STRENGTHS AND LIMITATIONS OF THIS STUDY
-
We used data collected from 23 173 workers at 367 small-sized and medium-sized enterprises in Japan to define workplace social capital.
-
The choice of the questions was not at our discretion, and we did not use a validated assessment of workplace social capital.
-
As the study companies were located in one area of Japan, the study participants might not have fully represented the whole working population of the country.
Introduction
Blood pressure is a health outcome that has been widely studied in relation to social capital (SC), defined as features of social organisation such as networks, norms and social trust that facilitate coordination and cooperation for mutual benefit.1 Previous studies have reported inverse associations between SC and blood pressure, both at the individual (using individual responses to questions regarding SC, eg, social support,2–4 social integration3 5–10 and trust8) and collective levels (eg, neighbourhood11 12) with some studies reporting no association.2 3 6 Several potential pathways have been suggested to interpret the protective effect of SC on health, including the acquisition of health-related information through a tightly knit social network, the receipt of instrumental and emotional support and informal social control (ie, the ability of members in a community to maintain social order).13
Since a workplace is a place where many people spend significant time interacting with others, it is often regarded as a source of social relations.12 Several studies have investigated the association between workplace SC and health outcomes, including elevated blood pressure.14–16 For example, Oksanen et al15 used information on individual-level SC indicators in the workplace in Finland, reporting that male employees with low workplace SC were at higher risk of developing chronic hypertension compared with those with high SC while they did not find any evidence of significant associations among female workers. At a collective level, Fujino et al16 examined the cross-sectional association between workplace SC and systolic blood pressure (SBP) among 5368 workers nested in 58 departments within a Japanese chemical company, reporting a significant association in relation to workplace SC among female workers.
Despite these studies, several issues remain addressed. First, we are unaware of any study examining the association between workplace SC assessed at a collective level and elevated blood pressure, except for Fujino et al,16 who conducted their study within departmental units of a single large corporate group. Identifying workplace-level factors associated with elevated blood pressure may facilitate our effort to mitigate the disease burden associated with elevated blood pressure. Second, few studies have focused on the role of workplace SC among those working for small-sized and medium-sized enterprises. This can be an important omission in our knowledge as 70% of workers in Japan work for small-sized and medium-sized enterprises, where people tend to work longer hours for less salary compared with those working for large companies.17 18 Some studies in Japan suggested that health outcomes tended to be worse among those working for small-sized and medium-sized companies than those working for larger companies.19 20 Third, limited studies have focused on gender differences in SC and health. Some existing literature has suggested that men and women may use SC differently due to variations in social networks, societal roles and responsibilities.21–24
Therefore, this study aimed to investigate the association between SC and SBP, using information collected among 23 173 Japanese workers nested in 367 small-sized and medium-sized corporates.
Methods
Study participants
Data were collected from the Seirei Medical Check-up Center (Shizuoka Prefecture, Japan), which provides health check-up services for employers who organise mandatory annual health check-ups for their employees. We obtained information on medical examinations as well as questionnaire information on psychosocial stress, which was collected as part of the Stress-Check Programme25 from 31 178 participants working for 382 companies, aged 18 years or older in the fiscal year of 2017 (April 2017–March 2018). We excluded those working for large companies (defined here as a company with more than 300 participants in the dataset) (n=8005 working in 15 companies), thus including 23 173 workers from 367 companies in this study.
Informed consent was obtained from the study participants to use their anonymised health check-up data and responses in the Stress-Check Programme for research purposes. Participants were informed that they could withdraw their participation at any time during the study. The study protocol was approved by the Ethics Committee of the National Federation of Industrial Health Organization (no approval number assigned) and the Ethics Committee of the National Center for Global Health and Medicine, Japan (approval number: NCGM-G-003189).
Outcome
Blood pressure was measured in a seated position using an automated sphygmomanometer. The measurement was a part of the annual health check-ups and was taken by trained nurses who were accustomed to conducting health check-ups. While a second measurement may have been taken for confirmation for some individuals, we only used information on SBP obtained in the first measurement as the primary outcome measure. The information on diastolic blood pressure (DBP) was used in the sensitivity analysis (described below).
Exposure (Social Capital)
The Brief Job Stress Questionnaire is a 57-item questionnaire developed for use in the Stress-Check Programme, an occupational health policy launched by the Ministry of Health, Labour and Welfare in December 2015 to screen for stressed individuals in workplaces. Under the Occupational Health and Safety Law, workplaces with 50 or more workers must implement a Stress-Check Programme annually.
