Related Experiment Video
Updated: Jun 26, 2025
08:26
Large-Scale SARS-CoV-2 Testing Utilizing Saliva and Transposition Sample Pooling
Published on: June 23, 2022
1.8K
COVID-19 hotspot detection in a university setting
Garrett Duncan1, William F Christensen1, Camilla Handley1
1Department of Statistics, Brigham Young University, Provo, Utah, United States of America.
Plos One
|May 16, 2024
Summary
This study introduces a novel approach for universities to identify COVID-19 hotspots using student data and an XGBoost model. This method helps balance in-person learning with public health by enabling targeted online transitions for at-risk groups.
Area of Science:
- Epidemiology
- Data Science
- Higher Education Administration
Background:
- The COVID-19 pandemic significantly disrupted academic institutions, necessitating adaptive policies to balance education and community health.
- Universities often implemented temporary shutdowns of in-person activities based on perceived COVID-19 transmission hotspots.
- Effective identification of these hotspots is crucial for maintaining academic continuity while mitigating disease spread.
Purpose of the Study:
- To develop and evaluate a data-driven approach for identifying COVID-19 hotspots within university settings.
- To provide administrators with a tool for making informed decisions about temporary shifts to online learning.
Main Methods:
- Utilized an XGBoost model incorporating student demographic data and weekly COVID-19 testing outcomes to estimate individual positive test probabilities.
- Engineered demographic variables to improve model predictive accuracy.
- Employed simulation experiments to establish a null distribution of positivity rates under no in-class transmission.
- Applied a modified false discovery rate procedure to identify statistically significant hotspots based on observed versus expected group positivity rates.
Main Results:
- The XGBoost model, enhanced with engineered demographic features, effectively estimated individual COVID-19 risk.
- Simulation-based analysis allowed for the identification of academic groups (classes, majors) with COVID-19 rates exceeding demographic expectations.
- The approach demonstrated practical utility in an anonymized Fall 2020 university case study.
Conclusions:
- The proposed method provides a robust framework for detecting infectious disease hotspots in interconnected populations, adaptable beyond university settings.
- This approach supports evidence-based decision-making for managing public health risks within organizations.
- It offers a scalable solution for monitoring and responding to disease outbreaks in academic and other communal environments.
