COVID-19: Forecasting confirmed cases and deaths with a simple time series model

Area of Science:

• Epidemiology
• Statistical modeling
• Public health

Background:

• Accurate outbreak forecasting is vital for policy decisions, but predicting pandemic severity is challenging.
• Current epidemiological models have limitations in long-term outbreak trajectory prediction.
• Interventions like lockdowns have significant socioeconomic consequences and inherent uncertainties.

Purpose of the Study:

• To develop a statistical, time series approach for short-term COVID-19 outbreak prediction.
• To model and forecast global confirmed cases and deaths, including uncertainty.
• To provide practically relevant forecast accuracy and uncertainty estimates.

Main Methods:

• Utilized a statistical time series model with a multiplicative trend assumption.
• Focused on short-term (10-day-ahead) forecasting of COVID-19 confirmed cases and deaths.
• Evaluated forecast performance over a four-month period.

Main Results:

• The proposed model demonstrated competitive forecast accuracy for COVID-19 cases and deaths.
• The model provided useful and practically relevant estimates of forecast uncertainty.
• Short-term forecasting of outbreak behavior was successfully achieved.

Conclusions:

• A simple statistical time series model can effectively forecast short-term COVID-19 trends.
• The model's uncertainty estimates aid in decision-making for outbreak management.
• This approach complements traditional epidemiological models for timely public health insights.