2.3.3. Comparison of the Two Groups

To compare the two groups, Wilcoxon's rank sum test and chi-square test were used for continuous variables and categorical variables, respectively.

#### 2.3.4. Multivariate Analysis

Logistic regression analysis was performed to assess the relationship between outcome and explanatory variables after adjustment for age, sex, and body mass index (BMI). To minimize the effects of outliers, a robust method was applied for the following regression tests.

The proportion of patients diagnosed as SARS-CoV-2 infection was very low. Thus, this study also used inverse probability weighting (IPW) to estimate the average treatment effect (ATE) of each item on SARS-CoV-2 infection and on the risk of infection symptom occurrence. In IPW, a propensity score is used to weigh each observed value in the sample. Two types of expected values are then estimated: the expected value of the outcome if the treatment is used for the overall sample (in this analysis, if individuals had travelled) and the expected value of the outcome if the treatment is not used. The ATE is estimated from the difference between these values.

Specifically, the inverse of the estimated propensity score (1/*∂*) is used for weighting. The inverse of a propensity score increases as the propensity score decreases. Therefore, a smaller weight is given to an observed value with a larger propensity score in the treated group, and a larger weight is given to an observed value with a larger propensity score in the control group. In other words, calculation is done with more weighting for an observed value that is rarer or accounts for a smaller proportion of the sample for each of the treated group and control group.

The statistical analyses were carried out using Stata/SE 16.0 (StataCorp LLC, College Station, TX, USA).
