tetano
Editor, Senior Moderator
PloS One
. 2026 Jul 10;21(7):e0349030.
doi: 10.1371/journal.pone.0349030. eCollection 2026.
Exploring the predictability of distributed lag nonlinear models using SARS-CoV-2 wastewater-based surveillance in multiple communities in Alberta, Canada
Rhonda J Rosychuk[SUP] 1 2 3 4 [/SUP], Bonita E Lee[SUP] 1 4 [/SUP], Judy Y Qiu[SUP] 4 5 6 7 [/SUP], Tiejun Gao[SUP] 6 [/SUP], Michael D Parkins[SUP] 8 [/SUP], Casey R J Hubert[SUP] 9 10 [/SUP], Xiaoli Pang[SUP] 6 7 [/SUP]
Affiliations
Background: Wastewater-based surveillance can be an important part of pandemic management, especially when testing capacity of individuals is limited. Statistical modeling can be used to examine the relationship between wastewater pathogen levels and clinical cases. The objective of this study was to examine the utility of distributed lag nonlinear modeling to derive the relationship between wastewater SARS-CoV-2 RNA levels and COVID-19 clinical cases across communities in Alberta, Canada when clinical testing was comprehensive.
Methods: This retrospective cohort study used data from 24-hour composite wastewater collected and tested two to three times per week from 11 wastewater treatment plants (WWTPs) in Alberta, Canada during May 10, 2020, to March 15, 2022. The number of daily new cases of COVID-19 downloaded from Alberta Health's centralized dataset of clinical surveillance of COVID-19 were mapped to each sewershed. Distributed lag nonlinear models were fit to describe the exposure-response relationship between the 7-day rolling average of SARS-CoV-2 RNA and daily new cases for each WWTP separately.
Results: The 11 WWTPs served a population of 3,422,062 (77% of Alberta's population) and 386,528 cases were documented during the study period. From 2021 onward, peaks in both wastewater viral RNA levels and cases tracked reasonably well. For almost all WWTPs, the best fitting model was a Poisson additive model with a P-spline for time. Models for the larger communities had better fits than smaller communities as represented by adjusted pseudo-R2 ranging from 80.7% to 94.4%. Models followed the same general trends as the actual COVID-19 cases over time.
Conclusions: With relationships between wastewater viral RNA levels for SARS-CoV-2 and COVID-19 cases expected to vary over time and to be non-linear, distributed lag nonlinear models are promising. While the form of the models was similar across WWTPs, the resulting estimates were different among sites suggesting site-specific analyses are essential.
. 2026 Jul 10;21(7):e0349030.
doi: 10.1371/journal.pone.0349030. eCollection 2026.
Exploring the predictability of distributed lag nonlinear models using SARS-CoV-2 wastewater-based surveillance in multiple communities in Alberta, Canada
Rhonda J Rosychuk[SUP] 1 2 3 4 [/SUP], Bonita E Lee[SUP] 1 4 [/SUP], Judy Y Qiu[SUP] 4 5 6 7 [/SUP], Tiejun Gao[SUP] 6 [/SUP], Michael D Parkins[SUP] 8 [/SUP], Casey R J Hubert[SUP] 9 10 [/SUP], Xiaoli Pang[SUP] 6 7 [/SUP]
Affiliations
- PMID: 42430452
- PMCID: PMC13354072
- DOI: 10.1371/journal.pone.0349030
Background: Wastewater-based surveillance can be an important part of pandemic management, especially when testing capacity of individuals is limited. Statistical modeling can be used to examine the relationship between wastewater pathogen levels and clinical cases. The objective of this study was to examine the utility of distributed lag nonlinear modeling to derive the relationship between wastewater SARS-CoV-2 RNA levels and COVID-19 clinical cases across communities in Alberta, Canada when clinical testing was comprehensive.
Methods: This retrospective cohort study used data from 24-hour composite wastewater collected and tested two to three times per week from 11 wastewater treatment plants (WWTPs) in Alberta, Canada during May 10, 2020, to March 15, 2022. The number of daily new cases of COVID-19 downloaded from Alberta Health's centralized dataset of clinical surveillance of COVID-19 were mapped to each sewershed. Distributed lag nonlinear models were fit to describe the exposure-response relationship between the 7-day rolling average of SARS-CoV-2 RNA and daily new cases for each WWTP separately.
Results: The 11 WWTPs served a population of 3,422,062 (77% of Alberta's population) and 386,528 cases were documented during the study period. From 2021 onward, peaks in both wastewater viral RNA levels and cases tracked reasonably well. For almost all WWTPs, the best fitting model was a Poisson additive model with a P-spline for time. Models for the larger communities had better fits than smaller communities as represented by adjusted pseudo-R2 ranging from 80.7% to 94.4%. Models followed the same general trends as the actual COVID-19 cases over time.
Conclusions: With relationships between wastewater viral RNA levels for SARS-CoV-2 and COVID-19 cases expected to vary over time and to be non-linear, distributed lag nonlinear models are promising. While the form of the models was similar across WWTPs, the resulting estimates were different among sites suggesting site-specific analyses are essential.