AMR does not have one data problem.

Antimicrobial resistance is often described as a data problem. That description is partly correct. Important surveillance gaps remain, data quality varies substantially between countries and sectors, and major blind spots persist across human health, animal health, agriculture and the environment.

But there is another problem that receives less attention. Even where data exists, it is often difficult to connect.

A national action plan may describe what a country intends to do. A surveillance system may report resistance patterns. A development programme may finance laboratory capacity. A ministry budget may fund infection prevention. A donor database may document another part of the response. Each piece can be valid. Yet answering a basic institutional question can still become difficult: what was planned, what was funded, what was implemented, and what evidence shows what happened next?

The AMR information base is becoming substantial

WHO's Global Antimicrobial Resistance and Use Surveillance System, GLASS, now supports standardized surveillance across a large number of countries. The 2025 global antibiotic resistance surveillance report analysed more than 23 million bacteriologically confirmed infections, with data reported by 104 countries for 2023 and 110 countries across the 2016-2023 period.

Countries also report policy and implementation progress through the Tracking AMR Country Self-Assessment Survey, or TrACSS. WHO maintains national action plan resources, implementation guidance, monitoring frameworks, surveillance systems and country-level AMR profiles. The global architecture therefore contains considerable information, and it continues to mature.

In May 2026, World Health Assembly Member States adopted the updated Global Action Plan on Antimicrobial Resistance 2026-2036. The new framework reinforces accountability, monitoring, reporting, sustainable financing and the implementation of multisectoral national action plans.

A plan is not implementation evidence

Consider a seemingly simple question: is a country's AMR national action plan being implemented? There is no single dataset that completely answers it.

At one level, a national action plan may be approved and reported as being implemented. At a higher level, that plan may have a costed and budgeted operational plan and a monitoring mechanism. At the highest level, financial provision for implementation is included in national plans and budgets. These are materially different states.

An approved plan is not the same as an operational plan. An operational plan is not the same as secured financing. Financing is not the same as implementation. Implementation is not the same as measurable outcome.

The evidence chain is distributed

Trying to reconstruct one AMR intervention can require moving across a national action plan, TrACSS, GLASS, government budgets, donor databases, development-bank project documents, procurement records, ministry updates, PubMed and institutional repositories. No individual source is necessarily deficient. The problem emerges when an analyst attempts to reconstruct the relationship between them.

A health-system investment may enable AMR surveillance while serving several other objectives. A surveillance programme may generate strong resistance data without providing a direct link to the financing that created the capability. A funding record may contain an AMR-relevant activity without proving what portion of the project capital should legitimately be attributed to AMR.

More data does not automatically produce decision intelligence

The purpose of AMR information is not merely to demonstrate that activity exists. Decision-makers need to understand relationships: which priorities are financed, where attribution remains uncertain, where capital stops being traceable to delivery, and which unresolved evidence could change the decision.

Those questions are not answered by adding another dataset. They require connecting existing evidence without overstating what the connection proves.

Not found is not absent

If an analyst cannot identify a publicly accessible AMR-specific national budget, the defensible conclusion is not that the country has no AMR budget. The defensible conclusion is narrower: within the public sources reviewed, no consolidated AMR-specific budget line was identified.

That distinction matters in global health, where internal ministry systems, donor reporting structures, subnational budgets and programme documents may not all be public or consistently classified. Unknowns should remain unknowns. They should not be silently converted into zeros.

From data collection to plan-to-proof

A useful test is whether an independent analyst can move through an evidence chain without making unsupported assumptions. Can the policy objective be identified? Can the corresponding activity be found? Can financing be attributed with the right confidence level? Can implementation be observed? Can outcomes be linked without overstating causality?

That is the difference between collecting AMR information and building decision-grade AMR intelligence. The global AMR ecosystem already contains substantial policy, surveillance, financing and programme information. The next challenge is to make those layers connect.