A human judgment approach to epidemiological forecasting.

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• Infectious diseases continue to pose a significant societal burden despite medical advancements.
• Effective epidemic preparedness relies on accurate forecasting, which has historically been challenging due to data uncertainty and complex disease dynamics.

Purpose of the Study:

• To evaluate the accuracy of collective human judgment in predicting epidemic trajectories.
• To compare the performance of human-driven forecasts against data-driven computational models.

Main Methods:

• Development of the web-based "Epicast" forecasting system to collect and aggregate real-time predictions from human participants.
• Assessment of human prediction accuracy for influenza and chikungunya using various metrics.
• Comparison of aggregated human forecasts with statistical forecasting systems for U.S. flu seasons (2014-2015 and 2015-2016).

Main Results:

• Human participants demonstrated measurable accuracy in predicting influenza and chikungunya trajectories.
• Combined real-time human predictions frequently surpassed the accuracy of several statistical forecasting systems, particularly for short-term predictions.
• Collective human judgment offers a viable alternative for epidemiological forecasting.

Conclusions:

• Collective human judgment possesses significant predictive power for epidemic forecasting.
• The Epicast system highlights the potential of integrating human insights into public health surveillance and response strategies.
• Further exploration of the benefits and limitations of human-based forecasting is warranted.