Health and Healthcare
The system that touches every life. The cost crisis, the aging math, the pandemic lessons mostly unlearned, and the AI revolution arriving in the clinic.
(about 10% of world GDP; World Bank figures)
(16.7% in 2023 by the World Bank measure, 18.0% in 2024 by US national accounts; roughly twice the developed-country average)
(UN estimate; about 51 in 1960; back near its pre-pandemic level after a dip to 71 in 2021)
A note on framing. Healthcare is one of the topics where almost every reader has personal experience that contradicts at least some of the policy story they have been told. There is no clean ideological answer to the healthcare question; there are real trade-offs that different societies have made differently. The data on which approaches deliver what is more useful than the slogans usually suggest, and the page below walks through it: what countries actually spend, what they get for it, and where the system is moving.
The size of the global healthcare bill
The world spent about $10.6 trillion on healthcare in 2023 - about a tenth of the total economy. That share has been rising in nearly every country for decades. There is no developed country where healthcare's share of national income has fallen meaningfully over the last fifty years. Healthcare is one of the few things people want more of as they get richer, and that weighs on every public budget.
Within that global total, the picture is sharply uneven. The United States alone spent about $4.5 trillion - more than two-fifths of the world's total - on about 4% of the world's population. The next-largest spenders (China, Japan, Germany, France, the UK, all collectively) make up roughly the next quarter. The rest of the world, including all of Africa, all of South Asia, most of Southeast Asia, and most of Latin America, splits the remaining 30% or so. By spending per person the differences are enormous: the US around $13,500 a year, most rich European countries between $5,000 and $7,000, China about $760, India about $85, much of Sub-Saharan Africa under $50. Those are all World Bank figures for 2023, so they compare like with like; the US national accounts put 2024 spending at $5.3 trillion, or about $15,500 a person, on the broader definition described below.
The gap between what a country spends and what it gets is the central puzzle of comparative healthcare economics. Spending more does not automatically buy more health. Costa Rica, spending about $1,150 per person, achieves life expectancy of about 81. Cuba is often cited alongside it, though its health statistics are state-reported and have been questioned in peer-reviewed work, and its life expectancy fell further than Costa Rica's during the pandemic. Several rich European countries with half of US per-person spending have longer average lives. Singapore spends about $4,000 per person on healthcare and has one of the world's longest life expectancies. The relationship between spending and outcome is loose at best, and that is one of the most important facts in the field.
Why the US number moves depending on where you read it. These charts use the World Bank series, which counts current health spending and puts the US at 16.7% of GDP in 2023. The US government's own national accounts, published by the Centers for Medicare and Medicaid Services, put it at 18.0% in 2024, because they also count capital investment in buildings and equipment, research, and public-health infrastructure. Neither is wrong. The World Bank figure is the one built for comparing countries, which is what these charts are for; the domestic figure is the one that matches the US budget debate. Both show the same shape, including the bump in 2020 when GDP fell faster than health spending did.
The aging math behind everything
The single most powerful force on global healthcare over the next thirty years is aging. The number of people over 65 is rising in every region (with a small lag in Sub-Saharan Africa). Medical spending per person rises sharply with age - roughly three to five times as much per year on care for someone over 75 as for someone in their 40s. As populations age, the per-person spending rises even when individual care patterns do not change.
The math runs into hard ceilings everywhere. Most pay-as-you-go pension and healthcare systems were built when there were four or five working-age adults per retiree. They are now drifting toward two and a half to three. By 2050 the developed-world ratio will be in the high 1s in many countries. The taxes and contributions that working-age adults pay cannot rise fast enough to keep the same per-retiree spending without a much larger underlying economy or sharply higher productivity. Something has to give: benefits, taxes, retirement age, immigration, or some combination.
Each developed country is solving this in its own way, with different politics. Germany, Japan, and Italy are slowly raising retirement ages and tightening benefits. France raised its pension age over mass protests in 2023, then suspended the rise in December 2025 until 2028 - a measure of how politically expensive the choices are. The US has not closed Medicare's projected shortfall: its hospital trust fund is due to run out in 2033, forcing an 11% payment cut unless Congress acts (2026 Trustees Report). The UK's National Health Service (NHS) has been under chronic funding stress as more of the population ages into heavier use. None of these is a crisis this year. Each is a slow-moving force that will dominate domestic politics for decades.
