Impact of COVID-19 Outbreak on Non-COVID-19 Patients

The analysis of the various results will be carried out on all HUGs and on the various hospitals / clinics in the canton. A survival analysis for the outcome of death or rehospitalization will be performed, with a comparison according to each period. After epidemy evolution, finally, the outcomes will be compared between periods pre-COVID (from 01 march 2019 to 28 february 2020) versus per-COVID (01 march 2020 to 28 february 2022), and versus post-COVID (01 march 2022 to 28 february 2023). And comparaison would be performed between periods during the wave (per-wave) versus periods inter-wave. A description will be made in number (%) for numerical data and in median (IQR) for quantitative data. Univariate comparisons between the different periods will be carried out by statistical tests, parametric or not, adapted according to the data (Chi2 or Fisher's test for qualitative data, Student's test or Mann-Whitney-Wilcoxon for quantitative data). Statistical significance will be retained in the event of p \<0.05. Multivariate analysis will be performed by logistic regression for the main outcome and by cox model for survival analysis. Different variables will be included in the models, including data on gender, age and comorbidity, as well as any variable having a difference with p \<0.2 in univariate analysis. Secondary analyzes will be carried out by pathology (as the main diagnosis) according to the specific results defined for each situation. In retrospective analysis, these specific data will be relatively limited on the HUG area of full analysis brings together around total of 240,000 hospital stays. The main outcome data will be complete with no missing data. On the other hand, since this is retrospective data, it is possible that some important variables are missing. In this case, other patient data with missing data will not be included in the multivariate analyzes. In the event of missing data greater than 10%, a second sensitivity analysis may be performed after replacing the missing data with a multiple imputation method.