Stressful working conditions and poor self-rated health among ﬁ nancial services employees

OBJECTIVE: To assess the association between exposure to adverse psychosocial working conditions and poor self-rated health among bank employees. METHODS: A cross-sectional study including a sample of 2,054 employees of a government bank was conducted in 2008. Self-rated health was assessed by a single question: “In general, would you say your health is (…).” Exposure to adverse psychosocial working conditions was evaluated by the effort-reward imbalance model and the demand-control model. Information on other independent variables was obtained through a self-administered semi-structured questionnaire. A multiple logistic regression analysis was performed and odds ratio calculated to assess independent associations between adverse psychosocial working conditions and poor self-rated health. RESULTS: The overall prevalence of poor self-rated health was 9%, with no signi ﬁ cant gender difference. Exposure to high demand and low control environment at work was associated with poor self-rated health. Employees with high effort-reward imbalance and overcommitment also reported poor self-rated health, with a dose-response relationship. Social support at work was inversely related to poor self-rated health, with a dose-response relationship. CONCLUSIONS: Exposure to adverse psychosocial work factors assessed based on the effort-reward imbalance model and the demand-control model is independently associated with poor self-rated health among the workers studied. DESCRIPTORS: Workload.

Self-rated health is a major robust indicator of overall health among general 15 and working populations. 11,19espite its subjective nature, longitudinal studies showed that poor self-rated health independently predicts the occurrence of future health events, including hospitalization and death, after adjusting for health variables and socioeconomic conditions. 8,9lso, it has been validated and widely used in different countries and populations. 19pulation studies show that self-rated health is infl uenced by demographic, socioeconomic, cultural, lifestyle, and work environment factors and health conditions such as chronic diseases and work-related health conditions. 11They found greater prevalence of poor self-rated health among subjects exposed to precarious working conditions 5 and adverse psychosocial conditions at work. 28e present study aimed to assess the association between poor self-rated health and exposure to adverse psychosocial working conditions among Brazilian fi nancial services employees.

