Data Challenge: Implementing Machine Learning for AMR trend recognition and projection

Umang Joshi, Sarah Harden, and Abby Sweet of the Lanzas Lab at NCSU would like to express their interest in the 2025 Vivli AMR Surveillance Challenge, particularly working with the Venatorx dataset. The primary research areas of our lab are understanding antimicrobial use, computational modeling for antimicrobial resistance in human and veterinary datasets, and modeling healthcare-associated infections. We believe this expertise aligns well with the objectives of this challenge. Currently, Umang and Sarah have been working on integrating machine learning tools for assessing risk factors and evaluating trends in antibiotic administration, respectively. Abby is leading efforts in agent-based modeling of healthcare-associated infections, with a particular focus on extended-spectrum beta-lactamases (ESBLs). For the Vivli Challenge, we propose using machine learning methods, specifically XGBoost or Gradient Boosting Machines, to the Venatorx dataset to assess trends in AMR for community-level infections. These tools are optimal for pattern recognition and offer explainable predictive capabilities. Our approach will have two main components: (1) evaluating existing trends in AMR rates for healthcare-associated infections and (2) predicting future trends in AMR rates for the pathogens represented in the study. For the first objective, we will begin by parsing and potentially subsetting the data to ensure a thorough understanding of its structure and variables. We will then analyze AMR trends by infection year, type, and country of origin. For the second objective, we will supplement the provided data with publicly available datasets and relevant literature to strengthen our predictive models, if necessary. Our main aim for this work is to contribute towards a deeper understanding of AMR rates, not just in the US, but worldwide for healthcare-associated infections. This challenge offers an excellent opportunity to expand our knowledge regarding a highly relevant topic that fits our research goals and interests.

Thank you for your consideration.
Umang Joshi, Sarah Harden, Abby Sweet, Sankalp Arya, and Cristina Lanzas