How country systems compare
The healthcare systems of different rich countries differ in ways that affect both costs and outcomes. The numbers below are rough - mostly World Bank data for 2023-24 - and meant to give a comparative sense rather than precise figures.
The takeaway: every developed country except the US has built a system that achieves universal coverage at substantially lower cost than the US, with comparable or better health outcomes. The specific structure varies (single-payer in Canada, multi-payer in Germany, mandatory insurance in Switzerland, mandatory savings in Singapore), but each covers essentially everyone. The US spends more per person than any peer country and still left 26.7 million people, 7.9% of the population, uninsured for all of 2025 (Census Bureau). The reasons are well documented, even where politically difficult to address.
Why US healthcare costs are different
The American cost gap is the single largest healthcare-policy puzzle in the developed world. Multiple research traditions have studied it, and the consensus is reasonably clear: the gap is mostly about prices, not utilization. Americans do not get more healthcare than residents of comparable countries; they pay more for each unit of care.
Some specific drivers. Brand-name drugs cost three to four times as much in the US as in other rich countries, though generics cost less (RAND, 2022 prices). One reason is that Medicare could not negotiate drug prices until a 2022 law; its first negotiated prices took effect in 2026, and in 2025-26 the Trump administration struck "most-favored-nation" pricing deals with 26 manufacturers. GLP-1 obesity drugs show how prices vary: the first US obesity pills, approved in December 2025 and April 2026, cost $149-299 a month in cash, while generic semaglutide has sold in India from under $25 a month since March 2026, when the patent there expired. Hospital charges are not standardized; the same procedure can cost wildly different amounts at different facilities and for different patients. Administrative overhead consumes roughly a quarter to a third of healthcare spending in the US versus a fraction of that in most single-payer systems, because every claim has to be negotiated separately between thousands of providers and hundreds of insurers. Specialist physicians earn more in the US than in any peer country. Medical malpractice and defensive medicine add costs that other countries handle differently.
None of these is the single decisive factor. Together they add up to a roughly $1.5 trillion a year cost premium that the US pays without obviously better outcomes. The political reason this is hard to fix is straightforward: every dollar of "waste" is revenue for someone - a hospital, an insurer, a drug company, a specialist - and each constituency has lobbyists, campaign contributions, and genuine arguments for why their slice of spending is justified. The system is sticky in the same way the wealth-concentration system covered elsewhere on this site is sticky: too many specific actors benefit from the current arrangement to make broad reform easy.
The pandemic and what was learned
The COVID-19 pandemic killed roughly seven million people worldwide by official counts. The World Health Organization (WHO) estimates 22.1 million excess deaths in 2020-23, including indirect ones. It produced the deepest global recession since World War II, the most rapid vaccine development in history, and a body of policy lessons that have been only partially absorbed.
What worked: rapid mRNA vaccine development built on years of underlying research, and emergency genomic sequencing networks. So did some national responses that combined competent public-health communication with adequate hospital capacity (Korea, Taiwan, New Zealand, Vietnam in the first year). The WHO coordinated much of the global response despite its political constraints, though the United States cited its handling of COVID-19 when it left the organization in January 2026. What did not work as well: the global cooperation needed to get vaccines to lower-income countries quickly, and the public-health communication in many countries. Preparedness investments promised after SARS-1, Ebola, and H1N1 were largely not made, and public-health measures proved politically fragile once the emergency phase passed.
The longer-run effects are still unfolding. Trust in public-health institutions has fallen in many countries (see the Trust and Legitimacy page); in the United States the argument now reaches the childhood vaccine schedule. After Health Secretary Robert F. Kennedy Jr. replaced the vaccine advisers of the Centers for Disease Control and Prevention (CDC) in June 2025, the agency in January 2026 cut the diseases it recommends vaccinating all children against from 18 to 11, saying a focused schedule would improve adherence and confidence. The WHO's director-general said the changes were "not aligned with decades of evidence," and the American Academy of Pediatrics kept its broader schedule and sued. In March 2026 a federal judge stayed the new schedule, the new advisers' appointments and their votes as likely unlawful; the government has appealed. An August 2026 executive order then called for a new schedule within 90 days and for splitting the combined measles, mumps and rubella shot into three. Measles has returned where vaccination fell below the 95% needed to stop it: the Americas lost their elimination status in November 2025, and US cases reached 3,659 by late September 2026, against 285 in all of 2024. The WHO Pandemic Agreement adopted in 2025 cannot be ratified until its annex on sharing pathogens is adopted, now sought for May 2027. Modeling cited by the Center for Global Development puts the chance of another pandemic as deadly as COVID-19 at about even over the next 25 years. Preparedness is better in some ways (genomic surveillance, vaccine platforms) and worse in others; whether that pandemic is handled better depends on choices that are not currently being made.
