Forecasting COVID-19 daily cases using phone call data.

Area of Science:

• Epidemiology
• Public Health
• Data Science

Background:

• Forecasting COVID-19 cases is crucial for managing the epidemic.
• Existing models (compartmental, statistical, AI) have limitations in accuracy, uncertainty handling, and local area focus.
• Previous forecasting efforts often provided disappointing accuracy and overlooked local needs.

Purpose of the Study:

• To develop a simple, interpretable, and reliable forecasting model for daily COVID-19 confirmed cases.
• To utilize readily available phone call data for enhanced predictive accuracy.
• To provide probabilistic forecasts for improved risk management by decision-makers.

Main Methods:

• A Multiple Linear Regression (MLR) model was developed and optimized.
• Phone call data was integrated as a key predictor variable.
• The model's performance was rigorously evaluated against benchmark models using multiple error and accuracy metrics.

Main Results:

• The proposed MLR model, incorporating phone call data, significantly outperformed ARIMA, ETS, Seasonal Naive, Prophet, and a non-call data regression model.
• The model demonstrated superior accuracy in point forecasts and probabilistic forecasting.
• The forecasting exercise yielded a simple, interpretable, and reliable model.

Conclusions:

• The developed MLR model offers a valuable tool for local decision-makers needing to forecast COVID-19 cases and manage strained health resources.
• Probabilistic forecasts enhance the ability to manage risk associated with epidemic uncertainty.
• This model can serve as a foundation for future forecasting efforts, improving pandemic response and preparedness.