COUnty aggRegation mixup AuGmEntation (COURAGE) COVID-19 prediction.

Area of Science:

• Epidemiology
• Data Science
• Public Health

Background:

• The COVID-19 pandemic poses a global health threat, necessitating accurate localized predictions for effective resource management.
• Predicting disease severity at a granular level is essential for proactive public health interventions and healthcare system preparedness.

Purpose of the Study:

• To develop a novel deep learning method, COURAGE, for short-term prediction of COVID-19 deaths at the county level in the United States.
• To leverage advanced AI techniques for accurate and efficient forecasting of localized COVID-19 mortality.
• To provide a tool for improved resource allocation and public health planning.

Main Methods:

• The COURAGE method utilizes a transformer-based deep learning model, adapted from Natural Language Processing, to analyze time-series data.
• The model incorporates publicly available data, including COVID-19 confirmed cases, deaths, community mobility trends, and demographic information.
• It captures both short-term and long-term dependencies in the data for robust prediction.

Main Results:

• The COURAGE model demonstrates state-of-the-art performance in predicting COVID-19 related deaths.
• Numerical experiments confirm the model's accuracy in forecasting county-level and aggregated state-level fatalities.
• The method proves computationally efficient while maintaining high predictive power.

Conclusions:

• The COURAGE model offers a powerful and accurate solution for short-term COVID-19 death prediction at the county level.
• Its reliance on publicly available data and deep learning makes it a valuable tool for public health agencies.
• Accurate localized forecasting facilitates better resource allocation and response strategies during the evolving pandemic.