Changes in the Employment Status and Risk of Stroke and Stroke Types
Background and Purpose—Because of limited evidence, we investigated a long-term impact of changes in employment status on risk of stroke.
Methods—This was a prospective study of 21 902 Japanese men and 19 826 women aged 40 to 59 years from 9 public health centers across Japan. Participants were followed up from 1990 to 1993 to the end of 2009 to 2014. Cox proportional hazard ratio of stroke (incidence and mortality) and its types (hemorrhagic and ischemic) was calculated according to changes in the employment status within 5 years interval between 1990 to 1993 and 1995 to 1998 (continuously employed, job loss, reemployed, and continuously unemployed).
Results—During the follow-up period, 973 incident cases and 275 deaths from stroke in men and 460 cases and 131 deaths in women were documented. Experiencing 1 spell of unemployment was associated with higher risks of morbidity and mortality from total, hemorrhagic, and ischemic stroke in both men and women, even after propensity score matching. Compared with continuously employed subjects, the multivariable hazard ratio (95% confidence interval) for total stroke incidence in job lost men was 1.58 (1.18–2.13) and in job lost women was 1.51 (1.08–2.29), and those for total stroke mortality were 2.22 (1.34–3.68) in men and 2.48 (1.26–4.77) in women. The respective hazard ratio (95% confidence interval) in reemployed men was 2.96 (1.89–4.62) for total stroke incidence and 4.21 (1.97–8.97) for mortality, whereas those in reemployed women were 1.30 (0.98–1.69) for incidence and 1.28 (0.76–2.17) for mortality.
Conclusions—Job lost men and women and reemployed men had increased risks for both hemorrhagic and ischemic stroke incidence and mortality.
Compared with the 1980s unemployment rate in Japan (2%), in the 1990s, the rate began to increase to 4.1% in 1998, reaching a peak of 5.4% in 2002.1 The health adverse effects of unemployment are well established2; job loss was associated with a higher proportion of depressive symptoms3 and increased risks of myocardial infarction,4–9 stroke,4–6,9–12 and all-cause mortality.6,13 However, a little is known about the effects of returning to work on health,14–16 that was focusing on mental health15,16 rather than on physical health.17 Apart from 2,10,16 the previous studies were conducted in Western countries, which have different social contexts in terms of labor market policies, perhaps making them less applicable to East Asian countries.
Some studies provided evidence that reemployed people enjoyed a partial psychological recovery from a job loss experience, but with no change on their physical health.15–17 Conversely, reemployed subjects had a high risk of death in another prospective study.13 Japanese employees who lost their jobs were at high chances to be reemployed in unsatisfactory jobs18 that might increase their stroke risk.19 No studies have yet been conducted to examine a sex-specific long-term effect of reemployment on risk of stroke incidence and mortality, which are known to be influenced by the employment status4–6,9–11,19 or the associations of job loss and reemployment with risk of stroke types (hemorrhagic and ischemic).
We hypothesized that changes in employment status may have an impact on risk of stroke, and we aimed to examine that in Japanese men and women.
Materials and Methods
Design and Population
The JPHC (Japan Public Health Center)–based prospective study was initiated in 1990 in 5 public health centers for cohort 1 and 6 public health centers in 1993 for cohort 2, with a total of 140 420 (men and women) aged 40 to 69 years. The details of the study cohort have been described elsewhere.20 Two public health center areas in metropolitan Tokyo and Osaka were excluded from the present analysis (n=23 524) because no data on stroke incidence were available. A self-administered questionnaire was distributed to all registered participants (response rate: 81.6%). A follow-up survey was conducted 5 years after the first survey (response rate: 84.9%). A total of 37 410 men and 38 401 women responded to both questionnaires. For this study, we set the dates of the first survey as our prebaseline and those of the second survey as our baseline to determine changes in the employment status between first and second surveys (Figure I in the online-only Data Supplement). For the purposes of this study, we limited our study population to those who were aged 45 to 59 years at baseline (the working age population; n=22 845 men and 22 122 women). We excluded 101 men and 1206 women with no information on their employment status and 842 men and 1090 women with a history of cancer or cardiovascular diseases at baseline. The final study population included 21 902 men and 19 826 women. The study was approved by the human ethics review committees of the National Cancer Center and Osaka University.
