Combined use of job stress models and self-rated health in nursing

OBJECTIVE: To identify combinations of two models of psychosocial stress at work among nursing teams and their associations with self-rated health. METHODS: This was a cross-sectional study among workers at three public hospitals in the municipality of Rio de Janeiro, Southeastern Brazil (N = 1307). In 2006, a multidimensional questionnaire including two scales for measuring stress at work (demand-control and effort-reward imbalance models) was administered. Partial and complete (including social support at work) demand-control models were considered, along with partial and complete (including excessive commitment to work) effort-reward models. Multiple logistic regression models were used to estimate adjusted odds ratios and their respective 95% confi dence intervals. RESULTS: The dimensions of both models were independently associated with self-rated health, with odds ratios between 1.70 and 3.37. The partial demandcontrol model was less associated with health (OR = 1.79; 95%CI 1.26;2.53) than was the partial effort-reward imbalance model (OR = 2.27; 95%CI 1.57;3.30). Incorporation of social support and excessive commitment to work increased the strength of the demand-control and effort-reward imbalance models, respectively. Increased strength of association was observed when the two partial models were combined. CONCLUSIONS: The results indicate that the effort-reward imbalance model performed better for this specifi c group and for the outcome evaluated, and that there was an advantage in using complete models or combinations of partial models. DESCRIPTORS: Nurses. Burnout, Professional. Working Conditions. Job Satisfaction. Cross-Sectional Studies.


INTRODUCTION
Interactions between different working conditions provide work environments that are more or less favorable to health.Assessment of these factors and their infl uence on workers' health has advanced considerably, but measuring them remains a challenge for social epidemiology. 4,22These measurements are often based on theoretical constructs that generate models that are tested in a variety of empirical fi elds. 4Two of these models are believed to have greater explanatory power and are widely used in the worldwide literature for measuring the psychosocial working environment: the demand-control (DC) model and the effort-reward imbalance (ERI) model.These models defi ne different stress factors in work that are potentially damaging to health, and they provide explanations regarding the relationship between stressful working conditions and physical and psychological wellbeing. 13,21he DC model, which was devised by Karasek, 13 originally considered interactions between two components that might promote "wear-and-tear" at work ("job strain"): psychological demands (pace and intensity of the work) and control (autonomy and skills relating to the process, required from the worker).Activities that involve high psychological demands and low control would promote job strain and consequently lead to physical and psychological illness.Subsequently, the model evolved to include a third dimension: the perception of social support for work elaborated by Johnson & Hall. 12According to these authors, social integration, trust within the group and help from colleagues and superiors in accomplishing task might act as protective (moderating) factors in relation to the effects of job strain on health.Two different instruments have been used in Brazil to measure the model, in translations with adaptations for Brazilian Portuguese: the "Job Content Questionnaire" 6 and the reduced scale known as the "Swedish Scale for Demand-Control-Social Support" (DCS), 2 which was developed by Theorell. 25e ERI model, which was developed in the mid-1990s, 21 makes the assumption that imbalances between excess effort at work and low reward or recognition of this effort generate stressful situations. 21The effort relates to the demands and obligations perceived by workers; the reward is composed of fi nancial recompense (adequate salary), self-esteem (respect and support from colleagues and superiors) and occupational status (promotion prospects, employment stability and social status). 23ust as in the DC model, a third dimension was incorporated into the original model: excess commitment to work.This is considered to be a dimension intrinsic to the worker, expressed through excessive endeavor at work combined with a strong desire for approval and esteem.Excess commitment to work is considered to be a factor that may interact with ERI to boost its harmful effects on health and wellbeing. 7,22th models have limitations with regard to capturing the complexity of the psychosocial aspects of work.Moreover, partial models focusing exclusively on the relationships between demand and control and between effort and reward 18,20 do not explain the contributions of social support and excess commitment to work.Recent studies have recommended that the two models should be used in combination, in relation to a given outcome. 4Studies on occupations have demonstrated that the predictive power of the combination (in comparison with the result relating to each model) is greater with regard to self-reported health, 3,18 insomnia, 19 mental stress at work, 4 acute myocardial infarction, 20 absenteeism due to illness 1,9 and depression. 26No investigations on the effect of the combined models on health-related outcomes in Brazil were identifi ed.
The association between psychosocial factors within nursing work and self-reported health is considered to be a summary measurement of health and an important mortality predictor. 10The objective of the present study was to identify combinations of two models for psychological stress at work among nursing teams, and their association with self-reported health.

