Modeling the number of people infected with SARS-COV-2 from wastewater viral load in Northwest Spain

Abstract

The quantification of the SARS-CoV-2 RNA load in wastewater has emerged as a useful tool to monitor COVID-19 outbreaks in the community. This approach was implemented in the metropolitan area of A Coruña (NW Spain), where wastewater from a treatment plant was analyzed to track the epidemic dynamics in a population of 369,098 inhabitants. Viral load detected in the wastewater and the epidemiological data from A Coruña health system served as main sources for statistical models developing. Regression models described here allowed us to estimate the number of infected people (R² = 0.9), including symptomatic and asymptomatic individuals. These models have helped to understand the real magnitude of the epidemic in a population at any given time and have been used as an effective early warning tool for predicting outbreaks in A Coruña municipality. The methodology of the present work could be used to develop a similar wastewater-based epidemiological model to track the evolution of the COVID-19 epidemic anywhere in the world where centralized water-based sanitation systems exist.

Keywords: COVID–19; Generalized Additive Models (GAM); Kernel smoothing; LOESS; SARS-CoV-2; Wastewater-based epidemiology.

Graphical abstract

Fig. 1

Map showing Galicia region in Spain, the metropolitan area of A Coruña including Oleiros, Cambre, Culleredo, Arteixo and A Coruña municipalities, and the Health Area of A Coruña-Cee, as well as the specific locations of the University Hospital of A Coruña (CHUAC) and WWTP Bens.

Fig. 2

Iceberg representing the overall health of the population of the metropolitan area of A Coruña infected by SARS–CoV–2, showing the real and official cases estimated in this study. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

Fig. 3

Estimation of the COVID–19 positive cases using simple linear regression models. A) ICU patients versus COVID–19 positive cases in A Coruña – Cee health area, and its linear fit for the period April 29th on. B) Relative cumulative cases versus relative positive cases in the health area of A Coruña – Cee and their linear fit.

Fig. 4

Viral load detected in the influent of the WWTP Bens (red line) and the estimated number of COVID–19 positive cases (blue line) in the metropolitan area of A Coruña.

Fig. 5

Correlation analysis between estimated COVID–19 positive cases and different variables. Scatterplot matrix shows fitted linear models and linear correlation coefficients of each pair of variables: time (measured in days from the beginning of reported COVID–19 cases), estimated COVID–19 positive cases (positive cases), daily mean viral load measured in Bens, mean flow of sewage water in Bens, mean rainfall, daily mean humidity and daily mean temperature.

Fig. 6

Smooth effect (solid line) and confidence band (shaded area) for the viral load (A) and for the mean flow (B) on the number of estimated COVID–19 active cases when fitting a GAM.

Fig. 7

Estimation of the number of COVID–19 real active cases using a linear regression model. Scatterplot represents the logarithm of the viral load measured in WWTP Bens and the estimated number of COVID–19 real active cases before (A) and after (B) removing the three outliers detected. The linear fit (solid line) and the confidence band (shaded area) are also included.

Fig. 8

Estimation of number of real COVID–19 active cases using the quadratic LOESS model. A) Estimated COVID–19 active cases in the metropolitan area vs the natural logarithm of the viral load. B) Estimated number of COVID–19 active cases in the metropolitan area vs the predicted number of COVID–19 real active cases.

Fig. 9

Estimation of number of real COVID–19 active cases using the linear model (in green), including the prediction confidence intervals. These estimates are compared with the number of cases of infected persons reported by the Galician health services (in blue). A) Trends corresponding to 2020. B) Trends corresponding to 2021. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)