2012 Volume 54 Issue 6 Pages 427-433
Objectives: Work stress is an emergent risk in occupational health in China, and its measurement is still a critical issue. The aim of this study was to examine the reliability and validity of a short version of the effort-reward imbalance (ERI) questionnaire in a sample of Chinese workers. Methods: A community-based survey was conducted in 1,916 subjects aged 30–65 years with paid employment (971 men and 945 women). Results: Acceptable internal consistencies of the three scales, effort, reward and overcommitment, were obtained. Confirmatory factor analysis showed a good model fit of the data with the theoretical structure (goodness-of-fit index = 0.95). Evidence of criterion validity was demonstrated, as all three scales were independently associated with elevated odds ratios of both poor physical and mental health. Conclusions: Based on the findings of our study, this short version of the ERI questionnaire is considered to be a reliable and valid tool for measuring psychosocial work environment in Chinese working populations.
Psychosocial stress at work has been found to predict a range of adverse outcomes, including impaired health among workers1) and poor organizational outcomes, such as reduced productivity and increased sickness absence2). Numerous conceptualizations of works stress have been developed. One of the most widely adopted and validated measures is based on the effort-reward imbalance (ERI) model. This model postulates that perceptions of high effort paralleled by low reward (e.g., salary, promotion, prospects and job security) induce feelings of stress. An additional intrinsic factor, termed overcommitment, is hypothesized to exert independent health effects but also to amplify the adverse effects of effort-reward imbalance3).
Most of the current body of work stress research stems from Western countries. Work stress is, however, not restricted to industrialized societies4) but represents an emerging public health concern in today's globalized economy, including in rapidly developing economies such as China5). Since 1990, dramatic social and economic changes have occurred in China, with rapid industrialization occurring that creates substantial challenges in terms of its impact on many aspects of employment and people’ well-being6). Much evidence has demonstrated that rapid social and economic changes are strongly associated with an increase in mental disorders7). Epidemiologic surveys in China over the last five decades (1960–2010) indicated the prevalence of mental disorders was markedly low (overall <3.2%) prior to 1990; however, the data after 2000 revealed a much higher rate (i.e., lifetime prevalence of any mental disorder of 21.9%), particularly for depression and anxiety disorders. In addition, China also has the largest number of reported suicides in the world8).
By the end of 2011, the economically active population in China exceeded 1 billion, accounting for one-seventh of the total population globally. A fifth of this population was found to suffer from mental disorders6,9). A large-scale survey in China showed 36.8% adults were severely stressed10), in line with our recent epidemiologic study among the general working population, and 32.4% of workers reported they had high or very high work stress5). To date, several laws have come into force concerning occupational safety and health in China6), such as the Law on the Prevention and Treatment of Occupational Diseases (since May 1, 2002), the Law on Work Safety (since November 1, 2002) and the Law on Employment Contracts (since January 1, 2008). Furthermore, basic occupational health services have been implemented in China “to prevent, control, and manage the harmful factors and harmful conditions arising from or existing in the workplace, as well as the risks endangering the health and safety of workers”9).
Given China's rapid and uneven transition from a state-planning-oriented economy to a market-driven economy, working life is accordingly characterized by high flexibility and mobility, by fixed contracts and by increased job insecurity for Chinese workers6,9). As several of these conditions are included in the ERI model, application of this model to the Chinese working population has been well justified11). Our previous research in China has identified effort-reward imbalance at work as one significant risk of cardiovascular diseases12,13), musculoskeletal disorders14), reproductive hazards15), mental distress16), impaired health functioning17), and professional turnover18).
The original ERI questionnaire is based on 23 items19). However, to reduce survey administration time in occupational settings for large-scale epidemiologic investigations, a shorter version of this questionnaire was recently developed in European studies20–22). Therefore, we set out to test the psychometric properties of the short measure of the 16-item ERI questionnaire in a Chinese sample.
