can widen health inequity while looking optimal. If equity is the point of global health, its main measuring instrument does not measure equity.

The fourth is practical. Cost-effectiveness league tables can crowd out considerations that do not reduce to health units, including rights, dignity, urgency, and the value of a functioning health system that can respond to problems nobody has modeled yet. Palliative care scores poorly on DALYs averted; that is an argument about the metric, not about palliative care.

Against all of this stands one blunt fact. Before the GBD, priority setting ran on advocacy, anecdote, and whichever disease had the most persuasive champion. Imperfect comparability is better than none, and the GBD's own findings, particularly on mental health and NCDs, have been the most powerful evidence against the priorities the old system produced. Use the numbers, quote the uncertainty, and never let a league table be the last word.

Key idea: The main criticisms of GBD are data sparsity behind confident point estimates, opacity and the displacement of national statistical authority, the embedding of value judgments including blindness to distribution, and the crowding out of non-quantifiable goods, all of which stand alongside the fact that the study made previously invisible burdens visible.

Common misconceptions

• A DALY is a death. A DALY is one lost year of healthy life; a single death of a young child generates many DALYs, and a chronic condition generates DALYs without any death.
• Disability weights measure the value of a person's life. They estimate health loss in a state relative to full health, although cost-effectiveness rankings built on them can still disadvantage people with disabilities.
• DALYs and QALYs are the same thing with different names. They run in opposite directions, use different elicitation methods, and are used for different purposes, burden description versus technology appraisal.
• An intervention is cost-effective if it costs less than GDP per capita per DALY averted. That rule of thumb has been disowned by its own originators' institution and bears no relation to what a health budget can afford.
• GBD numbers are measurements. For much of the world they are model outputs with wide uncertainty intervals that rarely survive the trip to a headline.

Recap

• DALYs equal years of life lost plus years lived with disability, and one DALY is one lost year of healthy life.
• YLL uses a single universal reference life table, so a death at a given age counts identically in every country.
• YLD is prevalence times a disability weight derived from large lay population surveys using paired comparisons.
• Age weighting and 3 percent discounting were part of the original DALY and were removed in GBD 2010.
• QALYs measure health gained and drive cost per QALY appraisal; the GDP-multiple cost-effectiveness threshold has been discredited.
• The GBD covers 371 diseases and 88 risk factors across 204 countries, and its serious critics point to data sparsity, opacity, embedded values, and blindness to distribution.

Sources

1. Institute for Health Metrics and Evaluation. (n.d.). Global Burden of Disease (GBD). healthdata.org
2. Institute for Health Metrics and Evaluation. (n.d.). GBD Compare data visualization. vizhub.healthdata.org
3. Anand, S., and Hanson, K. (1997). Disability-adjusted life years: A critical review. Journal of Health Economics, 16(6), 685-702. doi.org/10.1016/S0167-6296(97)00005-2
4. World Health Organization. (n.d.). Global Health Estimates. who.int
5. Wikipedia contributors. (n.d.). Disability-adjusted life year. en.wikipedia.org

The big picture

Two questions organize this lesson. First, why does the mix of diseases in a country change so predictably as it gets richer, from infections and famine toward heart disease and cancer? Second, if we know how to prevent most premature death, why do enormous gaps persist, not only between Chad and Switzerland but between two neighborhoods in the same city?

The first question has a famous answer, the epidemiologic transition, which you need to know and also need to know the limits of. The second question has no single answer, but it has a well-developed analytic apparatus, and learning to use it is one of the more transferable skills in this course. It is the difference between saying poor people are less healthy, which is an observation, and being able to say by how much, along which axis, whether the gap is widening, and which of it is remediable.

A warning before we start. Averages lie in a specific and predictable way. A national figure can improve every year while the gap inside the country widens every year, and a great deal of global health reporting misses exactly this. By the end of the lesson you will be able to compute both at once.

Key idea: The epidemiologic transition describes how a country's disease mix shifts as mortality falls, and health inequity analysis describes the gaps that persist within and between countries, which national averages routinely conceal.

Omran's transition

In 1971 the demographer Abdel Omran published a synthesis in the Milbank Memorial Fund Quarterly that named a pattern everyone had half-noticed. He proposed that populations move through three ages as mortality falls.

The age of pestilence and famine: mortality is high and volatile, epidemics and food crises drive death rates up and down, life expectancy sits somewhere between 20 and 40 years, and most deaths are of children and of women in childbirth. The age of receding pandemics: epidemics become less frequent and less lethal, mortality falls steadily, life expectancy rises toward 50 and beyond, and population growth accelerates because deaths fall before births do. The age of degenerative and man-made diseases: mortality stabilizes at a low level, life expectancy exceeds 70, and the dominant causes become cardiovascular disease, cancer, and the consequences of behavior and environment.

Later scholars added stages. Olshansky and Ault proposed a fourth age of delayed degenerative diseases, in which the same diseases still dominate but people die of them a decade or two later because of treatment and risk-factor control. Others have proposed a fifth stage of re-emerging and newly emerging infectious disease, prompted by HIV, drug-resistant tuberculosis, and pandemic influenza.

Alongside this runs the demographic transition, which is a different process with the same driver. Mortality falls first, fertility falls afterward with a lag, and the gap between them produces rapid population growth followed by a bulge of working-age adults, sometimes called the demographic dividend, and eventually an aging population. Understanding the age structure is essential to reading burden data, because a country with a young population will have low crude death rates from cancer without anyone being protected from cancer.

Key idea: Omran's three ages, pestilence and famine, receding pandemics, and degenerative and man-made diseases, describe a shift from infectious to chronic causes as mortality falls, with later authors adding stages for delayed degenerative disease and re-emerging infections.

Where the model breaks

The transition is a useful description and a poor prophecy. Three problems stand out.

It is not linear, and it reverses. Between the early 1990s and the mid-2000s, HIV drove life expectancy down by ten to twenty years in Botswana, Zimbabwe, Lesotho and Eswatini, reversing decades of gains before antiretroviral therapy turned it around. Male life expectancy in Russia fell from around 64 years in 1990 to about 57 by 1994 amid the post-Soviet collapse, driven substantially by alcohol and cardiovascular deaths. United States life expectancy declined in the mid-2010s and again during COVID-19, with drug overdose deaths exceeding 100,000 a year. Transitions are not ratchets.

It is not universal in sequence. Julio Frenk and colleagues argued in 1991 that Latin America did not fit the model, and proposed instead a protracted and polarized transition: countries carry infectious and chronic burdens simultaneously for decades rather than passing from one to the other, and different social groups within the same country sit at different stages at the same time. A wealthy district of Mexico City and a rural indigenous community can look like different centuries.

It smuggles in a theory of progress. The model was built on European and North American history, and reading it as the path every country will follow invites the assumption that low-income countries are simply behind rather than differently situated in a global economy that shapes their disease patterns. That is the same assumption this course flagged in Lesson 1.

What survives all this is the double burden, and it is the single most policy-relevant fact in the module. Low- and middle-income countries today face high rates of infectious disease and rising rates of noncommunicable disease at the same time, with health systems, budgets and workforces built for the first. You will see the consequences throughout Module 4.

Key idea: The epidemiologic transition reverses under shocks such as HIV, state collapse and pandemics, does not run in a fixed sequence, and in most low- and middle-income countries produces a double burden of infectious and chronic disease rather than a clean handover.

Between countries: the Preston curve

Plot every country with income per person on the horizontal axis and life expectancy on the vertical one. Samuel Preston did this in 1975 and the resulting shape is one of the most instructive pictures in the field. It rises very steeply at low incomes, so that moving from very poor to modestly poor buys many years of life, then flattens sharply, so that beyond roughly the income of a middle-income country, more money buys very little additional life expectancy.

The second finding is the important one. The whole curve has shifted upward over time. A country at a given income level today has substantially higher life expectancy than a country at that same real income had in 1930 or 1960. Preston estimated that income growth accounted for only a modest fraction, on the order of a tenth to a quarter, of the improvement in life expectancy over the period he examined. The rest came from things that travel independently of income: vaccines, antibiotics, oral rehydration, insecticide-treated nets, sanitation knowledge, literacy, and public health institutions.

Read the policy implication carefully, because it is the strongest argument for global health as a field. If health followed income tightly, the only route to better health would be economic growth, and health programs would be a sideshow. Because the curve shifts, knowledge and technology can deliver large health gains at any income level, which is exactly what Sri Lanka, Kerala, Costa Rica, Cuba, Vietnam and Rwanda have demonstrated at incomes far below what their health outcomes would predict.

Key idea: The Preston curve shows life expectancy rising steeply with income only at low incomes and flattening thereafter, and its upward shift over time shows that most historical health gains came from diffused knowledge and technology rather than from income growth.

Within countries: the social gradient

Now zoom inside a single country, and prepare to be surprised. Michael Marmot's Whitehall studies followed British civil servants, a group with no destitution, universal health coverage, and secure employment. The finding was not that the poorest clerks were sick and everyone else was fine. It was a gradient: at every step down the employment hierarchy, mortality rose, including from the second-highest grade to the highest. Health tracks relative position continuously, not merely absolute deprivation.

The magnitudes are large. The 2008 report of the WHO Commission on Social Determinants of Health, which Marmot chaired, contrasted male life expectancy in the Calton neighborhood of Glasgow with the suburb of Lenzie a few miles away, a difference reported at around 28 years, though later work has revised the specific figures and the comparison should be treated as illustrative rather than exact. In the United States, Raj Chetty and colleagues found in 2016 that men in the top 1 percent of the income distribution could expect to live about 14.6 years longer at age 40 than men in the bottom 1 percent, with a 10.1-year gap for women.

In lower-income countries the standard stratifier is the household wealth quintile from DHS and MICS surveys. Under-5 mortality in the poorest quintile typically runs about twice the rate in the richest quintile, and gaps by mother's education, by urban versus rural residence, and by region are often as large. Ethnicity and indigeneity produce comparable gaps in many countries, and where those data are not collected, the gap is invisible rather than absent.

Key idea: Health follows a continuous social gradient rather than a simple poor-versus-rich split, demonstrated in populations with no absolute deprivation, and the standard stratifiers for measuring it are income or wealth quintile, education, geography, sex, and ethnicity.

Inequality, inequity, and why the words matter

A health inequality is any measurable difference in health between groups. Men in most countries die earlier than women; that is an inequality. A health inequity is a subset: a difference that is unfair, avoidable, remediable, and systematically patterned along lines of social advantage. Margaret Whitehead's formulation is the standard one, and each of the four words does work. Unfair introduces a moral judgment that cannot be derived from the data. Avoidable and remediable exclude differences we do not know how to change. Systematic excludes random variation.

Test the distinction on cases. Higher rates of skin cancer in people with pale skin living in Australia is an inequality with a substantially biological cause. Higher infant mortality among Black women in the United States, persisting after adjustment for income and education, is an inequity, because it is patterned by social disadvantage and racism, and it is remediable. The distinction is not word games; it decides whether a gap is a policy failure or a fact of biology, and that decision determines whether anyone is accountable for it.

Key idea: A health inequality is any difference between groups, while a health inequity is a difference that is unfair, avoidable, remediable, and systematically patterned by social advantage, and only the second implies that someone is accountable.

The arithmetic that averages hide

Work this example carefully, because it explains a large share of global health reporting failures. A country reports under-5 mortality falling from 80 to 50 per 1,000 live births over a decade. Excellent news. Now stratify by wealth quintile. The richest quintile fell from 40 to 20. The poorest fell from 120 to 95.

Compute two things. The absolute gap between richest and poorest was 80 per 1,000 and is now 75 per 1,000, so it narrowed slightly. The ratio was 120 divided by 40, which is 3.0, and is now 95 divided by 20, which is 4.75, so relative inequality worsened substantially. The national average improved by 37 percent while the poorest quintile improved by 21 percent and the richest by 50 percent.

Every one of those statements is true, and a report can be written that emphasizes any of them. This is why serious equity analysis always reports both absolute and relative measures, and why WHO's Health Equity Assessment Toolkit exists. More formal summary measures, the slope index of inequality for absolute gaps across the whole distribution and the concentration index for how a health outcome is distributed across a ranked population, generalize the same idea beyond two groups.

There is also a pattern behind this arithmetic with a name: inverse equity. New health interventions typically reach advantaged groups first, which widens relative inequality in the early years, and only narrow it later as coverage spreads. That is an argument for designing delivery for the hardest-to-reach first, not an argument against the intervention.

Key idea: A national average can improve while relative inequality worsens, so equity analysis must report absolute gaps, relative ratios, and the change in each, and new interventions commonly widen relative gaps before narrowing them.

Where the gradient comes from

The Commission on Social Determinants of Health set out the framework the field still uses. Structural determinants, meaning governance, macroeconomic policy, social policy, and cultural and social norms, distribute people into socioeconomic positions defined by income, education, occupation, gender, race and ethnicity. Those positions then determine intermediary determinants: material circumstances such as housing and food, behaviors, psychosocial stressors, and access to and quality of health care. Health care sits near the end of that chain, which is why the Commission's recommendations were mostly not about health care.

How much of the gradient is material and how much is psychosocial remains genuinely contested. Marmot's line of work emphasizes control and status, drawing on the Whitehall finding that low job control predicted disease independently of income. Others argue that material conditions do most of the work and that the psychosocial account risks moving attention away from money and power. You do not have to resolve this, but you should notice that different answers imply very different policies: cash transfers and housing on one hand, workplace redesign and social participation on the other.

Key idea: The social determinants framework runs from structural drivers such as policy and social norms, through socioeconomic position, to material, behavioral, psychosocial and health-care pathways, which is why the strongest levers on health inequity usually sit outside the health sector.

Common misconceptions

• The epidemiologic transition is a law that every country follows in order. It reverses under shocks, runs in different sequences, and in most of the world produces a simultaneous double burden.
• Rich countries have chronic disease and poor countries have infections. Most low- and middle-income countries now carry heavy burdens of both, and the majority of NCD deaths occur in them.
• Health follows income, so growth is the answer. The Preston curve flattens at modest incomes and has shifted upward over time, meaning most gains came from knowledge and technology rather than income.
• Health inequality and health inequity mean the same thing. Inequity adds the judgments that a difference is unfair, avoidable, remediable, and systematically patterned.
• If a national average improves, everyone is better off. Averages can improve while relative gaps widen, which is why absolute and relative measures must both be reported.

Recap

• Omran described three ages of mortality, with later authors adding delayed degenerative disease and re-emerging infections.
• Transitions reverse, as HIV in southern Africa, post-Soviet Russia, and COVID-19 all demonstrated.
• Frenk's protracted and polarized model fits much of the world better: both burdens at once, and different social groups at different stages.
• The Preston curve is steep at low incomes, flat thereafter, and shifts upward over time, showing that technology and knowledge drive most gains.
• Marmot's Whitehall studies established a continuous social gradient in health, not a simple poor-versus-rich divide.
• Inequity means unfair, avoidable, remediable, systematic difference, and measuring it requires both absolute and relative comparisons.

Sources

1. Omran, A. R. (1971/2005). The epidemiologic transition: A theory of the epidemiology of population change. The Milbank Quarterly, 83(4), 731-757. doi.org/10.1111/j.1468-0009.2005.00398.x
2. Chetty, R., Stepner, M., Abraham, S., et al. (2016). The association between income and life expectancy in the United States, 2001-2014. JAMA, 315(16), 1750-1766. doi.org/10.1001/jama.2016.4226
3. World Health Organization. (n.d.). Social determinants of health. who.int
4. World Health Organization. (n.d.). Health equity. who.int
5. Wikipedia contributors. (n.d.). Preston curve. en.wikipedia.org

The big picture

Three diseases got their own multibillion-dollar global fund. It is worth asking why these three, because the answer is not simply that they kill the most people. Together HIV, tuberculosis and malaria caused roughly 2.5 million deaths in recent years, against about 20 million from cardiovascular disease alone. They were selected because around 2000 they were killing working-age adults at scale, concentrated in the poorest countries, spreading, and, critically, treatable with tools that existed but were not reaching anyone.

What followed is the largest deliberate mortality reduction effort in human history, and it worked. HIV deaths have fallen by roughly two thirds from their peak. Malaria deaths have fallen substantially since 2000. TB treatment has saved tens of millions of lives. It is also true that all three programs have stalled well short of their targets, that resistance is eroding the tools, and that in 2025 the funding underneath them contracted sharply. This lesson gives you the biology, the numbers, the program history, and the honest state of play for each.

Watch for one theme throughout. In each of these three stories, the decisive breakthroughs came at least as much from low- and middle-income country actors as from donors: Indian generic manufacturers, South African activists, Chinese pharmacologists, Brazilian health ministries. That is not a courtesy observation. It is the actual causal history.

Key idea: HIV, TB and malaria were targeted because they killed working-age adults at scale in the poorest countries with tools that existed but were not reaching people, and the resulting programs produced large measurable gains that have since stalled short of their targets.

HIV: the disease that made global health

HIV is a retrovirus that infects CD4 T cells, the coordinating cells of the immune system. It transmits through unprotected sex, blood, shared injecting equipment, and from mother to child during pregnancy, delivery or breastfeeding. Untreated, it depletes CD4 cells over roughly eight to ten years until the immune system can no longer control ordinary organisms, at which point opportunistic infections such as tuberculosis, cryptococcal meningitis and pneumocystis pneumonia cause death. That progression is what AIDS names.

The current numbers, from UNAIDS with the year attached because they move: about 40.8 million people were living with HIV in 2024, with roughly 1.3 million new infections and 630,000 AIDS-related deaths, against a peak near 2.1 million deaths around 2004. Roughly 31 million were receiving antiretroviral therapy. Eastern and southern Africa remains the epicentre with about half of all people living with HIV, and there the epidemic is markedly gendered, with women and girls accounting for the majority of new infections and adolescent girls carrying a disproportionate share. Globally, key populations, meaning sex workers, men who have sex with men, people who inject drugs, transgender people and prisoners, together with their partners, account for most new infections, which is why criminalization and stigma are epidemiological variables rather than side issues.

Treatment transformed everything. Combination antiretroviral therapy arrived in 1996 and turned HIV into a manageable chronic condition with near-normal life expectancy for people who start early and stay on it. Two further findings reshaped prevention. Treatment is prevention: a person with a sustained undetectable viral load does not transmit HIV sexually, the finding known as U equals U, established by the PARTNER studies. And pre-exposure prophylaxis works. WHO recommended daily oral PrEP in 2015, two-monthly injectable cabotegravir followed, and in 2025 WHO recommended twice-yearly injectable lenacapavir after trials showed extremely high efficacy, including one in young women in South Africa and Uganda with no infections in the treatment arm. A twice-yearly injection is a different prevention proposition from a daily pill, and the open question is price and delivery rather than efficacy.

UNAIDS frames progress with the 95-95-95 targets: 95 percent of people with HIV diagnosed, 95 percent of those on treatment, and 95 percent of those virally suppressed. As of 2023 the global figures were roughly 86, 77 and 72 percent when calculated across all people living with HIV. That cascade framing is useful because it localizes the failure: a country losing people at the diagnosis step needs different action from one losing them at retention.

