80 Chapter 4 Statistical analysis Mean and standard deviations (SD) were used to describe continuous normally distributed variables. Significance testing was done using the Welch two sample t-test for 2 groups (testing yes/no) and a one-way anova for viral co-infections. Median and interquartile range (IQR) were used to describe continuous non-normally distributed values. Significance testing was done using the Mann Whitney-U test for 2 groups and the Kruskal Wallis test for comparing the viral co-infections. Categorical variables were described as a frequency and percentage, and comparisons were done using a chi-squared test. A regression analysis was performed to analyse the effect of viral co-infection independent of other variables. Given the change in hospital mortality between the first and second wave of the COVID-19 pandemic21, we created a time trend variable that captures when in the pandemic a patient was admitted to the hospital, using the day of hospital admission. For both mortality and need of IMV, a separate univariable regression analysis was performed. The confounders used were age, sex, number of comorbidities, immunocompromised status, 4C Mortality score, presence of a viral respiratory co-infection, systemic treatment with corticosteroids during admission, including dexamethasone, hydrocortisone, prednisolone, prednisone and methylprednisolone (data on dosing regimen were not available) and the time trend variable. All clinically relevant variables and variables significantly associated with the outcome variable (IMV or mortality) were included in a multivariable regression analysis, to predict the odds of receiving IMV and mortality. IPW involved determining the probability of co-infection testing for each patient using a logistic regression model built using known confounders. The inverse of these probabilities was used in weighted analyses. Balance of patient characteristics after weighting was checked graphically using the standardised mean difference. A p value of 0.05 or less was considered to indicate statistical significance. Statistical analysis was performed using R Statistical Software (version 4.0.5). Results 212,466 participants met our inclusion criteria and were included in the analysis, with an outcome recorded on or before 8th December, 2021 (Figure 1). 17,011 (8%) underwent testing for additional respiratory viruses (influenza virus, adenovirus or RSV) in addition to SARS-CoV-2. Of these patients undergoing additional testing, 6965 had a documented result, including 583 who had a confirmed respiratory virus co-infection. Influenza virus was the co-infecting pathogen in 227/583 cases (39%), RSV in 220/583 (38%) and adenovirus in 136/583 (23%).
RkJQdWJsaXNoZXIy MTk4NDMw