Analyzing the European countries’ SARS-CoV-2 policies via Bayesian explainable deep learning and statistical inference

Abstract

Even when the SARS-CoV-2 pandemic recedes, evidence-based researches regarding the effectiveness of pharmaceutical and non-pharmaceutical government interventions (NPIs) remain important. In this study, SARS-CoV-2 data of 30 European countries from early 2020 up to mid-2022 are analyzed using Bayesian deep learning and statistical analysis. Four data sources containing each country’s daily NPIs (consisting of 66 government measures, virus variant distributions of 31 virus types, the vaccinated population percentages by the first five doses as well as the reported daily infections in each country) are concatenated to undertake a comprehensive assessment of the impact of SARS-CoV-2 influential factors on the spread of the virus. First, a Bayesian deep learning model is constructed with a set of input factors to predict the growth rate of the virus one month ahead of the time from each day. Based on the trained model, the importance and the marginal effect of each relevant influencing input factor on the predicted outcome of the neural network model is computed by applying the relevant explainable machine learning algorithms. Subsequently, in order to look at the problem from a different perspective and re-examine the influencing input factors inferred from the deep learning model, a Bayesian statistical inference analysis is performed within each country’s data. In the statistical analysis, for each influencing input factor, the distribution of pandemic growth rates, in the days where the selected explanatory factor has been active, is compared with the distribution of the pandemic growth rates, in the days where the selected explanatory variable has not been active. The results of the statistical inference approve the predictions of the deep learning model to a significant extent. Similar conclusions from the SARS-CoV-2 experiences of the thirty studied European countries have been drawn.