Monitoring and forecasting the SARS-Covid-19 pandemic in France (preprint)

Over the past year, many countries have resorted multiple times to drastic social restrictions to prevent saturation of their health care system, and to regain control over an otherwise exponentially increasing SARS-Covid-19 pandemic evolution. With the advent of data-sharing, computational approaches have gained a key role in evaluating future scenarios and offering predictions in a constantly evolving social environment. To design optimal social, hospitalization and economical strategies that guarantee control over the pandemic progression, we developed a data-driven modelling framework with the aim to provide reliable near future predictions under constantly evolving social and pandemic events. The framework is flexible enough to be used at a single hospital, regional or national level. We used a variety of data such as social, serological, testing and clinical data to compute the infection dynamics and the hospital workload for France. We developed inference methods to calibrate model parameters from observed hospitalization statistics over adjustable time periods. We applied our model to study the age stratified pandemic evolution inside and outside hospitals until February 2021, and the competition between vaccinations and the novel delta variant. We obtained several predictions about hidden pandemic properties such as fractions of infected, infection hospitality and infection fatality ratios. We show that reproduction numbers and herd immunity levels are not universal but strongly depend on the underlying social dynamics. We find that with normal social interactions the present vaccination status and rate is not sufficient to prevent a new pandemic wave driven by the delta variant.

Monitoring and predicting SARS-CoV-2 epidemic in France after deconfinement using a multiscale age-stratified rate model (preprint)

The entire world and France were strongly impacted by the SARS-COV-2 epidemic. Finding appropriate measures that effectively contain the spread of the epidemic without putting a too severe pressure on social and economic life is major challenge for modern predictive approaches. To assess the impact of confinement (March 17th till May 11th) and deconfinement, we develop a novel rate model to monitor and predict the spread of the epidemic and its impact on the health care system. The model accounts for age-dependent interactions between population groups and predicts consequences for various infection categories such as number of infected, hospitalized, load of intensive care units (ICU), number of death, recovered and more. We use online health care data for the five most infected regions of France to calibrate the model. At day of deconfinement (May 11th), we find that 13% (around 4.8M) of the population is infected in the five most affected regions of France (extrapolating to 5.8M for France). The model predicts that if the reproduction rate R0 is reduced by at least a factor of 2.5-3 for all age groups, which could be achieved by wearing masks and social distancing, a significant second peak could be prevented. However, if the reduction in R0 for the age group 0-25 would be less and below 2 (school openings), a second peak is unavoidable in which case the ICU will be saturated. In that context testing should be focused on children, but it will nevertheless have a limited impact on reducing the spread.