Dynamic causal modeling of the COVID-19 pandemic in northern Italy predicts possible scenarios for the second wave

The COVID-19 pandemic has sparked an intense debate about the factors underlying the dynamics of the outbreak. Mitigating virus spread could benefit from reliable predictive models that inform effective social and healthcare strategies. Crucially, the predictive validity of these models depends upon incorporating behavioral and social responses to infection that underwrite ongoing social and healthcare strategies. Formally, the problem at hand is not unlike the one faced in neuroscience when modelling brain dynamics in terms of the activity of a neural network: the recent COVID19 pandemic develops in epicenters (e.g. cities or regions) and diffuses through transmission channels (e.g., population fluxes). Indeed, the analytic framework known as "Dynamic Causal Modeling" (DCM) has recently been applied to the COVID-19 pandemic, shedding new light on the mechanisms and latent factors driving its evolution. The DCM approach rests on a time-series generative model that provides - through Bayesian model inversion and inference - estimates of the factors underlying the progression of the pandemic. We have applied DCM to data from northern Italian regions, which were the first areas in Europe to contend with the COVID-19 outbreak. We used official data on the number of daily confirmed cases, recovered cases, deaths and performed tests. The model - parameterized using data from the first months of the pandemic phase - was able to accurately predict its subsequent evolution (including social mobility, as assessed through GPS monitoring, and seroprevalence, as assessed through serologic testing) and revealed the potential factors underlying regional heterogeneity. Importantly, the model predicts that a second wave could arise due to a loss of effective immunity after about 7 months. This second wave was predicted to be substantially worse if outbreaks are not promptly isolated and contained. In short, dynamic causal modelling appears to be a reliable tool to shape and predict the spread of the COVID-19, and to identify the containment and control strategies that could efficiently counteract its second wave, until effective vaccines become available.