Accelerometer-Measured Diurnal Patterns of Sedentary Behavior among Japanese Workers: A Descriptive Epidemiological Study

Increased sedentary behavior (SB) can adversely affect health. Understanding time-dependent patterns of SB and its correlates can inform targeted approaches for prevention. This study examined diurnal patterns of SB and its sociodemographic associations among Japanese workers. The proportion of sedentary time (% of wear time) and the number of breaks in SB (times/sedentary hour) of 405 workers (aged 40–64 years) were assessed using an accelerometer. SB patterns and sociodemographic associations between each time period (morning, afternoon, evening) on workdays and nonworkdays were examined in a series of multivariate regression analyses, adjusting for other sociodemographic associations. On both workdays and nonworkdays, the proportion of sedentary time was lowest in the morning and increased towards evening (b = 12.95, 95% CI: 11.28 to 14.62; b = 14.31, 95% CI: 12.73 to 15.88), with opposite trend for breaks. Being male was consistently correlated with SB. Other sociodemographic correlates differed depending on time-of-day and day-of-the-week. For instance, desk-based workstyles and urban residential area were associated with SB during workday mornings and afternoons, being single was related to mornings and evenings, workdays and nonworkdays. Initiatives to address SB should focus not only on work-related but time-of-day contexts, especially for at-risk subgroups during each period.


Introduction
Longer time spent in sedentary behavior (SB), which is distinct from lack of physical activity, is associated with poor health outcomes, such as the higher risk of all-cause mortality, type 2 diabetes mellitus, some types of cancer, and cardiovascular diseases [1]. The recent technological revolution has promoted automatization and computerization of work, resulting in a reduction of physical demand and increase in sedentary time at work [2]. For example, the average working time among Japanese workers was approximately eight hours per day in 2017 [3]. Thus, workers are one of the at-risk populations for longer time spent in SB, and it is necessary to focus on reducing worker's SB as a public health strategy.

Sedentary Behavior and Physical Activity
SB and physical activity were measured by a triaxial accelerometer (Active style Pro HJA-350IT; Omron Healthcare Co. Ltd., Kyoto, Japan). This device was reported to have high validity [14] and reliability [15]. The participants were asked to wear the accelerometers during waking hours for seven consecutive days and to remove them during any water-based activities or contact sports. Nonwear time was defined as intervals of at least 60 consecutive minutes of no activity (0.9 or less Metabolic Equivalents (METs)) [14], with allowance for up to two minutes of observations of some limited movement (≤ 1.0 METs) within these periods [16]. Due to sparse data during the early morning and late evening, time spent in SB and engaged in physical activity were calculated between 06:00-23:59 and each period of the day (morning (06:00-11:59), afternoon (12:00-17:59), evening (18:00-23:59)) on workdays and nonworkdays. Days containing at least 10 h of wear time were regarded as valid [16]. Using the log diary, participants were asked to report the time when they put on and took off the accelerometer. The log diaries were checked along with accelerometer data to verify the time periods during which participants wore the accelerometer and days of the week. Those participants who completed at least four valid days, including at least one nonworkday, were included. Finally, participants with a minimum of 25% valid hours of wearing time in each of the three time periods [17] of at least three workdays and one nonworkday were included [7].
Three measures of SB and physical activity were designated for the total and each time periodsedentary, light-intensity physical activity (LPA), and moderate-to-vigorous physical activity (MVPA) time (minutes/period), the proportion of sedentary, LPA and MVPA time (% of wear time), and the number of breaks in SB (times/sedentary hour). SB, LPA, and MVPA were defined as wear time for any activity with an accelerometer-estimated intensity of ≤ 1.5, 1.5 and < 3.0, and 3.0 or more METs, respectively [18]. A break was defined as a period of nonsedentary minutes in between two sedentary bouts [19].