METHODS
This was a cross-sectional study among female nursing providers at three public hospitals in the municipality of Rio de Janeiro, Southeastern Brazil, in 2006.Out of 1,595 eligible workers, 1,307 (81.9%) took part in the study.
A self-applicable multidimensional questionnaire that had been refi ned through three rounds of pretesting (n = 50) and tested in a pilot study (n = 80) was administered by a trained team during working hours.
The outcome (self-reported health status) was obtained from the question: "In a general manner, compared with other people of your age, what would you say your own state of health was like?"The responses were dichotomized into two groups: good (very good/good) and poor (regular/poor).
The exposure variable DC was obtained using the Brazilian version of DCS. 2 This instrument was composed of two dimensions: psychological demands (fi ve questions) and control (six items).One item relating to control was removed ("In your work, do you often have to repeat the same tasks?")because of its poor psychometric performance in the context of the population investigated. 8The dimensions presented items with four response categories (from "frequently" to "never/almost never").The scores for each dimension separately, along with the ratio between the scores (D/C), were categorized in terciles.The highest tercile represented the highest level of psychosocial stress at work, according to the DC model.The fi rst (lowest) tercile, i.e. low psychosocial stress, was used as the reference category.Division into terciles was used as already done in other studies, 16 which also made this model comparable to ERI, which is often categorized in this manner. 4,7,22The dimension of social support at work was also evaluated in the model, with four response categories (from "totally agree" to "totally disagree").The score obtained (ranging from 6 to 24) was categorized into terciles, and the highest tercile (high social support) was used as a reference category.
The following covariables were taken into consideration: sociodemographic data (age, conjugal situation and schooling level), occupational data (professional nursing category, length of time in this activity, type of contract and number of jobs) and behavioral data and healthrelated problems (smoking, use of alcoholic drinks, physical activity practice and body mass index).
The ERI model was measured using the Brazilian version 5 of the questionnaire designed by Siegrist, 21 which presented adequate psychometric performance for the population of the present study. 9The effort dimension contained six items and the reward dimension was composed of 11 items: esteem (fi ve items), promotion prospects and salary (four items) and secureness at work (two items).The response categories for both dimensions were defi ned in two stages.The participants agreed or disagreed with statements relating to work situations and indicated the degree of stress attributed to that situation (from "not stressful" to "very stressful". For each participant, a ratio was constructed using the formula: e/(r*c), in which "e" was the score obtained from the effort questions, "r" was the score obtained from the sum of the reward questions and "c" was a correction factor (0.545454), considering the number of items in the numerator compared with the number in the denominator (6/11). 21,22The score for each dimension (effort and reward) and the ratio between them were categorized into terciles. 22The fi rst tercile (low stress level) was used as the reference category.
The dimension of excess commitment to work was composed of six items in which the workers indicated their degree of agreement with the statements (from "strongly disagree" to "strongly agree"), on a four-point scale.The score obtained was categorized into terciles and the fi rst tercile (low risk) was considered to be the reference category. 22e partial DC model (demand and control) and the complete DC model which added social support at work were both taken into consideration, as were the partial ERI (effort and reward) and the complete ERI, which added excess commitment to work to the partial model.
The association between the categories of psychosocial stress at work and self-reported health was obtained through descriptive statistics.Covariables in other studies 15,16,19,20,22,26 that were associated with the outcome and with the exposure variables, with a signifi cance level of 10% in bivariate analysis, were considered to be potential confounding variables and were included in the logistic regression models.SPSS v.14 software was used in the analyses.
Multiple logistic regression models were estimated in three stages: the association with the outcome shown by each dimension of the models, i.e. psychological demands, control and social support at work (DC) and effort, reward and excess commitment to work (ERI); the association with the outcome shown by each partial model (DC and ERI); the association with the outcome shown by each complete model (DC/social support and ERI/ excess commitment); and the association with the outcome shown by the combination of the partial models (DC and ERI).For each model/dimension combination, the workers were categorized into four groups.The group that was not exposed in any model/ dimension was taken to be the reference category.The model/dimension values (described earlier) were dichotomized according to terciles (fi rst and second terciles taken to indicate lack of exposure, and third tercile taken to be the presence of exposure), to construct categories.
This project was approved by the Research Ethics Committees of the hospitals involved, and by CONEP since foreign cooperation was involved (Procedural no.1318/2004).

