Forecasting efforts from prior epidemics and COVID-19 predictions

Area of Science:

• Epidemiology
• Public Health Modeling
• Infectious Disease Dynamics

Background:

• The COVID-19 pandemic highlighted the proliferation and societal impact of disease prediction models.
• Previous epidemics like SARS, H1N1, and Ebola also saw the development of such forecasting tools.
• Disease modeling research was less established during SARS and H1N1 compared to the Ebola epidemic.

Purpose of the Study:

• To evaluate the accuracy of disease prediction models used during past 21st-century epidemics.
• To assess the reliability of forecasts for epidemics preceding COVID-19.
• To inform the communication of uncertainty in epidemic forecasting.

Main Methods:

• Systematic review of published disease prediction models and their forecasts.
• Comparative analysis of forecast accuracy across different epidemics (SARS, H1N1, Ebola).
• Qualitative assessment of forecast trends, particularly regarding over-prediction.

Main Results:

• Disease prediction models were more prevalent during the Ebola epidemic than during SARS or H1N1.
• Forecasts of mortality for Ebola demonstrated significant deviations from actual outcomes.
• A consistent tendency for over-prediction of deaths was observed in Ebola forecasts.

Conclusions:

• Epidemic forecasting models, particularly for mortality, have historically shown limitations in accuracy.
• The tendency to over-predict outcomes in past epidemics necessitates cautious interpretation of current models.
• Effective communication of uncertainty in disease prediction models is essential for informed public health decision-making and societal response.