Sedentary Behavior and Physical Activity
SB and physical activity were measured by a triaxial accelerometer (Active style Pro HJA-350IT; Omron Healthcare Co. Ltd., Kyoto, Japan). This device was reported to have high validity [14] and reliability [15]. The participants were asked to wear the accelerometers during waking hours for seven consecutive days and to remove them during any water-based activities or contact sports. Nonwear time was defined as intervals of at least 60 consecutive minutes of no activity (0.9 or less Metabolic Equivalents (METs)) [14], with allowance for up to two minutes of observations of some limited movement (≤ 1.0 METs) within these periods [16]. Due to sparse data during the early morning and late evening, time spent in SB and engaged in physical activity were calculated between 06:00-23:59 and each period of the day (morning (06:00-11:59), afternoon (12:00-17:59), evening (18:00-23:59)) on workdays and nonworkdays. Days containing at least 10 h of wear time were regarded as valid [16]. Using the log diary, participants were asked to report the time when they put on and took off the accelerometer. The log diaries were checked along with accelerometer data to verify the time periods during which participants wore the accelerometer and days of the week. Those participants who completed at least four valid days, including at least one nonworkday, were included. Finally, participants with a minimum of 25% valid hours of wearing time in each of the three time periods [17] of at least three workdays and one nonworkday were included [7].
Three measures of SB and physical activity were designated for the total and each time period-sedentary, light-intensity physical activity (LPA), and moderate-to-vigorous physical activity (MVPA) time (minutes/period), the proportion of sedentary, LPA and MVPA time (% of wear time), and the number of breaks in SB (times/sedentary hour). SB, LPA, and MVPA were defined as wear time for any activity with an accelerometer-estimated intensity of ≤ 1.5, 1.5 and < 3.0, and 3.0 or more METs, respectively [18]. A break was defined as a period of nonsedentary minutes in between two sedentary bouts [19].

Sociodemographic Factors
Sex and age were obtained from residential registries. Height, weight, marital status (single, married), household income (< 5 million yen, ≥ 5 million yen), educational attainment (high school or less, two years of college or higher education), employment status (full-time, part-time), workstyle (desk-based, non-desk-based), smoking status (smoker, nonsmoker), and alcohol consumption (≤ 1-3 times/month, ≥ 1 time/week) were reported by participants. Body mass index (BMI; kg/m 2 ) was calculated from self-reported height and weight. The residential area was classified into urban (Koto) and suburban (Matsuyama).

Statistical Analyses
All analyses were separately conducted for workdays and nonworkdays. Overall average values of SB, LPA, and MVPA during 6:00-23:59 (whole day) and each period (morning, afternoon, and evening) were reported. The diurnal differences in the proportion of sedentary time (% of wear time) and the number of breaks (times/sedentary hour) were examined using the multilevel mixed-effects linear regression models (time periods as level 1 and individuals as level 2) with random intercepts, adjusting for sociodemographic variables. Differences in the proportion of sedentary time (% of wear time) and the number of breaks (times/sedentary hour) between subgroups in a whole day or each time period were examined using multiple regression analysis, adjusting for other sociodemographic variables. Interaction effects between sociodemographic variables and time period were examined using multilevel mixed-effects linear regression models with random intercepts. Stratified analysis was conducted for variables which demonstrated significant interactions. Several graphs were also created to facilitate the interpretation of the interaction effects. Statistical analyses were performed using STATA 15.0 (Stata Corp LLC, TX USA), and the level of significance was set at p < 0.05 for differences and p < 0.1 for interactions. Table 1 presents the characteristics of the final study sample of 405 participants. Mean age was 51.5 (standard deviation (SD), 7.0) years. Approximately half of the participants were women. Most participants were married; more than half had greater than or equal to 5 million yen of income, and two-thirds had graduated with two years of college or higher education. The participants were mainly full-time workers with a desk-based workstyle. A small number of participants were overweight (BMI ≥ 25 kg/m 2 ), smokers and half reported alcohol consumption one or more times per week. On workdays, the average length of time wearing the accelerometer was 937.0 (69.4) minutes overall, of which 352.4 (126.7) minutes were LPA and 77.4 (47.7) minutes were MVPA. On nonworkdays, mean time wearing the accelerometer was 880.6 (90.2) minutes throughout the day, of which 313.2 (109.5) and 56.7 (36.0) minutes were LPA and MVPA, respectively. Averaged minutes of accelerometer wearing time, LPA, and MVPA on three time periods of work and nonworkdays are presented in the Supplementary

Prevalence of Sedentary Behavior for the Whole Day and Three Time Periods
The mean (SD) sedentary time during the whole day (06:00-23:59) was 507.3 (140.2) minutes on workdays and 510.7 (127.5) minutes on nonworkdays, accounting for 54.2% (14.9%) and 58.0% (13.5%) of wearing time, respectively (Table 2). Both on workdays and nonworkdays, the average proportion of sedentary time was the lowest in the morning and highest in the evening. In the morning, the average proportion of sedentary time was 48.3% and 52.6%, respectively. These proportion were significantly higher in the afternoon (54.1%, b = 5.86, 95% CI: 4.19 to 7.53, p < 0.001; 55.6%, b = 2.97, 95% CI: 1.39 to 4.54, p < 0.001) and evening (61.2%, b = 12.95, 95% CI: 11.28 to 14.62, p < 0.001; 66.9%, b = 14.31, 95% CI: 12.73 to 15.88, p < 0.001). The difference between afternoon and evening was also significant (b = 7.09, 95% CI: 5.42 to 8.76, p < 0.001; b = 11.34, 95% CI: 9.76 to 12.91, p < 0.001). In contrast, the number of breaks on workdays and nonworkdays were the highest in the morning and lowest in the evening. The number of breaks in the morning was 12.   LPA, light-intensity physical activity; MVPA, moderate-to-vigorous intensity physical activity; SB, sedentary behavior; SD, standard deviation. Using multiple regression, the differences in the measures of the proportion of sedentary time and number of breaks between subgroups of each sociodemographic variable were examined, adjusting for other sociodemographic variables. *** p < 0.001, ** p < 0.01, * p < 0.05.

