%0 Journal Article
%T Time series modelling for wastewater-based epidemiology of COVID-19: A nationwide study in 40 wastewater treatment plants of Belgium, February 2021 to June 2022.
%A Bertels X
%A Hanoteaux S
%A Janssens R
%A Maloux H
%A Verhaegen B
%A Delputte P
%A Boogaerts T
%A van Nuijs ALN
%A Brogna D
%A Linard C
%A Marescaux J
%A Didy C
%A Pype R
%A Roosens NHC
%A Van Hoorde K
%A Lesenfants M
%A Lahousse L
%J Sci Total Environ
%V 899
%N 0
%D 2023 Nov 15
%M 37474075
%F 10.753
%R 10.1016/j.scitotenv.2023.165603
%X <B>BACKGROUND: </B>Wastewater-based epidemiology (WBE) has been implemented to monitor surges of COVID-19. Yet, multiple factors impede the usefulness of WBE and quantitative adjustment may be required.<BR><B>OBJECTIVE: </B>We aimed to model the relationship between WBE data and incident COVID-19 cases, while adjusting for confounders and autocorrelation.<BR><B>METHODS: </B>This nationwide WBE study includes data from 40 wastewater treatment plants (WWTPs) in Belgium (02/2021-06/2022). We applied ARIMA-based modelling to assess the effect of daily flow rate, pepper mild mottle virus (PMMoV) concentration, a measure of human faeces in wastewater, and variants (alpha, delta, and omicron strains) on SARS-CoV-2 RNA levels in wastewater. Secondly, adjusted WBE metrics at different lag times were used to predict incident COVID-19 cases. Model selection was based on AICc minimization.<BR><B>RESULTS: </B>In 33/40 WWTPs, RNA levels were best explained by incident cases, flow rate, and PMMoV. Flow rate and PMMoV were associated with -13.0 % (95 % prediction interval: -26.1 to +0.2 %) and +13.0 % (95 % prediction interval: +5.1 to +21.0 %) change in RNA levels per SD increase, respectively. In 38/40 WWTPs, variants did not explain variability in RNA levels independent of cases. Furthermore, our study shows that RNA levels can lead incident cases by at least one week in 15/40 WWTPs. The median population size of leading WWTPs was 85.1 % larger than that of non‑leading WWTPs. In 17/40 WWTPs, however, RNA levels did not lead or explain incident cases in addition to autocorrelation.<BR><B>CONCLUSIONS: </B>This study provides quantitative insights into key determinants of WBE, including the effects of wastewater flow rate, PMMoV, and variants. Substantial inter-WWTP variability was observed in terms of explaining incident cases. These findings are of practical importance to WBE practitioners and show that the early-warning potential of WBE is WWTP-specific and needs validation.