Application of Bayesian spatial modelling to uncover geographical disparities and improve antimicrobial resistant surveillance

IntroductionDisease surveillance is an essential element of an effective response to antimicrobial resistance (AMR). Associations between AMR cases and area-level drivers such as remoteness and socio-economic disadvantage have been observed, but spatial associations when modelling routinely collected surveillance data that are often imperfect or missing have not been previously possible. AimWe aimed to use spatial modelling to adjust for area-level variables and to enhance AMR surveillance for missing or sparse data, in an effort to provide clinicians and policy makers with more actionable epidemiological information. MethodsWe used monthly antimicrobial susceptibility data for methicillin-resistant Staphylococcus aureus (MRSA) from a surveillance system in Australia. MRSA was assessed for the effects of age, sex, socio-economic and access to healthcare services indices by fitting Bayesian spatial models. ResultsWe analysed data for 77, 760 MRSA isolates between 2016 and 2022. We observed significant spatial heterogeneity in MRSA and found significant associations with age, sex and remoteness, but not socio-economic status. MRSA infections were highest in adult females aged 16-60 living in very remote regions and lowest in senior males aged 60+ years living in inner regional areas.. ConclusionCurrent disease surveillance approaches for antimicrobial resistant infections have limited spatial comparability, are not timely, and at risk of sampling bias. Bayesian spatial models borrow information from neighbouring regions to adjust for unbalanced geographical information and can fill information gaps of current MRSA surveillance. Assessment of disease spatial variation is especially critical in settings which have diverse geography, dispersed populations or in regions with limited microbiological capacity.