CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Global space-time mapping of antimicrobial-resistance among selected priority bacterial pathogens, 2000-2019 Authors:  Benn Sartorius - University of Oxford (United Kingdom) [presenting]
Annie Browne - University of Oxford (United Kingdom)
Michael Chipeta - University of Oxford (United Kingdom)
Frederick Fell - University of Oxford (United Kingdom)
Sean Hackett - University of Oxford (United Kingdom)
Georgina Haines-Woodhouse - University of Oxford (United Kingdom)
Bahar Kashef Hamadani - University of Oxford (United Kingdom)
Emmanuelle Kumaran - University of Oxford (United Kingdom)
Catrin Moore - University of Oxford (United Kingdom)
Christiane Dolecek - University of Oxford (United Kingdom)
Abstract: Antimicrobial resistance (AMR) is a major and growing public health threat. Despite considerable global efforts, our understanding of AMR burden across space-time remains sparse. Understanding this is essential to better inform policy and help combat the further spread of AMR. The flagship Global Research on AntiMicrobial resistance (GRAM) Project, attempting to improve our understanding of global AMR burden, has compiled proportions of resistant isolates for priority bacterial pathogens to key antimicrobials by country/year. These data currently span ~2500 sources including patient-level AMR microbiology data, aggregated AMR data (e.g. surveillance), published studies and directly from collaborators. Influential covariates for each bacterial-antimicrobial combination were included in a machine learning stacked ensemble framework to improve predictive validity. We adapted and employed Bayesian space-time Gaussian Process Regression and conditional autoregressive modelling to help improve estimates in countries without data and uncertainty interval quantification. Initial results show large variations in resistance by pathogen, antimicrobial, location and year. Notably, particular bacteria-antimicrobial combinations have higher and/or increasing burden in many LMIC over the period. This will have important policy implications, especially in the context of the current COVID pandemic. The next stages will include refining/optimising the modelling framework.