Modelling of spatially misaligned wastewater-based surveillance data

Abstract

Wastewater-based epidemiology (WBE) has emerged as a promising approach to infectious disease modelling and early detection of disease outbreaks in general. Herein, the application of this approach to COVID-19 is explored through spatio-temporal models. The goal of predicting COVID-19 cases at a small administrative area level (sub-place) while using data collected at catchment area level introduces the issue of spatial misalignment. Spatial misalignment in the data must be accounted for by the modelling process, as is the case for many spatial disease models. Appropriate handling of the data requires disaggregation and matching of different spatial regions. The models are developed in a Bayesian framework using the INLA package in R, which is particularly intuitive in the context of latent Gaussian mixed models (LGMM). Misalignment is explored using a distance based approach which incorporates uncertainty as a component of the LGMM. COVID-19 cases were modelled at sub-place level and spatio-temporal trends were explored. The best performing model was able to utilise a link between the wastewater data and the COVID-19 cases. The modelling approach deals with data available at different spatial resolution, data covering misaligned and overlapping spatial areas, disaggregated and missing data, thereby accounting for complexity of real data. The methodology provides a novel structure to deal with the intricacy of the sampled wastewater and related data, and poses a proof of concept for wastewater-based spatial modelling for disease surveillance.