Wastewater-informed agent-based modelling of hepatitis E transmission dynamics

This study utilizes a wastewater-informed agent-based model calibrated with Approximate Bayesian Computation to demonstrate that pandemic lockdowns significantly reduced Hepatitis E transmission in Munich, revealing how combining wastewater surveillance with case data can uncover transmission dynamics masked in smaller datasets or by reporting biases.

The Gist

Imagine the city of Munich as a giant, bustling kitchen. In this kitchen, a sneaky little germ called Hepatitis E usually hides in the food people eat. Most of the time, when someone gets sick, it's because they ate something contaminated, like undercooked pork or shellfish.

Now, picture the city's sewage system as the kitchen's drain. Every time someone gets sick with Hepatitis E, the virus gets flushed down the drain and ends up in the wastewater. Scientists can look at this "drain water" like a detective looking at a crime scene to see how many germs are floating around, even if the sick people never went to the doctor.

The Mystery of the Lockdown

When the world went into lockdown during the pandemic, something strange happened in Munich's "drain water." The amount of Hepatitis E virus suddenly dropped by more than half.

This was a puzzle. If the virus is mostly spread through food, why would staying home and avoiding restaurants make the virus disappear from the sewers?

• Theory A: Maybe people stopped eating the risky food?
• Theory B: Maybe the virus is actually spreading in other ways (like person-to-person) that stopped when everyone stayed home?

Meanwhile, in a smaller town nearby, the "drain water" didn't show any drop. And the official list of sick people (who went to the doctor) in Munich didn't show a huge drop either. It was like looking at two different weather reports for the same storm: one said "heavy rain," the other said "sunny."

The Computer Detective

To solve this mystery, the researchers built a digital twin of the city. Think of this as a giant video game simulation where they created thousands of virtual people. They programmed these virtual people to:

1. Eat food.
2. Get sick (or not).
3. Flush the toilet.
4. Go to the doctor (or stay home).

They then ran this simulation millions of times, tweaking the rules until the virtual results matched the real-world data from 2020 to 2023. This is like trying to find the perfect recipe by tasting the soup over and over until it matches the flavor you remember.

The Big Reveal

The simulation told a clear story: The virus really did drop in the city during lockdowns.

The model estimated that transmission fell to about 35-40% of normal levels. This means the lockdowns did stop the virus from spreading, likely because people changed their habits (maybe cooking more at home or avoiding crowded places).

So, why didn't the other data show this?
The computer model acted like a pair of noise-canceling headphones that filtered out the confusion:

• The Doctor List: During lockdowns, people were more careful about going to the doctor. The model suggests that the few people who did get tested were actually more likely to be diagnosed, which masked the fact that fewer people were getting sick in the first place.
• The Small Town Drain: The smaller town's sewage system was like a tiny cup of water in a storm; the signal was too small and "noisy" to see the drop clearly.

The Takeaway

The main lesson here is that wastewater is a superpower for public health.

If you only look at the list of people who went to the doctor, you might miss the whole story. But if you look at the "drain water" and the doctor lists together, you get a crystal-clear picture. It's like trying to understand a movie by only looking at the subtitles (doctor reports) versus watching the whole film with sound (wastewater + reports).

By combining these clues, scientists can better understand how diseases spread, even when people aren't telling the whole truth or when the data is messy. This helps governments make smarter decisions to keep everyone safe.

Technical Summary

1. Problem Statement

Hepatitis E virus (HEV) is traditionally classified as a predominantly foodborne pathogen in developed nations, implying that transmission should be relatively stable and independent of human mobility or social distancing measures. However, during the COVID-19 lockdowns in Munich, Germany, a paradoxical observation emerged:

• Munich (Large Catchment): Wastewater concentrations of HEV at a treatment plant showed a significant decrease, suggesting that pandemic-related behavioral changes (e.g., reduced travel, altered food consumption) inadvertently reduced transmission.
• Smaller Catchment: In contrast, a smaller catchment area showed no comparable decline in either reported clinical cases or wastewater data.

The core scientific challenge was to determine whether the observed reduction in Munich was compatible with a near-exclusively foodborne transmission model and to reconcile the conflicting signals between the two surveillance modalities (clinical cases vs. wastewater) and the two geographic scales.

2. Methodology

To address these discrepancies, the authors developed a comprehensive modeling framework:

• Stochastic, Individual-Level Agent-Based Model (ABM): The researchers constructed a model simulating HEV transmission, viral shedding, and case ascertainment at the individual level within the Munich population. This allowed for the incorporation of heterogeneity in behavior and transmission routes.
• Approximate Bayesian Computation (ABC): Since HEV is often under-reported and likelihood functions can be complex, the team used ABC for parameter inference. This method allows for model calibration by comparing simulated data against observed data without requiring an explicit likelihood function.
• Calibration Strategy: The model was calibrated to data from 2020–2023 using three distinct approaches:
  1. Calibration against wastewater data alone.
  2. Calibration against reported clinical case data alone.
  3. Joint Inference: Simultaneous calibration against both wastewater and case data to leverage the strengths of both surveillance streams.

3. Key Contributions

• Integration of Multi-Source Data: The study demonstrates a novel framework for combining wastewater-based epidemiology (WBE) with traditional clinical surveillance to infer transmission dynamics for pathogens that are frequently under-reported.
• Resolution of Surveillance Discrepancies: It provides a mechanistic explanation for why different data sources (and different catchment sizes) yielded conflicting signals regarding HEV trends during the pandemic.
• Quantification of Lockdown Impact: The study moves beyond qualitative observation to provide quantitative estimates of how non-pharmaceutical interventions (NPIs) affected a pathogen not typically associated with person-to-person transmission.

4. Key Results

• Significant Transmission Decline: Posterior parameter estimates from the joint inference model revealed a substantial reduction in HEV transmission during lockdown periods. Transmission rates dropped to approximately 35–40% of the non-lockdown baseline.
• Statistical Significance: The 95% credible interval for this reduction was entirely below 1, confirming that the decline was statistically significant and not a random fluctuation.
• Explanation of Conflicting Signals:
  • Clinical Data: The lack of a visible decline in reported cases in the smaller catchment was attributed to a modest increase in diagnosis probabilities during lockdowns (likely due to increased healthcare seeking or testing focus), which masked the underlying drop in transmission.
  • Small Catchment Wastewater: The absence of a decline in the smaller catchment's wastewater data was attributed to higher measurement variability (noise) inherent in smaller sample sizes, which obscured the signal of reduced viral load.
• Foodborne Compatibility: The findings suggest that while HEV is primarily foodborne, the specific behavioral changes during lockdowns (likely related to food supply chains, dining habits, or reduced travel to endemic regions) were sufficient to disrupt transmission significantly.

5. Significance

This research highlights the critical value of wastewater-based surveillance as a complementary tool to clinical reporting. For pathogens like HEV, where clinical data is often sparse or biased by testing behaviors, wastewater data provides a more robust, population-level signal. By integrating these data streams into agent-based models, public health officials can achieve more accurate parameter inference, leading to better-informed risk assessments and more effective responses to future outbreaks or behavioral shifts in the population.