This research evaluates the effectiveness of county-specific Influenza-Like Illness (ILI) forecasting models. The following hypothesis is assessed: The model based on historical incidence, emergency department (ED) visits, local air/meteorological quality, and Google search trends will be significantly more accurate than the baseline model.
Three ILI-forecasting models specific to Kent County are created, using past cases incidence, ED visits, local air/meteorological quality, and Google search trends, using the following forecasting methods: seasonal integrated moving average modeling with exogenous regressors (SARIMAX), elastic net modeling, and random forest modeling. Their predictive utilities for ILI were assessed against a baseline model.
The elastic net and random forest models had fewer forecasting errors than the baseline model, with the random forest model emerging as most effective. Conversely, the SARIMAX model had the highest error of the four models.
This research aligns with the CDC Center for Forecasting and Outbreak Analytics 2023-2028 Strategic plan, highlighting promising potential for infectious disease modeling in local public health. The findings offer valuable methodological insights and lessons learned for capacity-building initiatives moving forward. Ultimately, this research establishes a strong argument for increased investment in disease forecasting and analytics capacity in local public health.
Disease forecasting, local public health, analytic methodology