Employment Status Change
Change of employment status was identified by responses to the questions on employment status (employed including self-employed or unemployed) in both the prebaseline and baseline surveys. We categorized subjects into 4 groups: (1) continuously employed (employed at both the prebaseline and baseline surveys times), (2) job loss (employed at prebaseline time to unemployed at baseline time), (3) reemployed (unemployed at prebaseline time to employed at baseline time), and (4) continuously unemployed (unemployed at both prebaseline and baseline surveys times). We considered those who reported no job at prebaseline and got a job in the baseline survey as reemployed because it is unlikely for Japanese people to get their first job at age 45 to 59 years.
Besides age and residential area, other hypothesized confounding factors were the prebaseline characteristics, including hypertension, diabetes mellitus statuses, and hypercholesterolemia which were determined from the responses to the prebaseline questions about medical histories and drug treatment, whereas the possible mediating factors were calculated from the responses to the baseline questions about the followings: perceived psychological stress evaluated via the responses to the question How much stress in your daily life? (little, average, large); perceived level of life enjoyment based on the subjective judgment on the participant’s sense of life enjoyment was evaluated via a score according to the responses to the question Are you enjoying your life? (low, medium, high); changes in living arrangement (no change versus changed living arrangement to live alone); quintiles of body mass index (BMI) calculated from reported height and weight (kg/m2); quintiles of metabolic equivalent in MET units for physical activity, smoking status (never, ex-smoker, current smoker of 1–19, 20–29, or ≥30 cigarettes per day); and ethanol intake (nondrinkers, former drinkers, and weekly ethanol intake of <150 g/wk, 150–<300 g/wk, 300–<450 g/wk, or ≥450 g/wk).
Confirmation of Stroke Incidence and Mortality
A total of 81 hospitals with the capability of treating stroke patients were registered within the administrative districts of the JPHC cohorts. Physicians blinded to the patients’ lifestyle data reviewed the medical records at each hospital. When subjects reported a history of stroke on a follow-up questionnaire (5 years after our baseline) and had not been registered as stroke cases, we asked by letter or telephone about the onset of stroke for permission to review medical records and then review them. Strokes were confirmed according to the criteria of the National Survey of Stroke,21 and the widespread use of computed tomographic scans even in Japanese local hospitals since the 1980s has probably made the medical records diagnosis of stroke and its types sufficiently accurate.22
All death certificates were forwarded centrally to the Ministry of Health, Welfare, and Labor; therefore, stroke death, defined according to the International Classification of Diseases, 10th Revision (codes I60 to I69), was confirmed even for relocated subjects. Detailed confirmation methods have been described elsewhere.20
Stroke incidence and mortality were analyzed separately. For the analysis of stroke incidence, person-years were calculated from the date of the baseline January 1, 1995 in cohort 1, and January 1, 1998 in cohort 2 until obtaining 1 of 4 possible end points: (1) incidence of first stroke, (2) relocation from the study area, (3) the end of the study; December 31, 2009 for cohort 1 and December 31, 2012 for cohort 2, or (4) death. For the analysis of stroke mortality, person-years were censored at the date of stroke death, emigration from Japan, or the end of the study (December 31, 2014 in both cohorts 1 and 2), whichever came first. For people who were lost to follow-up (n=168), the last confirmed date of their presence in the study areas was used as the date of censoring.
Mean values and proportions of participants’ characteristics were compared using the analysis of variance and the χ2 test. Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated using Cox proportional hazard regression analysis.
In the first model, we adjusted for age (continuous); in the second model, we further adjusted for the hypothesized confounding factors (prebaseline characteristics), and in the third model, we further adjusted for possible mediators (baseline characteristics); these characteristics might be considered consequences of changes in the employment status because such changes occurred before the baseline. Furthermore, to evaluate the effect of selection bias, we reanalyzed the data after running a propensity score matching for experiencing or not experiencing a spell of unemployment during the 5-year interval.23 Sensitivity analyses were done to guard against reverse causation by excluding early stroke incidence and mortality that occurred within 1 to 5 years of the baseline and to assess the effects of socioeconomic status by adjusting for education level for subjects in cohort 1 only, for whom information about educational level was available. All analyses were conducted using the SAS statistical package version 9.4 (SAS Institute Inc, Cary, NC).
