The six global health security agendas that will decide the next decade — and what they demand of a country like Nepal

Every year the global health community produces a list of priorities, and every year the distance between that list and a municipal health budget in a federal, low-income country stays roughly the same. I spend my working life on the second side of that distance: health systems strengthening under Nepal's federal structure, emergency preparedness, health information systems, and the translation of evidence into routine practice. This article is written from that vantage point. It takes the six agendas that genuinely dominate global health security in 2026, states where each one actually stands according to primary sources, and then asks what each one requires of the people who have to implement it.

I have tried to be disciplined about two things. First, every figure here is attributed to a named institutional source with a date, and the reference list at the end carries the full citations. Second, where a widely repeated number could not be verified against its primary source, I have left it out rather than launder it through another article. That is not caution for its own sake: the credibility of an implementation argument rests entirely on whether its numbers survive checking.

20 May 2025

WHO Pandemic Agreement adopted at WHA78

19 Sep 2025

Amended IHR (2005) entered into force

14.4 million

Dengue cases reported to WHO in 2024

1 in 6

Common bacterial infections resistant to treatment, 2023

1. Pandemic prevention, preparedness and response: a treaty adopted, an annex unfinished

The World Health Assembly adopted the WHO Pandemic Agreement by consensus on 20 May 2025 — the first legally binding instrument negotiated under WHO auspices since the Framework Convention on Tobacco Control. That was a genuine diplomatic achievement after three years of negotiation, and it is also where most public commentary stopped.

What happened afterwards matters more. The Agreement's operative mechanism for sharing pathogen samples and sequence data in exchange for equitable access to vaccines, therapeutics and diagnostics — the Pathogen Access and Benefit-Sharing (PABS) system under Article 12 — was deliberately left to a separate annex to be negotiated by an Intergovernmental Working Group. That annex is not finished. WHO Member States agreed in May 2026 to extend the PABS negotiations, and a further negotiating session was held in Geneva in July 2026. Until the annex is concluded and the Agreement has attracted the ratifications it needs, the treaty's central equity bargain is text rather than practice.

The instrument that is legally live is the older one. The 2024 amendments to the International Health Regulations (2005), adopted by WHA77 under resolution WHA77.17 on 1 June 2024, entered into force on 19 September 2025 for States Parties that did not reject them. This is the change that has immediate consequences for national systems: the IHR are binding on 196 States Parties, and the amendments sharpen obligations around national authorities, core capacities and event notification.

The capacity gap the amendments do not close

Legal obligation and operational capacity are different things, and the gap between them is where preparedness fails. Nepal's Joint External Evaluation — conducted from 28 November to 2 December 2022 using the third-edition tool and published in May 2023 — returned uneven scores across capacities, with recurring weakness at points of entry, in the multidisciplinary workforce, and in risk communication. Nepal's 2022 IHR self-assessment (SPAR) scored the legislation-and-financing capacity at 50% on average, with legislation itself at 40%.

The sequence since then is instructive. Nepal ran a national IHR planning workshop using the NAPHS tool in Kathmandu in November 2025 and finalised its IHR annual State Party report in February 2026, with support from the Pandemic Fund. That is real progress. It is also a country receiving external technical and financial support to complete a reporting obligation that has existed since 2007 — which tells you something structural about how IHR core capacities have been financed globally.

The financing itself is modest against the task. The Pandemic Fund's second call, approved by its Governing Board in October 2024, allocated US$418 million across 50 countries from a US$500 million envelope announced in December 2023; FAO alone is co-implementing 21 of those projects, worth US$109 million across 32 countries, which reflects how central animal-health systems have become to the preparedness agenda.

The pathogens that are actually testing the system

"Disease X" is a useful planning abstraction, but the events of 2025 were concrete. Tanzania declared the end of a Marburg virus disease outbreak on 13 March 2025 after 42 days with no new cases — two confirmed and eight probable cases, all fatal. On 14 November 2025 Ethiopia confirmed its first-ever Marburg outbreak, in Jinka. In December 2025 WHO reported broader local transmission of clade Ib mpox in countries that had previously reported only travel-associated cases or none. And on 15 November 2025 WHO was notified of a human case of avian influenza A(H5N5) in the United States — the 71st confirmed human influenza A(H5) case reported there since early 2024, and a subtype distinct from the H5N1 strain circulating in dairy cattle. The FAO/WHO/WOAH joint assessment of July 2025 continued to rate the overall public-health risk from these H5 events as low, rising to low-to-moderate for occupationally exposed people.