Workdays by Whole Day
For the whole day, the proportion of sedentary time was significantly higher in men, singles, those with higher educational attainment, desk-based workstyles, and those living in the urban area. On the other hand, women, participants who are married, those with non-desk-based workstyles, participants whose weight was in the normal range, and those participants who live in the suburban area reported higher numbers of breaks.

Workdays by Time Period
In analyses stratified by the time periods, the proportion of sedentary time was significantly higher in men and those with desk-based workstyles, during all three time periods. Additionally, being single and living in an urban area were significantly associated with a higher proportion of sedentary time during the morning period. Those participants with higher educational attainment, individuals with a full-time job, and those living in the urban area were more likely to spend time in SB during the afternoon period. There were no additional sociodemographic attributes significantly associated with the proportion of sedentary time during the evening period. Regarding the number of breaks, woman, non-desk-based workstyles, and the suburban residential area were significantly associated with taking more breaks during both the morning and the afternoon periods. Married participants and those with normal weight status (BMI < 25 kg/m 2 ) were also significantly associated with higher numbers of breaks during morning and afternoon periods, respectively. During the evening period, smokers were the only sociodemographic group that was significantly associated with more breaks.

Nonworkdays by Whole Day
For the whole day, men, being single, and those who consumed alcohol less frequently had a significantly higher proportion of sedentary time than participants in other sociodemographic groups. The number of breaks was significantly higher in women and married individuals.

Nonworkdays by Time Period
Analyses of nonworkdays stratified by time period showed a significantly higher proportion of sedentary time observed for men during all three time periods. Additionally, participants who were single, those with desk-based workstyles, and those who had a lower frequency of alcohol consumption had a significantly greater proportion of sedentary time during the morning period. On the other hand, there were no additional sociodemographic attributes significantly related to the proportion of sedentary time during the afternoon period. During the evening period, those whose age was ≥ 50 years and those who were single showed significantly higher proportions of the sedentary time. In terms of breaks, women and married participants had significantly more breaks during the morning period. Women, during the afternoon period, and married participants and those aged < 50 years during the evening period also had significantly more breaks, respectively.

Discussion
The present study examined the diurnal patterns of SB and its associations with sociodemographic attributes among Japanese workers. This study found that the proportion of sedentary time among day-time workers was lowest in the morning, increased over the time periods, and was the highest in the evening on both workdays and nonworkdays. This trend was less clear on nonworkdays due to the higher proportion of SB during the morning period. This is similar to some

