Governments attempting to monitor COVID-19 outbreaks usually have to make difficult tradeoffs.
Testing individual members of the population at scale is expensive and logistically difficult: limited laboratory equipment and medical facilities make it almost impossible to test an entire country’s population. A cheaper and logistically simpler strategy is to analyze sewage. Because COVID-19 and other viruses can be detected in human excrement, wastewater treatment plants offer aggregated information.
At the moment, wastewater monitoring is slower and less accurate than individual testing: it takes longer to get results, and it’s harder to determine the exact neighborhood where the outbreak occurred.
But we already have nearly all the knowledge we need to make wastewater monitoring faster and more accurate. We could detect and respond to COVID-19 (and other virus) outbreaks far more efficiently by:
- Driving a set of open standards;
- Developing miniaturized equipment; and
- Creating interoperable data platforms.
We recently outlined a five-part solution that could scale up to monitoring COVID-19 across an entire continent – or even globally. Since we believe many people, countries, and organizations could benefit from these recommendations, here’s a summary of the potential we’ve identified.
How wastewater monitoring works
Wastewater systems around the world work similarly. Water is dumped from individual houses into a larger centralized pipe system. The sewage then flows to a central water treatment plant.
A virus can often be detected in an infected person’s fecal matter. That’s why looking at the virus concentration detected at a particular water treatment plant allows us to estimate the number of infected people in the area served by the plant.
But wastewater can take days or weeks to reach a centralized treatment plant, since there are often intermediary cleaning steps along the way. So monitoring wastewater at the plant often means that while we know an outbreak happened, and we know how big it was, we don’t know exactly when or where in the catchment area it occurred (experts refer to this as poor spatial and temporal resolution).
For this reason, wastewater monitoring is often overlooked as a potential solution to detect viral outbreaks.
It would take a large-scale, coordinated effort to improve the current technology and processes in wastewater monitoring. But given the value of pinpointing new viral outbreaks, and compared to the cost and effort required to scale up individual testing, it’s clear this work needs to be done.
Our five-part solution to improve wastewater monitoring
Our proposed five-part solution to improve and scale wastewater monitoring includes:
- Standardized guidelines on how to set up monitoring equipment and take samples: All parties would need to agree to this set of standards, because in order to have comparable data, everyone has to do this in the same way.
- Reference methods for preparing samples and quantifying viruses: Even with standard guidance on setting up monitoring equipment, everyone further needs to agree on how to prepare samples in the lab and how to measure the concentration of specific viruses in wastewater.
- Miniaturized instrumentation to measure wastewater closer to the source: We need to manufacture smaller versions of certain monitoring equipment.
- Interoperable data platforms to share data across departments and countries: These need to be developed.
- Predictive analysis with machine learning: This would enable us to detect new viruses or new outbreaks of known viruses early enough to coordinate an appropriate response.
Some of these components already exist or are already being developed. We look at each of these five recommendations in detail to show where we are and where we could be (in 3–24 months’ time).
Solution 1: Monitoring guidelines
Producing a set of guidelines for treatment plants on how to extract samples and optimize their monitoring processes would support ensuring consistent data in all later steps.
Where we are:
- We already use wastewater monitoring to analyze and describe outbreaks retrospectively. Different cities do this in slightly different ways, and many steps are done manually. For example, there are two different methods of sewage sampling:
Grab sampling: taking a single sample at a single point in time;
Composite sampling: collecting and combining samples over a 24-hour period.
- Current processes often require highly trained experts onsite at each treatment plant.
- It can take weeks to get useful results, and the data isn’t always comparable between plants or countries.
Where we could be:
- With improved monitoring equipment and processes, we could proactively analyze wastewater to predict upcoming outbreaks.
- By automating as many steps as possible, standardizing equipment and protocols, and ensuring consistency across plants and countries, we could reduce the cost and improve the quality of this monitoring.
Solution 2: Reference methods for preparing samples and quantifying viruses
Creating reference samples and protocols for laboratories would ensure that sample preparation and concentration measurements are consistent.