In this study, to define individual-level SC, we selected the following eight questions in the Brief Job Stress Questionnaire25: (1) how freely can you talk with superiors? (2) how freely can you talk with coworkers? (3) how reliable are superiors when you are troubled? (4) how reliable are coworkers when you are troubled? (5) how well will superiors listen to you when you ask for advice on personal matters? (6) how well will coworkers listen to you when you ask for advice on personal matters? (7) to what extent do you agree with the following statement?: ‘There are differences of opinion within my department’ (reversely coded) and (8) to what extent do you agree with the following statement?: ‘The atmosphere in my workplace is friendly’. Response options were presented as a Likert scale (1=not at all, 2=somewhat, 3=very much, 4=extremely). After confirming that these questions had a high internal consistency (Cronbach’s alpha=0.86), we summed them to create a composite score ranging from 8 to 32, with a higher value indicating higher individual SC.
Aggregate-level SC (ie, workplace SC) was assessed using individual-level responses to questions regarding SC, following the lead of the previous study.16 More specifically, the average score of individual-level responses was calculated for each company (n=367). The individual-level and worksite-level SC scores were standardised to a mean of 0 and a SD of 1 when we used a multilevel linear regression model to examine the association.
Covariates
Information on the following covariates was collected during the health check-up: age (in years), sex (male/female), current smoking status (currently smoking/not currently smoking), daily alcohol consumption, physical activity, job strain, sleeping quality, body mass index (BMI), antihypertensive medication use (yes/no), and diabetes and dyslipidaemia history.
Alcohol consumption was estimated based on the questionnaire information on consumption frequency (ie, everyday; sometimes; and seldom) and amount per occasion (ie, <1; 1–1.9; 2–2.9; and ≥3 go; go is a traditional Japanese unit of volume equivalent to approximately 23 g of ethanol). We assigned the following values to each response option: everyday=7 days, sometimes=3.5 days and seldom=0 days; <1 go = 0.5 go, 1–1.9 go=1.5 go; 2–2.9 go=2.5 go; and ≥3 go = 3.5 go. We then computed daily alcohol consumption and classified the participants into the following categories: do not drink; <1; 1–1.9; and ≥2 go/day. Two questions were used to assess physical activity: (1) moderate-to-vigorous intensity physical activity: ‘Are you in the habit of doing exercise (lasting for 30 min and making you sweat) twice a week or more frequently for longer than 1 year?’; (2) light physical activity: ‘In your daily life, do you exercise which is equivalent to walking more than 1 hour a day?’ Four groups were created based on the combination of responses to the two questions.
Job strain was defined when individuals simultaneously reported psychological demands to be high and job control to be low.26 27 More specifically, psychological demands were assessed using the responses to the question regarding the extent to which they agreed with the following statements (presented as a Likert scale: 1=not at all, 2=somewhat, 3=very much, 4=extremely): ‘I have an extremely large amount of work to do; I cannot complete work in the required time; I have to work as hard as I can; I have to pay very careful attention; My job is difficult in that it requires high level of knowledge and technical skill; I need to be constantly thinking about work throughout the working day (score range: 6–24)’. Job control was defined using the following three questions (ie, I can work at my own pace; I can choose how and in what order to do my work; I can reflect my opinions on workplace policy) (score range: 3–12). We then dichotomised psychological demands and job control at the median values (low/high) and defined job strain. Sleep quality was assessed by using the responses to the question ‘Do you get enough rest by sleeping?’ (yes/no).
Body weight and height were measured to compute BMI (kg/m2). Antihypertensive medication status was based on self-reports (yes/no). Diabetes was defined as fasting plasma glucose ≥126 mg/dL, HbA1c ≥6.5% or current use of antidiabetic medication.28 Dyslipidaemia was defined as any of the following criteria: triglycerides ≥150 mg/dL, low-density lipoprotein ≥140 mg/dL, high-density lipoprotein <40 mg/dL (male) and <50 mg/dL (female), or treatment of dyslipidaemia.29
The number of workers for each company was computed by counting the number of individuals available in the dataset.
Statistical analysis
To account for missing information on SC indicators (n=192), SBP (n=154) and other covariates (n=144), multiple imputations were used to generate 20 datasets.30 More specifically, we used the chained equation method with 200 iterations.31 We used linear, logistic and multinomial logistic regression models for continuous, binary or categorical variables, respectively. Using Rubin’s rules, we combined imputation estimates.32
Given that previous studies on this topic suggested a sex-related difference in the association, we stratified the analyses by sex after confirming that sex and workplace SC interacted in relation to SBP (p<0.001). A multilevel linear regression model (ie, a random intercept model) was used to investigate the association between SC indicators and SBP while accounting for the possible heterogeneity at worksite level (Level 1: individual; Level 2: worksite). We examined if workplace-level and individual-level SC interacted in terms of SBP and did not find any evidence of significant cross-level interaction.
