Real-time parameter estimation of Zika outbreaks using model averaging.

Area of Science:

• Epidemiology
• Mathematical Modeling
• Public Health

Background:

• Accurate prediction of epidemic final size is crucial for public health response planning.
• The Richards model is commonly used for arboviral disease outbreaks, but other non-linear models may also fit data.
• Traditional methods often ignore model uncertainty by selecting a single best-fitting model.

Purpose of the Study:

• To estimate the final size and turning point of epidemics using model averaging.
• To conduct real-time predictions of epidemic final size by incorporating uncertainty from multiple models.
• To apply and evaluate a model averaging approach for Zika outbreak data.

Main Methods:

• Utilized several non-linear models to estimate epidemic parameters.
• Employed model averaging to combine predictions from multiple models, accounting for model uncertainty.
• Applied the methodology to Zika outbreak data from four Colombian cities (2015-2016).

Main Results:

• Model averaging provided more robust estimates for epidemic final size and turning points compared to single-model approaches.
• Real-time predictions of the final Zika outbreak size were generated.
• The approach demonstrated the utility of considering multiple models in epidemiological forecasting.

Conclusions:

• Model averaging is a valuable technique for improving epidemic forecasting and understanding model uncertainty.
• The proposed method enhances the ability of health authorities to plan responses to outbreaks like Zika.
• This approach offers a more reliable framework for real-time epidemic size prediction.