A multi-city COVID-19 forecasting model utilizing wastewater-based epidemiology

The COVID-19 pandemic highlighted shortcomings in forecasting models, such as unreliable inputs/outputs and poor performance at critical points. As COVID-19 remains a threat, it is imperative to improve current forecasting approaches by incorporating reliable data and alternative forecasting targets to better inform decision-makers. Wastewater-based epidemiology (WBE) has emerged as a viable method to track COVID-19 transmission, offering a more reliable metric than reported cases for forecasting critical outcomes like hospitalizations. Recognizing the natural alignment of wastewater systems with city structures, ideal for leveraging WBE data, this study introduces a multi-city, wastewater-based forecasting model to categorically predict COVID-19 hospitalizations. Using hospitalization and COVID-19 wastewater data for six US cities, accompanied by other epidemiological variables, we develop a Generalized Additive Model (GAM) to generate two categorization types. The Hospitaization Capacity Risk Categorization (HCR) predicts the burden on the healthcare system based on the number of available hospital beds in a city. The Hospitalization Rate Trend (HRT) Categorization predicts the trajectory of this burden based on the growth rate of COVID-19 hospitalizations. Using these categorical thresholds, we create probabilistic forecasts to retrospectively predict the risk and trend category of six cities over a 20-month period for 1, 2, and 3 week forecasting windows. We also propose a new methodology to measure forecasting model performance at change points, or time periods where sudden changes in outbreak dynamics occurred. We also explore the influence of wastewater as a predictor for hospitalizations, showing its inclusion positively impacts the model's performance. With this categorical forecasting study, we are able to predict hospital capacity risk and disease trends in a novel and useful way, giving city decision-makers a new tool to predict COVID-19 hospitalizations.