Key idea: HIV is now a treatable chronic condition, treatment also prevents transmission, and long-acting prophylaxis has arrived, so the remaining problem is overwhelmingly one of reaching people, especially criminalized and stigmatized populations, rather than of biomedical capability.

How the price of a life fell from ten thousand dollars to a hundred

This is the most important political story in modern global health, and it was not driven by donors.

In 2000, antiretroviral therapy cost roughly 10,000 to 15,000 US dollars per patient per year at patented prices. At Sub-Saharan African health budgets of a few dollars per person per year, the arithmetic said that treating people was impossible, and a great many respectable officials said exactly that in public. Brazil disagreed first, committing in 1996 to provide antiretrovirals free through its public system, manufacturing locally through public laboratories and using the credible threat of compulsory licensing to force price reductions from patent holders. Then in 2001 the Indian generic manufacturer Cipla offered a triple-therapy regimen at 350 dollars per patient per year, and later less. India's patent law at the time did not grant product patents on pharmaceuticals, which is why Indian firms could do this legally.

In South Africa, the Treatment Action Campaign turned the price question into a legal and moral confrontation, joining the government's defense when 39 pharmaceutical companies sued over medicines legislation in 2001. The companies withdrew. TAC then sued its own government to compel provision of nevirapine to prevent mother-to-child transmission, and won in the Constitutional Court in 2002. The South African government's own AIDS denialism under President Thabo Mbeki, which delayed treatment rollout for years, has been estimated in peer-reviewed work to have cost over 300,000 lives. That episode is essential to the honest version of this story: the obstacles were not only external.

The pieces then combined. The Global Fund arrived in 2002 and PEPFAR in 2003 with money to buy the now-cheap drugs. WHO's 3 by 5 initiative set a public target of three million people on treatment by 2005. The Medicines Patent Pool, established in 2010, institutionalized voluntary licensing so that generic manufacturers could produce newer patented medicines for low- and middle-income markets. First-line antiretroviral regimens now cost on the order of tens of dollars per person per year.

Key idea: The collapse in antiretroviral prices from over 10,000 dollars to under 100 dollars a year came from Brazilian state manufacturing, Indian generic competition under a permissive patent regime, and South African activist litigation, with donor money arriving to purchase drugs that others had already made affordable.

Tuberculosis: the oldest and most neglected

Mycobacterium tuberculosis spreads through the air when someone with active pulmonary disease coughs. WHO's standard estimate is that about a quarter of the world's population carries latent infection, of whom roughly 5 to 10 percent will develop active disease in their lifetime; that risk rises enormously with HIV co-infection, diabetes, malnutrition, silicosis and smoking. TB is a disease of crowding, poor ventilation and undernutrition, which is why it tracks poverty so tightly and why it fell steeply in Europe before any drug existed.

In 2023, about 10.8 million people fell ill with TB and about 1.25 million died, which returned TB to its position as the world's leading infectious cause of death, ahead of COVID-19. Eight countries account for roughly two thirds of cases: India, Indonesia, China, the Philippines, Pakistan, Nigeria, Bangladesh and the Democratic Republic of the Congo. Around 2.7 million of those 10.8 million people were never diagnosed or reported, which is the central operational problem: the missing millions.

Diagnosis has been the bottleneck for a century. Sputum smear microscopy, the standard tool in much of the world until recently, is nineteenth-century technology that misses roughly half of cases and cannot detect drug resistance. Molecular testing changed this: the Xpert MTB/RIF assay, endorsed by WHO in 2010, gives a result in under two hours and simultaneously detects rifampicin resistance. Computer-aided reading of chest X-rays has since been recommended for triage. The obstacle is not knowledge; it is that molecular tests cost more per test than microscopy and require reliable power and cartridge supply chains.

Treatment for drug-sensitive TB takes six months of four then two drugs, is highly effective, and is difficult to complete. The DOTS strategy of the 1990s built its whole architecture around directly observed therapy for that reason. Drug-resistant TB is the harder problem. About 400,000 people developed multidrug-resistant or rifampicin-resistant TB in 2023, and fewer than half were diagnosed and treated. Until recently, treating them meant 18 to 24 months of toxic drugs including daily injections that caused permanent deafness in a substantial minority, with cure rates around 60 percent. The BPaLM regimen, four oral drugs for six months, was recommended by WHO in 2022 and achieves success rates near 90 percent. That is one of the largest clinical advances in this entire course and it is still not universally available.

Prevention is the weak point. BCG, licensed in 1921, is the most widely administered vaccine in the world and protects infants against disseminated TB and TB meningitis, but it does not reliably prevent adult pulmonary disease, which is the form that transmits. The M72/AS01E candidate entered phase 3 trials in 2024 and is the first serious prospect of a replacement in a century. TB research remains underfunded relative to its burden by any measure, for a reason worth naming plainly: it is a disease of poor people in poor countries, with no wealthy market to pay for innovation.

Key idea: TB is again the leading infectious killer, its central failures are the millions never diagnosed and the drug-resistant cases never treated, and although the six-month oral BPaLM regimen transformed resistant-TB treatment in 2022, prevention still rests on a vaccine from 1921 that does not stop adult transmission.

Malaria: the arms race

Malaria is caused by Plasmodium parasites carried by female Anopheles mosquitoes. Plasmodium falciparum causes almost all severe disease and death and dominates in Africa; Plasmodium vivax, dominant in much of Asia and Latin America, forms dormant liver stages that cause relapses months later and needs a different cure.

The World Malaria Report for 2023 counted an estimated 263 million cases and 597,000 deaths across 83 countries. The concentration is extreme: the African region accounts for about 94 percent of cases and 95 percent of deaths, children under five for roughly three quarters of African deaths, and Nigeria alone for around a quarter of global deaths. Deaths fell substantially from about 2000 through the mid-2010s and have since plateaued, which is the fact that should worry you most in this section.

The toolkit is unusually well evidenced. Insecticide-treated nets are the single most cost-effective intervention in this course by most estimates. Indoor residual spraying suppresses transmission where housing and logistics allow. Rapid diagnostic tests moved treatment from presumptive to confirmed. Artemisinin-based combination therapies cure quickly and are the backbone of case management. Seasonal malaria chemoprevention now protects tens of millions of children across the Sahel during transmission peaks, and intermittent preventive treatment in pregnancy protects mothers and birthweight.

Artemisinin deserves a paragraph of its own, because it is a story global health tells badly. It came from Project 523, a secret Chinese military research programme begun in 1967. The pharmacologist Tu Youyou and her team screened traditional Chinese medical texts, identified Artemisia annua, and found that a low-temperature extraction preserved the active compound that boiling destroyed. She won the Nobel Prize in 2015. The drug that anchors global malaria treatment was developed in China during the Cultural Revolution, entirely outside the Western pharmaceutical system.

Now the arms race. Pyrethroid resistance in Anopheles is widespread, which is why new dual-active-ingredient nets combining a pyrethroid with chlorfenapyr or pyriproxyfen have been rolled out after trials showed better protection. Partial artemisinin resistance emerged in the Greater Mekong around 2008 and has now been confirmed independently in East Africa, including Rwanda, Uganda, Eritrea and Tanzania, which is the single most dangerous development in malaria control. Deletions of the pfhrp2 and pfhrp3 genes cause the most common rapid diagnostic tests to return false negatives, and have spread far enough in the Horn of Africa to force test switching. And Anopheles stephensi, an efficient urban-breeding vector from South Asia, has invaded Djibouti, Ethiopia, Sudan, Nigeria, Ghana and Kenya, raising the prospect of epidemic malaria in African cities that have never had much of it.

Against that, two vaccines finally exist. RTS,S was recommended by WHO in 2021 after pilot implementation in Ghana, Kenya and Malawi; R21 with Matrix-M adjuvant followed in 2023, developed at Oxford and manufactured at very large scale and low cost by the Serum Institute of India. Routine rollout began in Cameroon in January 2024 and has since extended across many African countries. Efficacy is moderate rather than sterilizing, roughly comparable to what a good bed net delivers, but a moderately effective vaccine deployed to tens of millions of children is a large absolute gain. Elimination is also real: Sri Lanka was certified malaria-free in 2016, China in 2021 after eliminating a disease that once caused 30 million cases a year, and Cabo Verde and Egypt in 2024.

Key idea: Malaria deaths fell sharply after 2000 and have since plateaued, artemisinin came from Chinese research outside the Western system, and progress is now threatened simultaneously by pyrethroid resistance, artemisinin partial resistance in East Africa, diagnostic-evading parasites, and an invasive urban vector, while two moderately effective vaccines have finally reached routine use.

What the three programs teach

Four transferable lessons. Price is politics: nothing about the biology of HIV changed between 2000 and 2003, but the price fell by 97 percent because of legal regimes, generic competition and activism, so when you meet an unaffordable intervention, treat the price as a variable rather than a constraint. Vertical works and vertical costs: concentrating money and accountability on one problem delivered results and built parallel systems, salary scales and supply chains that weakened general services. Communities deliver: community health workers, peer navigators and treatment literacy activists did much of the work clinics could not. And funding is political: the 2025 contraction in United States support showed that programs dependent on a foreign legislature's annual appropriation are structurally fragile however good their evidence.

Key idea: The big three programs show that drug prices are political rather than fixed, that vertical concentration buys speed at the cost of system integration, that community delivery is often decisive, and that donor-financed programs remain hostage to foreign domestic politics.

Common misconceptions

• HIV is still a death sentence. With early and sustained treatment, life expectancy approaches normal, and a person with an undetectable viral load does not transmit sexually.
• Donors made HIV drugs affordable. Prices collapsed because of Brazilian state manufacturing, Indian generic competition, and South African litigation; donor money then bought the cheaper drugs.
• TB is a disease of the past. It killed about 1.25 million people in 2023 and is again the world's leading infectious cause of death.
• BCG protects adults from tuberculosis. It protects infants against severe disseminated disease but does not reliably prevent the adult pulmonary disease that transmits.
• A malaria vaccine will end malaria. RTS,S and R21 offer moderate protection comparable to a good bed net, valuable at scale but not a replacement for vector control and treatment.

Recap

• About 40.8 million people were living with HIV in 2024, with 1.3 million new infections, 630,000 deaths, and roughly 31 million on treatment.
• Treatment is prevention, and long-acting injectable prophylaxis, recommended by WHO in 2025, changes what prevention can look like.
• Antiretroviral prices fell from over 10,000 dollars a year to under 100 through generic competition, compulsory licensing threats, and activism.
• TB caused about 10.8 million illnesses and 1.25 million deaths in 2023, with 2.7 million people never diagnosed and most drug-resistant cases untreated.
• The six-month all-oral BPaLM regimen raised drug-resistant TB cure rates from around 60 percent toward 90 percent.
• Malaria caused 263 million cases and 597,000 deaths in 2023, deaths have plateaued, and resistance in both parasite and mosquito now threatens the toolkit.

Sources

1. UNAIDS. (n.d.). Global HIV and AIDS statistics fact sheet. unaids.org
2. World Health Organization. (n.d.). Tuberculosis (fact sheet). who.int
3. World Health Organization. (n.d.). Malaria (fact sheet). who.int
4. Centers for Disease Control and Prevention. (n.d.). Global HIV and TB. cdc.gov
5. Wikipedia contributors. (n.d.). Artemisinin. en.wikipedia.org

The big picture

In 1974, when WHO launched the Expanded Programme on Immunization, fewer than 5 percent of the world's children were vaccinated against the six diseases it targeted. Today about 84 percent receive three doses of diphtheria-tetanus-pertussis vaccine. Measles vaccination alone is estimated to have prevented roughly 60 million deaths between 2000 and 2023. If you are looking for the single largest health return on investment in this course, it is here.

And yet. About 14.5 million children in 2023 received not a single routine vaccine, the so-called zero-dose children. Measles caused an estimated 10.3 million cases and about 107,500 deaths in 2023, almost all in unvaccinated children, in a world that has had a highly effective measles vaccine since 1963. Polio was supposed to be eradicated by the year 2000 and still is not.

This lesson uses the two eradication campaigns as the clearest teaching cases in global health. Smallpox is the field's greatest success and it explains what has to be true for eradication to work. Polio is the field's longest-running unfinished project and it explains what happens when biology, security, and trust each turn out to be harder than expected. The comparison teaches more than either story alone.

Key idea: Immunization is the highest-return intervention in global health, coverage rose from under 5 percent to about 84 percent since 1974, and the two eradication campaigns for smallpox and polio between them explain both what makes eradication possible and what makes it fail.

Herd immunity, worked

You need one piece of arithmetic to read this whole field. The basic reproduction number, written R nought, is the average number of new infections caused by one infectious person in a fully susceptible population. If R nought is greater than 1, an epidemic grows; if it is less than 1, it dies out.

Vaccination reduces the pool of susceptible people. The fraction of the population that must be immune to push effective transmission below 1 is 1 minus 1 divided by R nought. Work two cases. Measles has an R nought usually estimated between 12 and 18, so with R nought of 15 the threshold is 1 minus 1 divided by 15, which is 0.933, or about 93 percent. Polio has an R nought usually estimated between 5 and 7, so with R nought of 6 the threshold is 1 minus 1 divided by 6, which is 0.833, or about 83 percent.

Two consequences follow immediately. First, measles is the most demanding vaccine-preventable disease in the world; it requires more than 90 percent coverage with two doses, so it is the first disease to come back when a system falters. Measles outbreaks are therefore the standard early-warning signal that a country's routine immunization has weakened. Second, thresholds are population averages, and averages hide clusters. A country at 95 percent national coverage with one district at 60 percent will have outbreaks in that district, because transmission is local.

Also note what vaccine efficacy does to the arithmetic. If a vaccine is 90 percent effective and the herd immunity threshold is 93 percent, you must vaccinate 93 divided by 0.9, which is more than 100 percent of the population, meaning the threshold cannot be reached with one dose. That is exactly why measles requires two doses.

Key idea: The herd immunity threshold equals 1 minus 1 divided by R nought, giving about 93 percent for measles and about 83 percent for polio, and because coverage clusters geographically, high national averages routinely conceal districts below threshold.

Smallpox: how it was actually done

Smallpox killed an estimated 300 million people in the twentieth century alone, with a case fatality around 30 percent for variola major and permanent scarring or blindness for many survivors. It is the only human disease ever eradicated.

The campaign is often told as a triumph of Western technical assistance, and that telling is wrong in instructive ways. The proposal came from the Soviet Union: Viktor Zhdanov, deputy health minister, put it to the World Health Assembly in 1958, and the USSR supplied hundreds of millions of doses. The Intensified Eradication Programme launched in 1967 under D. A. Henderson with a budget most people considered inadequate. And the actual work, the house-to-house searching in Indian villages, Bangladeshi river districts, Ethiopian highlands and Somali towns, was done overwhelmingly by national health workers and local volunteers, tens of thousands of them, whose names are not in most accounts.

Two innovations mattered more than the vaccine itself. The bifurcated needle, developed by Benjamin Rubin, held a precise dose of vaccine between its two prongs, could be boiled and reused, required minimal training, and cut vaccine consumption by about three quarters. And surveillance and containment, sometimes called ring vaccination, replaced the original plan of mass vaccination. William Foege, working in eastern Nigeria in 1966 with insufficient vaccine, decided to find every case and vaccinate everyone around it rather than everyone in the region. It worked far better and far cheaper than blanket coverage, because it targeted the actual chains of transmission.

The last naturally occurring case of variola major was in a Bangladeshi child, Rahima Banu, in 1975; the last of variola minor was Ali Maow Maalin, a hospital cook in Merca, Somalia, in October 1977, who survived. One further death followed: Janet Parker, a medical photographer in Birmingham, England, died in 1978 after a laboratory escape, a reminder that stored pathogens are a residual risk. Eradication was certified in 1979 and declared by the World Health Assembly in May 1980.

Key idea: Smallpox eradication was proposed by the Soviet Union, executed largely by national health workers in affected countries, and made feasible by the bifurcated needle and by surveillance and containment, which found and encircled cases rather than trying to vaccinate everyone.

Why smallpox and not everything else

Smallpox had a rare combination of properties, and the standard criteria for eradicability, set out most clearly by Walter Dowdle, fall into three groups: biological feasibility, cost and benefit, and political commitment.

Property	Smallpox	Polio	Measles
Animal reservoir	None	None	None
Silent infection	Essentially none; every case visible	About 199 of 200 infections cause no paralysis	Rare; nearly all infections symptomatic
Diagnosis in the field	Distinctive rash, recognizable from a photograph	Requires stool culture and sequencing	Rash plus fever, fairly recognizable
Vaccine	Heat-stable, one dose, lifelong immunity	Multiple doses; live oral vaccine can revert	Two doses; needs cold chain
Transmissibility (R nought)	About 5 to 7	About 5 to 7	About 12 to 18

Read the silent infection row twice, because it is the single most important difference in this lesson. Every smallpox infection announced itself with an unmistakable rash, so finding cases meant finding transmission. In polio, only about one infection in 200 produces paralysis, so a single reported case means hundreds of silent infections already circulating, and the absence of cases never proves the absence of virus. That one biological fact explains most of the difference between a campaign that finished in thirteen years and one that has run for thirty-seven.

Key idea: Eradication requires no animal reservoir, a case that can be recognized in the field, and an effective vaccine, and polio's crucial disadvantage is that roughly 199 of every 200 infections are silent, so surveillance can never confirm the virus is gone.

Polio: the campaign that will not end

When the Global Polio Eradication Initiative launched in 1988, polio paralyzed about 350,000 children a year across 125 countries. Wild poliovirus type 2 was certified eradicated in 2015 and type 3 in 2019. Wild type 1 now circulates endemically in only two countries, Afghanistan and Pakistan. That is more than 99.9 percent of the way there, and the last fraction has consumed three decades and roughly 20 billion dollars.

Case counts bounce year to year in ways that make trend-reading hazardous: single digits in 2021, around 30 in 2022, roughly a dozen in 2023, and close to 99 in 2024, which is why you should always check the current figure rather than relying on the last one you read. Four obstacles explain the difficulty.

The first is the silent transmission problem already described. The second is the vaccine itself, and it is genuinely paradoxical. Oral polio vaccine, developed by Albert Sabin, uses live attenuated virus. It is cheap, needs no needle, can be given by volunteers, and produces gut immunity that blocks transmission, which is why it, rather than Jonas Salk's injected inactivated vaccine, was chosen for eradication. But the attenuated virus replicates in the gut and is shed, and in under-immunized populations it can circulate long enough to regain neurovirulence. The result is circulating vaccine-derived poliovirus, which now causes more paralytic cases in most years than wild virus does. A novel oral vaccine against type 2, engineered to be genetically more stable, has been deployed since 2021 to reduce this. Understand the position clearly: the campaign must now eradicate both the wild virus and the virus its own principal tool creates.

The third obstacle is insecurity and mistrust, and here global health has to look at its own conduct. In 2003 and 2004, several northern Nigerian states boycotted polio vaccination for about a year following claims that the vaccine caused infertility; virus from that period seeded outbreaks across more than a dozen countries. In Pakistan and Afghanistan, vaccinators and their police escorts have been killed repeatedly. That violence has many causes, but one of them is documented and self-inflicted: in 2011 the United States Central Intelligence Agency ran a fake hepatitis B vaccination campaign in Abbottabad, Pakistan, as a cover for collecting DNA to confirm the location of Osama bin Laden. The operation gave every conspiracy theory about vaccination campaigns a true story to point at. The White House pledged in 2014 that the agency would not use immunization programs for intelligence again. The damage was not reversible on that timescale.

