Europe PMC

This website requires cookies, and the limited processing of your personal data in order to function. By using the site you are agreeing to this as outlined in our privacy notice and cookie policy.

Applications of wastewater-based epidemiology as a leading indicator for COVID-19

Preprint from arXiv, 01 Apr 2021
PPR: PPR343017 

Preprint v1

This article is in the Europe PMC Open access subset. Refer to the copyright information in the article for licensing details.
Free full text in Europe PMC

This article is a preprint. It may not have been peer reviewed.
A preprint is a complete scientific manuscript that an author uploads on a public server for free viewing. Initially it is posted without peer review, but may acquire feedback or reviews as a preprint, and may eventually be published in a peer-reviewed journal. The posting of preprints on public servers allows almost immediate dissemination and scientific feedback early in the 'publication' process.

Preprint version history

  • Version 2 [23 Jun 2021]
  • Version 1 [01 Apr 2021]

Abstract 

Individuals infected with SARS-CoV-2, the virus that causes COVID-19, may shed the virus in stool before developing symptoms, suggesting that measurements of SARS-CoV-2 concentrations in wastewater could be a "leading indicator" of COVID-19 disease activity. Multiple studies have corroborated the leading indicator concept by showing that the correlation between wastewater measurements and COVID-19 case counts is maximized when case counts are lagged. However, wastewater-based epidemiology for COVID-19 encompasses many possible applications, and the meaning and quantification of a "leading indicator" may be different for different applications. Here we enumerate three different applications of wastewater-based epidemiology for COVID-19: a qualitative "early warning" system; an independent, quantitative estimate of disease activity; and a quantitative alert of bursts of disease activity. We find that the leading indicator concept has different implications and utility for each application and that wastewater's lead time will vary depending on epidemiological, biological, and health systems factors. There is no one single lead time for wastewater-based COVID-19 monitoring, and, at least in the case of using wastewater to inform estimates of population-wide disease activity, wastewater's lead time is not important to its public health utility.

Free full text 

arXiv

PPRID: PPR343017
EMSID: EMS125739ArXiv preprint, version 1, posted 2021 April 01

Applications of wastewater-based epidemiology as a leading indicator for COVID-19

Affiliations

  1. 1.Biobot Analytics, Inc., Cambridge, MA, USA

Copyright and license information

Copyright notice

This work is licensed under a CC BY-SA 4.0 International license.

This article is a preprint. It may not have been peer reviewed.
A preprint is a complete scientific manuscript that an author uploads on a public server for free viewing. Initially it is posted without peer review, but may acquire feedback or reviews as a preprint, and may eventually be published in a peer-reviewed journal. The posting of preprints on public servers allows almost immediate dissemination and scientific feedback early in the 'publication' process.

Abstract

Individuals infected with SARS-CoV-2, the virus that causes COVID-19, may shed the virus in stool before developing symptoms, suggesting that measurements of SARS-CoV-2 concentrations in wastewater could be a “leading indicator” of COVID-19 disease activity. Multiple studies have corroborated the leading indicator concept by showing that the correlation between wastewater measurements and COVID-19 case counts is maximized when case counts are lagged. However, wastewater-based epidemiology for COVID-19 encompasses many possible applications, and the meaning and quantification of a “leading indicator” may be different for different applications. Here we enumerate three different applications of wastewater-based epidemiology for COVID-19: a qualitative “early warning” system; an independent, quantitative estimate of disease activity; and a quantitative alert of bursts of disease activity. We find that the leading indicator concept has different implications and utility for each application and that wastewater’s lead time will vary depending on epidemiological, biological, and health systems factors. There is no one single lead time for wastewater-based COVID-19 monitoring, and, at least in the case of using wastewater to inform estimates of population-wide disease activity, wastewater’s lead time is not important to its public health utility.

Keywords: wastewater-based epidemiology, SARS-CoV-2, COVID-19, surveillance

Introduction

Wastewater-based epidemiology (WBE), the use of measurements from wastewater to aid public health surveillance, is increasingly used in the COVID-19 pandemic as a complement to more traditional monitoring methods like diagnostic testing (National Wastewater Surveillance System 2020; Larsen & Wigginton 2020). WBE has gained particular attention in part because wastewater concentrations of SARS-CoV-2 have been found to be a “leading indicator” of reported COVID-19 case counts (Table 1 and reference therein). The biological principle behind wastewater as a leading indicator is that many infected individuals shed the virus in stool before or around the time they develop symptoms and thus before they seek medical care (Daughton 2020; Zhu et al. 2020).

