Below, we explain how we gather and process data from international health organizations to present a comprehensive picture of AMR changes throughout the last few years.
Country-level data were extracted from the EU CDC’s 2022 supplemental data report,[1] the PAHO public dashboard, [2] and the WHO GLASS public dashboard (2020-2022).[3] While the EU CDC and PAHO reported data from 2019-2022, the WHO GLASS public dashboard only included data from 2020-2022, so 2019 data and subsequent calculations are only reported for Central and South America and Europe. CRE, ESBL-producing Enterobacterales, MRSA, and CRA were selected as the MDROs of interest as all were robustly reported on by the three data sources. The EU CDC and WHO GLASS did not distinguish between community onset and hospital onset infections, so our analysis did not distinguish between the location of onset. The PAHO also did not distinguish between community and hospital onset except for in the case of MRSA which were combined in our analysis. While the PAHO and WHO GLASS data sets reported cultures resistant to Imipenem and Meropenem separately which were summed to calculate total carbapenem resistant cultures and total cultures sent, the EU CDC reported Imipenem and Meropenem together. To account for this difference, we multiplied the EU CDC carbapenem resistant culture counts by two to approximate as if the EU CDC had also reported resistant Imipenem and Meropenem culture counts separately. While this does not affect the AMR proportion, it does allow for more accurate direct comparisons of AMR surveillance intensity. A similar approach was used for ESBL-producing Enterobacterales culture data with an adjustment multiplier of three, as the EU CDC reported Cefotaxime, Ceftriaxone, and Ceftazidime together, while the PAHO and WHO GLASS reported culture data for each 3rd generation cephalosporin separately. The EU CDC and WHO GLASS data sets did not include intermediate resistance data while the PAHO data set did. To provide the most conservative interpretation of the data, only PAHO reported cultures which were unambiguously resistant were counted as AMR positive cultures.
Finally, the United States Centers for Disease Control and Prevention’s (US CDC) models MDRO incidence using cases per hospital discharge based on proprietary data sources, the Premier Healthcare Database[4] and the BD Insights Research Database[5] which were not publicly available. The US CDC most recently published an AMR fact sheet in July of 2024[6] although this fact sheet only reported “increase, decrease or stable” for each MDRO, lacking effect sizes, measures of statistical significance, or details on the methods used to calculate these trends. The US CDC’s most recent report which included effect sizes and measures of statistical significance was published in 2022 and provided data for 2019 and 2020. [5] In this report, the US CDC’s model controlled for hospital characteristics, month of discharge, age group of patients, and sex of patients. While few specific details are publicly known of the US CDC’s model, we outline what is known in the supplemental materials. Because the US CDC does not report raw culture data unlike the EU CDC, the PAHO, or the WHO GLASS, the United States could not be included in our analysis.
Mathematical analysis of raw data involved multiplying each hospital's AMR percentage by its district's population relative to the population of the country. Furthermore, a threshold is applied to only consider countries with robust sampling, yielding a composite metric: population-weighted, surveillance-intensity-weighted mean percentages (PSIMP). Further details of the statistical analysis and justifications for our methods will be available in a research paper in the final process of editing.