*2.4. Statistical Analysis*

IBM SPSS 26.0 (Armonk, NY, USA) was used for data analyses. The diagnostic testing (e.g., outlier screening and distribution checking) was first conducted, and all data adhered to the normal distribution that the absolute values of skewness and kurtosis were <2. Descriptive statistics (e.g., mean, standard deviation, percentage) were used to describe baseline characteristics. The characteristics of depression were examined by independent Ttest and one-way analysis of variance (ANOVA). Hierarchical linear regression models were used to explore the association of precautionary behaviors with depression. To control the influence of past precautionary behaviors, residualized change scores (RCS; calculated by conducting linear regression between past behaviors and current behaviors) were used [36]. In Model 1, the demographic variables were set as predictors for the depression level. Subsequently, two covariates were added to the regression analysis in Model 2. Finally, the RCS of COVID-19 precautionary behaviors was included as a predictor in Model 3, controlled for the significant demographics and covariates. The role of the SES indicators in moderating the behavior–depression association was examined using IBM SPSS Process (Model 1), and the 95% confidence intervals (CIs) of the standardized effects were estimated using the bias-corrected bootstrap approach (5000 resample). The 5% level (two-tailed) was taken as the statistical significance cutoff point.

#### **3. Results**

#### *3.1. Sample Characteristics*

The descriptive characteristics of the sample are presented in Table 1. The data of 516 eligible older adults were included in the analysis. Most participants were females (57.9%) and were aged between 60 and 69 years (68.6%). The majority of the older adults were married (83.7%) and lived with their spouse, partners, or children (90.7%). In terms of the medical histories, about half of the participants have suffered from chronic diseases (e.g., heart diseases, diabetes, or cancer). For SES indicators, only a small percentage of participants were illiterate or semi-illiterate (8.7%), the majority of participants were pensioners/retired (92.6%), and more than half of the sample indicated an average level of household income (57.9%). In terms of BMI, a considerable proportion of elderly participants were overweight or obese (52.1%). In addition, most participants perceived their health status as good or excellent (52.7%), and only 9.7% of participants reported that their family members, friends, or neighbors had been infected with the COVID-19. According to the cutoff point for depression (CESD-10 ≥ 10) [35], 30.8% of the participants indicated significant depressive symptoms during the outbreak of COVID-19.

**Table 1.** Descriptive characteristics of the study sample (*n* = 516).


**Table 1.** *Cont.*


Note. SD = standard deviation.

#### *3.2. Characteristics of Depression*

As shown in Table 2, older adult's depression differed significantly for different characteristics. There were no significant differences in depression levels for gender, medical history of chronic diseases, occupational status, and BMI intervals (*p* = 0.10–0.95). The results indicated that older adults who were divorced/widowed and lived alone showed significantly higher depression levels than those who were married (*p* < 0.001) and lived with their spouse, partners, and children (*p* = 0.035). The depression level was significantly lower for participants who had higher educational levels (*p* = 0.001) and higher household income (*p* < 0.001) relative to those with poorer socioeconomic status. In addition, older adults who perceived their health status as bad and poor (*p* < 0.001) and who had acquaintances infected with COVID-19 (*p* = 0.003) indicated significantly higher depression levels than those in the other categories.

**Table 2.** Characteristics of depression (*n* = 516).




Note. SD = standard deviation.

#### *3.3. Association of Individual Precautionary Behaviors with Depression*

Based on the characteristics of depression, all demographic variables that have shown significant differences in the depression levels were included as predictors in the hierarchical linear regression models [31]. Dummy variables were generated for all polynomial predictors. Results revealed that two demographic variables were significant predictors for older adult's depression levels, including education level and household income, which aggregately accounted for 7% of the variance in the depression level (*p* < 0.001). In terms of the covariates, both subjective health status and infected cases of participants' acquaintances significantly predicted the depression level among participants, coupled with the demographics contributing to the explanation for 12 % of the variance in the depression levels (*p* < 0.001). After controlling for the demographic factors and covariates, changes in COVID-19 precautionary behaviors significantly predicted the depression of older adults (β = −0.18, 95%CI = −1.24 to −0.62, *p* < 0.001), contributing to a significant improvement in the variance explanation (ΔR<sup>2</sup> = 0.03, *p* < 0.001). The total model accounted for 15% of the variance in depression level (*p* < 0.001). Details of multiple linear regression analyses are shown in Table 3.


**Table 3.** Results of hierarchical linear regression models (*n* = 516).

Note. <sup>a</sup> Residualized change scores were used for the calculation.