Discussion
The present study examined the diurnal patterns of SB and its associations with sociodemographic attributes among Japanese workers. This study found that the proportion of sedentary time among day-time workers was lowest in the morning, increased over the time periods, and was the highest in the evening on both workdays and nonworkdays. This trend was less clear on nonworkdays due to the higher proportion of SB during the morning period. This is similar to some previous studies which have shown such a trend by the graphical illustrations of diurnal SB patterns among workers [10][11][12][13]. There were no previous studies identifying diurnal SB patterns in worker populations with quantitative analyses. Only a limited number of studies exist examining the differences in the work context, and they have found that SB during work hours was higher than those during nonwork hours [5,10,13]. The current findings imply that SB could be influenced by not only the work context but also the time period of the day in the worker population.
Our findings showed that even during nonworking (leisure) time, the patterns of SB were different-the least SB was observed in the morning, whereas most were in the evening. These patterns may be partly due to the interaction of physical and psychological conditions (less fatigued and more energetic in the morning after sleep), basic lifestyle (engaging in a household, social, and working tasks by daylight), and social structure (scheduled working, business, or store hours), which could be affected or set by general human circadian rhythms [20,21]. This interpretation may be confirmed by interaction effects found in this study-even though the patterns of SB during the morning and afternoon periods were different between subgroups in some of the sociodemographic associations, they reached approximately the same levels by the evening. On nonworkdays, this timing seems to be even earlier, in the afternoon. Considering such daily patterns in behavior and physical and psychological functions generally regulated by the circadian system, an effort to sleep earlier (to replace SB with equal amounts of sleep) might be a more feasible and beneficial approach to reduce SB rather than breaking up SB at night among the worker population, in addition to workplace interventions (e.g., sit-stand workstations for desk-based workers) conducted in previous studies [4,[22][23][24][25]. Two previous studies found that sleeping less (more awake time) was associated with increased screen time (e.g., TV viewing and computer use), which is one of the sedentary domains [26,27]. Avoiding or decreasing screen time at night may have double health benefits from reducing SB and improving sleep quality as well as alertness while awake [28], especially for at-risk subgroups in the morning.
This study also demonstrated different associations in different sociodemographic attributes for each time period of SB. Regardless of the time of day and days of the week, men spent more time in SB and took fewer breaks than women throughout the day. This may reflect the difference in social and domestic roles by sex in Japan. The social situation for Japanese women, even if they are working, is characterized by a higher engagement rate for household tasks and childcare [29], which may lead them to break up SB and to be less sedentary. Additionally, it has been reported that Japanese men have long commuting and working hours [29], which may limit their opportunities for being physically active.
Regarding mornings and evenings, which are mostly categorized as leisure time, marital status was consistently associated with total sedentary time and breaks (except during the evening on workdays). These results may reflect the social roles as a parent and partner. Single adults have higher levels of sedentary leisure activities [30,31], which could be due to fewer domestic responsibilities and greater free time. Also, workers aged 50 years or older had a longer total sedentary time and fewer breaks on nonworkday evenings, which may be partly due to decreased social and parenting activities [32,33], or increased fatigue [34]. These results may indicate that sociodemographic correlates of SB are largely related to social and domestic roles during the morning and evening, in addition to work-related attributes of longer SB such as workstyles [10].
During the workday afternoon, persons with desk-based workstyles, those with higher educational attainment, those in full-time employment, and those living in a urban residential area were related to longer time spent in SB and/or fewer breaks. These results indicate that the SB on the workday afternoon (most of which is working time) could be more influenced by work-related sociodemographic attributes. Furthermore, there are previous studies that reveal positive associations of occupational sedentary time with office workers [10] and full-time employment [35]. White-collar workers [9,35] and those having a private office [36], also found in the previous studies as correlates of higher occupational SB, are partly but closely related to educational attainment. Thus, it may be necessary to target full-time and desk-based workers in Japan, similar to many previous SB intervention studies conducted in Australia and Western countries [4]. Also, the differences in SB and breaks between two residential areas may be related to the work-related sociodemographic attributes such as the kind of occupation. The Matsuyama city had a higher proportion of those who are engaged in "agriculture and forestry" or "medical, healthcare and welfare" than the Koto ward [37]. This study, however, did not directly examine occupation. Further study would need to consider more specific work-related attributes such as occupation and job position. Using sociodemographic factors that were not work-related, this study found that higher BMI was associated with fewer breaks, consistent with previous studies [35]. Thus, approaches aiming to break up prolonged sedentary time may be beneficial in overweight and obese workers.
This study had several limitations. First, having a cross-sectional design, this study is unable to determine causal relationships between variables. Second, the accelerometer devices used in this study were unable to measure water-based activity or activity that involves the movements of legs and arms and to distinguish sitting or static standing postures accurately. Finally, since there were variations in participants' accelerometer wearing time (due to different times of waking up or going to sleep), this study cannot describe the diurnal variation of absolute time spent in SB. Future studies can also conduct cluster analysis to identify distinct subgroups of individuals with similar characteristics of diurnal sedentary patterns.
This was the first study, to our knowledge, to explore the diurnal patterns of objectively measured SB among workers and their sociodemographic associations. Another strength was the use of objectively measured SB that enabled us to investigate the accurate proportion of sedentary time and breaks during each time period.

Conclusions
In summary, the proportion of sedentary time was lowest in the morning (48.3%; 52.6%) and highest in the evening (61.2%; 66.9%) both on workdays and nonworkdays, with an inverse trend for breaks. Although being male was a consistent factor related to SB and breaks throughout the day and week, associations between other sociodemographic attributes and these SB measures differed depending on time periods of the day and the days of the week. SB and/or breaks seem to be mainly associated with work-related attributes such as workstyle, educational attainment, and employment status on workday afternoons, whereas social and domestic-related attributes such as marital status are related to mornings and evenings of workdays and nonworkdays.