The fourth is the endgame economics. The target date has moved from 2000 to 2005, 2012, 2018, 2023, and beyond. Each extension consumes resources that could go elsewhere, and some economists argue the money would now save more lives spent on routine immunization or on measles. The counterargument is that eradication is the only permanently free outcome, that stopping means permanent vaccination costs and eventual resurgence, and that the marginal cost of finishing is small against what has been spent. Both positions are defensible, which is exactly why the debate has not resolved.

Key idea: Polio eradication is more than 99.9 percent complete after 37 years and 20 billion dollars, and its remaining obstacles are silent transmission, vaccine-derived virus generated by the campaign's own principal tool, insecurity compounded by the CIA's 2011 fake vaccination operation, and an unresolved argument about whether finishing is worth the marginal cost.

Guinea worm: eradication without a vaccine

Dracunculiasis is worth a paragraph because it breaks the assumption that eradication needs a biomedical tool. There is no vaccine and no drug. The parasite is acquired by drinking water containing infected copepods, and a metre-long worm emerges through the skin about a year later.

The campaign, led by the Carter Center with ministries of health, works entirely through behavior and simple technology: cloth filters, pipe filters worn around the neck, larvicide in water sources, health education, and above all case containment, meaning keeping an infected person out of the water source while the worm emerges. Village volunteers do nearly all of it. Cases fell from about 3.5 million a year in 1986 across 21 countries to roughly a dozen to fifteen human cases a year recently. The final obstacle was unexpected: infections in dogs, particularly in Chad, created an animal reservoir that the original eradicability assessment did not anticipate.

Key idea: Guinea worm has been driven from 3.5 million cases a year to a handful using only filtration, education and case containment delivered by village volunteers, and its unexpected obstacle, infection in dogs, shows that eradicability assessments can be wrong about reservoirs.

The coverage problem underneath everything

Global immunization coverage fell during COVID-19 in the largest sustained decline in about thirty years, and recovery has been incomplete. The 14.5 million zero-dose children of 2023 are not randomly distributed: they are concentrated in conflict zones, urban slums, remote rural areas, and among migrant and marginalized populations, which means the last children to be reached are the hardest and most expensive to reach.

One correction is needed here. In high-income countries the dominant discussion is vaccine hesitancy, and it is a real and growing problem. Globally it is not the main driver of under-vaccination. Most unvaccinated children live in places where a vaccinator has not reliably come, where the clinic is far, where the cold chain fails, or where fighting makes outreach impossible. Access, supply and conflict explain more of the world's missed children than refusal does, and confusing the two leads to campaigns that address the wrong problem.

Key idea: Coverage fell sharply during the pandemic and has not fully recovered, and globally the main causes of under-vaccination are access, supply and conflict rather than refusal, even though hesitancy dominates the conversation in wealthy countries.

Common misconceptions

• Smallpox was eradicated by mass vaccination of everyone. Blanket vaccination was abandoned for surveillance and containment, which found cases and vaccinated the rings around them.
• Smallpox eradication was a Western gift. It was proposed by the Soviet Union, vaccine came largely from the USSR, and the field work was done overwhelmingly by health workers in affected countries.
• Polio is nearly gone, so it must be nearly over. Because only about one infection in 200 causes paralysis, absence of cases never proves absence of virus, and vaccine-derived poliovirus now causes most paralytic cases.
• The oral polio vaccine is simply better than the injected one. It gives superior gut immunity and needs no needle, and it can revert to neurovirulence, which is why the endgame requires switching away from it.
• Unvaccinated children are mostly the children of refusers. Globally, access, supply failure and conflict account for far more missed children than hesitancy does.

Recap

• The herd immunity threshold is 1 minus 1 divided by R nought, about 93 percent for measles and 83 percent for polio.
• Smallpox was eradicable because every infection was visible, there was no animal reservoir, and one dose of a heat-stable vaccine gave lasting immunity.
• The bifurcated needle and surveillance and containment, not blanket vaccination, delivered eradication by 1980.
• Polio's 199-in-200 silent infection rate makes confirming absence nearly impossible, and vaccine-derived poliovirus now causes most paralytic cases.
• Insecurity, the 2003-2004 northern Nigeria boycott, and the CIA's 2011 fake vaccination operation in Pakistan all set the campaign back.
• Guinea worm has fallen from 3.5 million cases to a handful using filtration, education and containment, with no vaccine or drug at all.

Sources

1. World Health Organization. (n.d.). Immunization coverage (fact sheet). who.int
2. World Health Organization. (n.d.). Measles (fact sheet). who.int
3. Global Polio Eradication Initiative. (n.d.). This week: polio now. polioeradication.org
4. The Carter Center. (n.d.). Guinea worm eradication program. cartercenter.org
5. Encyclopaedia Britannica. (n.d.). Smallpox. britannica.com

The big picture

This lesson covers three subjects that never make the front page and that determine more of the world's health than most things that do. Neglected tropical diseases disable more than a billion people. Water and sanitation sit underneath a large fraction of the infectious burden in this entire module. Antimicrobial resistance quietly removes the tools that all the rest of medicine depends on.

They share a structural property worth naming at the start. Each is a problem of the poorest people in the poorest places, each is unglamorous, and each requires infrastructure and systems rather than a single purchasable product. That combination is exactly what the global health funding architecture in Lesson 2 is worst at handling. When you finish this lesson, you will understand why these three topics keep appearing in strategy documents and keep not being solved.

Key idea: Neglected tropical diseases, water and sanitation, and antimicrobial resistance are the unglamorous foundations of infectious disease control, and all three are structurally disadvantaged by a funding system that rewards discrete measurable products.

What makes a disease neglected

WHO designates about 21 conditions as neglected tropical diseases, and roughly 1.5 billion people require treatment or care for at least one of them. The list includes lymphatic filariasis, onchocerciasis or river blindness, schistosomiasis, soil-transmitted helminths, trachoma, Chagas disease, leishmaniasis, human African trypanosomiasis, dengue, rabies, leprosy, Buruli ulcer, and more recently snakebite envenoming, added in 2017, and noma, added in 2023.

Neglected is a description of attention, not of biology. Three things produce it. First, most of these conditions disable rather than kill, so in a world that counted deaths they were nearly invisible; it took the DALY, and the years lived with disability half of it, to make them countable. Second, they occur almost exclusively among people with no purchasing power, so there is no commercial market and no research incentive. Third, they are diseases of stigma and disfigurement, which suppresses reporting and political demand. Someone with lymphatic filariasis or noma is often hidden by their family.

The workhorse tool is preventive chemotherapy, usually called mass drug administration: give a safe drug to everyone in an endemic community at regular intervals rather than testing and treating individuals. It works because the drugs are safe enough to give to people who may not be infected, and because treating everyone breaks transmission in a way that treating the diagnosed does not. At its peak the approach has reached roughly a billion people a year.

The drugs are largely donated, and this is a genuine and underappreciated corporate contribution. Merck has donated ivermectin for river blindness since 1987 with no end date, through the Mectizan Donation Program. GSK donates albendazole, Pfizer donates azithromycin for trachoma. Donation solves the price problem entirely and creates a different one: a global program whose central input depends on the continued goodwill of private firms, with no contractual guarantee and no fallback.

Progress is real and should be stated with dates. The London Declaration of 2012 aligned companies, donors and endemic countries behind elimination targets; the Kigali Declaration of 2022 renewed them alongside WHO's 2021-2030 road map. More than fifty countries have now eliminated at least one NTD, with India certified free of trachoma as a public health problem in 2024, and guinea worm nearly gone as covered in the previous lesson.

The critique deserves equal weight. Mass drug administration treats infection without touching the conditions that cause reinfection, so a village can be dewormed annually forever without its water improving. Trachoma control makes this explicit: the SAFE strategy stands for Surgery, Antibiotics, Facial cleanliness and Environmental improvement, and two of the four letters are water and sanitation. A drug-centric program is cheaper to run and easier to count, which is precisely why it is often what gets funded.

Key idea: Neglected tropical diseases affect about 1.5 billion people and became visible only when non-fatal burden became countable, mass drug administration with donated medicines has driven elimination in over fifty countries, and it treats infection without changing the water and sanitation conditions that cause reinfection.

Water, sanitation and hygiene: the ladders

The WHO and UNICEF Joint Monitoring Programme measures WASH using service ladders rather than a yes-or-no question, and understanding the rungs is essential to reading any statistic in this area.

Rung	Drinking water	Sanitation
Safely managed	Improved source on premises, available when needed, free from contamination	Improved facility not shared, with excreta safely disposed or treated
Basic	Improved source within 30 minutes round trip	Improved facility not shared with other households
Limited	Improved source over 30 minutes round trip	Improved facility shared between households
Unimproved	Unprotected well or spring	Pit latrine without slab, hanging or bucket latrine
No service	Surface water	Open defecation

The distinction between basic and safely managed is where most of the argument lives. A village can be reported as having improved water because a borehole exists, while the water is contaminated, the pump is broken half the year, and the round trip takes an hour. As of the 2022 estimates published by the Joint Monitoring Programme, roughly 2.2 billion people lacked safely managed drinking water, about 3.5 billion lacked safely managed sanitation, some 2 billion lacked a basic handwashing facility with soap and water at home, and around 419 million people still practised open defecation.

The health stakes are concrete. Diarrhoeal disease still kills on the order of 450,000 children under five each year, and it also drives undernutrition, which increases the severity of every other infection, a loop that makes attribution genuinely difficult.

Key idea: WASH is measured on service ladders from no service to safely managed, the gap between basic and safely managed is where most real deprivation sits, and around 2.2 billion people lacked safely managed drinking water as of 2022.

Oral rehydration: the intervention that came from Dhaka and Calcutta

This is the story to remember when anyone describes low-income countries as recipients of global health knowledge.

Cholera kills through dehydration. Until the 1960s the only reliable treatment was intravenous fluid, which requires sterile equipment, trained staff and a clinic, none of which exist in a village or a refugee camp. The breakthrough came from physiology: researchers established that the coupled transport of sodium and glucose across the intestinal wall remains intact during cholera infection, so a solution containing both salt and sugar in the right proportions is absorbed even when the gut is pouring out fluid. The work was done at the Cholera Research Laboratory in Dhaka, now icddr,b, and at institutions in Calcutta, by teams including David Nalin, Richard Cash, Norbert Hirschhorn and Bangladeshi and Indian colleagues.

The decisive demonstration came in 1971. During the Bangladesh Liberation War, cholera swept refugee camps across the Indian border and intravenous fluid ran out. Dilip Mahalanabis, working at the Bangaon camp, mixed oral rehydration solution in drums and had families administer it themselves. Case fatality fell from roughly 30 percent to around 3 percent. In 1978 The Lancet described oral rehydration therapy as potentially the most important medical advance of the century, and the claim has held up.

Scale-up was also led locally. In the 1980s BRAC, the Bangladeshi organization that became the largest NGO in the world, sent workers door to door to teach mothers to mix the solution from household salt and sugar, reaching over ten million households and paying workers partly on the basis of whether the mother could still make it correctly weeks later. That is one of the largest health education programs ever conducted, designed and delivered by a Bangladeshi organization.

Key idea: Oral rehydration therapy, which cut cholera case fatality from about 30 percent to around 3 percent, was developed by researchers in Dhaka and Calcutta, proven in Bangladeshi refugee camps in 1971, and taken to national scale by BRAC's door-to-door teaching.

When the evidence complicated the story

Global health needs a habit of reporting the trials that did not work, so here is an important one. It was long assumed that household water treatment, sanitation and handwashing would reduce childhood stunting by reducing infection and gut damage. Three large randomized trials tested it directly: WASH Benefits in Bangladesh and Kenya, and SHINE in Zimbabwe, all reporting around 2018 and 2019.

The nutrition arms improved growth modestly. The WASH arms largely did not improve linear growth, and combining WASH with nutrition generally added nothing beyond nutrition alone. This was a genuine surprise and it prompted a serious rethink rather than a retreat. The leading interpretations are that household-level interventions cannot protect a child in a community where contamination is everywhere, that the service levels delivered were basic rather than safely managed, and that transmission routes such as animal faeces in the compound were not addressed.

What survives is worth stating carefully, because it is easy to overread this result. WASH still prevents diarrhoea and cholera and trachoma; the trials did not show otherwise. What they showed is that modest household improvements are not a sufficient tool against stunting, and that transformative, community-wide, higher-level services are probably what the job requires. That is an argument for more ambitious infrastructure, not for less WASH.

One influential approach, Community-Led Total Sanitation, was developed by Kamal Kar in Bangladesh in 1999 and works through collective behavior change rather than subsidized latrines, triggering communities to end open defecation themselves. It has produced rapid change in many settings and drawn serious criticism for tactics that rely on shame, which raises a real ethical question about how far a public health method may go in using social pressure.

Key idea: The WASH Benefits and SHINE trials found that household-level water, sanitation and hygiene interventions did not reduce childhood stunting, which shifted the field toward community-wide, higher-service-level approaches rather than away from WASH.

Antimicrobial resistance: two problems, not one

Bacteria acquire resistance through mutation and through horizontal transfer of resistance genes on plasmids, and antibiotic exposure selects for the resistant survivors. Nothing about that is new; Alexander Fleming warned about it in his 1945 Nobel lecture. What is new is the scale.

The Global Research on Antimicrobial Resistance study, published in The Lancet in 2022, estimated that in 2019 bacterial antimicrobial resistance was directly attributable to about 1.27 million deaths and associated with about 4.95 million. Later analyses from the same group estimated that more than a million deaths a year have been directly attributable in every year since 1990, and projected the figure could reach around 1.9 million directly attributable and 8.2 million associated by 2050. The highest death rates are in sub-Saharan Africa and South Asia, not in wealthy countries, which is the opposite of most public assumptions.

Now the point that most coverage misses entirely. In many low-income settings, more people die from lack of access to antibiotics than from resistance to them. A child with bacterial pneumonia who never receives amoxicillin dies of a fully treatable infection. Ramanan Laxminarayan and colleagues have argued for years that the framing must be access as well as excess, because a policy that only restricts antibiotic use, applied to a country where most people cannot get antibiotics at all, will kill more people than it saves.

The drivers are correspondingly varied: sale of antibiotics without prescription, use in livestock which accounts for a large share of global consumption by volume, weak infection prevention in hospitals, substandard and falsified medicines delivering sub-therapeutic doses, and the absence of diagnostics that would let a clinician distinguish bacterial from viral illness and prescribe accordingly. The last of these is a technology gap rather than a behavior problem.

The response has three parts. Surveillance, through WHO's Global Antimicrobial Resistance and Use Surveillance System and the priority pathogen lists first issued in 2017 and updated in 2024. Stewardship, meaning better prescribing, infection control, vaccination which prevents the infections that prompt antibiotic use, and clean water which does the same. And innovation, which is where the market fails hardest: a new antibiotic is used briefly, held in reserve, and priced low, so its commercial return is poor, and several companies that successfully developed new antibiotics went bankrupt afterward. Proposed fixes are pull incentives such as subscription payments that pay for availability rather than volume, piloted by the English health service. In September 2024 the UN General Assembly held a high-level meeting on AMR and adopted a political declaration including a target to reduce bacterial AMR deaths by 10 percent by 2030.

Key idea: Bacterial antimicrobial resistance was directly attributable to about 1.27 million deaths in 2019 with the highest rates in sub-Saharan Africa and South Asia, and because lack of access to antibiotics kills more people than resistance in many poor settings, any credible response must expand access while improving stewardship.

Common misconceptions

• Neglected tropical diseases are neglected because they are rare. They affect about 1.5 billion people; they are neglected because they disable rather than kill and occur among people with no purchasing power.
• Mass drug administration solves NTDs. It suppresses infection but does not change the water, sanitation and housing conditions that cause reinfection, which is why trachoma control includes environmental improvement.
• Access to an improved water source means safe water. The ladders distinguish basic from safely managed; a functioning borehole an hour away with contaminated water counts as improved and is not safe.
• The WASH trials proved water and sanitation do not matter. They found household-level interventions insufficient against stunting; WASH still prevents diarrhoeal disease, cholera and trachoma.
• Antimicrobial resistance is mainly a rich-country hospital problem. Death rates are highest in sub-Saharan Africa and South Asia, where lack of access to antibiotics is often the larger killer.

Recap

• WHO lists about 21 neglected tropical diseases affecting roughly 1.5 billion people, made visible only when non-fatal burden became measurable.
• Mass drug administration with donated medicines has helped more than fifty countries eliminate at least one NTD.
• WASH is measured on service ladders, and about 2.2 billion people lacked safely managed drinking water as of 2022.
• Oral rehydration therapy was developed in Dhaka and Calcutta, proven in 1971 refugee camps, and scaled by BRAC door to door.
• The WASH Benefits and SHINE trials found household WASH did not reduce stunting, redirecting the field toward community-wide services.
• AMR was directly attributable to about 1.27 million deaths in 2019, and lack of antibiotic access kills more than resistance in many low-income settings.

Sources

1. World Health Organization. (n.d.). Neglected tropical diseases. who.int
2. WHO/UNICEF Joint Monitoring Programme for Water Supply, Sanitation and Hygiene. (n.d.). JMP data and service ladders. washdata.org
3. Antimicrobial Resistance Collaborators. (2022). Global burden of bacterial antimicrobial resistance in 2019: A systematic analysis. The Lancet, 399(10325), 629-655. doi.org/10.1016/S0140-6736(21)02724-0
4. World Health Organization. (n.d.). Antimicrobial resistance (fact sheet). who.int
5. Wikipedia contributors. (n.d.). Oral rehydration therapy. en.wikipedia.org

The big picture

Here is the fact that most people find hardest to absorb about modern global health. Noncommunicable diseases kill about 43 million people a year, roughly three quarters of all deaths worldwide, and about three quarters of those deaths occur in low- and middle-income countries. Some 18 million of them happen before the age of 70. Against that, as you saw in Lesson 2, NCDs receive under 2 percent of development assistance for health.

The mental model that has to go is the phrase diseases of affluence. It was never accurate and it is now actively harmful. A person in Ghana or Bangladesh with hypertension or diabetes typically develops it younger, is diagnosed later, is treated less consistently, faces higher complication rates, and pays for care out of pocket. NCDs in low-income settings are diseases of working-age adults, and they impoverish households in a way that a short infectious illness usually does not.

There is also a systems mismatch at the heart of this module. Health systems in much of the world were designed for acute episodes: a person arrives sick, is treated, and leaves. NCDs require the opposite, continuous care over decades, with medicines that must never run out and appointments that must be kept for forty years. Almost nothing about a system built for acute care handles that well.

Key idea: Noncommunicable diseases cause about three quarters of the world's deaths, most of them in low- and middle-income countries and 18 million before age 70, while receiving under 2 percent of health aid, and they demand continuous chronic care from systems built for acute episodes.

Cardiovascular disease: the largest single killer

Cardiovascular disease, meaning ischaemic heart disease, stroke and related conditions, killed an estimated 19.8 million people in 2021 by GBD accounting, making it the leading cause of death worldwide by a wide margin. The great majority of those deaths occur outside high-income countries.

