The proportion of deaths cases in confirmed patients of COVID-19 is still increasing for cumulative cases reported up to 14 April 2020

Thomas Scheuerl

The copyright holder for this preprint (which was not peer-reviewed) is . https://doi.org/10.1101/2020.04.17.20068908 doi: medRxiv preprint Data were analysed in R statistical environment [4] . A time-series analysis was conducted to track changes over time using generalized additive mixed effects models (GAMMs) following ref. [5] . This approach smoothes changes over time tracking non-linear trends. Change in proportion and count data were modelled using `Days` as continuous variable building a global model for all countires applying a smooth term with family either binomial for proportion or poisson for count data. The change over time per country was introduced into the random effect. The models were build using the R packages nlme [6] and mgcv [7] . The change over time was smoothed using 10 knots and a cubic regression spline. Model convergence was controlled using the lmeControl function, with the maximum number of iterations set to 500 and using optimisation opt=optim function. Temporal autocorrelation was introduced to account for violation of independence.