AMROrbit: a trajectory-based scorecard for antimicrobial stewardship using routine susceptibility testing data

Area of Science:

• Microbiology and Infectious Diseases
• Computational Biology and Bioinformatics
• Public Health and Epidemiology

Background:

• Antimicrobial resistance (AMR) is a critical global health challenge.
• Current surveillance methods using static antibiograms fail to capture dynamic resistance trends.
• Proactive identification of emerging AMR patterns is crucial for effective intervention.

Purpose of the Study:

• To introduce AMROrbit, an explainable, trajectory-based scorecard for proactive AMR pattern identification.
• To utilize routinely generated antimicrobial susceptibility data for dynamic resistance monitoring.
• To develop a system that overcomes the limitations of static antibiograms.

Main Methods:

• AMROrbit was applied to the Pfizer ATLAS dataset, analyzing ESKAPE pathogens and Escherichia coli.
• Resistance percentages were modeled using rolling windows to estimate baseline resistance (amplitude) and rate of change (velocity).
• Trajectories were classified based on movement within a 2D resistance space (amplitude vs. velocity) into categories like spiral-in and spiral-out.

Main Results:

• Over 62% of antibiotic-microorganism-country combinations showed increasing baseline resistance.
• A concerning global imbalance was observed, with 17.7% of trajectories spiraling out (worsening) versus 13.4% spiraling in (improving).
• Significant heterogeneity in resistance trajectories was noted across different settings and pathogen-antibiotic combinations.

Conclusions:

• AMROrbit offers a novel approach by integrating resistance levels and their rate of change.
• This framework enables earlier detection of emerging resistance, supporting timely stewardship interventions.
• The system's reliance on routine data ensures scalability across various healthcare settings.