Forecasting the future number of pertussis cases using data from Google Trends

Area of Science:

• Epidemiology
• Public Health Surveillance
• Computational Biology

Background:

• Re-emerging infectious diseases like pertussis require advanced monitoring and forecasting methods.
• Traditional epidemiological models rely on reported case numbers, which may have limitations in timeliness and completeness.

Purpose of the Study:

• To assess if Internet search data can complement traditional pertussis modeling.
• To evaluate the added value of Google Trends data (GTD) in forecasting pertussis cases.

Main Methods:

• SARIMA models were developed using reported weekly pertussis cases in Germany over four years.
• Pertussis-related Google Trends data (GTD) was incorporated as an external regressor.
• Model predictions with and without GTD were compared against validation data for one-year and two-week periods.

Main Results:

• The model incorporating GTD showed improved forecasting accuracy, with lower RMSE, MAPE, and MAE compared to the traditional model.
• The GTD-expanded model achieved a lower Corrected Akaike Information Criteria value.
• Statistical tests confirmed a significant improvement in predictive performance when using GTD for the one-year forecast period.

Conclusions:

• Internet-based surveillance data, such as GTD, demonstrably enhances the predictive capabilities of traditional disease models.
• Integrating digital data sources is a valuable strategy for improving the accuracy and timeliness of infectious disease forecasting.
• These findings support the consideration of internet-based data in future public health surveillance and disease modeling efforts.