AI and the new clinic
Artificial intelligence is reaching healthcare delivery at scale, and the implications are starting to be visible. Large-language-model assistants are being deployed in clinical documentation (where doctors spend a substantial portion of their time on paperwork) and patient triage (where AI-aided early diagnosis is being trialed in primary care). Other AI tools support diagnosis, where image analysis has matched specialists in some narrow domains, and drug discovery, where AlphaFold has changed how protein structures are computed.
The promise is real: documentation burden reduction would free clinician time for actual care; better diagnostic support could reduce errors that cause measurable harm; drug discovery acceleration could meaningfully lower costs. The risks are also real. AI errors in clinical settings have specific liability and safety implications, the data on most AI medical tools is still thin, and there is a meaningful chance of substituting AI's confident wrongness for human uncertainty in ways that produce worse outcomes rather than better. The deployment so far has been faster in administrative and supportive functions and slower in directly clinical ones, where the stakes are higher.
Two specific changes are likely over the next decade. The first is dramatic improvement in radiology, pathology, and dermatology - fields that involve pattern recognition on images where AI is already approaching specialist-level performance on common cases. The second is a steady expansion of AI-aided primary care in lower-income countries where specialist scarcity is acute. Combined with the global rise of telemedicine since the pandemic, this could meaningfully improve healthcare access for people who rarely see a specialist. Whether the developed-world clinical workflow changes as fundamentally is less clear; the regulatory, liability, and cultural barriers are higher.
The paths from here
Dated September 2026, with US Medicaid cuts phasing in, donor health aid at its lowest since 2009, and an Ebola emergency in the Democratic Republic of the Congo. Healthcare is unusual among the topics on this site in that it changes slowly and predictably until it changes suddenly. Each path below is one realistic shape the next two decades could take.
US healthcare slowly evolves
The US system continues its slow accretion of partial reforms - drug-price negotiation expanded, more states using their authority to control costs, more employers experimenting with reference-pricing and direct-contracting, gradual expansion of Medicare-style options for under-65 adults. No comprehensive overhaul; many incremental changes. By 2040 the system is materially different from today without ever being formally restructured.
Will it happen? This is the base case for the US given its political structure. The Affordable Care Act of 2010 was the largest single change in 50 years and was incremental within the existing structure. The changes of 2025-26 fit the pattern, in both directions: Medicare began paying for obesity drugs in July 2026 under a temporary program, at $50 a month to the patient, while enhanced marketplace subsidies expired at the end of 2025. The 2025 budget law adds Medicaid work requirements from 2027, which its supporters in Congress say will "restore the dignity of work"; the Congressional Budget Office estimates that it cuts federal Medicaid spending by close to $1 trillion over ten years and will leave 10 million more people uninsured in 2034. Future changes are most likely to be similar - real and consequential without being transformational.
The aging cost wave hits
By the early 2030s the cost pressures on Medicare, the NHS, and equivalent systems become impossible to defer. Some combination of higher contribution rates, reduced benefits, raised retirement ages, and rationing of expensive end-of-life care becomes the actual policy in most developed countries. The political conflict over which group bears which cost is the dominant domestic policy story of the decade.
Will it happen? Some version is virtually certain; the demographics are already set. The form the response takes will vary enormously by country - some through formal policy, some through gradual erosion, some through crisis-driven reform after a system is forced to default on commitments. Few countries have started the politics of this seriously.
Pandemic risk grows quietly
Without sustained investment in pandemic preparedness, the world becomes more rather than less vulnerable to the next major outbreak. Genomic surveillance networks fray. Antibiotic resistance worsens. Investment in vaccine platforms slows. When the next pandemic arrives, the world handles it less well than it did COVID-19.
Will it happen? The trajectory points this way absent serious investment that is not currently happening. The WHO has cut its two-year base budget by about 14%, and an Ebola epidemic in the Democratic Republic of the Congo, a WHO emergency since May 2026, had killed 3,799 people by September 23. The good news is that the technical capacity built during COVID-19 (mRNA platforms, sequencing networks, diagnostic ramping) does not fully disappear; the bad news is that the political infrastructure for using it has weakened. The probability of a worse-handled next pandemic is meaningful and rising.
