Watching the guards: a data-driven method to trigger warnings in national wastewater surveillance networks

Abstract

Surveillance networks have been established in many countries worldwide to monitor SARS-CoV-2 in sewage and to estimate the communal prevalence of COVID-19 cases. Despite their popularity, gaining a rapid understanding of how infectious diseases spread across the territory covered by a network is difficult because of the many factors involved. To improve the detection of warning signals within the territory, we propose to apply principal component analysis (PCA) to screen time-series data generated from wastewater treatment plants (WWTPs) under surveillance. Our analysis allows us to identify single WWTPs deviating from the normal behavior as well as deviations of a cluster of WWTPs (indicative of an intermunicipal outbreak). Our approach is illustrated through the analysis of the dataset generated by the Catalan Surveillance Network of SARS-CoV-2 in Sewage (SARSAIGUA). Using 10 principal components, we captured 78.6% of the variance in the original dataset of 51 variables (WWTPs). Our analysis identified exceedance of the Q-statistic threshold as evidence of anomalous performance of a single WWTP, and exceedance of the T2-statistic as a sign of an intermunicipal outbreak. Our approach provides a comprehensive picture of the spread of the COVID-19 pandemic, enabling decision-makers to make informed decisions and better manage future pandemics