Combining wastewater surveillance and case data in estimating the time-varying effective reproduction number

Wastewater surveillance has been increasingly acknowledged as a useful tool for monitoring transmission dynamics of infections of public health concern, including the coronavirus disease (COVID-19). While a range of models have been proposed to estimate the time-varying effective reproduction number (R_t) utilizing clinical data, few have harnessed the viral concentration in wastewater samples to do so, leaving uncertainties about the potential precision gains with its use. In this study, we developed a Bayesian hierarchical model which simultaneously reconstructed the latent infection trajectory and estimated R_t. Focusing on the 2022 and early 2023 COVID-19 transmission trends in Singapore, where mass community wastewater surveillance has become routine, we performed estimations using a spectrum of data sources, including reported case counts, hospital admissions, deaths, and wastewater viral loads. We further explored the performance of our wastewater model across various scenarios with different sampling strategies. The results showed consistent estimates derived from models employing diverse data streams, while models incorporating more wastewater samples exhibited greater uncertainty and variation in the inferred R_ts. Additionally, our analysis revealed prominent day-of-the-week effect in reported case counts and substantial temporal variations in ascertainment rates. In response to these findings, we advocate for a hybrid approach leveraging both clinical and wastewater surveillance data to account for changes in case-ascertainment rates. Furthermore, our study demonstrates the possibility of reducing sampling frequency or sample size without compromising estimation accuracy for R_t, highlighting the potential for optimizing resource allocation in surveillance efforts while maintaining robust insights into the transmission dynamics of infectious diseases.