Forecasting the final disease size: comparing calibrations of Bertalanffy-Pütter models.

Area of Science:

• Epidemiology
• Mathematical Modeling
• Infectious Disease Dynamics

Background:

• The 2013-2016 Ebola outbreak in West Africa highlighted the need for accurate epidemic forecasting.
• Standard least-squares (SSE) and weighted least-squares (SWSE) are common data-fitting methods.
• Accurate prediction of final disease size is crucial for public health response.

Purpose of the Study:

• To compare the performance of SSE and SWSE calibrations for fitting growth models to epidemic data.
• To evaluate the accuracy of disease size forecasts generated by these calibrations.
• To assess the reliability of confidence intervals provided by each method.

Main Methods:

• Utilized monthly data from the 2013-2016 West Africa Ebola outbreak.
• Applied Bertalanffy-Pütter growth models fitted to truncated initial epidemic data.
• Compared SSE and SWSE calibrations, with SWSE weights reciprocal to new infections.
• Analyzed forecast errors and fit residual distributions.

Main Results:

• SWSE calibration resulted in smaller forecast errors for final disease size compared to SSE.
• Forecasts using 16 or more months of data achieved relative errors below 10%.
• SWSE provided reliable upper and lower bounds for forecasts, while SSE systematically underestimated the final size.
• SWSE fit residuals did not violate the normal distribution hypothesis, unlike SSE.

Conclusions:

• Weighted least-squares (SWSE) is recommended over standard least-squares (SSE) for epidemic forecasting.
• SWSE offers improved accuracy and more reliable uncertainty quantification in disease size predictions.
• This method can enhance preparedness and response strategies for future outbreaks.