The dominant modifiable risk factor is high blood pressure. WHO's 2023 global report on hypertension estimated that about 1.3 billion adults aged 30 to 79 have hypertension, that roughly 46 percent of them do not know it, that around 42 percent are diagnosed and treated, and that only about 21 percent have it controlled. Read those numbers as a cascade, exactly like the HIV cascade in Lesson 6, because the shape of the loss tells you what to fix.

Cascade step	Approximate global position	What failure at this step means
Aware of diagnosis	About 54 percent	No routine blood pressure measurement in ordinary care contacts
On treatment	About 42 percent	Diagnosis made but medicines unavailable, unaffordable, or not started
Blood pressure controlled	About 21 percent	Treatment started but doses not titrated, or people lost to follow-up

Hypertension is nearly the ideal target for population health action. The medicines are off-patent and cost a few dollars a year. The measurement is a two-minute task requiring no laboratory. The condition is silent, so it is found only if someone looks. WHO's HEARTS technical package packages this into a standard protocol, and the practical innovations that make it work are unglamorous: single-pill fixed-dose combinations so patients take one tablet instead of three, treatment protocols simple enough for nurses and community health workers to run without a physician, and drug supply arranged so a refill is never missed. WHO has estimated that scaling effective hypertension treatment could avert on the order of 76 million deaths between 2023 and 2050.

Prevention runs mostly through salt. WHO recommends under 5 grams of salt per day; global average intake is roughly double that. Sodium reduction is one of the classic best buys, and yet only a small minority of countries have mandatory comprehensive sodium policies, which is a policy failure rather than a knowledge failure.

One cardiovascular condition deserves separate mention because it is purely a disease of poverty. Rheumatic heart disease follows untreated streptococcal throat infection in childhood, damages heart valves, and affects tens of millions of people, almost entirely in low-income settings and among disadvantaged populations in wealthier ones. It has been essentially eliminated in high-income countries by antibiotics and housing. Its persistence is a direct index of who does not get a doctor for a sore throat.

Key idea: Cardiovascular disease is the world's leading killer at about 19.8 million deaths a year, its dominant modifiable driver is hypertension in which only about a fifth of cases are controlled, and the tools to fix it are cheap off-patent drugs, nurse-led protocols, and salt policy rather than new technology.

Diabetes: a number that was revised upward dramatically

Diabetes is a good case study in how estimates move. The widely quoted figure for years was 422 million adults, based on 2014 data. A pooled analysis published in 2024 by the NCD Risk Factor Collaboration, using far more measurement data, put the number at about 830 million adults in 2022, up from roughly 200 million in 1990. That is not a sudden epidemic in 2022; it is a better measurement, driven partly by the inclusion of people with elevated HbA1c who had never been diagnosed. When you see a figure jump like that, check whether the world changed or the method did.

The treatment gap is the headline finding. More than half of adults aged 30 and over living with diabetes, on the order of 445 million people, were not taking any medication for it. Coverage is heavily concentrated: treatment rates are high in high-income countries and low across much of sub-Saharan Africa and South Asia, where prevalence has risen fastest.

Insulin access is the sharpest illustration of a market failure in this entire course. Insulin was discovered in Toronto in 1921 and the discoverers sold the patent for a nominal sum precisely so it would be affordable to everyone. A century later, three manufacturers control the large majority of the global insulin market, prices in many countries remain out of reach, and people with type 1 diabetes, who die without it within weeks, ration doses. WHO added long-acting insulin analogues to the Essential Medicines List in 2021 and launched a prequalification process to widen the supplier base, which is the standard tool for breaking a concentrated market.

The complications matter as much as the diagnosis. Diabetic retinopathy causes blindness that screening and laser treatment prevent. Diabetic foot ulcers lead to amputations that podiatry and footwear prevent. Diabetic kidney disease leads to kidney failure, and here the equity gap is absolute: dialysis and transplantation are simply unavailable across much of the world, so kidney failure is a death sentence in one country and a chronic condition in another.

Key idea: About 830 million adults had diabetes in 2022, more than half of adults over 30 with the condition take no medication for it, and insulin, given to the world in 1921 for a nominal patent fee, remains unaffordable in many countries because three firms dominate the market.

Cancer: prevention, diagnosis, treatment, and pain

The International Agency for Research on Cancer estimated about 20 million new cancer cases and 9.7 million cancer deaths in 2022, and projects around 35 million new cases a year by 2050 as populations grow and age. Cancer control has four distinct fronts and low-income countries face barriers at every one.

Prevention first, because a much larger share of cancer in low- and middle-income countries is caused by infection. Human papillomavirus causes nearly all cervical cancer, hepatitis B and C cause most liver cancer, and Helicobacter pylori causes most stomach cancer. That means vaccination and infection treatment are cancer prevention, which is a genuinely different strategic picture from the tobacco-and-ageing profile of wealthy countries.

Cervical cancer is the clearest case and the clearest opportunity. It caused roughly 660,000 new cases and 350,000 deaths in 2022, around 90 percent of the deaths in low- and middle-income countries, and it is almost entirely preventable. WHO's 2020 elimination strategy sets the 90-70-90 targets: 90 percent of girls fully vaccinated against HPV by age 15, 70 percent of women screened with a high-performance test twice by ages 35 and 45, and 90 percent of women with disease receiving treatment. Two developments made this far more achievable: WHO's 2022 recommendation that a single HPV dose can be sufficient, which roughly halves cost and logistics, and HPV DNA self-sampling, which removes the pelvic examination that suppresses screening uptake.

Rwanda deserves the credit here rather than any donor. It launched a national HPV vaccination programme in 2011, delivered through schools with community mobilization, and achieved coverage above 90 percent of eligible girls almost immediately, at a time when many wealthy countries were struggling to reach 60 percent. The lesson usually drawn is about the vaccine; the lesson actually available is about a government that ran a competent national delivery system.

Diagnosis and treatment are where the abyss opens. Radiotherapy is required for a large share of cancer treatment and many low-income countries have no radiotherapy machine at all, or one machine for tens of millions of people. Pathology capacity, without which treatment cannot be targeted, is similarly scarce. Childhood cancer makes the gap starkest: survival for children with cancer runs around 80 percent in high-income countries and often 20 to 30 percent in low-income ones, for the same diseases with the same biology. WHO's Global Initiative for Childhood Cancer, launched in 2018, targets 60 percent survival worldwide by 2030.

The fourth front is the one most often forgotten. Pain relief for terminal cancer requires oral morphine, which is cheap, off-patent and on the Essential Medicines List. The Lancet Commission on Palliative Care and Pain Relief documented in 2018 what it called an access abyss: a very large majority of the world's medical opioid supply is consumed in high-income countries, while most people dying of cancer in low-income countries receive nothing. The causes are regulatory fear of diversion, restrictive narcotics laws, and untrained prescribers. This is not a resource problem. Morphine costs almost nothing. It is a policy problem, and it is one of the most straightforwardly fixable injustices in global health.

Key idea: Cancer control fails in low-income countries at prevention, diagnosis, treatment and pain relief simultaneously, and the pain relief gap is the least excusable because oral morphine is cheap and the barrier is purely regulatory.

Why chronic care is structurally hard

Step back and notice what all three diseases demand. Lifelong medication with no interruption. Regular measurement and dose adjustment. Records that follow a patient across years and facilities. A supply chain that never stocks out. Affordable access at the point of care, because a person who must pay each month will stop paying. And a workforce that can manage all of this without a specialist, because there are not enough specialists and never will be.

Health systems in most low- and middle-income countries were not built for this, and the donor architecture reinforced the mismatch by funding episodic disease campaigns. There is, however, one large existing model of exactly the required kind: HIV programs. They already deliver lifelong medication, retention support, cohort monitoring and community follow-up to tens of millions of people. Integrating hypertension and diabetes care into HIV platforms is now being tried in several countries, and it is one of the more promising ideas in the field, precisely because it reuses a system that already works rather than building a fourth parallel one.

Key idea: Noncommunicable disease care requires uninterrupted lifelong treatment, continuous records and reliable supply, which acute-care systems handle badly, and the strongest existing platform for that kind of care in many countries is the HIV program.

Common misconceptions

• NCDs are diseases of affluence. About three quarters of NCD deaths occur in low- and middle-income countries, typically at younger ages and with worse outcomes.
• Treating hypertension requires expensive drugs and specialists. The medicines are off-patent and cost a few dollars a year, and nurse-led protocols with fixed-dose combinations deliver most of the benefit.
• The jump in diabetes estimates to 830 million means the epidemic suddenly exploded. Most of the change came from better measurement that captured previously undiagnosed cases.
• Cancer in poor countries is the same disease profile as in rich ones. A far larger share is caused by infections, which makes HPV vaccination, hepatitis control and H. pylori treatment cancer prevention.
• Pain relief for dying patients is unaffordable. Oral morphine is cheap and off-patent; the barrier is regulatory restriction and prescriber training, not cost.

Recap

• NCDs cause about 43 million deaths a year, three quarters of the global total, mostly in low- and middle-income countries.
• Cardiovascular disease kills about 19.8 million people a year, and only about 21 percent of the 1.3 billion adults with hypertension have it controlled.
• Sodium reduction is a proven best buy that few countries have implemented with mandatory policy.
• About 830 million adults had diabetes in 2022, and more than half of those over 30 receive no medication.
• Cervical cancer is nearly fully preventable, and Rwanda exceeded 90 percent HPV vaccination coverage from 2011 through its own national delivery system.
• The palliative care access abyss leaves most people dying of cancer in low-income countries without morphine that costs almost nothing.

Sources

1. World Health Organization. (n.d.). Noncommunicable diseases (fact sheet). who.int
2. World Health Organization. (n.d.). Hypertension (fact sheet). who.int
3. World Health Organization. (n.d.). Diabetes (fact sheet). who.int
4. International Agency for Research on Cancer. (n.d.). Global Cancer Observatory. gco.iarc.fr
5. World Health Organization. (n.d.). Cervical cancer (fact sheet). who.int
6. NCD Risk Factor Collaboration. (n.d.). Data and publications on diabetes and obesity. ncdrisc.org

The big picture

Four subjects in this lesson, and one structure underneath all of them. Tobacco, alcohol, ultra-processed food and unsafe roads together kill several million people a year. In every case the science is essentially settled, the effective policies are known, and they are cheap. What stands between the evidence and the policy is not knowledge. It is the organized political and legal activity of industries whose revenue depends on the current arrangement.

The field's name for this is the commercial determinants of health: the products, practices and political activity of private actors that shape population health. That framing is not a claim that companies are villains. It is a claim about incentives. A tobacco company that voluntarily halved its sales would be replaced by its board, so expecting voluntary restraint is a category error, and the policy question is always about rules rather than about goodwill.

Watch for something else in this lesson too. On food policy, the innovation came from Latin America and moved north. Chile and Mexico wrote the playbook that wealthier countries later copied, which reverses the usual assumed direction of global health knowledge.

Key idea: Tobacco, alcohol, unhealthy diets and road danger are cases where the science and the effective policies are known and the obstacle is the political and legal activity of the industries that profit, which is what the commercial determinants of health describes.

Tobacco: the treaty that worked

Tobacco kills more than 8 million people a year, of whom around 7 million are users and about 1.3 million are non-smokers exposed to secondhand smoke. There are roughly 1.25 billion tobacco users, and about 80 percent live in low- and middle-income countries, which is where the industry's growth strategy has been directed for decades as wealthy markets shrank.

Tobacco produced the only global health treaty in existence. The WHO Framework Convention on Tobacco Control was adopted in 2003 and entered into force in February 2005, and it now has about 183 Parties, making it one of the most widely adopted treaties in the United Nations system. It obliges countries to act on price and taxation, smoke-free public places, health warnings and packaging, and bans on advertising, promotion and sponsorship. Article 5.3 is the quiet masterpiece: it obliges governments to protect health policy from the commercial interests of the tobacco industry, which turns lobbying itself into a treaty compliance question.

WHO's MPOWER package translates the treaty into six measures: Monitor use, Protect from smoke, Offer help to quit, Warn about dangers, Enforce advertising bans, and Raise taxes. Of these, taxation is by a wide margin the most effective, and the evidence is unusually clean. Demand for cigarettes responds to price with an elasticity around minus 0.4 in high-income countries and larger, roughly minus 0.5 to minus 0.8, in low- and middle-income countries. Work the arithmetic: a 10 percent price increase reduces consumption by about 4 percent in a wealthy country and by 5 to 8 percent in a poorer one, with the largest responses among young people and low-income smokers, who are exactly the groups with most to gain.

The industry's response to effective regulation is the clearest available demonstration of the playbook. When Australia introduced plain standardized packaging in 2012, Philip Morris attempted to sue Australia using a bilateral investment treaty between Hong Kong and Australia, a case dismissed on jurisdictional grounds in 2015. Separately, Ukraine, Honduras, the Dominican Republic, Cuba and Indonesia challenged the measure at the World Trade Organization, and Australia prevailed. Uruguay, a small country with a small budget, was sued by Philip Morris over its packaging rules and won at the World Bank's investment arbitration body in 2016, with legal costs substantially covered by philanthropic support. The point of such cases is often not to win but to make a small country's health ministry contemplate a multi-year international arbitration before it legislates.

One genuinely contested area deserves an honest treatment. Electronic cigarettes, heated tobacco and nicotine pouches are less harmful than combustible cigarettes for an individual smoker who switches completely. They also recruit adolescents who would never have smoked, and they are now largely owned or distributed by the same tobacco companies. Countries have landed in very different places: the United Kingdom has treated them as a cessation tool, while India, Brazil and others have banned their sale. Both positions are defensible from the evidence, and the honest summary is that the individual-level harm reduction case and the population-level initiation case point in opposite directions, and which dominates depends on how many smokers switch versus how many non-smokers start.

Key idea: Tobacco kills over 8 million people a year and produced global health's only treaty, the FCTC, whose most effective single measure is taxation, and whose implementation has been met with international trade and investment litigation designed to deter smaller countries from legislating.

Alcohol: the same evidence, no treaty

Alcohol was responsible for around 2.6 million deaths a year in WHO's 2024 assessment using 2019 data, roughly 4.7 percent of all deaths, with a disproportionate share among young adults. It is a causal factor in liver disease, cardiovascular disease, injury, violence, and at least seven cancers, including breast cancer at levels many drinkers consider moderate.

The effective policies are as well established as for tobacco, and WHO's SAFER package names them: Strengthen restrictions on availability, Advance drink-driving countermeasures, Facilitate screening and treatment, Enforce bans on marketing, and Raise prices through taxes. Price, availability and marketing restrictions are the three best buys, and the evidence base is strong. Scotland's minimum unit pricing, introduced in 2018, was followed by a measurable fall in alcohol-attributable deaths, with published evaluations estimating a reduction in the order of 13 percent, concentrated in the most deprived areas.

So why is there no alcohol framework convention? Three reasons, and none of them is scientific. Alcohol is culturally embedded and economically important in many member states, including major exporters who would have to negotiate against their own industries. The industry is politically well organized and has positioned itself as a partner in responsible drinking initiatives, which occupy the policy space that regulation would otherwise fill. And unlike tobacco, alcohol has a defensible case for moderate use in most societies, which makes prohibition-adjacent framing easy to deploy against any regulation. The result is the WHO global alcohol action plan for 2022 to 2030, which is non-binding, and the widespread adoption of the least effective interventions, education and voluntary codes, over the most effective ones.

Key idea: Alcohol causes about 2.6 million deaths a year and the effective policies of price, availability and marketing restriction are as well evidenced as for tobacco, but political and economic embeddedness has produced only a non-binding action plan rather than a treaty.

The nutrition transition and the double burden

Barry Popkin's nutrition transition describes a shift in diets and activity that accompanies urbanization and rising income: away from staple grains and vegetables toward energy-dense foods high in fat, sugar and refined carbohydrate, increasingly ultra-processed and commercially marketed, alongside falling physical activity as work and transport mechanize. The crucial global health observation is that this transition is now happening faster and at much lower income levels than it did historically in Europe and North America, so populations acquire obesity and diabetes before they acquire the health systems to manage them.

The result is the double burden of malnutrition, and it is one of the most misunderstood ideas in this course. It means undernutrition and overnutrition coexisting: in the same country, in the same community, in the same household where a stunted child and an overweight mother sit at one table, and even in the same person across a lifetime. That last case has a mechanism. A child who is undernourished in early life develops metabolic adaptations for scarcity; when that person later encounters an energy-dense adult diet, the risk of obesity and diabetes is elevated. Undernutrition in infancy is therefore a risk factor for noncommunicable disease in adulthood.

The numbers, with vintage attached. In 2022, joint UNICEF, WHO and World Bank estimates put about 148 million children under five as stunted, 45 million as wasted, and 37 million as overweight. Among adults, more than a billion people worldwide were living with obesity as of 2022, including roughly 879 million adults and 159 million children and adolescents, with the fastest increases in low- and middle-income countries.

Now the policy, and here the leadership is unambiguously Latin American. Mexico introduced a tax on sugar-sweetened beverages in 2014, and evaluations found purchases fell by around 6 to 12 percent with the effect sustained over years and largest among lower-income households. Chile's 2016 Law of Food Labelling and Advertising went further than any high-income country had dared: mandatory black octagonal warning labels on products high in sugar, salt, saturated fat or calories, a ban on marketing those products to children, removal of cartoon characters from packaging, and a ban on their sale in schools. Evaluations found substantial declines in purchases of labelled products, with sugary drink purchases falling by roughly a quarter. Several countries, including Peru, Israel, Uruguay, Argentina, Colombia and Mexico, have since adopted variants of the octagon, and wealthy countries have been studying it rather than the reverse.

One older instrument belongs here too. The International Code of Marketing of Breast-milk Substitutes, adopted by the World Health Assembly in 1981 after the Nestlé boycott, restricts the promotion of infant formula, because aggressive marketing in settings without safe water contributed directly to infant deaths. Enforcement remains partial four decades later, and violations are still routinely documented.

Key idea: The nutrition transition now arrives faster and at lower incomes than it did historically, producing a double burden in which undernutrition and obesity coexist and early-life undernutrition raises later NCD risk, and the most effective policy responses, sugar taxes and front-of-pack warning octagons, were pioneered in Mexico and Chile.

Road traffic injury: designing for human error

Road traffic crashes kill about 1.19 million people a year according to WHO's 2023 global status report, and they are the leading cause of death for children and young people aged 5 to 29. Around 92 percent of those deaths occur in low- and middle-income countries, which hold roughly 60 percent of the world's vehicles. Vulnerable road users, meaning pedestrians, cyclists and motorcyclists, account for over half of deaths, and in some regions motorcyclists alone approach a third.

The framing shift that made progress possible is the Safe System approach, and it is worth understanding because it generalizes far beyond roads. The old model treated crashes as the result of bad drivers and responded with education and punishment. The Safe System begins from the premise that humans will always make mistakes, and therefore designs the system so that a predictable human error does not produce a dead human. Speeds are set below the threshold at which a crash kills. Roads are designed to separate incompatible users. Vehicles are built to absorb energy. Post-crash care is organized to reach people fast. Responsibility shifts from the individual road user to the people who design the system.