AI quietly transforms primary care
By the early 2030s, AI-aided diagnosis, decision support, and patient interaction become routine across most healthcare systems. Specialist-quality care becomes available remotely in lower-income countries. Documentation burden falls and clinician time shifts toward direct patient interaction. Rising productivity in healthcare partially offsets the aging cost wave.
Will it happen? Some version is likely. The pace depends on regulatory choices, on whether AI systems can demonstrate the safety records that medicine reasonably demands, and on whether the savings flow through to lower costs or get absorbed by other parts of the system. The most plausible version is uneven - faster in some specialties and countries than others.
Global healthcare divergence widens
The healthcare gap between richer and poorer countries widens further as donor cuts and aging-related cost pressures crowd out international health spending. The next set of expensive medical breakthroughs (gene therapies, advanced cancer treatments, precision medicine) becomes available primarily to wealthy patients. The Sub-Saharan African and South Asian healthcare gap with the rest of the world grows even as overall outcomes continue to improve.
Will it happen? The trajectory points this way absent specific countervailing policy. Development assistance for health fell from $80 billion in 2021 to an estimated $38 billion in 2025, the lowest since 2009, as the United States and other donors cut aid (Institute for Health Metrics and Evaluation). One Lancet modeling study projects more than 14 million additional deaths by 2030 if the US cuts are not reversed; the US government says its new bilateral health agreements keep paying for frontline medicines and health workers while cutting what it calls overhead. The next generation of expensive treatments will be largely available only to wealthy patients in wealthy systems. The global health story over the next twenty years has both real progress and real divergence in it.
The base case is slow drift in the US system, aging cost pressures hitting most rich countries by the late 2020s, and AI gradually reshaping clinical practice. The downside scenarios (worse pandemic handling, widening global divergence) are real but not certain. The healthcare story is more determined by demographics and politics than by technology, and the demographics are largely fixed for the next twenty years.
Where serious analysts disagree
Healthcare is one of the topics where careful researchers using similar data reach different conclusions about what to do. Each reading below is held by named scholars worth engaging directly.
The US system is structurally exceptional and very hard to fix
The US healthcare cost premium is not a bug that can be fixed through better policy at the margins. It is the predictable result of a system in which most spending happens through private intermediaries with strong political constituencies and prices are not set by any central authority. The political coalitions needed to reform it have repeatedly failed for sixty years. Significant reform requires a level of political consensus that has not existed in living memory and probably will not appear without a major crisis.
Held by: Aaron Carroll (AcademyHealth, previously Indiana), Atul Gawande (Harvard), and a substantial fraction of US health-policy researchers. Their position is uncomfortable politically; they rest it on the comparative data and the history of US reform attempts.
The cost crisis is mostly about prices, not utilization
Americans do not get more healthcare than people in peer countries; they pay more for each unit. Reform that focuses on reducing utilization (managed care, prior authorization, deductibles, copays) misses the actual problem and produces its own harms. Reform that focuses on prices (Medicare drug negotiation, hospital price transparency, all-payer rate setting) addresses the actual cost driver.
Held by: Uwe Reinhardt (the late Princeton health economist whose framing has shaped the field), Sherry Glied (NYU), and a body of price-focused health economics research. Their case is supported by the international comparison data and the limited success of utilization-focused reforms. US drug policy has moved their way under both parties: Medicare price negotiation under Biden, "most-favored-nation" deals under Trump.
Healthcare is fundamentally improvable through technology
The pessimism about healthcare reform misses the most important variable: the field is on the cusp of substantial productivity gains from AI, genomics, advanced imaging, and personalized medicine. Treating the cost-and-access problem as a static distributional fight ignores the possibility that improving the underlying technology meaningfully shifts the constraint set. The next 20 years will look more different from the previous 20 than the previous 20 looked from the 20 before that.
Held by: Eric Topol (Scripps), and a tradition of technology-optimist healthcare analysis. Their case is strongest on diagnostic and informational dimensions and weakest on the parts of healthcare that depend on human-time-intensive care that AI does not directly substitute for. Both sides are partly right.
