Speaker
Description
Throughout the pandemic, countries and scientists rushed to predict the impact of sequentially emerging variants by, in part, trying to predict the size of the wave likely to hit their local community. Our study aims to predict the amplitude of forthcoming COVID-19 waves using statistical modelling and a unique global dataset. Utilizing publicly available datasets, we compiled relevant features from demographics, geographic distribution, mobility patterns, epidemiological data, and immunity levels, which influence the spread and severity of COVID-19 waves. Moreover, through collaboration with the Public Health Agency of Canada, we developed viral genomic evaluation metrics, which allow for a nuanced understanding of the dynamic landscape of COVID-19 variants. To train our model, we used the amplitude of consecutive variant waves, specifically Delta, BA.1, and BA.2 waves. For feature selection, we employed elastic net regression, given the model’s effectiveness in handling diverse and multi-scale data types, with an emphasis on predictive accuracy. We implemented a leave-one-out cross-validation strategy that respects the time series nature of the data from each country. We then applied a random forest model using the same cross-validation strategy. Our model demonstrates robust performance in several scenarios, accurately predicting the relative size of forthcoming waves. Preliminary results underscore the critical role of ongoing epidemiological and genomic surveillance in enhancing our model's predictive capabilities. By integrating real-time data, we are able to refine our forecasts and provide insights crucial for public health planning and response strategies. This work complements and will be referenced in the presentation "Real-time tracking of variant evolution and dynamics for public health decision-making: Progress and challenges," as part of a broader research program focused on the urgent public health need to accurately monitor and forecast wave dynamics of rapidly evolving viruses.
