A new study links country-level antibiotic consumption with patterns of bacterial resistance across 114 countries. Its model estimates that a scenario with 10% lower antibiotic use across drug classes would be associated with about 38,000 fewer annual deaths attributable to resistance, with a 95% confidence interval of 36,000–41,000. This is a modelled counterfactual based on observational relationships, rather than the measured result of an antibiotic-reduction programme. Original PLOS Medicine study
The analysis, published on 26 October 2026, also places that estimate beside a wider problem: deaths from bacterial infections in settings where effective treatment can be hard to obtain. Reducing unnecessary exposure and improving access to needed antibiotics are related policy tasks, but the study does not collapse them into a single target for using less.
The unit of observation is a country and a year
The researchers used resistance surveillance from 2000–2018, covering more than 30 million bacterial isolates, alongside country-level antibiotic-consumption estimates. The analysis addressed 16 bacterial pathogens and seven drug classes, adjusting for country-level health and development indicators.
Its central measure was how a proportional difference in antibiotic use related to a proportional difference in resistance prevalence. Associations varied by organism and drug class; a relationship found for one combination was not assumed to apply equally to all others. Methods and results
That broad coverage is useful for examining patterns across settings. It also sets an important limit. A relationship observed between national consumption and resistance does not identify what happened to an individual patient, whether a particular prescription was appropriate or what a specific hospital intervention would achieve.
The authors explicitly describe the risk of the ecological fallacy: population-level associations may not hold at smaller levels of analysis.
What the 10% scenario changes
The model asks how resistance-related mortality might differ under lower antibiotic consumption, using the estimated relationships. The 38,000 figure concerns deaths attributable to resistance—the excess mortality compared with a counterfactual in which those infections were susceptible to drugs. It is not an estimate of all deaths involving a resistant infection, or all deaths from bacterial disease.
The authors estimate that this difference represents about 0.50% of total bacterial-disease mortality. That denominator explains why a substantial number of resistance-attributable deaths can still be a comparatively small share of the wider bacterial-disease burden. Counterfactual estimates and interpretation
The model does not demonstrate that an intervention reducing antibiotic consumption by 10% would save exactly that number of lives. The estimate depends on interpreting the observed relationships, the completeness of surveillance and assumptions about how resistance responds when use falls.
The authors also considered alternative scenarios. Their existence reinforces the point that the headline is conditional on a modelling choice, not a fixed return from removing a set number of prescriptions.
The UN’s 10% target measures something else
The same percentage appears in a different context in international policy. WHO’s account of the 2024 UN political declaration describes a goal to reduce deaths associated with bacterial antimicrobial resistance by 10% by 2030. That is a mortality target, not a universal instruction to reduce human antibiotic consumption by 10%. WHO’s declaration announcement
The declaration also sets a target for at least 70% of human antibiotic use globally to come from the WHO Access group and includes commitments to equitable access, diagnostics and infection prevention. The Access-group share describes the composition of use. It is different again from either a reduction in total consumption or a reduction in deaths.
| Measure | What changes in the measure? |
|---|---|
| Study’s 10% consumption scenario | Antibiotic use in a modelled counterfactual. |
| UN declaration’s 10% mortality target | Deaths associated with bacterial antimicrobial resistance. |
| 70% Access-group target | The share of human antibiotic use belonging to that antibiotic group. |
Sources: PLOS study and WHO policy announcement. These measures should not be exchanged simply because they appear in the same policy discussion.
Why access stays inside the interpretation
The study’s discussion stresses that lower-resource settings are underrepresented in surveillance and that country-level data can conceal local differences. It also notes that its static statistical model cannot predict long-term intervention effects or fully represent changing transmission and bacterial evolution.
Where access to effective antibiotics is limited, an intervention can affect both resistance selection and the treatment of bacterial disease. The model does not directly weigh every benefit and harm of a particular access policy, so it cannot supply a universal prescribing rule. That is a boundary of the analysis, rather than a reason to ignore resistance prevention.
The paper’s contribution is a quantified relationship to examine alongside infection prevention, appropriate treatment and access evidence. A policy assessment still needs to ask whose use changes, which drugs are affected, what infections are prevented and what happens to people who need treatment.