Pandemic preparedness has not improved and is in some ways worse
The political coalition for sustained pandemic-preparedness investment has eroded since 2022. The WHO Pandemic Agreement adopted in 2025 is weak on enforcement and cannot be ratified until its annex on sharing pathogens is adopted, now sought for May 2027. Genomic-surveillance networks have fragmented. The institutions that worked partially well during COVID-19 (the WHO, the CDC, the European Medicines Agency) have lost trust and resources. The next pandemic will arrive in worse strategic conditions than COVID-19 did.
Held by: Larry Brilliant (epidemiologist who helped eradicate smallpox), Bill Gates (in his foundation work), and a body of careful pandemic-policy analysis. Their warnings have been consistent. Others point to faster vaccine platforms, wider genomic sequencing, and amended International Health Regulations, in force since September 2025, that add a "pandemic emergency" alert level (the United States was among 11 states that rejected them).
Universal healthcare is achievable politically in the US given the right framing
Polling consistently shows majority US support for universal healthcare in the abstract. The political failure has been in translating that abstract support into specific reforms that survive the lobbying response. With the right framing - specifically, building on Medicare expansion rather than replacing the employer system, and framing the reform as cost reduction rather than coverage expansion - meaningful progress is more politically possible than a generation of failed reform attempts suggests.
Held by: Paul Krugman, Henry Aaron (Brookings), and a strand of progressive health-policy thinking. Their argument has been around for decades; whether the political conditions for it are improving or worsening is contested.
What the available evidence supports: the US system is exceptional and hard to fix at scale, and price-focused reform is more promising than utilization-focused reform. Technology will help, but not as much as enthusiasts hope. Pandemic preparedness has weakened in money and cooperation even as the tools have improved, and universal coverage in the US is more politically constrained than the abstract poll numbers suggest. Healthcare is one of the topics where the underlying conditions matter more than the policy details.
What this means for you
Healthcare is among the most personally relevant topics on this site. A few practical observations:
If you live in the United States
Your healthcare is more expensive and more uncertain than residents of any peer country face. Maintaining continuous insurance coverage, understanding your specific plan, building a relationship with a primary-care physician who knows your history, and saving against the high deductibles and out-of-pocket exposures American plans typically include are all worthwhile in ways that they would not be elsewhere. If you buy marketplace coverage or rely on Medicaid, recheck your costs and eligibility each year. Enhanced marketplace subsidies ended with 2025, and Medicaid work requirements for many adults begin by January 2027. The system is not going to fix itself in time to help you personally; planning around its actual structure is more useful than waiting for reform.
If you are aging or care for an aging parent
Long-term care is the largest healthcare cost most families encounter, and it is underinsured in almost every system. Long-term care insurance is expensive, has narrowed in availability, and produces mixed results when it pays out. Government programs pay for some long-term care but typically only after assets are exhausted. Family caregiving (mostly by adult daughters and daughters-in-law) absorbs much of the burden invisibly. Talking about long-term care preferences with aging parents before a crisis - including end-of-life choices - is uncomfortable and is one of the most consequential family conversations most people have.
If you have employer healthcare
Most American workers have insurance tied to their job, with implications most do not fully think through. Job changes mean insurance changes (and possibly new deductibles starting over). Self-employment, retirement before Medicare age, and gaps between jobs all carry healthcare risk that is hard to plan around. The structure subsidizes larger employers (who can negotiate better insurance) over small ones, and rewards staying in jobs that are otherwise no longer the best fit. None of this is your fault as an individual; understanding it helps you plan around it.
If you live outside the United States
Your country probably delivers healthcare more efficiently and more universally than the US, but each system has its own pressures - the NHS waiting lists, the European cost escalations, the Asian aging math. Engaging with the specific debates in your country (rather than the simpler "we're better than the US" framing) is more useful. Most of the rich-country healthcare systems are under stress that will produce real changes over the next decade. What those changes look like depends on local politics that benefit from informed citizen engagement.
If you work in technology or healthcare
The intersection of AI and healthcare is one of the largest single opportunities in technology over the next decade, with serious challenges around safety, regulation, and clinical workflow. Clinical training programs, healthcare-management roles, and the technology-meets-medicine intersection are all growing areas of need. The supply of trained clinicians is tight in most rich countries; the supply of people who can responsibly bridge clinical and technical worlds is even tighter. Anyone with both skill sets is in unusually high demand.
The mechanics behind this
The healthcare story sits on top of three deeper mechanisms covered elsewhere on this site. If the analysis above depends on ideas you want to understand first, these fundamentals make the conversation more legible:
