84 Chapter 4 Inverse probability weighting To correct for the observed differences between the tested and the untested populations (supplementary table 3), an inverse probability weighting analysis was performed. This weight was based on the probability to have undergone additional testing. Standardised mean differences before and after weighting for the tested and not tested group can be found in supplementary figure 1. The multivariable regression analysis with IMV as the outcome variable was repeated with the same predictor variables as in the unweighted analysis (Table 2). The OR for IMV in influenza co-infection was 4.32 (95% CI 2.06-9.04, p=<0.0001). Similarly, the multivariable regression analysis with in-hospital mortality as the outcome variable was repeated. The OR for in-hospital mortality in influenza co-infected patients was 2.49 (95% CI 1.11-5.56, p value 0.025) (Table 3). OR for receiving IMV (95% CI) p value SARS-CoV-2 mono-infection 1 Adenovirus co-infection 0.61 (0.17-1.62) 0.377 Influenza virus co-infection 4.32 (2.06-9.04) <0.0001 RSV co-infection 0.76 (0.14-2.61) 0.697 Table 2: Weighted multivariable logistic regression analysis with IMV as the outcome variable OR for in-hospital mortality (95% CI) p value SARS-CoV-2 mono-infection 1 Adenovirus co-infection 1.49 (0.66-3.24) 0.324 Influenza virus co-infection 2.49 (1.11-5.56) 0.025 RSV co-infection 0.60 (0.12-2.06) 0.459 Table 3: Weighted multivariable logistic regression analysis with in-hospital mortality as the outcome variable Discussion In this study of hospitalised people with COVID-19, we demonstrated that influenza co-infection was associated with increased odds of receiving IMV, and both adenovirus and influenza co-infection were associated with increased in-hospital mortality. The number of viral co-infections in COVID-19 differs widely between studies, and
RkJQdWJsaXNoZXIy MTk4NDMw