METHODS
All employees of a large Brazilian government bank working at branches in all state capitals and the Federal District at the end of 2007 -a total of 40,005 workers -were eligible to participate in the study.A crosssectional study was conducted in a random sample of 2,500 workers stratifi ed by gender (1,250 males and 1,250 females).Details about the population and sampling method have been published elsewhere. 31self-administered questionnaire was used covering fi ve areas of interest: sociodemographic, behavioral, health, psychosocial and work-related factors.The sociodemographic characteristics included gender, age, marital status, education, race/skin color, having children, household ownership, and being the head of the household.
The behaviors assessed were smoking, alcohol use, and physical activity.Regarding health, the questionnaire collected information about physician-diagnosed chronic diseases (hypertension, diabetes, asthma, bronchitis, myocardial infarction, stroke, and musculoskeletal disorders), hospitalizations during the preceding 12 months, sleep problems, and regular use of medications.General psychosocial factors included exposure to stressful situations and experience of any form of discrimination.
Work-related factors included length of service in the company, current job, and psychosocial characterization of the job, which is the exposure of interest in the current study.All covariates described above were considered as potential confounders in the assessment of the relationship between exposure to adverse psychosocial working conditions and poor self-rated health.
The independent associations with the outcome were assessed using the demand-control model 13 and effortreward imbalance (ERI) model. 26To assess adverse psychosocial working conditions we used the short version of the Job Content Questionnaire (JCQ) adapted to the Brazilian Portuguese by Araújo et al (2003)  1 and the Brazilian Portuguese version of the Effort-Reward Imbalance Scale. 30e JCQ consists of 22 questions with Likert-type responses that range from "strongly agree" to "strongly disagree."The responses were coded according to the JCQ User's Guide. 14Based on the assumptions of Karasek's model, variables were dichotomized at their median and grouped into four distinct categories.Employees whose work was performed under conditions of exposure to a high demand and low control (high strain) combination were included in the group with the greatest exposure to stressful conditions.Employees whose work exposed them to high demand but high control (active work) were included in the intermediate exposure group.Workers exposed to low control and low demand (passive work) were also included in the intermediate exposure group.Employees who worked under low demand and high control (low strain) were regarded as not exposed to stress and were the reference category in the statistical analysis.Cronbach's alpha values for the demandcontrol scale were 0.80, 0.86 and 0.90 for demand, control and social support at work, respectively.
The ERI scale has three subscales: effort (6 items), reward (11 items) and overcommitment (6 items).The fi rst two subscales comprise questions with Likerttype responses with varying degrees of agreement or disagreement, and scores ranging from 1 to 5. The overcommitment subscale also has Likert-type responses ranging between agree and strongly disagree, with scores from 1 to 4. Using the theoretical assumptions of the model, a work-related stress index for the ERI model was constructed using cutoffs based on the tertiles of the distribution. 12The highest one was the most exposed group, the second the intermediate exposure, and the fi rst tertile were those with the lowest exposure and the reference category in the analysis.Overcommitment was similarly categorized into tertiles.Cronbach's alpha values for the ERI scales were 0.82, 0.80 and 0.85 for effort, reward and overcommitment, respectively.
Self-rated health, the dependent variable in this analysis, was obtained by answering a single question: "In general, compared to people of your age, would you say your health is (…)," with fi ve response options (excellent, very good, good, poor, and very poor).For the analysis, responses were grouped into "good" and "poor.""Good" groups together responses "excellent," "very good," and "good" while "poor" aggregates responses "poor" and "very poor." The magnitude of the statistical association between poor self-rated health (poor/very poor) and psychosocial working conditions was determined by odds ratio (OR) obtained by multiple logistic regression, with a 95% confi dence interval.Different weights were assigned to male and female participants refl ecting the difference in the probability of each gender participating in the study, as the eligible population comprised 61.1% males and 38.9% females, and the sample 50% of each gender.The analysis was performed using Stata Statistical Package version 9.2.
All variables associated with poor self-rated health in the univariate analysis at p<0.20 were included in the multivariate analysis carried out in two stages.First, the sociodemographic variables associated with poor self-rated health were adjusted for other covariates in this domain.Next, the variables in the behavioral domain were added to the model, then the variables of the health domain, followed by psychosocial exposures.Finally, work-related variables were added, including exposure to adverse psychosocial working conditions.This fi nal exposure variable was assessed separately, with two models, one based on the demand-control scale and the other based on the ERI scale.All variables that remained associated with poor self-rated health at p<0.05 were retained in the fi nal models.
The study was approved by the Ethics Research Committee of the Federal University of Minas Gerais (process nr.350/07, Aug 2007) and all participants signed an informed consent form.

RESULTS
Of the 2,500 workers enrolled to participate in the study, 163 were ineligible because they were retired, on medical leave, or their employment contract was temporarily discontinued.There were no statistically signifi cant differences between these 163 ineligible workers and those eligible to participate in the study with regard to gender, age, marital status, schooling, or length of services in the company.Of the 2,337 eligible, 2,054 (88%) were included in the study, being 49.7% male and 50.3% female.The mean age was 40 years (SD = 9.17), and the average length of service in the company was 15 years (SD = 9.64).The majority were married, college-educated, self-referred as white, and non-smokers.The most commonly reported diseases were hypertension (28%), bronchitis (23%), and workrelated musculoskeletal disorders (23%).
In the univariate analysis, age over 40 years, being a smoker or former smoker, physical inactivity, moderate alcohol consumption, sleep problems, use of medications, or having one or more chronic diseases were each signifi cantly associated with poor self-rated health (Table 1).
Working at the company for more than fi ve years, as well as exposure to adverse psychosocial working conditions, assessed by the demand-control and ERI models, were also signifi cantly associated with poor self-rated health (Table 2).Tables 3 and 4 present models with progressive adjustments for the variables identifi ed in the univariate analysis as associated with poor self-rated health at p<0.20, for the demand-control and ERI models, respectively.After adjustments for the variables from all domains (model 4, Table 3), older age, physical inactivity, presence of one or more chronic diseases, sleep problems, regular use of medications, length of service in the company between six and 14 years, as well as exposure to high strain work, and lack of social support at work remained statistically associated with poor self-rated health.Cigarette smoking and moderate use of alcohol did not remain statistically associated with poor self-rated health.The fi nal model showed that after adjusting for all confounders high strain working conditions was statistically associated with poor self-rated health.High social support at work reduced by about 4-fold the likelihood of reporting poor self-rated health.
Table 4 shows that older age, physical inactivity, smoking, presence of chronic diseases, sleep problems and regular use of medications were signifi cantly associated with poor self-rated health in the fi nal model.This model shows that the exposure to effort-reward imbalance as well as being overcommitted to work increased the likelihood of reporting poor self-rated health, and these associations were independent of all confounding variables included in the analysis.

