Methods
We built a network of six academic hospitals in California (Stanford University)/USA, Frankfurt (Goethe University), Giessen (Justus Liebig University) and Marburg (Philipps University)/Germany, and Moscow (Clinical Hospital 52 in collaboration with Sechenov University)/Russia. The German and US participating centers collected clinical and laboratory data for all patients hospitalized due to COVID-19 since the beginning of the pandemic until end of 2020 and September 2020 respectively. Moscow delivered case-control type of data and included patients hospitalized during 23.03.-16.05.2020. Reported comorbidities reported in the present study were based on the 2020 International Classification of Diseases-10th Revision codes as described in Supplementary Table 1. Laboratory values were calculated and expressed in the same units for direct comparison whenever applicable.
Statistical Analysis
All statistical analyses were conducted in R (R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/; version 4.1.2).
The prevalence of asthma patients among the hospitalized COVID-19 patients was calculated from the number of hospitalized asthma patients divided by the reported number of total hospitalized patients. While Germany and Stanford data correspond to a cohort of all hospitalized COVID-19 patients at these hospitals in the indicated time interval, Moscow data corresponds to all asthma cases and a set of control cases without asthma. The prevalence in Moscow was calculated based on all hospitalized COVID-19 cases in the Moscow hospitals (4,549). To compare the prevalence of asthma to the corresponding prevalences in the general population, a binomial test was used. The 95% confidence interval was calculated by the method of Clopper-Pearson .
To test whether asthma patients were over-represented among the ICU admitted patients we used Fisher’s exact test and show the estimated odds ratios and 95% confidence intervals between Odds (ICU|asthma) and Odds (ICU|no asthma) reported by the R-function fisher test.
To test whether any of the additional preconditions are over-represented in the asthma compared to the non-asthma patients we used Fisher’s exact test. We corrected for multiple testing using the method of Benjamin-Hochberg and report significant differences at a false discovery rate of 10%.
To test whether the number of additional preconditions is different between patients with asthma and without asthma we used a Wilcoxon rank sum test. To adjust for confounders explaining the overrepresentation of asthmatics in Stanford, we performed logistic regression using the center, age group, sex and the 11 comorbidities to predict whether a patient was asthmatic. This analysis was restricted to the centers Moscow and Stanford, for which we had microdata available. From the so-fitted model, we calculated the Odds to be asthmatic given that the patient was from Stanford, female and had no comorbidity over the different age groups.
To identify possible confounders for the laboratory measurements a linear regression model was fitted with the predictors age group, sex, preconditions (including asthma). These analyses revealed a strong and significant effect of the precondition COPD and the eosinophil count at admission, during the hospital stay, and at discharge. Thus, we removed patients that had a precondition of COPD and recalculated averages and standard errors.