Multi-Level Biosensor-Based Epidemic Forecasting in Small Areas

• Activity Number: 165 - SLDS CSpeed 2
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Salvador Balkus* and Hua Fang and Honggang Wang
• Companies: University of Massachusetts Dartmouth and University of Massachusetts Dartmouth and University of Massachusetts Dartmouth
• Keywords: Epidemic Model Simulation; Biosensor Modeling; Biosensor Analysis; eHealth; mHealth; COVID-19

Abstract

Statistical models have been developed for predicting COVID-19 in individuals, but less work has been performed on how to leverage these predictions in forecasting cases at various geographic hierarchies. We have proposed a Multi-Level Adaptive and Dynamic Biosensor Epidemic Model (m-ADBio) for use in small areas such as municipalities, neighborhoods, or school districts. This new model attains higher accuracy than traditional SEIR models. m-ADBio can leverage real-time epidemic data streams to dynamically update predicted cases, avoiding underestimation problems from delayed testing reports. This model also leverages commuting data to estimate additional cases carried in from outside the region, which results in greater accuracy. These advances allow m-ADBio to adapt to arbitrarily small geographic regions. In turn, this provides more accurate forecasts at higher levels such as communities, counties, and states. We evaluate the model implementation through a case study of COVID-19 using real-world college-level data and simulated sensor data. Our m-ADBIO model, fully implemented as an R package, will be available for more accurate and detailed epidemic forecasts.