DISCUSSION
The results of this study further support the relationship between exposure to adverse psychosocial work environment and poor self-rated health.This effect was seen in both demand-control and ERI models.In the demand-control model, it was stronger among workers exposed to high strain and low control and those who lacked social support at work.In the ERI model, high effort-reward imbalance and high level of commitment at work increased the likelihood of poor self-rated health.To our knowledge, this is the fi rst study to investigate the association between exposure to adverse psychosocial environment and poor self-rated health among fi nancial services employees.
The overall prevalence of self-rated health in the present study was greater than that found by Barros  et al (2006), 2 using a similar question, among active adults living in Brazilian state capitals.Dachs & Santos  (2006)  4 analyzed data from the Brazilian National Household Survey and also found a lower prevalence of poor self-rated health among Brazilians aged 15 and older (6%).However, Giatti et al (2010) 6 found a much higher prevalence of poor self-rated health (20%) among adults aged 15 to 64 years living in four Brazilian cities included in the National Household Survey on Risk Behaviors and Reported Morbidity from Non-Communicable Disease.Nonetheless, considering that we studied a sample of relatively more educated, active young individuals, the observed prevalence of poor self-rated health is quite high.This may be explained by the fact that fi nancial service employees experience high levels of psychological stress, 31 and have high rates of absenteeism, especially due to workrelated musculoskeletal and mental disorders. 29ternational studies 20 have shown that the prevalence of poor self-rated health increases with age.This fi nding is consistent with that found in the present study that a higher proportion of individuals aged 40 and above rated their health status as poor.
Considering the adjusted models using both the demand-control and the ERI models age, physical inactivity, presence of one or more chronic conditions, sleep problems, and regular use of medications were statistically associated with poor self-rated health.Having one or more chronic diseases, regular medication use, and sleep problems were positively associated with poor self-rated health according to fi ndings of other studies. 3,9,10Medication use -inasmuch as it suggests the presence and severity of a condition-would be expected to be associated with poorer self-rated health.
Physical inactivity was positively associated with poor self-rated health in the demand-control analysis, corroborating international evidence. 21,32Smoking, on the other hand, was only associated with poor self-rated health in the ERI model, which is a fi nding that does not agree with that reported by Kouvonen et al (2005)  16 that found smoking related to stress when evaluated using both models.Physical inactivity and smoking are behavioral factors associated with many physical and mental health problems.Both are also related in different ways to stress.The reason why in the present study each of them remained statistically signifi cant only in one of the fi nal models may be lack of power or related to the constructs of the demand-control and ERI models as the other confounding variables in these models are the same.For instance, only a weak association was found between exposure to high occupational effort and low reward and sedentary lifestyle. 17nsidering the demand-control model, the odds that workers exposed to a hostile work environment (high demand and low control) reported poor self-rated health was about six (unadjusted) and two times (after adjustment) higher than that of unexposed workers.
Social support, as proposed in Karasek's model (1979), 13 showed an inverse association, reducing by half the odds of poor self-rated health.These results are corroborated by fi ndings of prospective, cross-sectional studies.Niedhammer et al (2003) 22 found a similar relationship regardless of potential confounders in a longitudinal analysis.Among individuals who worked in an unfavorable environment, Borg et al (2000) 3 found a two-fold higher prevalence rate of poor selfrated health.Molarius et al (2006)  21 found that workers exposed to high demand and low control also have a higher prevalence of poor self-rated health.In this study and that of Hasson et al (2006)  9 lack of social support at work also had a harmful effect on perceived health, doubling the odds of poor self-rated health.