2. Climate change and One Health: the vectors have already moved

The institutional architecture here is the Quadripartite One Health Joint Plan of Action, led by FAO, UNEP, WHO and WOAH, whose mandate was extended to 2029 in a Quadripartite statement of August 2026 following the One Health Summit and executive meeting earlier that year. Extending a plan of action is a signal of continued political commitment. WHO's own technical review of the enablers and barriers to implementing that plan is the more honest document: it points to financing and national multisectoral coordination as the constraints, which is to say the plan's problems are the same problems it had at the start.

Meanwhile the epidemiology is not waiting for the coordination mechanism. WHO's Weekly Epidemiological Record update on dengue reported 14,434,584 cases for 2024, of which 7,718,585 were laboratory-confirmed, alongside 52,738 severe cases and 11,201 deaths, across all six WHO regions — a level WHO describes as unprecedented worldwide. An independent analysis in the International Journal of Infectious Diseases put 2024 at roughly 14.1 million cases and around 9,000 recorded deaths, a twofold increase over a 2023 figure that had itself broken records. The two death counts differ because of reporting cut-offs and case definitions; I cite both rather than choosing the more dramatic one.

Malaria shows the other half of the pattern — hard-won progress that has stalled. The World Malaria Report 2024 estimated 263 million cases and 597,000 deaths in 2023, with cases up by roughly 11 million on 2022 and about 95% of deaths in the WHO African Region. Cumulatively since 2000, an estimated 2.2 billion cases and 12.7 million deaths have been averted, and 44 countries plus one territory have been certified malaria-free. Both facts are true at once: the tools work, and the trajectory is going the wrong way.

Vector-borne disease is where climate change stops being an environmental abstraction and becomes a case definition in a district hospital register.

For Nepal the operational implication is specific. A country with tropical Tarai plains, mid-hills and high mountains has, in effect, a vertical climate gradient along which vector habitat can migrate. Dengue transmission in Nepal is no longer confined to the historically endemic lowland districts, and hill and valley districts have recorded transmission in recent years. I would not quote a national case figure here without taking it directly from the current EDCD or WHO Nepal surveillance bulletin, and neither should any programme document — the numbers move seasonally and secondary sources lag. The design conclusion does not depend on the exact figure: entomological surveillance, clinical case management protocols and municipal vector control need to exist in districts that have no institutional memory of dengue at all.

• Treat altitude as a moving boundary, not a fixed protection: plan surveillance for districts that are currently marginal, not only those currently endemic.
• Pair human case surveillance with entomological surveillance, or you will detect expansion only after an outbreak.
• Give One Health a budget holder. A memorandum between ministries that no one funds produces coordination meetings, not shared data.
• Write heat into health service planning — cooling, water, triage load and outdoor-worker protection — as a seasonal operational plan, not a climate policy statement.

3. Antimicrobial resistance: the one where the surveillance is finally good enough to be frightening

The UN General Assembly held its high-level meeting on antimicrobial resistance on 26 September 2024, and the resulting political declaration — annexed to a General Assembly resolution adopted in October 2024 — committed to reducing global deaths from bacterial AMR by 10% by 2030, alongside commitments on funded national action plans and sustainable financing. Notably, the declaration's quantified content is largely that single mortality target; I could not verify a specific headline financing figure inside it, and I will not invent one.

The surveillance evidence published a year later is the part that should change behaviour. WHO's Global antibiotic resistance surveillance report 2025, released on 13 October 2025, found that one in six laboratory-confirmed bacterial infections causing common infections worldwide in 2023 were resistant to antibiotic treatment, and that resistance increased in more than 40% of the pathogen-antibiotic combinations monitored between 2018 and 2023. The Lancet's GRAM analysis of the global burden of bacterial AMR from 1990 to 2021, with forecasts to 2050, published in September 2024, remains the reference source for burden estimates and projections — and if you are citing its headline mortality numbers in a policy document, take them from the paper itself rather than from any summary, including this one.