Speed is the master variable, and the arithmetic is unforgiving because kinetic energy rises with the square of velocity. The risk that a pedestrian dies when struck rises very steeply above about 30 kilometres per hour, which is why WHO recommends 30 kilometres per hour limits wherever motor vehicles mix with pedestrians and cyclists. That single measure, applied in urban areas, does more than any driver education campaign ever devised.

The other levers are well evidenced and unevenly applied: seatbelt and helmet laws with real enforcement, blood alcohol limits, child restraints, and vehicle safety standards. That last one has a specific global inequity. Independent crash testing has repeatedly shown that models sold in African, Latin American and South Asian markets have been built without safety features that are mandatory in Europe, sometimes the same nameplate with a weaker structure. The Decade of Action for Road Safety 2021 to 2030 targets a 50 percent reduction in deaths, after the previous target of halving deaths by 2020 was missed.

Key idea: Road crashes kill about 1.19 million people a year and are the leading killer of people aged 5 to 29, with 92 percent of deaths in low- and middle-income countries, and the Safe System approach responds by designing speeds, roads and vehicles so that inevitable human error is not fatal.

The playbook, named

Across all four subjects, the same sequence recurs often enough to be predictable. Dispute the science and fund research that manufactures doubt. Reframe the harm as a matter of personal responsibility and individual choice. Offer voluntary codes and self-regulation to occupy the space a law would fill. Lobby directly and through front organizations and trade associations. Litigate under trade and investment law, especially against smaller countries. And when a market restricts, shift marketing effort to markets that have not yet legislated.

Recognizing the sequence is a practical skill rather than a rhetorical one. If you know that voluntary self-regulation typically appears precisely when binding regulation becomes plausible, you will read a company's announcement of a new responsible marketing pledge differently. And you will notice that the countries that legislated successfully, Australia on packaging, Uruguay on tobacco, Chile on food labels, Mexico on sugar, all did so with strong domestic coalitions rather than because an international body told them to.

Key idea: The industry response to health regulation follows a recognizable sequence of doubt, personal-responsibility framing, voluntary codes, lobbying, trade litigation and market shifting, and the countries that overcame it did so through domestic political coalitions.

Common misconceptions

• Health education is the main tool against tobacco and alcohol. Taxation, availability restriction and marketing bans consistently outperform education, which is why industries prefer education.
• Sugar taxes do not change behavior. Evaluations of Mexico's 2014 tax found sustained purchase reductions, largest among lower-income households.
• Obesity and undernutrition are opposite problems in different countries. They coexist in the same countries, households and life courses, and early undernutrition raises later obesity and diabetes risk.
• Road deaths are caused by bad drivers. The Safe System approach starts from the certainty of human error and holds system designers responsible for making errors survivable.
• Global health policy innovation flows from rich countries outward. Front-of-pack warning labels and beverage taxes were pioneered in Chile and Mexico and copied elsewhere.

Recap

• Tobacco kills over 8 million people a year and 80 percent of users live in low- and middle-income countries.
• The FCTC, in force since 2005 with about 183 Parties, is global health's only treaty, and Article 5.3 obliges governments to shield policy from industry.
• Tobacco taxation is the most effective measure, with larger consumption responses in poorer countries and among the young.
• Alcohol causes about 2.6 million deaths a year, and the evidence for price, availability and marketing controls has produced only a non-binding action plan.
• The double burden of malnutrition coexists within countries, households and individuals, with 148 million children stunted and over a billion people living with obesity as of 2022.
• Road crashes kill about 1.19 million people a year, mostly in low- and middle-income countries, and the Safe System approach designs for inevitable human error.

Sources

1. World Health Organization. (n.d.). Tobacco (fact sheet). who.int
2. WHO Framework Convention on Tobacco Control Secretariat. (n.d.). The WHO FCTC. fctc.who.int
3. World Health Organization. (n.d.). Alcohol (fact sheet). who.int
4. World Health Organization. (n.d.). Road traffic injuries (fact sheet). who.int
5. World Health Organization. (n.d.). Obesity and overweight (fact sheet). who.int
6. UNICEF. (n.d.). Malnutrition in children. data.unicef.org

The big picture

Roughly one in eight people alive is living with a mental disorder, on the order of a billion people. Depression affects around 280 million and anxiety disorders around 300 million. Mental and substance use disorders account for a very large share of the world's years lived with disability, placing them among the leading causes of disability on the planet. About 727,000 people died by suicide in 2021, and suicide is among the leading causes of death in people aged 15 to 29, with roughly three quarters of suicides occurring in low- and middle-income countries.

Now the number that defines this lesson. In many low- and middle-income countries, more than three quarters of people with a mental disorder receive no treatment of any kind. The median government allocation to mental health is around 2 percent of the health budget, and in low-income countries median spending has been under one dollar per person per year. Mental health has also received on the order of one percent or less of development assistance for health.

This is the largest gap between burden and response anywhere in global health, and the reasons are structural rather than accidental. No death certificate says depression, so a field organized around counting deaths could not see it. Stigma suppresses demand and disclosure. There is no vertical fund and no emergency framing. And the constituency that would advocate is, by the nature of the conditions, often least able to organize. If you want a single illustration of how measurement shapes priorities, this is it.

Key idea: Mental disorders affect about one in eight people and cause an enormous share of global disability, yet most people in low- and middle-income countries receive no treatment, mental health receives about 2 percent of health budgets, and the neglect follows directly from a field that historically counted deaths rather than disability.

The human rights dimension

Before treatment, an uncomfortable topic. In many countries, people with severe mental illness are held in conditions that would be recognized as abuse in any other context. Human Rights Watch documented in 2020 the practice of shackling, people chained or locked in confined spaces, in some sixty countries across every region, often in the absence of any services and frequently in religious or traditional healing facilities that families turn to because nothing else exists. Long-stay psychiatric institutions in many countries operate with minimal oversight.

The Convention on the Rights of Persons with Disabilities, and particularly its Article 12 on equal recognition before the law, has reframed the debate. The strongest reading holds that substituted decision-making should be replaced by supported decision-making and that involuntary treatment cannot be justified on the basis of disability. WHO's guidance on community mental health services, published in 2021, promotes explicitly non-coercive, rights-based models.

There is genuine disagreement here that a course should not smooth over. Many clinicians argue that a complete prohibition on involuntary intervention would abandon people in acute psychosis or severe suicidal crisis who cannot at that moment make decisions, and that capacity-based frameworks are more defensible than disability-based ones. Disability rights advocates respond that capacity assessments have historically been the vehicle for exactly the abuses at issue, and that services built around coercion never develop the alternatives. Both sides are arguing in good faith from real evidence, and the practical work is happening in the middle: reducing coercion, building crisis alternatives, and ending shackling, which essentially nobody defends.

Key idea: Shackling and institutional abuse of people with mental illness are documented across some sixty countries, and the CRPD has driven a live and unresolved debate about whether involuntary treatment can ever be justified, within which the reduction of coercion and the ending of shackling command wide agreement.

Task-sharing: what actually works with no psychiatrists

Many low-income countries have fewer than one psychiatrist per million people. Any model that depends on specialists is therefore not a plan. The response, developed largely by researchers in low- and middle-income countries, is task-sharing: training non-specialists, including nurses, community health workers and lay counsellors, to deliver structured brief psychological interventions with supervision. The evidence base is now substantial, and it is one of the strongest examples of knowledge flowing from poorer countries to richer ones.

WHO's Mental Health Gap Action Programme, launched in 2008, provides the framework and its Intervention Guide gives non-specialist primary care workers decision protocols for depression, psychosis, epilepsy, self-harm and substance use. But the persuasive evidence comes from specific trials.

The Thinking Healthy Programme, developed by Atif Rahman and colleagues in rural Pakistan and published in The Lancet in 2008, trained Lady Health Workers, the country's existing community health cadre, to deliver cognitive behavioural techniques to mothers with perinatal depression during routine home visits. The trial found prevalence of depression roughly halved at six and twelve months compared with control. It has since become a WHO-endorsed package used internationally.

The Friendship Bench, developed by the Zimbabwean psychiatrist Dixon Chibanda, is the case most worth knowing in detail. Facing a national shortage of psychiatrists and having lost a patient who could not afford bus fare to a clinic, Chibanda trained grandmothers, older lay women already embedded in their communities, to deliver structured problem-solving therapy on wooden benches placed in clinic grounds. A randomized trial published in JAMA in 2016 found substantial reductions in common mental disorder symptoms compared with enhanced usual care. The programme scaled across Zimbabwe and has been adapted in other countries, including in high-income cities.

In Goa, India, the organization Sangath, with Vikram Patel and colleagues, developed the Healthy Activity Program, in which lay counsellors deliver behavioural activation for depression in primary care. Trials published in The Lancet found it effective and cost-effective, with benefits sustained at twelve months. Group interpersonal therapy delivered by trained lay facilitators in rural Uganda showed similar results as far back as 2003.

Notice the pattern. Zimbabwe, Pakistan, India and Uganda produced these models; they were tested rigorously, and they are now exported to wealthy health systems facing their own workforce shortages. If you were looking for a counterexample to the assumption that global health knowledge flows outward from rich countries, this is the strongest one in this course.

Key idea: Task-sharing with trained lay counsellors and community health workers is the evidence-based answer to a world with almost no psychiatrists, and the leading models, the Thinking Healthy Programme in Pakistan, the Friendship Bench in Zimbabwe and the Healthy Activity Program in India, were developed and tested in low- and middle-income countries and are now exported to wealthy ones.

Suicide prevention and the logic of means restriction

Here is the finding that surprises almost everyone, and it is one of the most important ideas in preventive public health. The single most effective population-level suicide prevention measure is not counselling or screening. It is restricting access to the means.

The reasoning rests on two empirical facts. Suicidal crises are often acute and short-lived, sometimes measured in minutes to hours. And methods differ enormously in lethality, so whether a person survives a crisis depends heavily on what was within reach. Remove or make less lethal the most available method, and a large fraction of attempts do not become deaths, and most people who survive an attempt do not go on to die by suicide.

Sri Lanka provides the clearest demonstration ever documented. Through the 1980s and early 1990s Sri Lanka had one of the highest suicide rates in the world, driven overwhelmingly by the ingestion of highly toxic agricultural pesticides that were stored in ordinary rural homes. Beginning in 1995 the government progressively banned the most acutely toxic compounds, starting with the WHO Class I pesticides and later specific highly lethal products. The national suicide rate fell by roughly 70 percent from its peak, and researchers estimated that on the order of 93,000 deaths were averted between 1995 and 2015.

Two details make the case decisive. Agricultural yields were not measurably harmed, because less toxic alternatives did the same job, so the intervention had almost no cost. And no mental health services were involved at all. A pesticide regulation delivered one of the largest mortality reductions in the history of public health.

The principle generalizes. Barriers on bridges, restrictions on firearm access, and limits on pack sizes of paracetamol have all shown effects. There is also a legal dimension: suicide remains a crime in roughly twenty countries, which deters help-seeking and corrupts the data, and several countries have decriminalized it recently, including Pakistan and Guyana in 2022 and Ghana in 2023.

Key idea: Means restriction is the most effective population suicide prevention strategy because suicidal crises are often brief and method lethality varies enormously, demonstrated by Sri Lanka's pesticide bans which cut the national suicide rate by about 70 percent and averted an estimated 93,000 deaths without harming agricultural yields.

Substance use and harm reduction

Substance use disorders sit inside the mental health burden and carry their own access problem. Opioid agonist therapy with methadone or buprenorphine is among the best-evidenced treatments in medicine for opioid dependence, reducing mortality substantially. It is unavailable or illegal in a number of countries, including some with significant injecting drug use, on the grounds that it substitutes one drug for another. That framing is not supported by the evidence.

The same applies to needle and syringe programmes and to naloxone distribution for overdose reversal. Both have strong evidence for reducing HIV transmission and death, and both remain politically contested in ways that track moral framing rather than data. This matters for HIV control as well: as you saw in Lesson 6, people who inject drugs are a key population, and criminalization is an epidemiological variable.

Key idea: Opioid agonist therapy, needle and syringe programmes and naloxone distribution are well evidenced and remain restricted in many countries on moral rather than empirical grounds, with direct consequences for both overdose mortality and HIV transmission.

The economic argument, and its limits

Advocates frequently deploy an economic case, and you should know both its content and its risks. WHO and World Bank analyses have estimated that depression and anxiety cost the global economy on the order of a trillion dollars a year in lost productivity, and that scaled-up treatment returns roughly four dollars for every dollar invested. That is a genuinely strong return by the standards of anything in Lesson 4.

The limit is worth naming. An argument grounded in productivity implies that the value of treating a person depends on their economic output, which quietly excludes the elderly, the severely disabled, and anyone outside the labour market. It is a useful argument in a finance ministry and a corrosive one if it becomes the only argument. The stronger foundation is that untreated suffering at this scale is itself the problem, and the economics is a supporting fact rather than the reason.

Key idea: Mental health treatment returns roughly four dollars per dollar invested against a trillion-dollar annual productivity loss, and relying solely on that argument implicitly devalues people outside the labour market, so it should support rather than replace the case from suffering itself.

Common misconceptions

• Mental illness is a problem of wealthy societies. The burden is global, about three quarters of suicides occur in low- and middle-income countries, and the treatment gap is largest where resources are lowest.
• Treating mental illness requires psychiatrists. Trained lay counsellors and community health workers deliver structured brief therapies with demonstrated effect in randomized trials.
• Suicide prevention means counselling and screening. Means restriction is the most effective population intervention, as Sri Lanka's pesticide bans demonstrated.
• People who survive a suicide attempt will simply try again. Most do not go on to die by suicide, which is precisely why interrupting access during an acute crisis saves lives.
• Opioid agonist therapy just swaps one addiction for another. It is among the best-evidenced treatments in medicine for reducing mortality in opioid dependence.

Recap

• About one in eight people lives with a mental disorder, and most people in low- and middle-income countries receive no treatment.
• Mental health receives around 2 percent of government health budgets and about one percent or less of health aid.
• Shackling and institutional abuse are documented in some sixty countries, and the CRPD has opened a live debate about coercion in treatment.
• Task-sharing models from Pakistan, Zimbabwe, India and Uganda have strong randomized evidence and are now exported to wealthy countries.
• Means restriction is the strongest suicide prevention strategy; Sri Lanka's pesticide bans cut suicide rates by around 70 percent.
• Opioid agonist therapy and harm reduction are well evidenced and politically restricted in ways that track moral framing rather than data.

Sources

1. World Health Organization. (n.d.). Mental disorders (fact sheet). who.int
2. World Health Organization. (n.d.). Suicide (fact sheet). who.int
3. Chibanda, D., Weiss, H. A., Verhey, R., et al. (2016). Effect of a primary care-based psychological intervention on symptoms of common mental disorders in Zimbabwe. JAMA, 316(24), 2618-2626. doi.org/10.1001/jama.2016.19102
4. Rahman, A., Malik, A., Sikander, S., Roberts, C., and Creed, F. (2008). Cognitive behaviour therapy-based intervention by community health workers for mothers with depression and their infants in rural Pakistan. The Lancet, 372(9642), 902-909. doi.org/10.1016/S0140-6736(08)61400-2
5. World Health Organization. (n.d.). Mental health (health topic). who.int

The big picture

Take one year, roughly 2023, and lay out the deaths this lesson is about. About 260,000 women died from causes related to pregnancy and childbirth. About 1.9 million babies were stillborn. About 2.3 million died in their first 28 days of life. About 4.8 million children died before their fifth birthday. And 1.5 million or so adolescents died, mostly from causes that have nothing to do with disease.

Now the fact that should stay with you. Almost none of these deaths required a scientific breakthrough to prevent. The interventions exist, most are old, and many cost less than a cup of coffee. What is missing is delivery: getting a known thing to a specific person in a specific place at a specific moment, usually within hours. That makes this lesson less about medicine than about logistics, geography, money and power, which is precisely why it belongs in a global health course rather than an obstetrics one.

One organizing framework runs through the maternal half of the lesson and generalizes well beyond it. Deborah Maine and Sereen Thaddeus proposed in 1994 that a maternal death typically involves one or more of three delays: delay in deciding to seek care, delay in reaching a facility, and delay in receiving adequate care once there. Each delay has different causes and different remedies, so the first analytic task in any maternal death review is to establish which delay killed the woman.

Key idea: Maternal, newborn and child deaths are overwhelmingly preventable with existing cheap interventions, so the binding constraint is delivery, and the three delays model, in deciding to seek care, in reaching care, and in receiving adequate care, is the standard tool for locating where a system failed.

Why risk screening fails and emergency care works

For decades, safe motherhood programs tried to identify high-risk pregnancies during antenatal care and refer those women for facility delivery. It did not work, and understanding why is the single most important insight in maternal health.

The complications that kill are largely unpredictable. Postpartum haemorrhage, the leading direct cause at roughly a quarter of maternal deaths, most often occurs in women with no risk factors whatsoever, and it can kill a healthy woman in under two hours. Eclampsia, sepsis and obstructed labour are similarly poor at announcing themselves in advance. A screening approach therefore misses most of the deaths while sending many low-risk women for unnecessary intervention.

The strategy that replaced it is different in kind: assume any woman may need emergency care, and organize so that every woman gives birth with a skilled attendant within reach of a facility that can actually treat the emergency. The technical standard is emergency obstetric and newborn care, defined by signal functions. Basic care includes parenteral antibiotics, uterotonic drugs, anticonvulsants for eclampsia, manual removal of the placenta, removal of retained products, assisted vaginal delivery, and newborn resuscitation. Comprehensive care adds the two things that require a hospital: caesarean section and blood transfusion.

Notice what this reframes. The question is no longer whether a woman attended antenatal care but whether, at 3 a.m., there is a functioning theatre, a surgeon or trained clinical officer, blood, and a road. Countries that cut maternal mortality quickly, including Sri Lanka, Malaysia, Thailand and Rwanda, did so by building that referral capacity and by making professional attendance at birth the norm, not by improving risk prediction.

Key idea: Most fatal obstetric complications are unpredictable, so risk screening fails, and mortality falls when every birth has a skilled attendant with functioning emergency obstetric care reachable in time, including caesarean section and blood transfusion.

The commodities that changed the arithmetic

A handful of specific drugs and practices carry a large share of the progress, and it is worth knowing them by name.

Problem	Intervention	Why it matters
Postpartum haemorrhage	Oxytocin at delivery; heat-stable carbetocin	Oxytocin needs a cold chain that often does not exist; carbetocin, shown non-inferior in a 2018 trial, does not
Established haemorrhage	Tranexamic acid	The 2017 WOMAN trial found roughly a third fewer deaths from bleeding when given within three hours
Eclampsia and pre-eclampsia	Magnesium sulfate	Cheap, off-patent, and definitively established by the 2002 Magpie trial as superior to alternatives
Detecting haemorrhage early	Calibrated drape plus treatment bundle	The 2023 E-MOTIVE trial found early objective blood loss measurement with a bundled response cut severe bleeding substantially
Preterm birth	Antenatal corticosteroids; kangaroo mother care	Immediate kangaroo mother care reduced mortality in low-birthweight infants by about a quarter in a 2021 trial

Two patterns are visible in that table. First, several advances are not new molecules but new formulations or delivery protocols that survive real conditions, which is what heat-stable carbetocin and the calibrated drape represent. Second, kangaroo mother care, skin-to-skin contact with the mother providing warmth and feeding, was developed in Bogota, Colombia in the late 1970s by Edgar Rey Sanabria in response to incubator shortages, and it is now recommended worldwide including in wealthy neonatal units. That is another instance of the reverse flow you saw in the mental health lesson.

