County augmented transformer for COVID-19 state hospitalizations prediction

Area of Science:

• Public Health
• Data Science
• Epidemiology

Background:

• The COVID-19 pandemic has strained medical resources, necessitating effective public health decision-making.
• Accurate forecasting of COVID-19 hospitalizations is critical for efficient medical resource allocation.

Purpose of the Study:

• To propose a novel method, County Augmented Transformer (CAT), for predicting COVID-19 hospitalizations.
• To generate accurate four-week-ahead hospitalization predictions for all US states.

Main Methods:

• Utilized a transformer-based deep learning model, a self-attention mechanism adept at capturing temporal dependencies.
• Incorporated publicly available data: COVID-19 confirmed cases, deaths, hospitalizations, and household median income.
• Developed a data-driven approach for time series forecasting.

Main Results:

• The CAT model demonstrated strength and usability in predicting COVID-19 hospitalizations.
• Numerical experiments validated the model's accuracy and efficiency.
• The model effectively captures both short-term and long-term patterns in hospitalization data.

Conclusions:

• The County Augmented Transformer (CAT) model shows promise as a tool for medical resource allocation.
• Accurate hospitalization predictions can significantly improve public health response strategies.
• This data-driven approach offers a scalable solution for pandemic resource management.