Short-Term Forecasting of Respiratory Virus Transmission and Severe Disease Risk Among Pediatric and Adult Populations for Public Health Planning in Ontario

Background: Influenza A presents a major public health issue in Ontario, affecting the population with regular seasonal outbreaks that lead to health and economic burdens. This study uses a collection of historical data on Influenza A cases which dates back to 2015. Objectives: Utilizing a data-driven approach, this study aims to enhance short-term forecasting of Influenza A cases. By predicting the number of cases over a 14-day horizon, we intend to provide actionable insights for public health officials to implement timely and effective interventions. The model seeks to differentiate impacts on pediatric and adult populations, thereby optimizing resource allocation and intervention strategies tailored to demographic-specific needs. Methods: Utilizing the R package tscount, tailored to count time series data, we modeled the number of Influenza A cases. This package supports generalized linear models, which are pivotal for predicting time-dependent phenomena and allow for the inclusion of past values and covariates into the models. The analysis was segmented by age group to refine predictions. Results: The model accurately forecasted 14-days of Influenza A cases for both pediatric and adult populations. Predictive performance was assessed using the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), which upheld the model's efficacy in short-term forecasting within acceptable error thresholds. Conclusions: The developed forecasting model proves effective for short-term predictions, providing insights into outbreak trends and severity. These predictions are crucial for optimizing resource distribution, informing public health advisories, and enhancing community preparedness.