We modelled workplace SC in Model 1 while adjusting for age. We further adjusted for the number of study participants in each worksite, BMI, antihypertensive medication, diabetes, dyslipidaemia, current smoking status, daily alcohol consumption, physical activity, job strain and sleeping quality in Model 2. To examine if workplace SC is independently associated with SBP over and above individual SC, we adjusted for individual SC in Model 3. Results are shown in the form of a change in SBP per 1SD increase in SC indicators.
We conducted a series of sensitivity analyses in which we (1) excluded those taking antihypertensive medication, (2) used DBP as the outcome and (3) used hypertension as the outcome. We defined hypertension when SBP ≥140 mmHg, DBP ≥90 mmHg or the use of antihypertensive medication and used a multilevel logistic regression model. The models were built in the same manner as in the main analysis. All the statistical analyses were conducted using Stata V.16.0 (StataCorp, College Station, Texas, USA).
Patients and public involvement
Patients and the public were not involved in this study.
Results
Table 1 shows the basic characteristics of the study participants after multiple imputations. The mean age was 42.7 for males and 41.7 for females. Those categorised as overweight and obese comprised 21.6% and 5.3% of male participants, while the corresponding figures were 12.5% and 4.4% for females, respectively. The proportion of those who currently smoked was higher in males (36.2%) than in females (11.9%). Those who consumed alcohol (≥1 go per day) included 28.1% male participants and 8.4% female participants. The use of antihypertensive medication was reported in 9.8% and 5.8% of male and female participants, respectively. The basic characteristics of study participants before multiple imputations are shown in online supplemental table 1.
Supplemental material
Basic characteristics of study participants in Shizuoka Prefecture, Japan (2017) after multiple Imputations
Table 2 shows the results of a multilevel linear model investigating the association between SC indicators and SBP. Among male participants, we did not find any evidence of a significant association between workplace SC and SBP. For example, a 1SD increase in workplace SC was associated with a change of −0.12 (95% CI=−0.46 to 0.21) in Model 3. Individual-level SC was rather linked with higher SBP (coef.=0.19, 95% CI=−0.01 to 0.39). Conversely, 1SD increase in workplace SC among female participants was linked with lower SBP (coef.=−0.53, 95% CI=−1.02 to −0.05), which remained statistically significant after adjusting for covariates and possible mediators in Model 2. Further adjustment for individual-level SC in Model 3 attenuated the association, which became non-significant (coef.=−0.41, 95% CI=−0.87 to 0.05), while individual-level SC itself was inversely associated with SBP (coef.=−0.43, 95% CI=−0.73 to −0.13).
Results of multilevel linear model investigating the association between SC measured at individual and workplace levels and systolic blood pressure among a Japanese working population in Shizuoka, Japan (2017)
In the sensitivity analyses in which we confined our participants to those not taking antihypertensive medication (table 3) or the analysis using DBP as the outcome (online supplemental table 2), the results did not materially change. More specifically, significant associations were observed among female participants regarding workplace SC before adjustment for individual-level SC. These associations in relation to workplace SC became non-significant after adjusting for individual SC, while individual SC was inversely associated with SBP. We did not observe any evidence of a significant association among male participants. When we used hypertension as the outcome (online supplemental table 3), we did not find strong evidence of significant association in relation to hypertension, but the associations trended in the same direction as observed in the main analysis.
Results of multilevel linear models investigating the association between SC measured at individual and workplace levels and systolic blood pressure among a working population in Shizuoka, Japan, confining the analytic sample to those without antihypertensive medication (2017)
Discussion
Studying 23 173 participants working for 367 small-sized and medium-sized companies in Japan, we found that workplace SC was inversely associated with SBP among females but not among males. The significant association among female participants was attenuated, and it became non-significant when the individual-level response to SC questions was incorporated into the model, indicating that we did not find strong evidence that workplace SC exerted contextual effects on blood pressure such that working at a workplace with high SC protects one from having raised blood pressure, irrespective of their individual SC.
The significant association between workplace SC and SBP among female participants is in line with several studies showing that aggregate-level SC was associated with lower blood pressure.11 12 14–16 As mentioned earlier in the Introduction section, several mechanisms have been suggested to link SC and blood pressure, including the acquisition of health-related information, the receipt of instrumental and emotional support and informal social control.13 In our specific context, workplace SC may also directly have reduced psychological stress as a result of increased social support,33 which in turn decreased blood pressure.34 In addition, health-related information (eg, lifestyle recommendations for preventing high blood pressure) might have diffused more easily among female employees working at a workplace with high versus low SC.35 It is also possible that female workers with a high versus low workplace SC obtained behavioural guidance from their colleagues.