Table 1. Reported lead times for wastewater SARS-CoV-2 concentrations in the context of ongoing COVID-19 activity.
Study Study location Study period Lead time (days) Methodology for quantifying lead time
D’Aoust et al. 2021 Ottawa, Canada Jun-Aug 2020 2 Maximum Pearson correlation between wastewater and number of new cases
Feng et al. 2021 12 WWTPs covering 10 cities in Wisconsin, USA Aug 2020-Jan 2021 0 to 6 (different for each WWTP) Maximum Spearman correlation between wastewater and smoothed (7-day average) number of new cases
Kumar et al. 2021 Gandhinagar, Gujarat, India Aug-Sep 2020 14 Visual inspection of percent change in wastewater concentration and number of new cases
Nemudryi et al. 2020 Bozeman, Montana, USA Mar-Jun 2020 2 (mid-March); 4 (May) Maximum Pearson correlation between wastewater measurements and number of positive tests
Peccia et al. 2020 New Haven, Connecticut, USA Mar-Jun 2020 6-8 Distributed lag time series model linking wastewater measurements and number of positive tests by report date
Wu et al. 2020b Greater Boston, MA, USA Jan-May 2020 4 (maximum correlation); 4-10 (range) Pearson correlation between unsmoothed viral titers in wastewater and number of new cases
Wurtzer et al. 2020 Paris, France Mar 2020 8 Visual inspection of wastewater and number of positive tests

However, we suspect the term “leading indicator” is being used in multiple senses in the context of COVID-19 WBE, perhaps because the term, originally used in economics and business (Bloom et al. 2007), has not seen widespread use in infectious disease public health. 1 Here we aim to define what it means for wastewater to be a “leading indicator” for COVID-19 and how wastewater’s lead time can be quantified. Here we review three main applications of WBE for COVID-19 and discuss what “leading indicator” means in the context of each application.

Application 1 - Outbreak detection on a background of little disease

The first application is a qualitative “early warning” system, testing for a detection or nondetection of SARS-CoV-2 in wastewater (Hassard et al. 2020; National Wastewater Surveillance System 2020; Zhu et al. 2020). In other words, the goal in this application is to answer the question: are there currently more than zero infected individuals in the surveilled population?

This application is relevant at the level of individual facilities, such as a college dormitory, correctional institution, or nursing home (Betancourt et al. 2020; Harris-Lovett et al. 2021; Peiser 2020; Targeted Wastewater Surveillance at Facilities, Institutions, and Workplaces 2020), as well as larger wastewater catchments, such as a city with little ongoing transmission of SARS-CoV-2 (COVID-19 weekly surveillance reports; Fongaro et al. 2021; Jørgensen et al. 2020; Medema et al. 2020; Randazzo et al. 2020). If the surveilled population has had zero (or very few) actively infected individuals and no detectable SARS-CoV-2 in wastewater, then the detection of SARS-CoV-2 in wastewater is an indication that one or more individuals in the population have become infected, hopefully triggering interventions to prevent the spread of disease.

If there is no active case finding in the relevant population, then wastewater, insofar as it can detect the presence of infected individuals before they present with symptoms, is a “leading indicator” with a lead time of 1 to 6 days (Zhu et al. 2020). If there is active monitoring for SARS-CoV-2, then wastewater’s lead time is the delay between when an infected individual begins shedding and when they would be detected by other means. Thus, this lead time depends on public health interventions in the relevant population. If there are sufficient resources to allow screening everyone in the monitored population for SARS-CoV-2 every day, then wastewater will likely not provide additional useful information (Peccia et al. 2020). Conversely, if there is little active case finding, wastewater will be a leading indicator.

Application 2 - Independent estimate of community-level disease activity and trends

The second application of WBE for COVID-19 is as a population-level estimate of disease activity. Rather than measuring SARS-CoV-2 activity using just those individuals who tested positive and were formally counted as a case, wastewater detects virus in a sample formed from the pooled excretions of many individuals, regardless of whether they are symptomatic, have access to healthcare, or whose healthcare system has abundant resources for testing (Medema et al. 2020). Thus, this application is appealing because it provides estimates that are potentially less biased and less resource-intensive compared to traditional disease monitoring using diagnostic testing (National Wastewater Surveillance System 2020; D’Aoust et al. 2021; Daughton 2020; Larsen & Wigginton 2020; Wu et al. 2020a).

