*3.2. Multivariate Logistic Regression Analysis*

Multivariate logistic regression analysis was conducted (Table 2, left column and Figure 1).

**Figure 1.** Odds ratio of COVID-19 infection by multivariate logistic regression for pre-existing conditions and lifestyle factors. \* *p* < 0.05.

**Table 2.** Comparisons of COVID-19 and control groups using two statistical methods. Multiple logistic regression (left column) and inverse-probability weighing method (IPW, right column) were conducted. IPW was controlled for age and coexistence with heart disease. Odds ratio and average treatment effects of being in the COVID-19 group are shown.


IPW: inverse probability weighting analysis, OR: odds ratio; CI: confidence interval; N.A.: not applicable, ATE: average treatment effect. \* Disease due to which you were prohibited by a doctor from exercising, or disease or injury due to which you have major difficulties in walking (e.g., rheumatoid arthritis and bone fracture).

> Age was negatively correlated with diagnosis (odds ratio (OR) 0.94 per year, 95% confidence interval (CI) 0.91–0.98, *p* < 0.01 in multiple regression), while coexistence of heart disease (OR 11.33, 95%CI 2.50 to 51.25, *p* < 0.01) and other conditions (OR 6.03, 95%CI 1.41 to 25.77, *p* = 0.02) were positively associated with SARS-CoV-2 infection. As for lifestyle

factors, washing hands (OR 0.10, 95%CI 0.02 to 0.56, *p* = 0.01) was negatively associated with infection. Interestingly, the diagnosis was significantly and positively correlated with changing clothes frequently (OR 2.96, 95%CI 1.08 to 8.15, *p* = 0.04), sanitizing belongings (OR 3.78, 95%CI 1.37 to 10.44, *p* = 0.01), and avoiding outings (OR 3.20, 95%CI 1.03 to 9.88, *p* = 0.04).

#### *3.3. Analysis Using Inverse Probability Weighting Method*

As sample size of the SARS-CoV-2 group was small, an IPW analysis was also conducted, controlling for background factors that showed significant differences in multiple regression, that is, age and coexistence of heart disease (Table 2, right column).

Habit of crowd avoidance (ATE −62.2, 95%CI −115.6 to −8.7, *p* = 0.02) and hand washing (ATE −84.8, 95%CI −155.8 to −13.7, *p* = 0.02) were negatively correlated with SARS-CoV-2 infection. In contrast, habits of changing clothes frequently (ATE 274.4, 95%CI 113.2 to 435.6, *p* < 0.01) and sanitizing their belongings (ATE 100.0, 95%CI 55.9 to 144.1, *p* < 0.01) were positively associated with the infection, which was consistent with the results of the logistic regression. In addition, no or rare remote work (ATE −77.1, 95%CI −143.4 to −0.2, *p* = 0.02) were negatively correlated with infection compared with almost daily remote work, which was contrary to the common thinking that remote working is effective in infection prevention.

#### *3.4. Sensitivity Analysis*

For sensitivity analysis, the same analysis in Table 2 was conducted among those who responded as being diagnosed as SARS-CoV-2 in the third survey (thus, their answers were not fully validated). In total, 110 were included in the SARS-CoV-2 group and 16,365 in the control group. In this analysis, habit of changing clothes frequently and sanitizing their belongings were consistently and positively associated with SARS-CoV-2 infection in IPW analysis (Table S2).
