Software code

Make R code available for estimations of disease frequency using Bayesian hierarchical spatial approaches


This initiative is linked to a study about tuberculosis (TB) and COVID-19 in Lima, Peru. The incidence of notified cases of active TB (2018-2019) and the incidence of COVID-19 deaths (in wave 1 and in wave 2) vary across the 50 districts of Lima. Our hypothesis is that we observe a geographical overlap in the burden of TB and COVID-19 because the underlying factors for these conditions overlap. We use a Bayesian hierarchical spatial approach (shared component Besay, York and MolliƩ (BYM) 2 model) to estimate risk ratios per district for TB and for COVID-19, split variance into three components (TB-specific, COVID-19-specific, and shared variance), and explore correlations with covariates. All analyses are done in R (nimble package). Through this initiative, we intend to document our approach in detail and share all related materials, including datasets (if allowed) and annotated R code. These materials will be useful as supplementary files with publications, as a basis for a comparison with the heterogeneity of TB or COVID-19 burden in other cities, and as teaching material for spatial epidemiology (to illustrate among others the ecological study design, standardisation, collinearity, neighbourhood structure, spatial autocorrelation, BYM2 model, exceedance probability, and shared component models).


  • Ongoing initiative

  • Started in 2021, in collaboration with colleagues from UHasselt and KULeuven