Influenza forecasting in human populations: a scoping review

Area of Science:

• Epidemiology
• Public Health
• Biostatistics

Background:

• Influenza activity forecasts are vital for public health preparedness and response.
• Methodological approaches to influenza forecasting require characterization to identify research gaps.

Purpose of the Study:

• To conduct a scoping review of influenza forecasting methodologies.
• To identify research gaps and areas for improvement in influenza prediction models.

Main Methods:

• A systematic literature search was performed using PRISMA methodology across multiple databases (PubMed, CINAHL, Project Euclid, Cochrane).
• Included studies published since January 1, 2000, focused on influenza forecasting validated against independent data and using surveillance data.
• 35 publications were included, analyzing population-based, medical facility-based, and pandemic spread forecasts across different global regions.

Main Results:

• Forecasting models included statistical (N=18) and epidemiological (N=17) approaches, utilizing diverse data sources like virological, syndromic, and internet search queries.
• Five studies employed data assimilation methods for real-time forecast updates.
• Significant variation was observed in forecasting outcomes and validation metrics, with limited direct comparisons of modeling approaches, calibration assessments, or systematic incorporation of expert input.

Conclusions:

• There is a need for enhanced influenza forecasting practices, including sensitivity analysis, model calibration, and direct comparisons of diverse modeling approaches.
• Integration of expert input and operational research in real-world applications are recommended.
• Improved communication and mutual understanding between modelers and public health officials are essential for advancing influenza forecasting capabilities.