Two further topics belong here, stated factually. Unsafe abortion causes roughly 8 percent of maternal deaths, on the order of tens of thousands a year, and those deaths occur overwhelmingly in countries with restrictive laws; WHO's 2022 abortion care guideline recommends removing medically unnecessary barriers on the evidence that restriction does not reduce abortion rates but does raise the share performed unsafely. And a large body of research, including a multi-country study published in 2019, documents that mistreatment during childbirth, including physical and verbal abuse and non-consented procedures, is common, which matters both in itself and because women who expect mistreatment avoid facilities.

Key idea: Progress in maternal survival has come from heat-stable and protocol-based innovations such as carbetocin, tranexamic acid, magnesium sulfate and calibrated blood loss measurement, alongside kangaroo mother care developed in Colombia, and it is undermined where abortion is restricted and where women are mistreated in facilities.

Family planning, and the ethics that have to come with it

Roughly 250 million women who want to avoid pregnancy are not using a modern contraceptive method. Meeting that unmet need reduces maternal deaths directly, because fewer pregnancies means fewer chances to die, and indirectly through birth spacing, which improves both maternal and child survival. On cost-effectiveness grounds contraception ranks among the strongest interventions in this course.

And here the course must slow down, because the history is genuinely ugly and it explains present-day suspicion. During India's Emergency in 1975 to 1977, a mass sterilization campaign performed millions of procedures in a single year under quotas and coercion, with documented deaths. Between 1996 and 2000 Peru's government sterilized well over 200,000 people, disproportionately poor and indigenous Quechua-speaking women, many without meaningful consent. China's one-child policy involved coercive enforcement over decades. Sterilization abuse against poor, minority and disabled women is documented in the United States, Puerto Rico and elsewhere. In every case, population control framing, the idea that the number of poor people is itself the problem, supplied the justification.

The 1994 International Conference on Population and Development in Cairo is the hinge. It replaced demographic targets with reproductive rights and health as the organizing principle: the goal is that individuals can decide freely how many children to have and when, not that a country hits a fertility number. That shift is why modern programs are measured by unmet need and informed choice rather than by acceptors recruited, and why quotas and incentives for providers are considered a red flag rather than good management.

Carry this into how you read contemporary advocacy. When you encounter an argument for contraception framed around population growth, climate impact, or national development rather than around what women themselves want, you are hearing the pre-Cairo frame in modern clothing, and the historical record on where that leads is not ambiguous.

Key idea: About 250 million women have an unmet need for modern contraception and meeting it saves lives, and because coercive population control programs in India, Peru, China and elsewhere caused serious harm, the post-Cairo standard measures success by informed individual choice rather than by demographic targets.

Child survival: the cheapest lives in global health

The causes of under-5 death are concentrated. Preterm birth complications and intrapartum events such as birth asphyxia dominate the newborn period. Pneumonia, diarrhoea and malaria dominate after it. Undernutrition is an underlying contributing factor in something like 45 percent of all under-5 deaths, which means nutrition programs prevent deaths that will be recorded as infections.

The Lancet's Child Survival Series in 2003 made the case that has driven policy ever since: around 63 percent of child deaths could be prevented with interventions already available and already cheap. The list has been stable for two decades. Oral rehydration salts with zinc for diarrhoea. Vaccines. Insecticide-treated nets. Exclusive breastfeeding for the first six months, which globally still covers under half of infants. Dispersible amoxicillin for pneumonia. Vitamin A supplementation. Chlorhexidine cord cleansing. Antenatal corticosteroids and kangaroo care for small babies.

What makes these reach children is usually a community cadre rather than a clinic. Integrated community case management trains community health workers to assess and treat pneumonia, diarrhoea and malaria in the village, with referral for danger signs. The two most cited national programs were built by the countries themselves: Ethiopia's Health Extension Programme, which from 2003 deployed tens of thousands of salaried health extension workers, mostly women, to every rural community, and Nepal's Female Community Health Volunteer programme, which has operated a network of roughly 50,000 volunteers since the late 1980s.

Key idea: Around 63 percent of child deaths are preventable with interventions that already exist and cost very little, undernutrition underlies roughly 45 percent of them, and delivery usually depends on community health worker cadres of the kind Ethiopia and Nepal built themselves.

Adolescents: the missing middle

There are about 1.3 billion adolescents aged 10 to 19, and roughly 1.5 million die each year. Look at what kills them and you will see why the health system is poorly configured to help: road traffic injury, drowning, interpersonal violence, self-harm, and in girls, complications of pregnancy and childbirth. These are not infectious diseases and they are mostly not treated in clinics.

Adolescents fall structurally between programs. They are too old for maternal and child health services, which typically end at five, and they are not the target of adult services either. Very few countries have adolescent-specific health policy with a budget attached. Where services exist, confidentiality rules, parental consent requirements and provider attitudes frequently prevent access to exactly what adolescents most need, which is contraception, mental health support and treatment for sexually transmitted infection.

Three facts frame the agenda. Around 12 million girls aged 15 to 19 give birth each year in low- and middle-income countries, and pregnancy and childbirth complications are among the leading causes of death for girls in that age band in some regions. UNICEF estimates that about 640 million women and girls alive today were married before their eighteenth birthday, roughly one in five young women. And about half of all mental disorders begin by age 14, which means that adolescence is where prevention has its largest possible leverage and receives almost none of the resources.

Key idea: Adolescents number 1.3 billion, die mainly from injury, violence, self-harm and pregnancy rather than disease, fall between child and adult health programs, and represent the period of maximum leverage for mental health prevention and the largest gap in services.

Common misconceptions

• Antenatal risk screening identifies the women who will die. Most fatal complications occur without warning in low-risk women, which is why universal access to emergency obstetric care replaced screening.
• Maternal deaths need hospitals and specialists everywhere. They need a functioning referral chain with the signal functions available in time, which mid-level clinicians and district facilities can provide.
• Family planning programs are straightforwardly good. They save lives and have a documented history of coercion; the post-Cairo standard is informed individual choice, not demographic targets.
• Child deaths require new technology. Around 63 percent are preventable with interventions that already existed in 2003 and cost very little.
• Adolescents are healthy and need little attention. About 1.5 million die each year, mostly from injury, violence and self-harm, and half of all mental disorders begin by age 14.

Recap

• The three delays model locates a maternal death in the decision to seek care, the journey, or the quality of care on arrival.
• Risk screening fails because fatal complications are unpredictable; emergency obstetric and newborn care with skilled attendance is what works.
• Heat-stable carbetocin, tranexamic acid, magnesium sulfate and calibrated blood loss measurement are the key haemorrhage and eclampsia tools.
• Kangaroo mother care was developed in Bogota in response to incubator shortages and is now recommended worldwide.
• Roughly 250 million women have unmet need for modern contraception, and the Cairo conference in 1994 replaced demographic targets with reproductive rights.
• Adolescents are 1.3 billion people who die mainly from injury and self-harm and fall between child and adult health programs.

Sources

1. World Health Organization. (n.d.). Maternal health. who.int
2. WOMAN Trial Collaborators. (2017). Effect of early tranexamic acid administration on mortality in women with post-partum haemorrhage. The Lancet, 389(10084), 2105-2116. doi.org/10.1016/S0140-6736(17)30638-4
3. UNICEF. (n.d.). Neonatal mortality. data.unicef.org
4. United Nations Population Fund. (n.d.). Family planning. unfpa.org
5. World Health Organization. (n.d.). Adolescent pregnancy (fact sheet). who.int

The big picture

Everything in this course so far has been an intervention: a vaccine, a drug, a bed net, a counselling session. A health system is the machinery that turns an intervention into a person's health. It is also where most global health failure actually happens, because the gap between what medicine can do and what people receive is a systems gap rather than a scientific one.

Here is the statistic that should reorganize how you think about this. The Lancet Global Health Commission on High Quality Health Systems estimated in 2018 that about 8 million deaths a year in low- and middle-income countries are from conditions treatable by health care. Roughly 5 million of those people did use health services and died anyway because the care was poor quality, and roughly 3.6 million never accessed care at all. Read that again. More people die from bad care than from no care. Access alone is not the goal, and a system measured only by coverage will report success while people die inside it.

This lesson covers the anatomy of a health system, how it is paid for, who works in it, and how medicines reach it. It is the least glamorous lesson in the course and the one most directly relevant if you ever work in the field.

Key idea: Health systems are where interventions become health or fail to, and because roughly 5 million deaths a year in low- and middle-income countries occur among people who did receive poor-quality care, coverage without quality is not a solution.

The six building blocks, and what they miss

WHO's 2007 framework names six building blocks of a health system: service delivery; the health workforce; health information systems; access to essential medicines, vaccines and technologies; financing; and leadership and governance. It is the standard vocabulary and you should know it because everyone uses it.

You should also know its limits. The framework is a checklist rather than a theory: it lists components without explaining how they interact, and systems fail mostly through interactions. It has almost nothing to say about people as users rather than recipients, about demand, trust and health-seeking behavior, all of which determine whether a service is used. And it implies a static machine, when a health system behaves more like a complex adaptive system in which an intervention in one place produces compensating adjustments elsewhere. Introduce performance pay for one service and staff shift effort away from unpaid services. Add a vertical program with better salaries and the general wards lose their nurses. None of that is visible in a list of six boxes.

Key idea: The six building blocks of service delivery, workforce, information, medicines, financing and governance are the field's standard vocabulary, and they describe components without capturing the interactions, trust, and adaptive behavior through which systems actually succeed or fail.

Financing: three functions and one principle

Health financing does three things. Revenue collection raises money, from taxes, insurance contributions, donors or households. Pooling combines it so that the healthy subsidize the sick and the risk is shared. Purchasing allocates it to providers and services, and how you purchase determines what providers do.

One principle dominates: prepayment and pooling. Paying at the point of service is the worst way to finance health care, and the reasoning is straightforward once you see it. It charges people precisely when they are least able to pay, because illness reduces income and increases cost simultaneously. It deters people from seeking care at exactly the moment care would be cheapest. And it provides no risk-sharing at all, so a single serious illness is a household catastrophe.

The measurement concept is catastrophic health expenditure, conventionally defined as out-of-pocket health spending exceeding 10 percent of a household's total budget, with a more severe threshold at 25 percent. WHO and World Bank monitoring has estimated that around 2 billion people face financial hardship from health spending, that roughly 1 billion face catastrophic expenditure at the 10 percent threshold, and that on the order of 344 million people are pushed into extreme poverty by health costs. That last group is the mechanism by which illness manufactures poverty, and it is invisible in any mortality statistic.

Out-of-pocket payment is roughly 18 percent of global health spending, but the average conceals the pattern: in a number of low- and lower-middle-income countries it exceeds 40 or even 50 percent of total health expenditure. The user fees introduced across many countries during structural adjustment, described in Lesson 2, are the historical source of much of this, and most countries that introduced them have since abolished them for at least primary care after evidence showed they suppressed use by the poor without raising meaningful revenue.

Key idea: Health financing collects, pools and purchases, and because point-of-service payment charges people when they are least able to pay, catastrophic health expenditure affects around a billion people and pushes hundreds of millions into extreme poverty.

Universal health coverage, honestly assessed

Universal health coverage means that all people receive the health services they need, of sufficient quality to be effective, without suffering financial hardship. It is Sustainable Development Goal target 3.8, and it is usually pictured as a cube with three dimensions to expand: which people are covered, which services are covered, and what proportion of costs are covered. Any real system trades these off; a government that promises everything to everyone free typically delivers a poor version of it.

Progress has stalled. The UHC service coverage index, scored from 0 to 100, reached about 68 globally by 2021 and has moved little since 2015, while financial hardship has worsened. Coverage improved; protection did not.

The instructive cases are middle-income countries that did it themselves. Thailand introduced its Universal Coverage Scheme in 2002, extending coverage to the roughly 30 percent of the population previously uninsured, financed from general taxation rather than contributions, at a total health spend of a few percent of GDP. Crucially, Thailand paired it with a health technology assessment agency, HITAP, to decide what the scheme would and would not buy, which is the discipline that keeps a universal promise solvent. Brazil's Sistema Unico de Saude, created by the 1988 constitution, is the largest public health system in the world, and its Family Health Strategy of community-based primary care teams has been associated in published evaluations with substantial reductions in infant and amenable mortality. Rwanda's community-based health insurance covers the large majority of the population with income-scaled premiums. Ghana established a national health insurance scheme in 2003, and Costa Rica has run universal coverage with outcomes rivalling far wealthier countries for decades.

Now the critiques, at full strength. Insurance-based schemes struggle where most people work informally, because there is no payroll to deduct from, and they often begin by covering formal-sector workers, who are already better off, so early UHC expansion can widen inequity before narrowing it. Enrolment on paper is not access: schemes routinely report high coverage while patients still pay informally, travel far, or find the medicine out of stock. Benefit packages exclude what is expensive, which frequently means the chronic and mental health care that Module 4 showed to be most neglected. And the quality problem returns: a scheme that pays for a consultation without ensuring the consultation is competent buys attendance rather than health.

Key idea: Universal health coverage means needed, effective services without financial hardship, its global service coverage index has stalled near 68 since 2015, and the achievements of Thailand, Brazil, Rwanda, Ghana and Costa Rica sit alongside real failures of informal-sector exclusion, paper coverage, narrow benefit packages, and unmeasured quality.

The workforce: shortage, distribution, and migration

WHO projects a shortfall of roughly 10 million health workers by 2030, revised down from an earlier estimate of 18 million, concentrated in low- and lower-middle-income countries. The distribution problem is even sharper than the absolute one: a long-cited figure from WHO's 2006 world health report held that the African region carried about a quarter of the global disease burden with around 3 percent of the world's health workers, and while the specific numbers have shifted, the imbalance has not.

International recruitment is where this becomes politically charged, and the honest analysis has to hold two things simultaneously. Individual health workers have every right to migrate, and it is not defensible to trap a Zimbabwean nurse in Zimbabwe for the good of the health system. Remittances sent home are substantial and often exceed aid flows. At the same time, when a high-income country recruits from a low-income one, it acquires a trained professional whose education was financed by a poorer public, which is a transfer of resources in the wrong direction and at a scale that dwarfs many aid programs.

The instrument is the WHO Global Code of Practice on the International Recruitment of Health Personnel, adopted in 2010. It is voluntary, and its practical teeth are the health workforce support and safeguards list, which names countries facing the most pressing workforce shortages, 55 of them in the 2023 update, and discourages active recruitment from them. Compliance is uneven, and recruitment from listed countries has continued.

Migration abroad is also not the largest flow. Internal movement matters more in many countries: from rural to urban areas, from the public sector to the private and NGO sectors, and from clinical work into better-paid donor project management, the internal brain drain described in Lesson 2.

The most effective response has been to build cadres that do not exist elsewhere and therefore do not migrate. Mozambique's surgical technicians, trained over three years to perform caesarean sections and other emergency surgery, have been studied repeatedly and show outcomes comparable to physicians at far lower cost, with dramatically higher retention in rural districts, because their qualification is not recognized abroad. Malawi's clinical officers, Ethiopia's health officers and health extension workers, and Zambia's medical licentiates follow the same logic. These are not second-best improvisations; they are deliberate workforce designs that fit the actual labour market.

Key idea: The world faces a projected shortfall of about 10 million health workers concentrated in poorer countries, international recruitment transfers publicly financed training from poor to rich countries under a voluntary code, and the most durable response has been mid-level cadres such as Mozambique's surgical technicians whose qualifications are not portable abroad.

Essential medicines and the last mile

In 1977 WHO published something quietly radical: a list of about 200 medicines judged sufficient to meet the priority health needs of a population, selected on evidence of efficacy, safety and comparative cost-effectiveness. The Model List of Essential Medicines is now revised every two years and contains roughly 500 items for adults, with a separate children's list. Its power is that it lets a health ministry with a small budget make defensible choices and negotiate as a single large purchaser rather than buying whatever is marketed to it. Most countries maintain national lists derived from it.

Having a list is not having the medicine. Facility surveys in low- and middle-income countries have repeatedly found essential medicine availability in public facilities in the range of half to two thirds, with patients then buying at much higher prices in the private sector. Stockouts are usually failures of the last mile rather than of national procurement: the drugs arrive in the capital and do not reach the district, because of forecasting errors, weak logistics information systems, financing gaps, or transport. This is one area where genuine innovation has come from low-income countries, most visibly Rwanda's use of autonomous drone delivery for blood and emergency supplies to remote facilities from 2016, later adopted in Ghana and elsewhere.

Quality is the harder problem. WHO has estimated that around one in ten medical products in low- and middle-income countries is substandard or falsified, meaning either that it failed to meet quality specifications or that it is deliberately misrepresented. The consequences are treatment failure, drug resistance, and death, and they are largely invisible because a patient who fails treatment on a fake antimalarial is recorded as a treatment failure. The response runs through national regulatory authority strengthening, WHO prequalification, and the African Medicines Agency, whose founding treaty entered into force in 2021.

Finally, pooled procurement and local manufacturing. PAHO's Revolving Fund, UNICEF's Supply Division and the Global Fund's pooled procurement mechanism all use aggregated demand to lower prices, the same logic as Gavi. On manufacturing, the pandemic exposed how dangerous it is for a continent to import almost all its vaccines: the African Union has set a target for 60 percent of vaccines used in Africa to be manufactured in Africa by 2040, and the mRNA technology transfer hub established in Cape Town in 2021 was created precisely because the technology was not shared voluntarily.

Key idea: The WHO Essential Medicines List lets poor health systems choose and negotiate rationally, but availability in public facilities often runs near half, stockouts are usually last-mile failures, and roughly one in ten medical products in low- and middle-income countries is substandard or falsified.

Common misconceptions

• Universal health coverage means free health care for everything. It means needed, effective services without financial hardship, which always requires explicit decisions about which services are covered.
• If coverage rises, health improves. Roughly 5 million deaths a year in low- and middle-income countries occur among people who did receive care that was poor quality.
• Out-of-pocket payment at least raises revenue. User fees raised little, deterred the poor from seeking care, and have been abolished for primary care in most countries that tried them.
• The health workforce crisis is caused by emigration. Internal movement from rural to urban and from public to private and donor-funded work is often the larger flow.
• If a medicine is on the national essential medicines list, patients can get it. Availability in public facilities frequently runs near half, and roughly one in ten products may be substandard or falsified.

Recap

• WHO's six building blocks are service delivery, workforce, information, medicines, financing and governance, and they omit demand, trust and system dynamics.
• Financing collects, pools and purchases, and prepayment with pooling is the core principle because point-of-service payment charges the sick.
• Catastrophic health expenditure affects about a billion people and pushes roughly 344 million into extreme poverty.
• The UHC service coverage index has stalled around 68 since 2015 while financial protection has worsened.
• Thailand, Brazil, Rwanda, Ghana and Costa Rica built universal systems themselves, with real limits around informal employment and quality.
• About 10 million health workers will be missing by 2030, and mid-level cadres such as Mozambique's surgical technicians are the most durable response.

