AMR (R package)

Lore & Background

The AMR package was initially developed at the University of Groningen and the University Medical Center Groningen (UMCG), in the Netherlands. Its work was published as a peer-reviewed paper in the Journal of Statistical Software and formed the basis of two PhD theses at the University of Groningen. According to CRAN, the package depends on other R packages like dplyr, tidyr, ggplot2, and rlang, and recent versions require R >= 3.5 or higher.

Antimicrobial resistance is recognized by the World Health Organization as one of the leading public health threats globally. Prior to the AMR package, there was a lack of standardized open-source tools to process and analyze AMR data in

Reader's Guide

The AMR package represents a significant contribution to the field of antimicrobial resistance data analysis by providing a free, open-source, and standardized toolbox. Its integration of international guidelines, such as CLSI and EUCAST breakpoints, and reference data like microbial taxonomy and drug coding systems, directly addresses the need for reproducible and clean analysis in clinical, veterinary, and environmental microbiology. Since its first public release on CRAN in early 2018, the package has been downloaded from over 175 countries, indicating widespread global adoption. Studies employing the package have covered diverse areas including national paediatric bloodstream infection surveillance, population-wide AMR burden analysis, companion animal MRSA surveillance, and environmental resistance gene screening. The package's design, with no external dependencies and compatibility with older R versions, lowers the barrier to entry. The availability of a Python wrapper and the release of reference data in multiple formats extend its utility beyond the R ecosystem. By enabling standardized interpretation of laboratory measurements and facilitating reproducible workflows, the AMR package has become a key tool for researchers and clinicians working to understand and combat antimicrobial resistance.

Did You Know?

• The AMR package has been downloaded from over 175 countries, as measured by CRAN download logs.
• It has no external dependencies and is compatible with R versions from R-3.0 (April 2013) onwards.
• A Python wrapper is available that runs the AMR R package natively in Python.
• The package's work formed the basis of two PhD theses at the University of Groningen.

Origin & the Gap It Filled

The AMR package emerged from a specific institutional context: it was first built at the University of Groningen and the University Medical Center Groningen in the Netherlands. The intellectual work behind it was substantial enough to form the foundation of two doctoral theses at Groningen, and the development was documented in a peer-reviewed article in the Journal of Statistical Software. The package addressed a clear gap in the antimicrobial resistance data ecosystem. While the World Health Organization had long flagged antimicrobial resistance as a top-tier global health threat, and while international standards existed for laboratory testing and reporting, there was no standardized, open-source tool for reproducibly processing the resulting data. Laboratory information system exports often needed heavy cleaning, and interpretation depended on keeping pace with shifting taxonomy and clinical breakpoint guidelines. AMR was conceived as an integrated toolbox that folds those reference materials directly into the analytical workflow, removing the manual stitching that previously separated raw lab output from meaningful surveillance results.

Core Functionality & Reference Data Integration

At its heart, AMR bundles several interlocking capabilities that cover the full arc of susceptibility data analysis. A built-in taxonomic database draws on the List of Prokaryotic names with Standing in Nomenclature, the Global Biodiversity Information Facility, and MycoBank, letting users standardise microorganism names and properties with a single lookup. A comprehensive catalogue of antimicrobial and antiviral agents is linked to multiple international coding systems, including ATC, EARS-Net, PubChem, LOINC, and SNOMED CT. Perhaps most critically, the package integrates clinical breakpoint guidelines from both CLSI and EUCAST, spanning multiple years of published guidance and extending to veterinary breakpoints and epidemiological cut-off values. These resources let raw laboratory measurements—minimum inhibitory concentrations and disk diffusion diameters—be translated into standardised susceptibility categories. Beyond interpretation, the package supports identifying first isolates per patient, generating antibiograms including WISCA, flagging multi-drug resistant organisms, and computing resistance proportions. It can ingest data in any format, including exports from the WHO's WHONET software, and all reference data sets are downloadable in Excel, Apache Parquet, SPSS, Stata, and plain text.

Custom Data Types & Technical Architecture

One of AMR's most distinctive technical contributions is the introduction of a family of new S3 data classes into the R ecosystem. These include mo for microorganism codes, ab for antimicrobial drug codes, sir for susceptibility interpretation results, mic for minimum inhibitory concentrations, and disk for disk diffusion diameters. Each class carries its own validation logic and intelligent coercion rules, so that a mis-typed species code, for example, is caught early rather than propagating silently through a pipeline. The classes also integrate smoothly with common R workflows, meaning users can pipe AMR objects through tidyverse or base R operations without losing type information. Architecturally, the package is deliberately lightweight: according to CRAN, it carries no external dependencies and remains compatible with R versions dating back to R-3.0, released in April 2013. This minimal footprint makes it easy to install in resource-constrained settings. For users who prefer Python, a wrapper package runs the AMR R package natively within Python, follows identical version numbering, and is built automatically alongside the R release. The entire project is distributed under the GNU General Public License version 2.0.

Global Adoption & Real-World Applications

Since its first public appearance on CRAN in early 2018, AMR has achieved a geographic reach that is remarkable for a specialised statistical tool. CRAN download logs record installations from more than 175 countries, and the package has been adopted across clinical, veterinary, and environmental microbiology settings worldwide. The breadth of published work using it illustrates this range: national paediatric bloodstream infection surveillance programmes have relied on it, as have population-wide analyses of AMR burden. In veterinary medicine, companion animal MRSA surveillance spanning multiple practices has employed the package, while in environmental science it has been used for resistance gene screening in fresh produce. A perspective article on clinical AMR surveillance tools also featured AMR as a reference implementation. The package's design philosophy—free, open-source, reproducible, and standardised—has made it a practical choice for laboratories and public-health agencies that need to move from raw susceptibility data to actionable surveillance outputs without assembling a bespoke pipeline. Its adoption across such diverse settings underscores how a well-constructed open-source tool can become infrastructure for a global public-health challenge.