Ceasing sampling at wastewater treatment plants where viral dynamics are most predictable

Area of Science:

• Environmental microbiology
• Infectious disease epidemiology
• Data science

Background:

• Wastewater sampling is a valuable tool for tracking infectious disease dynamics.
• The COVID-19 pandemic led to the expansion of wastewater sampling sites.
• As the pandemic subsides, there is a need to optimize sampling strategies and reduce operational costs.

Purpose of the Study:

• To develop and evaluate a method for identifying wastewater sampling sites that can be discontinued.
• To assess the information redundancy across different sampling locations.
• To optimize resource allocation for ongoing infectious disease surveillance.

Main Methods:

• Utilized machine learning models to predict daily mutation frequencies at one wastewater site based on data from other sites.
• Quantified the prediction error to assess the unique information content of each site.
• Applied the method to wastewater data from five locations in Switzerland.

Main Results:

• Identified wastewater sampling sites with the lowest prediction error as those containing the least unique information.
• Demonstrated a systematic approach to evaluating prediction errors and their interpretations.
• Found two out of five Swiss locations where sampling could be ceased with minimal loss of information.

Conclusions:

• Machine learning-based prediction error analysis is an effective method for optimizing wastewater surveillance networks.
• This approach enables data-driven decisions on discontinuing sampling sites, leading to resource efficiency.
• The findings support the adaptation of wastewater-based epidemiology for sustainable infectious disease monitoring.