Carbapenem Resistance: A Spatiotemporal Modeling Approach for the Vivli AMR Surveillance Data Challenge

By 2050, deaths attributed to antimicrobial resistance (AMR) will reach upwards of two million people globally. It will continue to be one of the leading causes of death around the world, with a disproportionate impact on developing nations. Many of the deaths associated with AMR can
be resolved through improved healthcare infrastructure, better reporting of local AMR incidence, and improved access to antibiotics specific for treating AMR. A 2025 study, published by Naghavi M. et al. at the Institute of Health Metrics and Evaluation, found that among gram-negative bacteria, resistance to carbapenems in particular increased more than any other
antibiotic class and will continue to remain a significant concern for future public health efforts. This study proposes three objectives: 1) identify the geographical and temporal trends associated with increased carbapenem-resistant infections. 2) Based on identified trends, develop an ensemble model made up of generalized additive models (GAMs), Random Forest,
XGBoost, and Bayesian hierarchical models to accurately forecast high risk areas and future trends in carbapenem-resistant infections (CRI). 3) Using the ensemble model performance and data quality evaluation, highlight areas for public health experts that merit targeted interventions of high-risk CRI. Optionally, if data permit, we also aim to examine the spatial
heterogeneity between general and nosocomial AMR related to CRI. We will leverage data from Innoviva, Sidero-WT, Venus Remedies, Paratek, and Atlas to access over one million global isolates of carbapenem-resistance while extracting a wide range of country-level information on potential drivers of this particular gram-negative resistance. This project allows us to use novel methodologies to create a prototype for future carbapenem-resistant pathogens outbreak simulation to strengthen health systems as well as inform public health access. It will also allow us to improve global understanding of resistance drivers and support targeted healthcare interventions.