Where we are:
- We know we can detect viruses – including SARS and SARS-CoV-2 – in wastewater, but there are several ways of measuring samples and quantifying virus concentrations.
- It’s not always clear what “normal” is.
- Samples might be prepared in different ways, which can affect the results and make it difficult to compare data across laboratories.
Where we could be:
- By creating standardized protocols for preparing samples and quantifying viruses, as well as agreeing on a set of reference samples, different labs in different countries can share and compare data. This standardization would enable faster, more sensitive analysis – both important factors, especially for an early warning system.
Solution 3: Miniaturized instrumentation for collecting and normalizing fecal samples
Manufacturing smaller versions of the monitoring equipment designed to work in large treatment plants would help us pinpoint outbreaks more accurately and cost-effectively.
Where we are:
- Wastewater treatment plants measure averaged information over their entire catchment area. There are logistical challenges to replicating this sampling process at a more fine-grained level, as some of the equipment is designed for large treatment plants.
- This means that an outbreak in a smaller part of a specific catchment area might be “smoothed out” by the average of the larger area, making it difficult to detect the outbreak.
Where we could be:
- Developing miniaturized sampling equipment would allow us to take more detailed measurements.
- More detailed measurements could tell us exactly where an outbreak occurred – down to a city block or even a specific building.
- By not waiting for the wastewater to hit a centralized plant, we could also obtain these measurements more quickly, giving us better spatial and temporal resolution – that is, allowing us to act to curb or even prevent an outbreak.
Solution 4: Developing an interoperable platform to integrate environment and health data
A single data platform used across departments, roles, and cities would get everyone on the same page.
Where we are:
- Many departments and countries are already sharing their data more widely. This allows for some collaboration among other countries and departments, and even lets members of the public get information directly.
- Health departments and water management departments don’t always collaborate effectively, and it’s sometimes difficult to use insights from wastewater management to drive health-related decisions.
- It’s still difficult for countries to share data.
Where we could be:
- Building an interoperable data platform to process and combine indicators from water management departments (e.g., a surge in viral concentration) and health departments (e.g., infection and hospitalization rates) would help us make more effective decisions.
- By encouraging multiple cities to adopt the same platform, we could easily see everything from the “bird’s-eye view” down to an outbreak in a specific postcode.
- Researchers, engineers, politicians, and decision-makers could all obtain their data from a reliable, central “source of truth.”
Solution 5: Predictive analytics
We can use machine learning to make automated predictions of when and where outbreaks will occur.
Where we are:
- Wastewater monitoring is currently too slow to be useful for an early warning system.
- Even if all the solutions above were implemented, it still takes time to gather and analyze the samples and data we need.
- We currently use this data mainly for retrospective analysis.
Where we could be:
- By using advanced machine learning, we could find new patterns in wastewater that indicate an outbreak – even before patients first report symptoms.
- We can use machine learning to correct for the factors that bias our measurements, such as the time of day, day of the week, or even the weather – all of which influence the number of people within a catchment area at any given time.
- Metagenomic analysis would also help us detect new viruses. We can monitor for new viral DNA that hasn’t previously been detected in wastewater and map it to known viruses to identify potentially dangerous outbreaks of new mutations.
Examples of promising wastewater monitoring projects
- France and the Netherlands are leading the way in Europe. In April 2020, France already saw the first results from a month of sampling, and the Netherlands scaled up monitoring to the national level beginning in August 2020.
- In the Netherlands, the RIVM takes weekly samples from over 300 sewage plants and makes this data public via an online dashboard.
- The EU has said that an EU-wide project is viable.
Would you like to bring this technology to your city?
If you’re involved in water management or viral outbreak monitoring and want to discuss bringing this technology to your city, we’d love to talk. If you’re a resident who would benefit from your city adopting this technology, please share this article with your representatives.
Book a call with our CEO if you’d like to discuss using machine learning for health monitoring.
