Laura Guzman-Rincon: Statistical framework for the nowcasting and forecasting of infectious disease growth rates

Abstract: The quantitative assessment of the recent state of an epidemic is a key public health tool that has substantial policy impacts, helping to determine if existing control measures are sufficient or need to be strengthened. Varying data quality, reporting biases and differences in disease characteristics all complicate our understanding of the current state of the epidemic and our ability to make short-term projections.

To tackle this problem, we propose a statistical framework to estimate and forecast the growth rates of any outbreak of an infectious disease spreading in a population. The statistical approach incorporates a Gaussian Process within a Bayesian hierarchical model to approximate the underlying dynamics of the data and produce short-term forecasts. This method has been applied to the SARS-CoV-2 pandemic in both the UK and Kenya and was used to routinely analyse spatial dynamics of SARS-CoV-2 during 2020 and 2021.

The proposed framework can be beneficial when utilised at the onset and throughout an epidemic. It is fast and applicable to various diseases and data sources. For instance, it can monitor the reported number of infections caused by specific variants or strains. This approach can be especially valuable during the initial stages of an epidemic, allowing for quantification of the epidemic’s state while more complex mechanistic and disease-specific models are still in development.