Our finding that workplace SC was significantly associated with lower blood pressure among females but not among males is in line with the study of Fujino et al,16 who found that the proportion of individuals who reported a lack of helpfulness and the proportion of individuals reporting a lack of trust in coworkers were both significantly associated with increased blood pressure (in models without adjusting for individual-level responses) among 4735 workers nested in 58 departments in a single large corporate group in Japan; they found a stronger association in females than in males. On the other hand, a study by Oksanen et al,15 that was conducted among 11 777 male and 49 145 female employees in Finland, reported that high workplace SC was linked to a reduction in the risk of developing chronic hypertension by 40%–60% in men, but not in women.
While several factors might have been involved in the difference in the sex-related discrepancy between the two countries, one possible explanation for the null finding among Japanese men is the ‘Enkai Effect’, which means ‘the effect of drinking at a party ‘.36 Enkai or Nomikai, which commonly occurred in the Japanese working culture, particularly before the COVID-19 pandemic, describes the after-hours drinking occasions among coworkers or friends. While this can be regarded as the consequence of workplace SC, which can provide an occasion for the exchange of social support, this event can be detrimental to the health of workers as some people feel stressed to participate in these events. It is also possible that these occasions can spread unhealthy behaviours among colleagues such as excessive alcohol drinking, and the consumption of food served at Enkai, usually high in sodium (ie, the dark side of SC37). They may offset the beneficial effect of workplace SC on blood pressure, particularly among males, who participate in Enkai more frequently than females.
When we adjusted for individual responses to the SC question, the association between workplace SC and SBP was attenuated and became statistically non-significant. Although previous studies that examined aggregate-level SC (eg, workplace SC and community SC) did not necessarily account for individual responses to SC indicators, it is essential to make adjustments for individual responses to differentiate contextual from compositional effects (ie, examining if the aggregate-level SC prevents one from having/developing raised blood pressure, irrespective of individual SC). As Unger et al38 pointed out, the mechanisms linking SC and cardiovascular disease risk might operate more at the individual level. It is also possible that the variation in workplace SC was limited such that it cannot manifest itself in terms of its contextual effect on hypertension. It would not be easy to observe a significant association if participants all lived in an environment with high SC or worked in companies with high SC.
In this study, individual-level SC was associated with lower blood pressure among females, but not among males. The gender difference in the association was more apparent at the individual level compared with the workplace level, as the association in relation to individual-level SC among male workers trended towards a positive association. One possible reason is that women are known to generally excel in building social connections with others and effectively navigating both formal and informal social networks.39 In addition, this can be also attributed to the aforementioned ‘Enkai effect’ among male workers. While much effort has been paid to improving SC in many settings, policymakers and practitioners should be aware that improving SC can lead to unpleasant health consequences, depending on the approach taken.
Several limitations should be addressed. First, the study’s cross-sectional nature does not infer a causal relationship of the association observed in this study. For example, it is possible that those with medical conditions tended to seek social support at the workplace (ie, reverse causation). Second, this is the secondary data analysis, and the choice of the questions was not at our discretion; we did not use a validated questionnaire regarding workplace SC . In addition, while we defined SC by asking participants questions regarding the situation at the workplace, it is also possible that individuals enjoy social resources accessed via other sources (eg, the neighbourhood or the internet). Third, we used information on SBP obtained in the first measurement, which might have been subject to a white-coat effect. Fourth, the study participants did not represent the whole working population of Japan as we used information collected in one area of the country. Fifth, selection bias might have existed as some employees did not participate in the health check-up; we did not have information on the non-participation rate. However, such cases were deemed rare as annual health check-ups are mandatory under the law. Sixth, several confounders were not available in our dataset (eg, educational attainment and income, specific types of jobs they engaged in and some lifestyle parameters including diet).
Conclusions
Studying 23 173 participants working for 367 small-sized and medium-sized companies in Japan, we found that workplace-level SC was inversely associated with SBP among females, partly explained by individual-level responses. However, we did not find any evidence of significant inverse associations among males.
Data availability statement
Data are available upon reasonable request. Data are available from the corresponding author upon reasonable request ([email protected]).
Ethics statements
Patient consent for publication
Ethics approval
The study protocol was approved by the Ethics Committee of the National Federation of Industrial Health Organization (no approval number assigned) and the Ethics Committee of the National Center for Global Health and Medicine, Japan (approval number: NCGM-G-003189). Participants gave informed consent to participate in the study before taking part.
This post was originally published on https://bmjopen.bmj.com