We conducted a survey in Kunming, a city with a population of 6.4 millions people located in Southwestern China. Residents aged 30–65 years with paid employment were invited to participate. Participants in the survey were selected using a multistage random sampling process. First, 15 communities were randomly selected. In each community, 150–200 subjects were planned to be included. Second, the first households were selected through random sampling of residential address numbers, and all household members present at the residence who met the inclusion criteria received a self-administered questionnaire. Third, neighbors living in a residence next door who met the inclusion criteria were also recruited as far as necessary to meet the fixed sample size. A total of 2,498 questionnaires were sent out, and 2,178 were returned (response rate 87%). Subjects with missing values on the ERI questionnaire were excluded (n=262), leaving a sample of 1,916 subjects. The excluded subjects did not differ significantly from the remaining participants with regard to sociodemographic or work-related characteristics. This study protocol was approved by the research ethics committee of Kunming Medical University.
MeasurementOur short measure of the ERI questionnaire was translated into Chinese (by Li J) and back-translated into English (by Shang L). Then the back-translation was reviewed to check its agreement with the original (by Siegrist J). The questionnaire consisted of three scales reflecting effort (3 items), reward (7 items, including 2 items for esteem, 2 items for job security, and 3 items for job promotion) and overcommitment (6 items)20). Item responses were scored on a 4-point Likert scale (1=strongly disagree, 4=strongly agree with statement) (see Appendix). The ratio between the effort and reward scales was calculated (weighed by numbers of items) in line with previous studies19,20). All scales were categorized by tertiles into three groups (high, intermediate and low) for the current analyses. The 8-item short-form health survey (SF-8) was applied to assess physical and mental health functioning. SF-8 is comprised of an 8-item subset of the SF-36, which has been used in China widely17). Poor health functioning was defined as the lowest tertile of the score. In addition, information on age, gender, education (high school or below vs. college or above), employment status (non-precarious vs. precarious work, which was defined on the basis of temporary or part-time work) was collected.
Statistical analysisMeans and standard deviations (SDs) were computed for each item and for the scales according to gender, education and employment status. Cronbach's alpha coefficients were examined to assess the internal consistency of the scales. Confirmatory factor analysis was performed to test structural validity. Three models were specified: Model 0 presupposed no separate factors whereby all items were assumed to load on a single factor; Model 1 assumed three first-order factors representing effort, reward and overcommitment. Finally, Model 2 provided the closest representation of the theoretical structure, with three second-order factors, effort, reward and overcommitment, loading on a third-order factor representing the latent ERI construct, and with the components esteem, job security, and job promotion identified as first-order factors loading on the reward factor20–22). Further, to assess the criterion validity of ERI with regard to poor health functioning, odds ratios (ORs) were calculated along with corresponding 95% confidence intervals (95% CIs) using logistic regression modeling. ORs were adjusted for age, gender, education, and employment status. All analyses were conducted with the SAS 9.2 software.
The mean age of the sample was 42.26 years (SD=3.77), 51% were men, all respondents were married, 42% had received college or higher education, and 35% were in precarious employment. Mean SF-8 scores equaled 51.43 (SD=6.20) for physical health and 50.66 (SD=6.63) for mental health.
Means and SDs of the ERI questionnaire scales and items are shown in Table 1. The mean scores for effort, reward and overcommitment, and the mean of effort-reward ratio were 7.65 (SD=1.55), 18.88 (SD=2.59), 15.32 (SD=2.38) and 0.97 (SD=0.26), respectively. Differences were demonstrated according to gender, education, and employment status where effort was significantly higher in men and in those with precarious work. Female gender, low education, and precarious work displayed significantly lower reward, whereas male gender was significantly associated with higher overcommitment. The Cronbach's alpha coefficients for effort, reward, and overcommitment were 0.67, 0.72, and 0.73, respectively.