This study also showed similar results in the analysis of adverse psychosocial exposure using the ERI model.Workers exposed to high effort-reward imbalance were about seven (unadjusted) and three times (after adjustment) more likely to self-rate their health as poor.These fi ndings are also consistent with those of prospective studies in other countries.Krause et al (2010) 18 reported poorer general health in workers exposed to a high effort-reward imbalance.Siegrist et al ( 2004) 27 found that the odds ratios for poor self-rated health was signifi cantly high among workers exposed to a high effort-reward imbalance in 14 of 16 studies in European countries.Pikhart et al (2001)  25 also reported that an effort-reward imbalance is a powerful determinant of poor self-perceived health, with increasing rates among most exposed workers.The same phenomenon was reported in Weyers et al study (2006). 33ercommitment increased the odds of poor self-rated health by 2.7 for workers in the highest tertile.Weyers et al (2006) 33 found a similar association with high level of job commitment.Niedhammer et al (2004) 23 also reported a similar fi nding in their cross-sectional study, but in the prospective component of the study the association between overcommitment and poorer self-rated health was seen among males only.Job overcommitment is a subjective measure, an intrinsic component, related to the individual's ability to cope with job demands (Siegrist 1996).As self-assessment of health is also very subjective, it is possible that this result is infl uenced by an intrinsic relationship of outcome and exposure in the study. 7 the present study we used two models because several studies have pointed to their complementarity as they emphasize different aspects of the work environment.Slightly stronger associations were found using the ERI model.This seems to be in accordance with de Jonge et al (2000) 12 fi ndings that high efforts and low occupational rewards were stronger predictors of poor well-being than low job control when both job stress models were simultaneously adjusted.Some authors  Notwithstanding the consistent relationships found in this study between an adverse psychosocial work environment and poor self-rated health, some limitations should be considered when interpreting the results.Workers absent for medical reasons did not participate in the study, which contributed to underestimating the prevalence of poor self-rated health and reduced the magnitude of the associations found with adverse psychosocial environment.The use of a self-administered questionnaire has disadvantages and advantages.On one hand, it made a national study feasible, increasing the representativeness of the study.On the other, it is possible that diffi culty in understanding the questionnaire might have affected the quality of responses.Considering that the average schooling level of the participants is high, it is unlikely that respondents had any major diffi culty with the questionnaire.Finally, owing to the cross-sectional design, we cannot exclude a reverse causality or make causal inferences.Despite these limitations, it is important to emphasize that the study included a signifi cant sample of bank employees and achieved a high response rate (88%).Moreover, the fi ndings are consistent with several international studies, both cross-sectional and prospective, as discussed before.
In conclusion, the present study shows that the prevalence of poor self-rated health is high among bank workers.These results corroborate the fi ndings of several cross-sectional and prospective studies about an independent association between adverse psychosocial work conditions and poor self-rated health.Finally, the study shows that these associations are present in the analyses based on the demand-control model as well as the ERI model.
Further investigation, especially with different study designs such as prospective studies, involving other fi nancial services workers is needed to clarify the relationships found here.

Table 1 .
Prevalence and odds ratio of poor self-rated health among bank employees according to sociodemographic, lifestyle, and health factors, and exposure to stressful conditions and discrimination.Brazil, 2008.(N=2054)

Table 2 .
Prevalence and odds ratio of poor self-rated health among bank employees according to job characteristics and psychosocial working conditions.Brazil, 2008.(N=2054)

Table 3 .
Factors associated with poor self-rated health among bank employees including job characteristics defi ned in the demand-control model.Brazil, 2008.(n=2054) a Adjusted for sociodemographic variables; b Adjusted for sociodemographic and behavioral variables; c Adjusted for sociodemographic, behavioral and health variables; d Adjusted for sociodemographic, behavioral, health, and working variables; e Final model

Table 4 .
Factors associated with poor self-rated health among bank employees including job characteristics defi ned in the effort-reward imbalance model.Brazil, 2008.(n=2054) Adjusted for sociodemographic, behavioral and health variables; d Adjusted for sociodemographic, behavioral, health, and working variables; a Adjusted for sociodemographic variables; b Adjusted for sociodemographic and behavioral variables; c e Final model