Background
Antimicrobial resistance is widely considered an urgent global health issue due to associated mortality and disability, societal and healthcare costs.
Aim
To estimate the past, current and projected future proportion of infections resistant to treatment for eight priority antibiotic-bacterium combinations from 2000 to 2030 for 52 countries.
Methods
We collated data from a variety of sources including ResistanceMap and World Bank. Feature selection algorithms and multiple imputation were used to produce a complete historical dataset. Forecasts were derived from an ensemble of three models: exponential smoothing, linear regression and random forest. The latter two were informed by projections of antibiotic consumption, out-of-pocket medical spending, populations aged 64 years and older and under 15 years and real gross domestic product. We incorporated three types of uncertainty, producing 150 estimates for each country-antibiotic-bacterium-year.
Results
Average resistance proportions across antibiotic-bacterium combinations could grow moderately from 17% to 18% within the Organisation for Economic Co-operation and Development (OECD; growth in 64% of uncertainty sets), from 18% to 19% in the European Union/European Economic Area (EU/EEA; growth in 87% of uncertainty sets) and from 29% to 31% in Group of Twenty (G20) countries (growth in 62% of uncertainty sets) between 2015 and 2030. There is broad heterogeneity in levels and rates of change across countries and antibiotic-bacterium combinations from 2000 to 2030.
Conclusion
If current trends continue, resistance proportions are projected to marginally increase in the coming years. The estimates indicate there is significant heterogeneity in resistance proportions across countries and antibiotic-bacterium combinations.
Linear Fractional Transformations for the Approximation of Various Uncertainty Sets
