Exposure notification system activity as a leading indicator for SARS-COV-2 caseload forecasting.

Area of Science:

• Epidemiology
• Public Health Informatics
• Computational Biology

Background:

• COVID-19 pandemic spurred development of digital Exposure Notification (EN) systems.
• These systems offer novel avenues for public health interventions.
• Limited research exists on utilizing real-time EN data for predictive epidemiological modeling.

Purpose of the Study:

• To assess the utility of real-time data from California's CA Notify (Google Apple Exposure Notification - GAEN platform) for short-term COVID-19 case forecasting.
• To determine if incorporating EN activity improves predictive accuracy compared to traditional models.

Main Methods:

• Extended a statistical model using historical case counts to predict future caseloads.
• Integrated anonymized EN activity data from CA Notify into the forecasting model.
• Compared model performance (with and without EN data) against actual reported caseloads for 1-7 day forecasts.

Main Results:

• Time series analysis revealed a temporal association between EN system activity and COVID-19 caseloads.
• Incorporating EN data significantly improved short-term caseload prediction accuracy.
• Bayesian inference confirmed a non-zero influence of EN terms, reducing Mean Absolute Percentage Error and Mean Squared Prediction Error by 5-32%.

Conclusions:

• Smartphone-based EN systems demonstrably enhance the accuracy of short-term epidemiological forecasts.
• These predictive models show promise for deployment as local early warning systems for resource allocation and intervention planning.