A statistical model for forecasting probabilistic epidemic bands for dengue cases in Brazil

Area of Science:

• Epidemiology
• Public Health
• Mathematical Modeling

Background:

• Dengue is a significant vector-borne disease in Brazil, posing a major public health challenge.
• The rising burden of dengue necessitates advanced modeling for disease preparedness and response.
• The Brazilian Ministry of Health sought modeling solutions to manage dengue outbreaks.

Purpose of the Study:

• To develop a Bayesian forecasting model to predict dengue cases 52 weeks in advance for all Brazilian health districts.
• To create probabilistic epidemic bands for effective dengue outbreak monitoring and assessment.
• To classify epidemic severity and deviation from historical patterns.

Main Methods:

• A Bayesian forecasting model was developed using historical dengue case data.
• The model predicts weekly cases for 118 health districts up to 52 weeks ahead.
• Probabilistic epidemic bands were defined using percentiles and interpreted based on historical occurrence.

Main Results:

• The model accurately captured epidemic curve shapes in validation seasons (2022-2023, 2023-2024).
• The 2022-2023 season was classified as "fairly high, atypical" with 1,436,034 cases.
• The 2023-2024 season was "exceptionally high, very atypical" with 6,454,020 cases, exceeding the 90% percentile.
• The 2024-2025 forecast estimates a median of 1,526,523 cases.

Conclusions:

• Probabilistic epidemic bands are effective tools for monitoring dengue outbreaks.
• The model facilitates prospective comparisons of current outbreaks against historical patterns.
• This approach aids in assessing the severity and atypicality of ongoing dengue epidemics.