Forecasting influenza epidemics in China using transmission dynamic model with absolute humidity

Area of Science:

• Epidemiology
• Environmental Health
• Mathematical Modeling

Background:

• Low temperatures traditionally linked to influenza epidemics do not explain tropical/subtropical summer peaks.
• Absolute humidity (AH) exhibits a U-shaped relationship with influenza transmission across diverse climates.
• A unified forecasting system is needed for China's varied climatic conditions.

Purpose of the Study:

• To develop and evaluate a unified influenza forecasting system for China.
• To assess the system's performance across different climate zones (northern and southern China).
• To determine the system's accuracy in predicting epidemic timing and magnitude.

Main Methods:

• Weekly influenza forecasts generated using an AH-driven susceptible-infected-recovered-susceptible (SIRS) model.
• Ensemble adjustment Kalman filter (EAKF) integrated for forecasting.
• Model performance evaluated by sensitivity, specificity, and prediction accuracy for peak timing and magnitude.

Main Results:

• The forecasting system demonstrated high predictive capability (mean sensitivity >87.5%; mean specificity >80%).
• Forecast accuracy for peak timing and magnitude was higher in northern China (82% and 60%) compared to southern China (42% and 20%) at 3-6 weeks ahead.
• Accuracy improved with forecasts made closer to the actual epidemic peak.

Conclusions:

• The developed AH-driven forecasting system reliably predicts seasonal influenza epidemics in China.
• The system's performance varies regionally, highlighting the influence of climate factors.
• This approach offers a valuable tool for influenza preparedness nationwide.