This is where AMR becomes a health information systems problem, which is the territory I work in. A national AMR strategy without a functioning laboratory information pipeline produces prevalence figures that are artefacts of which hospitals happen to culture. Add over-the-counter antibiotic dispensing and weakly regulated veterinary use — documented concerns across South Asia — and the resulting national estimate is not a measurement of resistance but a measurement of where testing happens.

1. 01Fix the denominator first: know how many specimens are cultured, in which facilities, for which syndromes.
2. 02Enrol sentinel laboratories with quality assurance before scaling coverage; wide-but-unreliable surveillance is worse than narrow-and-reliable.
3. 03Route susceptibility data into the routine health information system, not a parallel project database that ends with the project.
4. 04Make stewardship a measurable clinical process — prescription audit, restricted formulary, documented indication — not an awareness campaign.
5. 05Bring veterinary and agricultural antimicrobial use into the same reporting frame, or the One Health commitment stays rhetorical.

4. Global health architecture and financing: the base is contracting while the agenda expands

This is the agenda that determines whether the other five are affordable, and it moved sharply in the wrong direction in 2025. The OECD reports a 9% fall in official development assistance in 2024 — the first time in nearly three decades that France, Germany, the United Kingdom and the United States all cut ODA in the same year — and projects a further fall of 9% to 17% in 2025. IHME's Financing Global Health 2025 and its companion Lancet analysis of development assistance for health from 1990 to 2030 document what that means specifically for health aid.

WHO's own finances moved in parallel. The Organization launched its first Investment Round in 2024, with a pledging moment at the G20 Summit in Rio de Janeiro in November 2024, and raised at least a further US$210 million at a WHA78 pledging event in May 2025. Those sums sit against a programme budget for 2026-27 that was revised downward during the same period, following the announced withdrawal of its largest historical contributor. The reduced budget still leaves a financing gap, and specialist reporting has put figures on that gap that I would want to confirm against the Executive Board's own financing documents before quoting.

The reform agenda that anticipated some of this is the Lusaka Agenda, published on Universal Health Coverage Day in December 2023 as the conclusion of the Future of Global Health Initiatives process. Its logic is exactly right: shift global health initiatives towards country-led planning, one plan and one budget, strengthened domestic systems, and decision-making closer to the countries that carry the disease burden. Its weakness is accountability — as the Center for Global Development noted in 2024, metrics to track delivery on the Lusaka Agenda had not yet been established.

In a federal system this is not an abstraction. Nepal devolved substantial health responsibility to 753 local governments; when external financing tightens, the question becomes whether a municipality's conditional grant, its own revenue and its planning capacity can hold service delivery steady. That is precisely the work of municipal health planning and budgeting, and it is why I regard public financial management as a health security capability rather than an administrative one.

5. Digital health and AI: from pilot to infrastructure, with governance running behind

The World Health Assembly endorsed the extension of the Global Strategy on Digital Health from 2025 to 2027 in May 2025, and mandated the development of a successor strategy for 2028-2033. An extension is an admission: the original strategy's implementation targets were not met on schedule, and the reasons are the familiar ones — fragmented systems, weak interoperability, and workforce capacity.

On artificial intelligence, the governance texts exist and are usable. WHO issued its first global report on the ethics and governance of AI for health in June 2021, setting out six guiding principles for the design and use of AI in health, and followed it in January 2024 with dedicated guidance on large multi-modal models — the generative systems that the 2021 report could not have anticipated. If you are writing those six principles into a national policy, quote them from the WHO publication itself; they are paraphrased inaccurately more often than not, and I am not going to add another paraphrase to the pile.

The regulatory picture is a useful reality check on how fast "AI as core infrastructure" is actually arriving. Under the EU AI Act, which entered into force in August 2024 with prohibitions applying from February 2025, the high-risk obligations most relevant to clinical AI were subsequently deferred: Annex III systems, which cover many clinical decision-support tools, now apply from December 2027, and Annex I systems, which include AI-enabled medical devices, from August 2028. A jurisdiction with substantial regulatory capacity has already pushed its medical-AI deadlines back once. Any national plan that assumes faster institutional readiness in a lower-resource setting deserves scrutiny.

