EpiFlow Improves Disease Forecasting Using Wastewater Data

This paper introduces EpiFlow, a comprehensive framework designed to improve the utility of wastewater surveillance data for real-time disease forecasting. While wastewater viral loads (WVL) are known to correlate with disease burden, their full potential in predictive models has been underexplored, particularly when traditional indicators suffer from reporting fatigue or low prevalence. EpiFlow addresses this by providing principled methods for processing WVL data, analyzing its causal relationship with disease burden indicators, and integrating these insights into a time-varying forecasting model. The framework also considers the impact of reporting delays. The researchers validated EpiFlow by forecasting COVID-19 hospital admissions in Virginia. Their findings indicate that incorporating WVL through EpiFlow significantly boosts forecast accuracy compared to baseline models, achieving a 20 percentage point improvement in forecast coverage, even during periods of low prevalence or with delayed WVL reporting.