Related Experiment Video
Updated: Jul 10, 2025
10:53
Concentration of Virus Particles from Environmental Water and Wastewater Samples Using Skimmed Milk Flocculation and Ultrafiltration
Published on: March 17, 2023
1.6K
Optimizing campus-wide COVID-19 test notifications with interpretable wastewater time-series features using machine
Tuo Lin1, Smruthi Karthikeyan2, Alysson Satterlund3
1Department of Biostatistics, University of Florida, Gainesville, FL, 32608, USA.
Scientific Reports
|November 24, 2023
Summary
Wastewater surveillance for SARS-CoV-2 effectively predicts COVID-19 infections in residences. A positive wastewater sample on at least 3 of the past 7 days is the best indicator for individual infections.
Area of Science:
- Environmental Science
- Epidemiology
- Public Health
Background:
- Wastewater surveillance for SARS-CoV-2 is effective for population health monitoring.
- High-resolution wastewater surveillance at the University of California, San Diego (UCSD) aims to detect undiagnosed COVID-19 cases and trigger resident notifications.
- Optimal indicators for triggering these notifications remain undetermined.
Purpose of the Study:
- To develop a data processing pipeline to identify key features in wastewater surveillance data that predict COVID-19 presence in associated residences.
- To establish optimal determinants for triggering public health interventions based on wastewater data.
Main Methods:
- Utilized time series data of wastewater SARS-CoV-2 results and individual testing results from UCSD students (11/2020-11/2021).
- Developed hierarchical classification/decision tree models to identify informative wastewater features predicting individual infections.
- Compared decision tree model performance against other statistical and machine learning models.
Main Results:
- The most effective predictor of positive individual COVID-19 tests in residence buildings was wastewater samples testing positive on at least 3 of the preceding 7 days.
- Hierarchical decision tree models demonstrated superior predictive performance compared to other models while maintaining interpretability.
Conclusions:
- The study successfully identified key wastewater surveillance features for predicting individual COVID-19 infections within residential settings.
- Findings have informed the refinement of campus-wide public health guidelines and notification systems at UCSD to enhance early detection and response to potential outbreaks.
