Estimation of risk factors for COVID-19 mortality - preliminary results

Francisco Caramelo; Nuno Ferreira; Barbara Oliveiros

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.24.20027268 doi: medRxiv preprint A similar procedure is performed to attribute one comorbidity to each subject of the dataset: a random number with five possible outcomes (four for comorbidities and one for no disease) is drawn and the corresponding probability is given by an adequate distribution ( Figure 2 ). A CSV (Comma-Separated Values) file containing the comorbidity distribution by age and gender feeds the routine. Comorbidities' empirical prevalence can be computed from China CDC report for all subjects, but does not allow different values for each age interval and gender and, for that reason, we considered that prevalence values for each comorbidity were constant across age groups and gender. Despite this limitation, tests can be run using worldwide prevalence for each disease. Finally, comorbidities classes were assumed to be exclusive, i.e., no subject could have more than one comorbidity.