Sources

1. World Health Organization. (n.d.). Universal health coverage. who.int
2. Kruk, M. E., Gage, A. D., Arsenault, C., et al. (2018). High-quality health systems in the Sustainable Development Goals era. The Lancet Global Health, 6(11), e1196-e1252. doi.org/10.1016/S2214-109X(18)30386-3
3. World Health Organization. (n.d.). Health workforce. who.int
4. World Bank. (n.d.). Universal health coverage. worldbank.org
5. Wikipedia contributors. (n.d.). WHO Model List of Essential Medicines. en.wikipedia.org

The big picture

Almost everything in global health governance is voluntary. The International Health Regulations are the exception: a binding instrument accepted by 196 States Parties that tells governments what they must do when a dangerous disease appears. Understanding the IHR means understanding both the strongest legal tool this field has and why that tool did not prevent the worst public health catastrophe in a century.

There is one structural problem at the centre of everything in this lesson, and it is worth stating before the detail. A country that reports an outbreak promptly is rewarded with border closures, flight cancellations, collapsed exports and tourism losses, sometimes running to billions of dollars. A country that delays reporting suffers none of that immediately. The IHR asks states to act against their own short-term interests and offers them nothing in return. Every failure you are about to read has some version of that incentive underneath it.

Key idea: The International Health Regulations are global health's only binding legal instrument, and their central weakness is that prompt outbreak reporting triggers immediate economic punishment while offering the reporting country no compensation or protection.

What the IHR actually require

The lineage runs from the International Sanitary Regulations of 1951 through the IHR of 1969, which covered only three diseases: cholera, plague and yellow fever. That disease-list approach failed obviously when SARS emerged in 2002 and 2003 as something not on any list, and the resulting revision produced the IHR of 2005, in force since 2007.

Four features matter. The scope became all-hazards: states must notify WHO of any event that may constitute a public health emergency of international concern, whatever its cause, chemical and radiological included, rather than a fixed set of diseases. The notification deadline is 24 hours from assessment. States must build and maintain core capacities in surveillance, laboratory diagnosis, response, risk communication and control at points of entry. And critically, WHO gained the authority to consider information from non-state sources, including media and civil society, rather than being confined to what governments chose to tell it. That change ended a situation in which a government's silence was legally equivalent to nothing happening.

A public health emergency of international concern, or PHEIC, is declared by the Director-General on the advice of an Emergency Committee, and it triggers temporary recommendations. There have been eight: H1N1 influenza in 2009, polio and the West African Ebola epidemic in 2014, Zika in 2016, Ebola in the Democratic Republic of the Congo in 2019, COVID-19 in 2020, and mpox in 2022 and again in 2024.

The weaknesses are structural. There is no enforcement mechanism and no sanction for non-compliance. Core capacity requirements have gone widely unmet for nearly two decades. Article 43 says states should not adopt health measures more restrictive of international traffic than reasonably available alternatives, and states routinely ignore it, as they did comprehensively during COVID-19 and again during mpox. A binding treaty with no consequence for breach is, in practice, a strong recommendation.

Key idea: The IHR of 2005 replaced a three-disease list with an all-hazards duty to notify WHO within 24 hours, required core national capacities, and allowed WHO to act on non-governmental information, but it carries no enforcement mechanism and its restrictions on excessive travel measures are routinely ignored.

Ebola in West Africa: the rehearsal nobody learned from

The 2014 to 2016 epidemic in Guinea, Liberia and Sierra Leone caused about 28,600 cases and around 11,300 deaths. It is worth studying because every failure it exposed reappeared in 2020.

WHO was slow. The outbreak was recognized in March 2014, Medecins Sans Frontieres warned publicly that it was out of control, and the PHEIC was not declared until August. The subsequent independent assessment was blunt about institutional caution, weak country office capacity, and reluctance to declare an emergency partly out of concern for the economic consequences to affected states. The reforms that followed were real: WHO created its Health Emergencies Programme in 2016 with operational rather than purely normative capacity, and a Contingency Fund for Emergencies to release money without waiting for appeals.

The other half of the story is usually told badly. Nigeria contained an imported case in Lagos, a city of some 20 million people, and did it largely with its own institutions, repurposing the emergency operations centre built for polio eradication and conducting thousands of contact tracings. The physician Ameyo Adadevoh, who identified the index case and refused to discharge him under considerable pressure, contracted the disease and died. Senegal and Mali also contained importations. And in the three worst-affected countries, transmission ultimately fell when communities changed burial practices and organized their own case-finding, work led by local leaders and anthropologists rather than imported clinical teams. The epidemic ended through social change as much as through biomedical response.

Key idea: The West African Ebola epidemic exposed slow emergency declaration and weak WHO operational capacity, prompting the Health Emergencies Programme, and it was contained partly through Nigeria's own polio infrastructure and through community-led changes to burial and case-finding practice.

COVID-19: the numbers, stated plainly

WHO was informed of a cluster of pneumonia cases in Wuhan on 31 December 2019. The PHEIC was declared on 30 January 2020, and WHO began describing the situation as a pandemic on 11 March 2020.

Now the mortality, and the distinction that matters more than any other in this lesson. Roughly 7 million deaths were confirmed and reported to WHO. WHO's excess mortality estimate for 2020 and 2021 alone was about 14.8 million, with an uncertainty range of roughly 13.2 to 16.6 million. Excess mortality compares observed deaths from all causes against the number expected from historical trends, so it captures deaths never tested, deaths from disrupted health services, and deaths in places with weak registration. The gap between 7 million and 14.8 million is not a rounding difference; it means confirmed counts undercounted by roughly a factor of two.

India illustrates the point at national scale: officially reported deaths were on the order of half a million, while WHO's excess mortality estimate for India was around 4.7 million. The Indian government disputed the methodology, and the dispute was genuine rather than merely political, because modelling excess deaths where registration is incomplete requires strong assumptions. What is not in dispute is that confirmed case and death counts were never an adequate measure of this pandemic anywhere.

The scientific response was extraordinary and should be recorded as such. The genome sequence was published in January 2020 and vaccines were authorized within about eleven months, the fastest vaccine development in history by a wide margin. It was also heavily publicly financed: through the United States government's advance purchase and development programme, German federal funding to BioNTech, and United Kingdom public funding for the Oxford vaccine, among others. The private sector executed; the public sector carried much of the risk.

Key idea: COVID-19 caused about 7 million confirmed deaths and an estimated 14.8 million excess deaths in 2020 and 2021 alone, so confirmed counts undercounted by roughly half, and the record-fast vaccine development was underwritten substantially by public money.

The vaccine equity failure

This is the part of the pandemic that global health as a field has to own, because it was not a technical failure. Vaccines existed. They were allocated by purchasing power.

By the end of 2021, most high-income countries had vaccinated a large majority of their populations while coverage with even a first dose in low-income countries was around or below 10 percent. The mechanism was simple. Wealthy governments signed bilateral advance purchase agreements covering multiples of their populations before vaccines were proven, which pre-empted manufacturing capacity. COVAX, the multilateral pooled procurement facility created precisely to prevent this, was left bidding for supply behind those bilateral deals. It eventually delivered on the order of 2 billion doses, which is a real achievement, but late and far short of its targets. Donations, when they came, frequently arrived with short shelf lives and little notice, which imposed the cost of failure on recipient countries whose systems were blamed for wastage.

Two responses from low- and middle-income countries define the period. In October 2020, India and South Africa proposed a waiver of certain TRIPS intellectual property obligations for COVID-19 products at the World Trade Organization. It was supported by most developing countries and opposed by several high-income members. What eventually passed in June 2022 was a narrow decision on vaccines only, which most public health advocates judged too limited and far too late to affect the pandemic.

The second response was more consequential. When mRNA developers declined to share technology, WHO established an mRNA technology transfer hub in Cape Town in 2021, centred on the South African biotechnology company Afrigen, which reverse-engineered an mRNA vaccine candidate from published information and began training partners across Africa, Latin America and Asia. Whatever that hub eventually produces, it changed the strategic question from whether wealthy manufacturers would share to whether they could remain the only source. That is a shift in agency, and it came from Cape Town rather than Geneva.

Key idea: Vaccine scarcity during COVID-19 was allocated by purchasing power through bilateral advance purchase agreements that pre-empted COVAX, the TRIPS waiver arrived narrow and late, and the most significant structural response was the WHO mRNA technology transfer hub built around Afrigen in Cape Town.

What was reformed afterward

Three things changed, and you should know their status and their limits.

Amendments to the IHR were adopted at the World Health Assembly in June 2024. They created a new and higher category of pandemic emergency above the existing PHEIC, required each state to designate a National IHR Authority, added equity language on access to health products, and established a coordinating financial mechanism. They entered into force in 2025 for states that did not reject them. Note what did not change: there is still no enforcement.

The WHO Pandemic Agreement was adopted in May 2025 after more than three years of negotiation. Its most consequential element is a Pathogen Access and Benefit-Sharing system, under which manufacturers participating in the system would make a share of pandemic-related production, discussed as 20 percent with half donated and half sold at affordable prices, available to WHO for equitable distribution. The operational annex required further negotiation, and the agreement enters into force only after a threshold of ratifications, so at the time this course was written its practical effect was still unknown. Check the current status rather than assuming.

The Pandemic Fund, established in 2022 and hosted at the World Bank, provides grants for preparedness capacity in low- and middle-income countries. It has been persistently smaller than the need its own founding documents identified.

Key idea: Post-COVID reform produced IHR amendments in 2024 adding a pandemic emergency category and equity provisions, a Pandemic Agreement adopted in 2025 containing a pathogen access and benefit-sharing system that still required ratification, and a Pandemic Fund smaller than the identified need, with enforcement still absent throughout.

Why preparedness indices failed, and what predicted outcomes instead

This is the humility lesson, and it may be the most useful thing in the module.

In 2019, the Global Health Security Index scored countries on their pandemic preparedness. The United States ranked first and the United Kingdom second. Both then experienced among the highest COVID-19 death tolls per capita in the high-income world. The index measured stockpiles, plans, laboratories and formal capacities, and those turned out to be weak predictors of what actually happened.

Subsequent analysis, including a large cross-national study published in The Lancet in 2022, found that measures of trust, both trust in government and interpersonal trust between citizens, were associated with lower infection rates, while preparedness index scores were not. The mechanism is intuitive once stated: preparedness capacities only matter if people follow guidance, accept vaccination, report symptoms and comply with measures, and all of that runs on trust rather than on equipment.

Two implications follow. First, be sceptical of any index that scores complex system performance from documents and inventories. Second, the investments that most improve a country's pandemic response may be political and social rather than technical, which is uncomfortable for a field that prefers procurable solutions. Vietnam, Senegal, Rwanda, New Zealand and several others outperformed their index rankings substantially, and the explanations usually involve prior epidemic experience, decisive early action, and public trust rather than superior stockpiles.

Key idea: The 2019 Global Health Security Index ranked the United States and United Kingdom highest and predicted COVID-19 outcomes poorly, while measures of governmental and interpersonal trust were associated with better outcomes, indicating that preparedness is as much political as technical.

Common misconceptions

• The IHR let WHO order countries to act. WHO can declare emergencies and issue recommendations; it has no enforcement power and cannot compel any state.
• About 7 million people died of COVID-19. That is the confirmed count; excess mortality estimates for 2020 and 2021 alone are around 14.8 million.
• COVAX failed because of poor logistics in recipient countries. It was outbid by bilateral advance purchase agreements that pre-empted manufacturing capacity before doses existed.
• The TRIPS waiver was granted and made no difference. A narrow vaccine-only decision passed in June 2022, far later and much less extensive than the original 2020 proposal.
• Countries with the best preparedness scores handled the pandemic best. The 2019 index ranked the United States and United Kingdom first and second, and trust predicted outcomes better than preparedness scores did.

Recap

• The IHR of 2005 bind 196 States Parties to notify WHO within 24 hours of any event that may constitute a PHEIC, with no enforcement mechanism.
• Prompt reporting triggers trade and travel losses with no compensation, which is the incentive problem underlying most IHR failures.
• The West African Ebola epidemic prompted WHO's Health Emergencies Programme, and Nigeria contained its outbreak using its own polio infrastructure.
• COVID-19 caused about 7 million confirmed deaths and an estimated 14.8 million excess deaths in 2020 and 2021.
• Vaccine allocation followed purchasing power; the mRNA hub in Cape Town was the most significant structural response from affected countries.
• IHR amendments in 2024 and the 2025 Pandemic Agreement added equity provisions and a benefit-sharing system without adding enforcement.

Sources

1. World Health Organization. (n.d.). International Health Regulations. who.int
2. World Health Organization. (n.d.). Coronavirus disease (COVID-19). who.int
3. Bollyky, T. J., Hulland, E. N., Barber, R. M., et al. (2022). Pandemic preparedness and COVID-19: An exploratory analysis of infection and fatality rates. The Lancet, 399(10334), 1489-1512. doi.org/10.1016/S0140-6736(22)00172-6
4. World Trade Organization. (n.d.). TRIPS and public health. wto.org
5. Wikipedia contributors. (n.d.). COVAX. en.wikipedia.org

The big picture

Two forces now generate more health emergencies than anything else: armed conflict and a changing climate. They are converging. Around 123 million people were forcibly displaced worldwide by the end of 2024, roughly triple the figure of a decade earlier, and something like 300 million people have needed humanitarian assistance in recent years. Meanwhile the years since 2023 have been the hottest on record, and the health consequences are no longer projections.

Both subjects have a common structure worth naming at the outset. In each, the people harmed are overwhelmingly not the people who caused the harm, and the response is organized by institutions accountable to somebody else. That is the same problem you met in Lesson 1 in a different costume, which is why this lesson belongs in the governance module rather than in a disease module.

One correction before we begin. Most displaced people are not in camps and most are not in wealthy countries. Around three quarters of refugees are hosted in low- and middle-income countries, and most are in a country neighbouring their own. The mental picture of a tented camp administered by international agencies describes a shrinking minority of the actual situation.

Key idea: Conflict and climate now drive most health emergencies, about 123 million people were forcibly displaced by the end of 2024, and roughly three quarters of refugees are hosted by low- and middle-income countries, mostly outside camps.

How humanitarian health is measured and organized

Humanitarian health has one master indicator: the crude mortality rate, expressed as deaths per 10,000 people per day. The convention is that a rate of 1 per 10,000 per day signals an emergency out of control, because it is roughly double the baseline rate in most low-income settings, and for children under five the threshold is 2 per 10,000 per day. Those numbers are the trigger for scaling a response, and they are why mortality surveys, rather than facility statistics, are among the first things a humanitarian epidemiologist does.

The technical standards are set out in the Sphere Handbook, first published in 1998 and revised repeatedly, which specifies minimum standards across water and sanitation, food security and nutrition, shelter, and health, alongside a humanitarian charter grounded in the right to life with dignity. Sphere is the reason a field worker can say what an adequate litres-per-person-per-day figure is without inventing one. Coordination runs through the cluster system introduced in the 2005 humanitarian reform, in which WHO leads the health cluster and UNICEF leads water and sanitation and nutrition.

What kills people in acute displacement is predictable and has been for decades: diarrhoeal disease, measles, acute respiratory infection, malaria and malnutrition, interacting with each other. The Goma catastrophe of 1994, when cholera and dysentery killed on the order of 12,000 Rwandan refugees within about a month, remains the canonical case of what happens when a million people arrive somewhere with no water supply. The operational lesson is that in the first phase, water, sanitation, measles vaccination, food and shelter save more lives than clinical services do.

The model has shifted, though, and the shift matters. Displacement is now typically protracted, measured in years and often decades, and increasingly urban rather than camp-based. Building parallel camp health services for a population that will still be there in fifteen years, in a city where they live among the host population, makes progressively less sense than supporting and financing the host country's own health system to serve everyone. That is the direction policy has moved, and it is contested mainly on financing grounds, since host governments are being asked to absorb costs.

One more element belongs here. Attacks on health care, meaning strikes on hospitals, ambulances and health workers, are recorded in the hundreds to over a thousand per year by WHO's surveillance system, across conflicts including Sudan, Gaza, Ukraine, the Democratic Republic of the Congo and Myanmar. Such attacks are prohibited under international humanitarian law, and the UN Security Council adopted Resolution 2286 in 2016 demanding their end. The practical consequence for global health is that health system destruction outlives the fighting: a bombed hospital and a killed or emigrated workforce continue producing deaths long after a ceasefire.

Key idea: Humanitarian health uses crude mortality rate thresholds of 1 per 10,000 per day overall and 2 for under-fives, works to Sphere standards through the cluster system, prioritizes water, sanitation, measles vaccination and nutrition in the acute phase, and has shifted from camp services toward supporting host country systems as displacement becomes protracted and urban.

Refugee and migrant health

Two findings from WHO's 2022 world report on the health of refugees and migrants are worth carrying. First, migrants are frequently healthier at arrival than the populations they join, because migration selects for the young and physically capable, a pattern called the healthy migrant effect. Second, that advantage erodes, sometimes rapidly, under the conditions of the destination: hazardous work, poor housing, discrimination, and exclusion from services.

The barriers to care are mostly not clinical. Legal status determines entitlement in many systems, so an undocumented person may be entitled to emergency care only. Fear of immigration enforcement deters people from presenting even where they are entitled. Language and interpretation are chronically underfunded. Documentation requirements exclude people who fled without papers. And discrimination by providers is documented in many settings.

Two specific groups deserve mention. Internally displaced people, who are roughly half of the total displaced, remain within their own country and therefore fall under their own government's responsibility, which is a problem when that government is a party to the conflict displacing them. And migrant workers, particularly in construction and domestic work, face occupational hazards with limited legal protection; heat exposure among outdoor workers in the Gulf and South Asia is a growing and measurable cause of kidney injury and death, which connects this section directly to the next.

Key idea: Migrants often arrive healthier than host populations and lose that advantage to working and living conditions, and their barriers to care are legal, administrative and social rather than clinical, with internally displaced people particularly exposed because responsibility rests with the government displacing them.

Climate change as a health problem

Climate change affects health through at least seven distinct pathways, and keeping them separate prevents muddled analysis.

Pathway	Mechanism	Example
Direct heat	Heat stroke, cardiovascular and renal strain, occupational injury	Rising heat-related mortality in people over 65 and chronic kidney disease in outdoor labourers
Extreme events	Injury, drowning, displacement, service disruption	Floods and cyclones destroying facilities and cutting supply lines
Vector-borne disease	Range and season expansion for mosquitoes and other vectors	Dengue spreading to higher altitudes and latitudes, with record case counts in the Americas in recent years
Water-borne disease	Flooding contaminating supplies; drought concentrating pathogens	Cholera outbreaks following floods and water scarcity
Food systems	Yield losses plus falling micronutrient content of staple crops	Undernutrition and reduced zinc and iron in grains grown at higher carbon dioxide levels
Air quality	Shared sources with greenhouse gases; wildfire smoke; ozone	Air pollution deaths that mitigation would also prevent
Displacement and conflict	Livelihood loss and resource stress driving movement	Drought-linked migration adding to humanitarian caseloads

Now the numbers, handled carefully because this is an area where advocacy and evidence sometimes blur. WHO's most quoted estimate is that climate change will cause approximately 250,000 additional deaths per year between 2030 and 2050 from malnutrition, malaria, diarrhoea and heat stress. That figure is widely cited and widely misunderstood: WHO itself describes it as conservative, because it models only a handful of pathways and excludes most of the table above, along with displacement, conflict, mental health and health system disruption. More recent analyses produce substantially larger figures. The honest statement is that the 250,000 number is a floor derived from a narrow model, not a best estimate.