Compared with those who were continuously employed, men and women who experienced at least 1 spell of unemployment were more likely to smoke, to consume more alcohol, to be hypertensive and diabetics, and to live alone; jobless men and women were more likely to smoke and to be hypertensive and diabetics; reemployed men, in contrary to reemployed women, were more likely to live alone with higher perceived stress, whereas continuously unemployed men and women were more likely to be hypertensive. All such differences were statistically significant (Table 1).
During the mean follow-up period of 15 years, 973 cases of newly diagnosed stroke cases (396 hemorrhagic and 577 ischemic) in men and 460 cases (133 hemorrhagic and 219 ischemic) in women were documented, and during 17 years of mortality follow-up, 275 stroke deaths (169 hemorrhagic and 49 ischemic) in men and 131 deaths (96 hemorrhagic and 16 ischemic) in women were documented (Table 2).
With reference to continuously employed subjects, the multivariable HR (95% CI) for total stroke incidence was 1.58 (1.18–2.13) in men and 1.51 (1.08–2.29) in women who lost their jobs, and that for mortality was 2.22 (1.34–3.68) in men and 2.48 (1.26–4.77) in women. In reemployed men, the multivariable HR (95% CI) was 2.96 (1.89–4.62) for stroke incidence and 4.21 (1.97–8.97) for mortality, whereas in women was 1.30 (0.98–1.69) for incidence and 1.28 (0.76–2.17) for mortality. In continuously unemployed men, the multivariable HR (95% CI) was 1.36 (0.73–2.55) for incidence and 5.24 (2.66–10.31) for mortality, whereas in women was 1.58 (0.99–2.49) for incidence and 5.35 (2.44–7.76) for mortality. (Table 2, model 2) The multivariable HR (95% CI) in men who experienced at least 1 spell of unemployment was 1.76 (1.39–2.23) for incidence and 3.00 (2.06–4.37) for mortality, whereas in women was 1.38 (1.11–1.72) for incidence and 1.98 (1.35–2.88) for mortality. (Table 3, model 2). The associations remained unchanged after further adjustment for possible mediating factors (model 3), after excluding early incident and mortality cases (Table I in the online-only Data Supplement), after adjusting for educational level as an indicator for the socioeconomic status (Table II in the online-only Data Supplement), and after running a propensity score weighting for having an unemployment experience (Table III in the online-only Data Supplement).
Having an experience of unemployment was associated with risk of hemorrhagic stroke; the multivariable HR (95% CI) in men was 1.77 (1.21–2.59) for incidence and 2.28 (1.32–3.94) for mortality, whereas in women was 1.28 (1.06–1.63) for incidence and 1.91 (1.22–2.99) for mortality. The multivariable HR (95% CI) of ischemic stroke in men was 1.75 (1.30–2.37) for incidence and 3.45 (1.57–7.58) for mortality, whereas in women was 1.33 (1.07–1.83) for incidence and 1.61 (1.02–8.80) for mortality (Table 3, model 2).
In this cohort of middle-aged Japanese men and women, experiencing at least 1 spell of unemployment was associated with increased risks of stroke morbidity and mortality. Compared with continuously employed subjects; job lost men and women had increased risks of developing and dying from both hemorrhagic and ischemic strokes. Reemployed men, but not women, also had increased risks of stroke. Continuously unemployed men and women showed higher risks of stroke mortality. Our findings were generally consistent with findings from previous studies4–6,9–11; however, only a Korean study investigated sex-specific association between unemployment with risk of stroke mortality and described an increased risk among men but not women.10
The postulated mechanisms by which unemployment associates with increased risk of stroke can be explained by the causation and the selection hypotheses.12 Mechanisms in the causation hypothesis include depression,3 financial strain,24 stress, social stigma leading to social isolation, and the loss of psychosocial assets such as marital status and work relationships.25 This may also trigger subsequent destructive coping strategies such as increased tobacco and alcohol use.26 Job lost men in our study showed unhealthy profiles of behavior and psychological state than continuously employed men.
