Estimating the impact of non-pharmaceutical interventions on documented infections with COVID-19: A cross-country analysis

Nicolas Banholzer; Eva van Weenen; Bernhard Kratzwald; Arne Seeliger; Daniel Tschernutter; Pierluigi Bottrighi; Alberto Cenedese; Joan Puig Salles; Stefan Feuerriegel; Werner Vach

The introduced model links the number of the new cases to the number of the existing cases. The estimated coefficient for each NPI quantifies the relative reduction in new cases. We report posterior means and, in the figures, depict the posterior distribution that refers to the 80 % and 95 % credible intervals (CrI). The model takes into account day-of-the-week effects as the frequency of testing may depend on the day of the week and reporting may be delayed to a higher degree during weekends. The model allows the constant rate of new cases to vary across countries as this rate depends on the (unknown) age composition of the true cases, the population density, and other country-specific factors. We exclude the very early phase up to 100 documented cases as we have to expect that each country had to establish its documentation practice in the early phase (however, we vary the cases at start as part of the sensitivity analysis).