And there is a prior constraint that no model improves. Health facilities without reliable electricity, connectivity, device maintenance or local-language interfaces cannot run digital tools of any kind, and the workforce that would use them has other duties. The digital divide inside a country is usually larger than the divide between countries. A digital health strategy that does not begin with connectivity, power, maintenance and training budgets for the lowest-tier facilities is a strategy for the capital city.

• Interoperability before intelligence: standards-based data exchange is what makes any later analytics possible.
• Governance artefacts, not intentions: documented purpose, data provenance, named human accountability, bias assessment and an audit trail for every automated decision.
• Evaluate in the setting of use. Performance reported in a high-income validation set tells you very little about a rural district.
• Budget for maintenance at the same time as procurement, or the system degrades into a parallel paper process.

6. Non-communicable diseases: the largest burden, still financed as a secondary concern

The scale is not disputed. WHO's fact sheet, updated in September 2025, records that non-communicable diseases killed at least 43 million people in 2021, equivalent to 75% of non-pandemic-related deaths worldwide; 18 million people died from an NCD before the age of 70, and 82% of those premature deaths occurred in low- and middle-income countries. Cardiovascular disease accounts for the largest share, at least 19 million deaths. WHO's health-topic page states a slightly different set of percentages — 74% of all deaths, and 86% of premature deaths in LMICs — reflecting different update cycles; I cite both rather than pretending to a precision the source data does not support.

The political response arrived in 2025. The fourth UN High-Level Meeting on NCDs and mental health was held on 25 September 2025, and the General Assembly formally adopted its political declaration on 16 December 2025, reaffirming the SDG 3.4 commitment to reduce premature NCD mortality by one third by 2030. That is now the fourth such declaration since 2011, and the target it reaffirms is off track.

For Nepal, the most recent nationally representative risk-factor evidence remains the STEPS survey of 2019, published in 2021 — which means national NCD policy is being made against data that predates the pandemic. WHO's 2023 tobacco country profile documents Nepal's MPOWER policy scores and notes that cigarette affordability decreased relative to 2012. I have not been able to confirm a published successor to Nepal's multisectoral NCD action plan for 2014-2020, and that gap, if it is real, is the single most consequential item on this list for domestic policy: a reaffirmed global target with no current national plan and no current national risk-factor survey behind it.

What links all six: the implementation deficit is the agenda

Read together, these six agendas share a single failure mode. Each has strong normative content and adequate technical guidance. Each is constrained at exactly the same layer: legal mandate, financed workforce, functioning data, and accountable subnational institutions. A pandemic treaty cannot compensate for a district that cannot verify a signal. An AMR mortality target cannot be measured where cultures are not taken. A digital health strategy cannot run through facilities without power. An NCD declaration cannot bend a mortality curve without a supply chain and a follow-up register.

This is why I frame my own work as health systems strengthening and implementation science rather than as any single disease programme. The frameworks used to study implementation — the Consolidated Framework for Implementation Research, RE-AIM and PRISM, and the Proctor taxonomy of implementation outcomes — exist precisely because the failure is systematically located between the evidence and the practice, and because that space can be studied rather than merely lamented.

1. 01Assume less external financing, not more. Design core preparedness, surveillance and NCD functions to survive a project's closure.
2. 02Treat legislation and public financial management as health security capabilities. They score low in assessments because they are unglamorous, and they block everything else.
3. 03Invest in routine data before analytics. Completeness, timeliness and internal consistency are prerequisites for every AI ambition on the agenda.
4. 04Push capacity to the level that holds the mandate. In a federal system, that means municipal planning, procurement and supervision skills, not national workshops.
5. 05Publish uncertainty. A programme that states what it does not know is easier to correct than one that reports a confident number nobody can trace.

If you work on any of these six agendas in a low- or middle-income setting and want to discuss how the international commitments translate into a plan your institution can actually execute, my contact details are on the site. The reference list below carries every source used, so you can check the argument rather than take it on trust.