RESULTS
The participants' mean age was 40 years (SD 12.8 years); 57.2% had had university-level education and 27.9% were nurses.Around one third had more than one nursing job and almost half were federal public employees.With regard to health conditions, 26.3% said that they had hypertension, 44.6% were classifi ed as obese or overweight, 13.2% were smokers and more than two thirds said that they did not do physical activities.More than 80% of the interviewees said that their health was "good/regular" (Table 1).Non-completion and inconsistencies in fi lling out the questionnaire were identifi ed in the proportions of 8,3% (109), 4,3% (56), 1,4% (19) e 0,08% (11) in relation to the ERI model, the DC model, the dimension of social support at work and the dimension of excess commitment to work, respectively (data not shown in tables).High psychological demand and low social support at work were associated with low self-reported perceptions of health in the DC model.Weaker associations were observed with job control.There were associations with higher levels of excess commitment to work, followed by lower levels of reward and higher levels of effort (ERI model).In both models, there was a clear doseresponse gradient, even after adjusting for confounding variables.Stronger associations were observed as the exposure levels increased (Table 2).
In the partial models, higher levels of psychosocial stress were associated with poor health and presented higher odds ratios.Association measurements with higher values were observed in the ERI model.Both models gained additional strength of association through inclusion of the additional dimensions (social support at work in the DC model and excess commitment to work in ERI) (Table 3).
Combination of the two partial models led to a stronger association.The odds of reporting poor health were 1.60 times higher (95% CI: 1.77;3.83)among individuals classifi ed as presenting stress in both models, compared with the value observed among individuals who were not classifi ed in any of the models.In the case of stress classifi ed only in the DC model, the odds ratio was 1.55 (95% CI: 1.00;2.41),and in the ERI model alone, the odds ratio was 1.67 (95% CI: 1.08;2.59)(Table 4).

