*2.3. Statistical Analysis*

The patients' characteristics and clinical conditions were summarized. The mean (SD) was used for the continuous variable, and percentage was used for the categorical and binary variables for all patients. Non-normally distributed numerical variables were summarized as medians with interquartile range.

The data were compared between two groups of patients with symptomatic COVID-19 (hospitalized versus non-hospitalized) and based on the vital statuses of inpatients (dead versus alive). The chi-square test/Fisher exact test were used for categorical variables and a two-sample t-test was used for continuous variables to study the univariate effect of the variables on each outcome of interest. Stepwise logistic regression was preformatted to study the risk/association to hospitalization. Variables would retain in the model if there was a significant effect with *p*-value < 0.05 after adjusting for the other covariates with estimation of odds ratio (OR) and its 95% confident intervals (CI), where OR < 1 and 95% CI < = 1 indicates a proactive effect for hospitalization, while OR > 1 and 95% CI > 1 indicates a risk for hospitalization. A similar analysis was performed to study the risk/association to inpatient death.

The statistical analysis was performed with SPSS version 25.0 (IBM Corp, Armonk, NY, USA).