On the other hand, the selection hypothesis is encountered when jobless or reemployed subjects had lost their previous jobs because of a preexisting ill-health. It is possible that a prestroke, presurvey poor health predicted both unemployment and stroke. High proportions of job lost subjects in our study were hypertensive or diabetics; however, these risk factors are mostly asymptomatic and do not act as barriers to work.27 Thus, the causative effect may have a greater influence than the selection effect, although they both interact in a mutually reinforcing way to impact health.14
Although, it was seen in previous Western studies that reemployment provided favorable conditions for health, such as reduced psychological distress,15,18 a reduction in unhealthy reactive behaviors such as heavy drinking,28 and alleviation of financial strain. However, we observed a greater increased risk in the reemployed men than that found in the job lost men. In our study, the risk of incident total stroke was significantly higher in reemployed men when the job loss group was used as a reference; HR (95% CI) was 1.82 (1.07–3.08; not shown in table). No such an increased risk was seen in reemployed women.
At least 4 explanations can be proposed to explain the increased risk in the reemployed men. First, reemployment may carry a double burden, one because of the previous job loss and the other because of strain to keep the newly attained job. Reemployed men are more likely to have greater job insecurity.29 Men who experience 1 spell of unemployment are at high risk of experiencing another.13,30 Where unemployment was associated with poor psychological health, this job insecurity for reemployed was more strongly associated with poor physical health.31 Reemployed men might put pressure on themselves to keep their job, such as being reluctant to take a sick day off32,33 or to visit a doctor because of limited access to health insurance.34 Second, the life-term employment system in Japan enforces male employees to devote themselves to a stable job, once entered the labor market. Contrary, when they lose the job, they are more likely to be reemployed in lower positions, which imply having poorer financial conditions35 and lower work-related benefits.29 After subjects’ right to unemployment benefit ran out, the rate of reemployment (especially in part-time jobs) increased significantly in Japanese subjects who had lost their jobs.18 Interestingly, among 20 254 continuously employed men who have reported whether they changed their job or not in this study, the multivariable HR (95% CI) for incident stroke in those who changed their job (n=1743) referenced to those kept the same job (n=18 511) was 1.30 (1.06–1.60). Third, the changes in the employment status in our study had occurred (between 1990 and 2000), a period known as the lost-decade, in which Japan experienced economic recession after the economic crash of 1990. The association between job loss and health tended to weaken in periods of recession as the general unemployment rate increased, suggesting that keeping a newly attained job during recession, than during less uncertain times, may infer much stress and might be more harmful to health than losing a job.13 Finally, the selection bias might be encountered in reemployed men; reemployed men might have lost their previous jobs because of previous illness. Poor health does not necessarily act as a barrier to the return to work27; thus, with the proposed stress, those unhealthy individuals returned to work were at greater risk of stroke morbidity and mortality. However, reemployed men in this study did not show higher prevalence of preexisting comorbidities than continuously employed men.
For reemployed women, it is plausible not seeing an increased risk as that in men because the pattern of Japanese women’s participation in the labor force by age group is represented by an M shape, reflecting their tendency to have a career break during their 30s, in response to their family responsibilities, and then return to the labor force in their 40s,36,37 which is different from men in whom the proportion of working age men without job is relatively low in Japan.37 Findings among women should be interpreted carefully because while 10% of women reported to be housewives and 3.5% reported to have no job in our baseline data; these 2 categories were probably mixed because our prebaseline data could not identify housewives from unemployed women. Thus, reemployed women are likely to restart their career after their career break, whereas reemployed men are likely to regain other jobs after their job loss. The meaning of reemployment could differ between men and women, which could generate sex differences in stroke risk.