WBE and diagnostic testing can also provide synergistic insights. If trends in both case rates and wastewater change directions, public health officials can be more certain that disease activity has truly passed an inflection point. If wastewater measurements rise while case counts remain stable, then diagnostic testing is undercounting cases (Wu et al. 2020a). Conversely, if rates of positive diagnostic tests rise but wastewater measurements remain stable, then the apparent increase in cases may be due to increased testing (i.e., less undercounting) rather than to increased disease activity (Gerrity et al. 2021).

In principle, this second application of WBE could be a “leading indicator”. Various reports have suggested that correlations between wastewater concentrations of SARS-CoV-2 and reported COVID-19 cases are maximized when cases are lagged by 2 to 8 days (i.e., when wastewater leads cases by 2 to 8 days; Table 1). This interpretation, however, is subject to two important caveats.

First, to say that wastewater is a “leading indicator” with a lead time of D days does not mean that today’s wastewater measurement says exactly how many new cases will be reported D days from now (Figure 1). Of course, neither wastewater measurements nor reported case counts are perfectly accurate: the time series of new case counts, offset by D days, would never perfectly line up with the timeseries of WBE measurements because of measurement error. However, even if wastewater measurements and reported case counts were perfectly accurate, we would not expect the two time series to line up, since individuals who are shedding virus in stool today are at different places in their disease trajectory (Hoffman & Alsing 2021). Some will become symptomatic sooner, some later, and some not at all. Some may already have become counted as reported cases. Variability in multiple factors —disease time courses, healthcare seeking behavior, access to testing, and testing turnaround time— manifests as a shifting but also a smoothing of case count rates relative to shedding rates, thus creating a complex temporal relationship between wastewater measurements and case counts.

Figure 1
Open in new tabFigure 1

COVID-19 incidence, as measured by reported case counts, is not simply a translated (or “lagged”) copy of the time series of wastewater SARS-CoV-2 measurements. Black ticks mark 3 major US holidays (Thanksgiving, Christmas, New Year’s Day). Wastewater data were collected from the Deer Island North Wastewater Treatment plant, which serves Boston, Massachusetts, USA and its northern suburbs, as described elsewhere (Wastewater COVID-19 Tracking; Wu et al. 2020a). Case counts include Massachusetts’s Suffolk and Norfolk counties (US Coronavirus Cases & Deaths by State 2021).

Second, even insofar as a single lead time is meaningful for signals that do not perfectly align, that lead time will vary across populations and through time in a single population, and it will vary depending on the statistical methodology used (Figure 2). From a biological point of view, we would expect the lead time to change depending on who is being infected, their disease severity, and the precise SARS-CoV-2 variant that infected them (Cevik et al. 2020; Kissler et al. 2021). From a health systems point of view, the delay between shedding virus and becoming a reported case depends on the availability and accessibility of diagnostic tests. More testing and faster turnaround time means shorter lead time. Thus, we should not expect that studies will all report the same lead time for this application of WBE for COVID-19.

Figure 2
Open in new tabFigure 2

Correlations between measured wastewater SARS-CoV-2 concentrations and reported COVID-19 case counts (y-axis) vary depending on lead time (x-axis), correlation metric, incidence metric, and time. Zero lead time refers to the correlation between wastewater and the case counts on the day of the wastewater sampling. Positive lead times refer to wastewater correlated with later case counts (e.g., a lead time of +3 days refers to the correlation between wastewater and cases 3 days later). Negative lead times refer to wastewater correlated with earlier case counts. Reported correlations use the Spearman metric and the 7-day average of case counts unless noted otherwise. (Thus, the “Spearman” curve in the left panel and the “7-day average” curve in the middle panel are identical.) “Pearson (log)” means the Pearson correlation between 7-day average case counts and the logarithm of wastewater concentrations. Data used are the same as in Figure 1, using all available data from March 2020 to March 2021.

Given the complexity around wastewater’s lead time in this application, we hypothesize that WBE has more value as an independent measurement and additional public health data stream rather than as a precise leading indicator of population-level disease activity. As an analogy, we note that influenza monitoring systems include counts of laboratory-confirmed influenza-related hospitalizations as well as counts of outpatient visits for influenza-like illness. In principle, outpatient visits could be a leading indicator of hospitalizations because individuals with severe disease would visit a physician before becoming hospitalized. However, the main value of outpatient visits for influenza monitoring is not as a leading indicator but rather as a complementary window onto disease activity. A few days’ extra notice is not as important as having an independent confirmation that flu season is generally accelerating, or has peaked and is on the decline. Analogously, for this application, WBE’s primary value is in providing an independent confirmation that COVID-19 activity is generally high or low, or generally rising or falling, rather than having a lead time of 2 to 8 days (Gonzalez et al. 2020).