DISCUSSION
This study made it possible to identify different dimensions of psychosocial stress at work among nursing teams and their association with self-reported health, along with an estimate for the increase in the strength of association achieved by combining the scales.It was found that all of the separate dimensions of the DC model (demand, control and social support) and ERI model (effort, reward and excess commitment to work) were associated with self-reported health.The partial DC and ERI models were associated with self-reported health, with a stronger association observed with ERI.Incorporation of social support and excess commitment to work increased the strength of association in the DC and ERI models, respectively.Combination of the partial DC and ERI models increased the strength of the association.Social support and excess commitment to work were shown to be important in relation to the outcome analyzed, and the latter was more strongly associated with self-reported health.The partial models were suffi cient to detect the association with the outcome, but incorporation of these other dimensions contributed towards increasing the strength of association.For DC, this could be attributed to a limitation of the partial model, which did not consider relationships between people, 24 which are inherent to nursing work.Some authors have affi rmed the importance of excess commitment to work in the partial ERI model, 7,22 although others have disagreed. 18Critics have argued that excess commitment to work is a measurement of a subjective nature: an intrinsic component relating to individuals' capacity to face up to work demands.They also considered that the outcome evaluated was subjective, since it referred to workers' own assessments of their state of health.Part of this result may have been infl uenced by common method variance, in which the independent and dependent variables were hard to distinguish. 18In fact, a similar analysis with this same group of workers did not identify any association between excess commitment and absenteeism. 9Studies that evaluate the predictive capacity of incorporating excess commitment to work, into ERI in relation to different types of outcome, are recommended.
ERI was shown to be more strongly associated with health than was DC, in both the partial and the complete formulations.Greater explanatory power regarding the infl uence of psychosocial stress from work on health has been attributed to ERI. 4 Part of the better performance results from ERI can be attributed to the constructs of the dimensions that make up the models.For example, for certain occupations, reward may have a more important meaning than does control in the work process, within the context of the present-day world of work.In turn, control has been emphasized as an important dimension, particularly in studies evaluating cardiovascular outcomes. 3,18As in our results, another study among healthcare professionals 7 showed that the control dimension had lower predictive power in relation to workers' wellbeing, compared with the other dimensions.In our study, this fi nding may have been partially attributable to problems regarding the psychometric performance of this dimension, as reported in a recent paper. 8e results from the present study seem to corroborate investigations that have indicated that the two partial models are complementary in nature, with regard to associations with health outcomes, 3,4,18,20,26 in that the models emphasize different aspects of the working environment relating to stress and becoming ill.While DC considers specifi c characteristics of work tasks, ERI is based on reciprocity between the effort expended on tasks and the recompense in terms of salary, secureness of work, prospects and esteem.
The DC-ERI combination may reduce the limitations inherent to each model.In relation to DC, its applicability to certain occupations, especially those dealing with people in service sectors, like healthcare workers, has been questioned. 4,24This model was devised in the 1970s, within the context of industrial organizations, at a time of greater employment stability in developed countries, and for this reason, its emphasis is on the characteristics of work tasks.Thus, it does not take into consideration the recent changes in the nature of work relating to the global economy, which have added other factors such as precariousness (insecurity at work), rapid organizational changes and few prospects of promotion.Recognizing these limitations, the revised model of the Job Content Questionnaire 2.0 is currently undergoing tests in other countries. 14,17 the case of ERI, although it considers aspects of work within the context of a global economy, the questionnaire has a limited number of questions on social relationships within the working environment.Moreover, the score calculation favors classifi cation errors and lack of completion among the subjects (around 8% in the present study).This scale was recently revised by Siegrist et al, 23 who put forward a shorter version with a simpler format for obtaining the scores (from "strongly disagree" to "strongly agree").
Certain limitations can be highlighted in the present study.Although relatively large, the sample was restricted to a female group with a specifi c occupation in public institutions.Therefore, the ability to generalize from these fi ndings is limited.The data were crosssectional, which does not allow a time relationship between the events studied to be established.Reverse causality cannot be dismissed, since workers with worse health levels may have overestimated the psychosocial stress at work.In addition, studies based on selfreported measurements may be infl uenced by factors such as memory bias, socially desirable responses and diagnostic suspicion bias.The multiplicity of ways of categorizing the model scores makes comparisons between studies diffi cult, since there is no consensus about the right way to make them. 11Some authors have used the dimensions as continuous variables, or have categorized the D/C score ratio into terciles 4,15 or quartiles, 20 or dichotomized them using medians of the distribution. 3,18In ERI, the scores for the effort-reward ratio have been categorized into terciles 4 or used as a cutoff point for values greater than one. 20 conclusion, the results indicate that there was better performance from the ERI model for this specifi c group and for the outcome assessed, and that complete models or combination of the partial models were advantageous.

Table 1 .
Sociodemographic factors, work characteristics and health characteristics of the nursing workers.Rio de Janeiro,

Table 2 .
Self-reported poor health and crude and adjusted odds ratios, according to the dimensions of the demand-control and effort-reward imbalance models, among nursing workers.Rio de Janeiro, Southeastern Brazil, 2006.
a Adjusted for age, schooling level, type of work contract and number of jobs b Chi-square for linear association

Table 3 .
Crude and adjusted odds ratios for the association between the partial and complete demand-control and effort-reward imbalance models, in relation to self-reported poor health among nursing workers.Rio de Janeiro, Southeastern Brazil, 2006.

Table 4 .
Crude and adjusted odds ratios for the association between the combination of the demand-control and effort-reward imbalanced models and self-reported poor health among nursing workers.Rio de Janeiro, Southeastern Brazil, 2006.adjusted for age, schooling level, type of work contract, number of jobs, physical activity practice, smoking and consumption of alcoholic drinks.DC: demand-control model; ERI: effort-reward imbalance model a