The length of unemployment has been associated with stroke mortality7,12–14; the increased risk of stroke mortality in continuously unemployed subjects of this study was greater than that for incident stroke. Although we adjusted for history of hypertension and diabetes mellitus, other chronic morbidities or disabilities in the continuously unemployed subjects, such as renal or hepatic diseases, might have obliged them to stay jobless and might have put them at a higher risk of mortality.
This study is the first prospective study to provide evidence on the long-term effect of reemployment on risk of morbidity and mortality from stroke and stroke types and the first to use the propensity score for the probability of experiencing unemployment. However, the present analysis has several limitations. First, our variable of employment status could not differentiate between voluntary and involuntary job loss, including retirement. Findings from several previous studies showed that both voluntary and involuntary job loss were associated with a higher risk of stroke.4–6 However, to reduce this problem, we limited our population by age <60 years as early retirement is unusual in Japan. Also, we did not update the changes in the employment status using data collected 5 years after our baseline because any changes in the employment status after our baseline will be contaminated by the effect of losing a job because of retirement (31% of the studied participants). Second, the few cases of stroke events in the categories of reemployed and continuously unemployed subjects hindered us from examining the association of losing or gaining a job as separate exposure variables with risk of stroke types. Third, although the response rate to the prebaseline survey (81.6%) should be considered satisfactory, we do not know whom exactly the nonrespondents represented. What we found was that, 5% of the nonrespondents to our baseline survey (compared with 1.5% of the respondents) were unemployed at the time of the prebaseline survey. Last, our questionnaire did not give an indication of the employment status in the intervening 5 years between the prebaseline and the baseline surveys. To gain a complete picture of the effect of reemployment on health, it is necessary to follow the same people over consecutive employment transitions.
Middle-aged Japanese men and women who have experienced at least 1 spell of unemployment showed unhealthy profiles of behavior and psychological state and had an increased risk of stroke (both hemorrhagic and ischemic strokes) morbidity and mortality. Moreover, men, but not women, who have lost their jobs and then returned to the labor market were at a greater risk of stroke which may be attributed to greater job insecurity. Avoidance of unemployment and maintenance of job security during the productive ages could help reduce the stroke risk.
Dr Eshak participated in the study analysis and interpretation of data and drafting of the article and provided statistical expertise. Drs Honjo, Iso, Ikeda, Inoue, Sawada, and Tsugane participated in the study concept and design, acquisition of data and interpretation of data, and critical revision of the article.
Sources of Funding
This study was supported by National Cancer Center Research and Development Fund (23-A-31[toku] and 26-A-2) (since 2011) and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.016967/-/DC1.
- Received October 15, 2016.
- Revision received February 5, 2017.
- Accepted February 16, 2017.
- © 2017 American Heart Association, Inc.
- 1.↵Results of Labor Force Survey 1980–2013 [database online]. Tokyo, Statistical Bureau: Ministry of Internal Affairs and Communications; 2017. http://www.stat.go.jp/english/. Accessed January 31, 2017.
- Kawachi I,
- Berkman L
- Kasl SV,
- Jones BA
- Gallo WT,
- Teng HM,
- Falba TA,
- Kasl SV,
- Krumholz HM,
- Bradley EH
- Olesen K,
- Rugulies R,
- Rod NH,
- Bonde JP
- Noelke C,
- Avendano M
- Kang MY,
- Kim HR
- Franks PJ,
- Adamson C,
- Bulpitt PF,
- Bulpitt CJ
- Gebel M,
- Voßemer J
- Tsugane S,
- Sawada N
- Bartley M
- Waters LE
- Bonde JP
- Narendranathan W,
- Nickell S,
- Medcalf D
- Virtanen M,
- Kivimäki M,
- Elovainio M,
- Vahtera J,
- Ferrie JE
- Vahtera J,
- Kivimäki M,
- Pentti J,
- Linna A,
- Virtanen M,
- Virtanen P,
- et al
- Couch KA
- Tachibanaki T
- 37.↵Gender Equality Bureau Cabinet Office. The Present Status of Gender Equality and Measures. http://www.gender.go.jp/english_contents/about_danjo/whitepaper/plan2000/1999/p1c201.html. Accessed February 6, 2017.