Study design and participants
We conducted a longitudinal survey that consisted of 5 web-based surveys administered by an online research company, Macromill, Inc. in Japan. The 1st survey (T1) was conducted from 17 to 22 July 2020, the 2nd survey (T2) was conducted from 18 to 23 September 2020, the 3rd survey (T3) was conducted from 22 to 27 January 2021, the 4th survey (T4) was conducted from 17 to 21 September, and the 5th survey (T5) was conducted from 21 to 26 January 2022.
From a pool of approximately 10 million individuals registered with Macromill, Inc. and the companies with which the Macromill has partnerships, we recruited participants who were between 20 and 79 years old and who lived in the 13 prefectures that implemented special measures related to the first wave of the COVID-19 pandemic in Japan. A quota sampling method was used to obtain age groups of equal size (i.e., groups of individuals in their 20s, 30s, 40s, 50s, 60s, and 70s), participants of both sexes (male and female), and participants with different employment statuses (full-time worker; no regular employment; and unemployed, including homemaker, retired, and jobless). All the participants received Macromill points for their participation; Macromill points are associated with the original point service provided by Macromill, Inc., and participants can trade these points for prizes or cash.
The present study analyzed participants who completed all 5 surveys and identified their employment status as full-time worker, homemaker, no regular employment, or not working at T5 to investigate the relation between employment status and mental health.
Sample size and effect size
We planned to recruit 2,700 participants at T1 based on a calculation of the appropriate sample size [15]; however, the present study was a longitudinal study, and so it was difficult to control its sample sizes. We calculated effect sizes as follows: Cohen's D (d) for the t test, Phi (φ) and Cramer's V (V) for the chi-square test, and partial η2 (ηp2). With respect to d, 0.80 indicates a large effect size, 0.50 a medium effect size, and 0.20 a small effect size. Regarding φ and V, 0.50 represents a large effect size, 0.30 a medium effect size, and 0.10 a small effect size. Although no clear criteria are associated with ηp2, the higher the value of this measure is, the larger the effect size. For adjusted R2, 0.26 designates a large effect size, 0.13 a medium effect size, and 0.02 a small effect size.
Measurements
The present study collected participants’ sociodemographic characteristics and included 3 questionnaires pertaining to depressive symptoms, PTSD symptoms, and coping strategies. The sociodemographic characteristics investigated included age, sex, the presence or absence of chronic illnesses, marital status, the presence or absence of children, employment status, household income (< 4 million JPY, 4–8 million JPY, and > 8 million JPY), and the economic impact for the COVID-19 pandemic. In Japan, the average annual household income was 5.52 million JPY, and the median income was 4.37 million JPY [17].
Depressive symptoms were measured using the Japanese version of the Patient Health Questionnaire-9 (PHQ-9) [18, 19]. Participants were asked to indicate the frequency with which they had experienced depressive symptoms over the past 2 weeks. The PHQ-9 consists of 9 items scored on a four-point scale (0 to 3); the total score can range from 0 to 27, with higher scores indicating more depressive symptoms.
PTSD was measured using the Japanese version of the Impact of Event Scale-Revised (IESR) [20–22]. Participants were asked to indicate the frequency with which they had experienced PTSD symptoms over the past week. The IESR consists of 22 items scored on a five-point scale (0 to 4); the total score can range from 0 to 88, with higher scores indicating more PTSD symptoms.
Coping strategies were measured using the Japanese version of the Brief Coping Orientation to Problems Experienced (Brief COPE) [23, 24]. Participants were asked about the frequency with which they used various coping styles to deal with the social changes and inconveniences resulting from the COVID-19 pandemic at the time of completing the survey. The scale consists of 28 items and assesses 14 coping styles. Each coping style is evaluated by reference to two items that are scored on a four-point scale (1 to 4); the total scores for each coping style can range from 2 to 8. Higher scores indicate that these means of coping are used very frequently.
Statistical analysis
The present study analyzed participants who completed all 5 surveys and indicated their employment status as full-time worker, no regular employment, or unemployed at T5. First, we compared the demographic characteristics of the analyzed participants and to those of participants who were excluded from the analysis. Sociodemographic characteristics at T1, mean PHQ-9 scores at T1 and mean IESR scores at T2 were compared using the chi-square test and two-sample t test. Subsequently, to compare the sociodemographic characteristics and coping strategies of the analyzed participants by sex, two-sample t tests and chi-square tests were conducted to explore their sociodemographic characteristics at T5 and mean Brief COPE scores at T5.
Thereafter, to examine changes in the level of depressive and PTSD symptoms across the timepoints of the survey, a mixed-model analysis of variance (ANOVA) was conducted with respect to the mean PHQ-9 and IESR scores including survey time as a within-subject factor and sex as a between-subjects factor.
Finally, to examine the influences of sociodemographic characteristics and coping strategies on depressive and PTSD symptoms, multiple linear regression analyses were conducted by sex using PHQ-9 and IESR scores at T5 as the dependent variables and sociodemographic characteristics and Brief COPE scores at T5 as the predictor variables.
The statistical significance level was set at p < .05, and effect sizes, Cohen's D (d) for the t test, Phi (φ) and Cramer's V (V) for the chi-square test, partial η2 (ηp2) for ANOVAs, and adjusted R2 were used for multiple linear regression, as mentioned previously. All statistical analyses were conducted using IBM SPSS statistical software (Version 28).