Application 3 - Outbreak detection on a background of ongoing disease activity

The third application is to detect an outbreak or “burst” of disease activity using wastewater measurements. Unlike the first application, the goal here is to identify a burst on a background of ongoing disease activity. For example, given daily wastewater sampling through a holiday or special event, such as Thanksgiving or the Superbowl in the US, could the number of excess infections that occurred specifically because of that event be confidently estimated? Ideally, ongoing SARS-CoV-2 wastewater monitoring could detect the burst of infections and alert public health officials about the timing and magnitude of the resulting surge in cases and hospitalizations. In this application, WBE would be a “leading indicator” insofar as it would provide an alert about the burst in disease activity before other monitoring systems, like case counts, would.

In principle, this approach could produce an indicator with some days lead over case counts alone. However, to our knowledge, this application has seen only anecdotal and not quantitatively rigorous use (Fox 2020). In contrast to Application 2, which involves identifying trends in disease activity over weeks, this application requires quantifying changes in trends that occur rapidly, perhaps on the timescale of days rather than weeks. This short timescale exacerbates the analytical challenges faced by the other applications, including the variability in wastewater measurements and case counts. Importantly, validating a detection algorithm requires establishing a gold standard of true disease activity, to determine if wastewater measurements can reliably identify inflection points in that disease activity. Case counts, although an attractive indicator of disease activity, cannot themselves be the gold standard because of their weekly patterns (e.g., fewer cases counted on weekends) and unusual behavior during holidays and special events (e.g., the periods around major US winter holidays in Figure 1). Rolling 7-day averages of case counts are also an inappropriate gold standard because, although they are not subject to within-week variations in case reporting, the 7-day average obscures exactly the short-term bursts in activity this application aims to detect (Bloom et al. 2007). Thus, a robust burst-detection algorithm likely needs to include a sophisticated statistical inference of underlying disease activity, which remains an area of ongoing research.

Conclusions

When using WBE to detect new infections on a background of little or no disease activity, wastewater is a leading indicator insofar as it can identify new infections before those individuals would be identified by some other method. If there are sufficient resources to allow screening everyone in the monitored population for SARS-CoV-2 every day, then wastewater likely leads reported case counts by very little time, if any. If there is no active case finding, then WBE is a leading indicator of 2 to 8 days, that is, the time between when an individual begins shedding detectable virus in their stool and when they seek diagnostic testing.

When measuring disease activity in a larger community, wastewater’s lead time will depend on the population that is infected and the health systems that serve them, both of which vary from place to place and time to time. Reporting a lead time that is a single number is therefore misleading. Furthermore, there may be many lead times that describe the relationship between WBE measurements and reported case counts nearly equally well (Figure 2; D’Aoust et al. 2021; Graham et al. 2021). If a lead time of +4 days (i.e., in which WBE is a leading indicator) correlates only very slightly “better” than a lead time of -1 days (i.e., in which WBE is a lagging indicator), then why would we confidently say that WBE is leading by 4 days? What practical use would there be to propound a single lead time?

Given the complexity of WBE’s lead time at the community level, we emphasize the value of WBE during periods of sustained community transmission as an independent indicator of disease activity that is unaffected by access to testing, healthcare-seeking behaviors, or other socio-behavioral factors that are critical to pandemic response (D’Aoust et al. 2021; Zhu et al. 2021). WBE provides public health officials evidence, in real time, about whether testing is providing a reliable indicator of COVID-19 disease activity.

As the field of WBE for COVID-19 grows and matures, new applications will arise, and existing applications will be refined. Practitioners should recognize that the utility of WBE for public health will vary between applications and will change over time as our methodologies improve and as the epidemiology of COVID-19 changes. Lead times are only one part of that larger picture.

Acknowledgments

Helpful discussions with Eric Alm, Mary Bushman, Nour Sharara, and Amy Xiao.

Author Information

Correspondence: oi.toboib@erialc

Notes

1 A PubMed search on 10 Mar 2021 for “leading indicator” and any of “public health”, “infectious disease”, “SARS-CoV-2”, or “COVID-19” yielded 38 publications, many of which dealt with occupational health, predictors of an individual’s disease trajectory, or “leading” in the sense of “most important” rather than “temporally before”.

Statements

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

SO provides compensated consulting services to Biobot Analytics, Inc. MI and CD are employees of Biobot Analytics, Inc.

Data available upon request.

References

History

  • Posted April 01, 2021.

Full text links 

Free to read at arXiv
https://arxiv.org/abs/2104.00684v1