A1782
Title: Application of hidden Markov models to interpret the wastewater viral signal of rarely detected viruses
Authors: Johanna de Haan-Ward - University of Ottawa (Canada) [presenting]
Chandler Wong - University of Ottawa (Canada)
Elizabeth Renouf - University of Ottawa (Canada)
Shen Wan - University of Ottawa (Canada)
Xin Tian - University of Ottawa (Canada)
Robert Delatolla - University of Ottawa (Canada)
Elizabeth Mercier - University of Ottawa (Canada)
Abstract: Since the onset of the COVID-19 pandemic, wastewater-based surveillance has emerged as a critical tool for monitoring disease spread, including early detection and anticipating periods of increased disease burden. Statistical methodology has been essential in understanding the viral level of endemic viruses such as SARS-CoV-2 and influenza, with the goal of linking the wastewater viral signal to key outcomes for public health authorities, such as the number of hospitalizations. Rare viruses present a challenge for statistical modelling due to irregular sampling of wastewater, zero-heavy distributions, and in some instances missing information about clinical cases. Hidden Markov models have been widely applied in fields such as ecology, finance, and medicine, where time-series data are assumed to result from an underlying, unobservable process. Hidden Markov models were applied to two years of wastewater samples from Ottawa, Canada analyzed for the presence of Mpox, a zoonotic virus first detected in Canada in early 2022 and spread through close human contact. Hidden Markov models specified with two and three underlying states were compared, where the time-series data were assumed to derive from a hurdle model to account for excess zeros in the time series, including during times of high disease burden. The three-state model provides inference best aligned with discipline-specific beliefs about the disease burden of Mpox.
