Statistical analysis
We compared demographic characteristics between asthmatic and
non-asthmatic inpatients and between hospitalized and non-hospitalized
asthmatic patients using the Kruskal–Wallis test for continuous
variables and Chi-square test for categorical variables. Binary logistic
regression was used to determine the association between asthma and risk
of hospitalizations among SARS-CoV-2 positive patients, and ordinal
logistic regression was used to determine the association between asthma
and COVID-19 disease severity among hospitalized patients. To quantitate
the contributions of other known confounding factors to COVID-19
hospitalizations and COVID-19 disease severity, we used three models in
each analysis: the first model (unadjusted) univariately analyzed the
effects of asthma, the second model used a multivariate analysis
adjusted for demographics (age, BMI, ethnicity), and the third model
used a multivariate analysis adjusted for both demographics and known
COVID-19 related comorbidities associated with more severe disease
(diabetes, coronary heart disease, and
hypertension).18,19 Adjusted odds ratios (ORs) and
95% confidence intervals (CIs) were calculated for each model. We
evaluated the relationship between laboratory values at admission and
COVID-19 disease severity using ordinal regression models as an
exploratory analysis. The interaction between each laboratory value and
asthma status was also analyzed in the model. Bonferroni correction for
multiple comparisons was applied for laboratory values and false
discovery rate (FDR) was controlled at the 10% level. Absolute
eosinophil counts from hospitalization were grouped by collection time
into admission, during and discharge. Admission eosinophil counts were
collected within three days prior to admission, discharge eosinophil
counts were collected on the day of discharge, and during
hospitalization eosinophil counts were collected between admission and
discharge counts. The median value was chosen to use as the per-person
count if a patient had multiple values at any of the three timepoints. A
mixed effect model was constructed to assess change in eosinophil counts
during hospitalization in those patients that did not receive systemic
steroids. In the sub-cohort analysis among those enrolled in a
prospective, longitudinal study (NCT# 04373148), the time from symptom
onset to resolution was compared between asthmatic and non-asthmatic
patients using the Kaplan-Meier curve and log-rank test. All analyses
were performed in R language (version 4.0.3) and packages stats (version
4.0.3), ordinal (version 2019.12.10).