Identifying shifts in antimicrobial resistance (AMR) trends is crucial for guiding effective public health strategies. We recently showed that changes in COVID-19 growth rates could be detected early using the derivatives of generalised additive models (GAMs). We have now extended this methodological approach to include an explicit spatiotemporal component using Markov Random Field (MRF) smooths interacting with time smooths. By incorporating further interactions with other characteristics, such as sociodemographic covariates, this novel methodology can also detect changes in subgroups of the population that remain otherwise undetected. The approach we have developed could be adapted to detect relevant shifts in AMR trends, enabling prompt, targeted interventions and optimal resource allocation to significantly improve clinical outcomes. Our project aims to analyse pathogen-specific change points in AMR prevalence and growth rates in countries overall and, where patient-level data is available (e.g. age, sex), within relevant sub-populations using the spatiotemporal GAM approach. We will leverage data from the GSK SOAR and Shionogi SIDERO-WT datasets, which include date of collection, and further assess potential for expanding this approach to data with less granular time information (ATLAS). In addition, leveraging the flexibility of spatiotemporal GAMs and previous work for the recent Lancet Series in AMR, where we extracted a wide range of country-level information on potential drivers of AMR— including antibiotic use, vaccination, water, sanitation and hygiene (WASH), climate and demographics , — potential future AMR trajectories will be predicted from our models, enabling policy-makers to assess the potential impact of inaction and to set realistic targets for AMR National Action Plans. Using this project, we expect to introduce a novel methodology to more effectively monitor and predict future AMR trajectories at a national level, as well in specific subgroups of the population, enabling timely and targeted interventions to more effectively reduce AMR burdens.
Center for Global Research Data