What is measured rather than projected is more persuasive. The Lancet Countdown, which tracks indicators annually, has documented large increases in heat-related mortality among people over 65, comparing recent decades with the 1990s, along with rising heat exposure of outdoor workers and associated losses of labour capacity. The IPCC's sixth assessment concluded that between 3.3 and 3.6 billion people live in contexts highly vulnerable to climate change.

Key idea: Climate affects health through heat, extreme events, vector- and water-borne disease, food systems, air quality and displacement, and WHO's widely cited estimate of 250,000 additional deaths a year from 2030 to 2050 is explicitly a conservative floor covering only a few of those pathways.

The equity structure, and the co-benefits

Climate and health is the clearest case of a global health problem in which causation and consequence are almost perfectly inverted. The African continent accounts for a very small share of historical and current global emissions, in the low single digits, and bears a disproportionate share of the health impacts. Small island states face existential risk from decisions taken elsewhere. Within countries the same inversion repeats: the poorest households have the least air conditioning, the most outdoor work, the flimsiest housing and the least insurance.

That structure is why loss and damage finance, agreed in principle at COP27 in 2022 and operationalized at COP28 in 2023, is framed by many developing countries as compensation rather than aid, and why the amounts pledged relative to the estimated need remain a live grievance.

The most useful practical argument in this area is health co-benefits. Actions that reduce emissions also produce large, immediate, local health gains, which changes the political calculation from future global altruism to present domestic benefit. Air pollution is the clearest case: WHO attributes on the order of 6.7 million deaths a year to ambient and household air pollution combined, with other analyses giving higher figures, and roughly 2.3 billion people still cook with polluting fuels. Cleaner energy, cleaner cooking, active transport and reduced meat intensity all cut both emissions and disease, and the health benefits arrive within years while the climate benefits arrive over decades.

Adaptation is equally concrete and again includes leadership from lower-income countries. After a severe heat wave in 2010, the Indian city of Ahmedabad developed South Asia's first heat action plan in 2013, combining early warning, public messaging, cooling spaces, and training for health workers to recognize heat illness. Evaluations have credited it with saving on the order of a thousand lives a year, and it has been replicated across dozens of Indian cities. Climate-resilient health systems, meaning facilities that survive floods, keep cold chains running through power failures, and can surge during heat waves, are the corresponding infrastructure agenda.

Key idea: The populations least responsible for emissions bear the greatest health burden, which frames loss and damage finance as compensation rather than aid, and the strongest practical argument for action is health co-benefits, since cutting air pollution alone addresses on the order of 6.7 million deaths a year.

Common misconceptions

• Most refugees live in camps in wealthy countries. About three quarters are hosted in low- and middle-income countries, mostly neighbouring their own, and increasingly in cities rather than camps.
• In an acute emergency, clinical care saves the most lives. Water, sanitation, measles vaccination, food and shelter save more in the first phase, which is why crude mortality rate and coverage are measured first.
• Migrants arrive with poor health and burden host systems. Migration selects for the healthy, and the advantage erodes under destination conditions and access barriers.
• WHO's estimate of 250,000 climate deaths a year is the expected total. WHO describes it as conservative; it models only a few pathways and excludes displacement, conflict and health system disruption.
• Climate action is a cost with benefits only in the distant future. Cutting air pollution and improving cooking fuels deliver large health gains within years, independent of long-run climate effects.

Recap

• About 123 million people were forcibly displaced by end-2024, and roughly three quarters of refugees are hosted in low- and middle-income countries.
• A crude mortality rate of 1 death per 10,000 per day marks an emergency, with 2 per 10,000 for under-fives; Sphere sets the technical standards.
• Displacement is now protracted and urban, shifting policy from parallel camp services toward financing host country health systems.
• Attacks on health care number in the hundreds to over a thousand a year and destroy capacity that outlasts the fighting.
• Climate affects health through at least seven pathways, and WHO's 250,000 deaths a year estimate is an explicit conservative floor.
• Health co-benefits make mitigation locally attractive, and Ahmedabad's 2013 heat action plan is the model adaptation case, developed in India and replicated widely.

Sources

1. United Nations High Commissioner for Refugees. (n.d.). Refugee data finder. unhcr.org
2. World Health Organization. (n.d.). Climate change and health (fact sheet). who.int
3. The Lancet Countdown on Health and Climate Change. (n.d.). Annual indicator reports. lancetcountdown.org
4. Sphere Association. (n.d.). The Sphere Handbook: Humanitarian charter and minimum standards. spherestandards.org
5. World Health Organization. (n.d.). Air pollution. who.int

The big picture

This last lesson turns the course's methods on the course's own subject. Global health does research on people who cannot easily refuse, holds resources that people need, and describes populations that rarely control the description. Those three facts generate an ethics that goes beyond ordinary medical ethics, and a live argument about whether the field's structure is defensible at all.

Then we do something less comfortable than either: count. The Sustainable Development Goals gave the world explicit health targets for 2030 with numbers attached, and the honest scorecard shows most of them will be missed. Reading that scorecard correctly, as neither a vindication of cynicism nor a reason to look away, is the final skill this course is trying to teach.

Key idea: Global health's ethics arise from doing research on people who cannot easily refuse, controlling resources people need, and describing populations that do not control the description, and the field's own 2030 targets provide a concrete test of whether it is working.

The ethical foundations and where they came from

Research ethics is built from disasters. The Nuremberg Code of 1947 emerged from the trial of Nazi physicians and established voluntary consent as absolutely essential. The World Medical Association's Declaration of Helsinki, first adopted in 1964 and revised repeatedly since, translated that into a working code for medical research, and remains the most cited document in the field. The Belmont Report of 1979, written after the Tuskegee syphilis study was exposed, named three principles that still organize every ethics committee: respect for persons, beneficence, and justice. The CIOMS international ethical guidelines, most recently revised in 2016, adapt all of this specifically to research in low-resource settings.

Justice is the principle that does the most work in global health. It asks who bears the burdens of research and who receives its benefits. A trial conducted among poor people that produces a treatment only wealthy people will be able to afford satisfies respect for persons and beneficence while failing justice completely. That is the shape of most of the controversies below.

Key idea: Research ethics developed from Nuremberg, Helsinki, Belmont and CIOMS, and in global health the Belmont principle of justice, asking who bears the burdens and who receives the benefits of research, is the one most frequently violated.

The standard of care controversy, presented from both sides

In 1994 a United States trial, ACTG 076, showed that a long course of zidovudine given to pregnant women with HIV sharply reduced transmission to their infants. The regimen cost around 800 dollars per patient at the time, which was more than the entire annual per capita health budget of most affected African countries.

Researchers then began trials in Africa and Asia of a much cheaper short-course regimen, using placebo controls. In 1997 Peter Lurie and Sidney Wolfe published a critique in the New England Journal of Medicine, and Marcia Angell wrote an editorial comparing the design to Tuskegee. The objection is straightforward and serious: an effective intervention was known, so giving women a placebo meant knowingly allowing infant infections that could have been prevented, and no ethics committee would have approved such a design in the country funding the research.

The defence is also serious and you should be able to state it. The relevant comparison, defenders argued, is not what is available in Boston but what is available to these women, which was nothing. An equivalence trial against the expensive regimen would have required far more participants, taken much longer, and produced an answer about a drug the health system could not buy. The short-course trials produced a regimen that was affordable and was adopted, preventing a large number of infections quickly.

Both positions rest on real values. The universalist position holds that a person's entitlement in research does not depend on their country's wealth. The contextualist position holds that refusing to study affordable options in the name of principle leaves people with nothing at all. The compromises that emerged, and they are compromises rather than resolutions, run through the Helsinki revisions and CIOMS guidance: strong justification required for any placebo where effective treatment exists, obligations of post-trial access so participants continue to receive a proven benefit, and requirements that research be responsive to the health needs of the host community.

A less ambiguous case sits alongside it. In 1996, during a meningitis epidemic in Kano, northern Nigeria, Pfizer conducted a trial of the experimental antibiotic trovafloxacin in children. Allegations followed that consent was inadequate and that families did not understand they were in a trial. Litigation continued for over a decade and was settled in 2009. Its effect on trust was long-lasting, and it is part of the local background to the vaccine boycotts described in Lesson 7.

Key idea: The 1997 controversy over placebo-controlled perinatal HIV trials pits a universalist standard, that research entitlements do not depend on national wealth, against a contextualist one, that studying affordable options is what actually helps, and it produced compromise rules on placebo justification, post-trial access, and responsiveness to host community needs.

Consent, community, and who owns the samples

Individual informed consent, designed for an autonomous patient in a wealthy clinic, strains in practice. Therapeutic misconception, the belief that a research procedure is treatment personally chosen for one's benefit, is common everywhere and stronger where research is the only route to any care at all. Consent given to a doctor by someone with little formal education, in a setting where refusing an authority figure is socially costly, may be voluntary in form and not in substance. Signature requirements exclude people who cannot read and frighten those with reason to avoid documents.

The practical responses are community engagement and community advisory boards, which involve representatives in design and oversight, and the recognition of ancillary care obligations, meaning what researchers owe participants who turn out to have conditions the study is not about. None of these substitutes for individual consent; they surround it.

The sharpest current issue is what leaves the country. Samples and data collected in one place are frequently analysed, stored, published and commercialized elsewhere, sometimes for purposes participants never agreed to. The Havasupai case in the United States, where blood collected for diabetes research was used for studies on schizophrenia and population migration, is the canonical illustration and ended in a settlement and return of samples.

Internationally, this became a matter of law and sovereignty. The Nagoya Protocol, adopted in 2010 under the Convention on Biological Diversity, established that countries have rights over their genetic resources and that benefits arising from their use must be shared. The health precedent came earlier: in 2007 Indonesia stopped sharing H5N1 influenza virus samples with WHO collaborating centres, on the argument that samples from Indonesia were used to make vaccines Indonesia could not buy. The position, called viral sovereignty by its critics, was widely condemned and produced the 2011 Pandemic Influenza Preparedness Framework, which tied sample sharing to benefit sharing. The PABS system in the 2025 Pandemic Agreement descends directly from that argument. Whatever you conclude about the tactic, note that it worked: the norm changed because a middle-income country withheld something valuable.

The constructive response is capacity rather than restriction. Initiatives such as H3Africa, which built genomics research capacity and biobanks on the African continent rather than exporting samples, changed the default from extraction to local analysis.

Key idea: Informed consent strains under therapeutic misconception and social hierarchy, requiring community engagement and ancillary care obligations, and the export of samples and data has been countered by the Nagoya Protocol, by Indonesia's 2007 refusal to share influenza samples, and by capacity-building initiatives such as H3Africa.

Decolonizing global health: the claims and the objections

The argument has an empirical core, and you should engage with the evidence rather than the slogan. Audits of the literature repeatedly find that a substantial share of papers about African populations include no author based in Africa, and that first and senior authorship concentrate in high-income institutions even when the data, fieldwork and patients are entirely local. Editorial boards of leading journals are dominated by people in wealthy countries. Research funding flows overwhelmingly to northern institutions, which then subcontract, and overhead recovery rules often mean a southern partner cannot claim costs a northern one can. International and national staff doing similar work in the same organization are frequently on separate pay scales. Visa refusals fall on one set of passports. And the governing bodies of most global health institutions sit in Geneva, Washington, London and Seattle.

Those are facts, not interpretations, and they describe a structure in which the people closest to the problems have least control over the money, the agenda and the credit. That is the case Seye Abimbola, Madhukar Pai, Catherine Kyobutungi and others have pressed, and it is why the word colonial gets used: extracting raw material, processing it elsewhere and selling back the finished product is a recognizable pattern.

Now the objections, which deserve equal seriousness. The term is imprecise, and imprecision lets institutions perform commitment cheaply: a statement, a panel, a renamed centre, and no change to a budget line. Some argue that borrowing the moral vocabulary of anti-colonial struggle for what is sometimes a professional-class dispute about authorship and grants cheapens the original. Others warn that a framing organized around nationality risks substituting one crude sorting rule for another, and that elite researchers in capital cities may not represent rural populations any better than foreigners do. And a number of researchers in low- and middle-income countries have pointed out the irony of a decolonization debate conducted mainly in English, in journals owned in the global north, largely by people with northern affiliations.

The most useful move is to convert the argument into testable commitments. Does funding go directly to institutions in the country, with equal overhead recovery? Who holds the grant and who signs off on the budget? Are authorship criteria applied honestly? Where are the data stored and who may publish from them? Are national and international staff on one pay scale? Is the decision-making body located where the work happens? Africa CDC is the clearest institutional answer so far: established in 2017 and granted autonomous agency status by the African Union in 2022, it moved continental outbreak decision-making to Addis Ababa, and its role during COVID-19 and mpox demonstrated what that relocation buys.

Key idea: The decolonizing argument rests on documented patterns in authorship, funding flows, pay scales and governance location, and the strongest objections concern imprecision, symbolic substitution for budget change, and the risk of nationality replacing accountability, which is why testable commitments about money, data and decision rights are more useful than the label.

The scorecard

The Sustainable Development Goals set health targets for 2030. Here is the position, with the caveat that these figures move and should be checked against current sources.

Target	Goal by 2030	Approximate position	Verdict
3.1 Maternal mortality	Under 70 per 100,000 live births	Around 197 globally in 2023	Badly off track; needs roughly 15 percent annual decline against recent 1 to 2 percent
3.2 Child mortality	Under 25 under-5 and under 12 neonatal per 1,000	Around 37 under-5 globally	Partly on track; a majority of countries will meet the under-5 target, far fewer the neonatal one
3.3 AIDS, TB, malaria	End the epidemics	HIV deaths falling, TB rising again, malaria plateaued	Off track on all three, TB furthest behind
3.4 Premature NCD mortality	Reduce by one third	Slow decline, and a minority of countries on track	Off track
3.8 Universal health coverage	Coverage and financial protection	Service coverage index around 68 and stalled since 2015; hardship rising	Off track

Read this correctly, because two wrong readings are available and both are common. The cynical reading says the targets were always fantasy and nothing works. That is refuted by the data in this very course: child deaths more than halved since 1990, HIV deaths fell by roughly two thirds from their peak, and more than fifty countries eliminated at least one neglected tropical disease. The complacent reading says progress is broadly on track and needs only patience. That is refuted by the table.

The accurate reading is specific. Progress has been real, large, and uneven, it has slowed since about 2015, and it went backwards during and after COVID-19. Targets are useful because they force this kind of accounting, and missing a target is not the same as failing. The child mortality decline of the past three decades is one of the great achievements in human history and the SDG target for it will still be missed in many countries.

Key idea: Most health SDG targets will be missed, with maternal mortality furthest off track, and the correct reading is neither that nothing works nor that progress is adequate, but that large real gains slowed after 2015 and reversed during the pandemic.

Working in this field

If you want to work in global health, the most useful advice is unglamorous. Acquire a genuine skill that a health system needs: epidemiology and biostatistics, health economics, implementation science, supply chain and logistics, data engineering, health law and policy, clinical training, anthropology, or finance and audit. A general global health qualification with no underlying craft is the weakest position in the market and the least useful to any team.

Then ask one question about any role you consider, and ask it honestly. Does this job exist because a specific skill is genuinely scarce in this place, or because the funding was structured to employ someone from my country? If the second, you are occupying a post that a national professional could fill, and the appropriate response is to build that person's position rather than your own.

Short-term clinical volunteering deserves a direct warning. Students and untrained volunteers performing clinical procedures abroad that they would not be permitted to perform at home cause documented harm, undermine local providers, and disrupt continuity of care. The practice is now widely discouraged by professional bodies. Ethical alternatives exist and are better for everyone: longer placements under local supervision, work in your own area of actual competence, support functions such as data systems or logistics, and advocacy or research conducted from wherever you are.

Finally, be realistic about the sector as it stands. The contraction of major donor funding from 2025 has made global health employment smaller and more volatile than it was for the previous two decades, and it has concentrated the surviving work in national institutions, regional bodies such as Africa CDC, and domestic public health. That is arguably where the work should have been all along, and it is where the durable careers now are.

Key idea: Enter global health with a specific transferable skill rather than a general qualification, ask honestly whether a post exists because a skill is scarce or because funding was designed to employ foreigners, avoid short-term clinical volunteering, and expect the durable work to sit in national and regional institutions.

Common misconceptions

• Ethics committees settled the standard of care question. The universalist and contextualist positions were never resolved; Helsinki and CIOMS produced compromise rules on placebo justification and post-trial access.
• Written informed consent guarantees voluntary participation. Therapeutic misconception, social hierarchy and literacy barriers can make consent formal rather than substantive, which is why community engagement surrounds it.
• Decolonizing global health is only a rhetorical debate. Its empirical core includes measurable authorship patterns, funding flows, overhead recovery rules, pay scales and the location of governing bodies.
• Missing the SDG targets means global health has failed. Child deaths more than halved since 1990 and HIV deaths fell by roughly two thirds from their peak, while most 2030 targets will still be missed.
• Short-term volunteer clinical trips are a good way to contribute. Untrained volunteers performing procedures they could not perform at home cause documented harm and displace local providers.

Recap

• Research ethics runs from Nuremberg through Helsinki, Belmont and CIOMS, and justice is the principle most often at stake in global health.
• The perinatal HIV placebo trials of the 1990s set a universalist standard of care against a contextualist one, producing compromise rules rather than resolution.
• Sample and data export drove the Nagoya Protocol, Indonesia's 2007 influenza sample refusal, and the benefit-sharing systems that followed.
• The decolonizing argument rests on documented authorship, funding, pay and governance patterns, and its serious critics warn against symbolic substitution for budget change.
• Most health SDG targets will be missed, maternal mortality furthest, even though child mortality and HIV outcomes improved enormously since 1990.
• Careers should rest on a specific craft, and the durable work increasingly sits in national and regional institutions.

Sources

1. World Medical Association. (n.d.). Declaration of Helsinki: Ethical principles for medical research involving human subjects. wma.net
2. Council for International Organizations of Medical Sciences. (n.d.). International ethical guidelines for health-related research involving humans. cioms.ch
3. Abimbola, S., and Pai, M. (2020). Will global health survive its decolonisation? The Lancet, 396(10263), 1627-1628. doi.org/10.1016/S0140-6736(20)32417-X
4. United Nations. (n.d.). Sustainable Development Goal 3: Good health and well-being. sdgs.un.org
5. Africa Centres for Disease Control and Prevention. (n.d.). About Africa CDC. africacdc.org
6. Secretariat of the Convention on Biological Diversity. (n.d.). The Nagoya Protocol on access and benefit-sharing. cbd.int