| All | Gender | Education | Employment status | ||||
|---|---|---|---|---|---|---|---|
| (n=1,916) | Men (n=971) | Women (n=945) | High school or below (n=1,119) | College or above (n=797) | Non-precarious work (n=1,253) | Precarious work (n=663) | |
| Effort | 7.65 ± 1.55 | 7.75 ± 1.58 | 7.54 ± 1.52 | 7.60 ± 1.52 | 7.70 ± 1.59 | 7.70 ± 1.54 | 7.55 ± 1.57 |
| ERI 1 | 2.54 ± 0.64 | 2.58 ± 0.63 | 2.50 ± 0.65 | 2.53 ± 0.65 | 2.56 ± 0.61 | 2.56 ± 0.62 | 2.51 ± 0.67 |
| ERI 2 | 2.46 ± 0.70 | 2.52 ± 0.73 | 2.40 ± 0.67 | 2.43 ± 0.70 | 2.51 ± 0.69 | 2.50 ± 0.69 | 2.39 ± 0.71 |
| ERI 3 | 2.64 ± 0.66 | 2.66 ± 0.66 | 2.63 ± 0.65 | 2.65 ± 0.67 | 2.64 ± 0.64 | 2.64 ± 0.64 | 2.65 ± 0.68 |
| Reward | 18.88 ± 2.59 | 19.01 ± 2.59 | 18.74 ± 2.59 | 18.48 ± 2.56 | 19.43 ± 2.54 | 19.26 ± 2.60 | 18.14 ± 2.41 |
| ERI 4 | 2.89 ± 0.52 | 2.91 ± 0.51 | 2.86 ± 0.52 | 2.88 ± 0.55 | 2.91 ± 0.47 | 2.90 ± 0.51 | 2.86 ± 0.54 |
| ERI 8 | 2.89 ± 0.53 | 2.90 ± 0.54 | 2.87 ± 0.51 | 2.86 ± 0.53 | 2.93 ± 0.52 | 2.93 ± 0.52 | 2.80 ± 0.53 |
| ERI 6 | 2.44 ± 0.65 | 2.43 ± 0.64 | 2.46 ± 0.65 | 2.36 ± 0.62 | 2.56 ± 0.67 | 2.52 ± 0.66 | 2.31 ± 0.60 |
| ERI 7 | 2.73 ± 0.76 | 2.76 ± 0.75 | 2.69 ± 0.77 | 2.56 ± 0.76 | 2.96 ± 0.71 | 2.90 ± 0.72 | 2.41 ± 0.74 |
| ERI 5 | 2.43 ± 0.66 | 2.46 ± 0.65 | 2.41 ± 0.66 | 2.39 ± 0.66 | 2.50 ± 0.65 | 2.46 ± 0.65 | 2.39 ± 0.66 |
| ERI 9 | 2.76 ± 0.57 | 2.78 ± 0.57 | 2.74 ± 0.56 | 2.70 ± 0.59 | 2.84 ± 0.53 | 2.79 ± 0.56 | 2.69 ± 0.58 |
| ERI 10 | 2.73 ± 2.58 | 2.77 ± 0.58 | 2.70 ± 0.58 | 2.73 ± 0.59 | 2.74 ± 0.57 | 2.77 ± 0.56 | 2.67 ± 0.61 |
| Effort-reward ratio | 0.97 ± 0.26 | 0.97 ± 0.27 | 0.96 ± 0.24 | 0.98 ± 0.25 | 0.94 ± 0.26 | 0.95 ± 0.26 | 0.99 ± 0.26 |
| Overcommitment | 15.32 ± 2.38 | 15.48 ± 2.38 | 15.16 ± 2.39 | 15.39 ± 2.35 | 15.21 ± 2.43 | 15.26 ± 2.27 | 15.44 ± 2.59 |
| OC 1 | 2.44 ± 0.62 | 2.44 ± 0.62 | 2.45 ± 0.62 | 2.49 ± 0.63 | 2.38 ± 0.60 | 2.44 ± 0.61 | 2.46 ± 0.65 |
| OC 2 | 2.58 ± 0.60 | 2.63 ± 0.59 | 2.52 ± 0.60 | 2.59 ± 0.60 | 2.56 ± 0.60 | 2.56 ± 0.59 | 2.61 ± 0.61 |
| OC 3 | 2.44 ± 0.67 | 2.47 ± 0.69 | 2.42 ± 0.65 | 2.45 ± 0.69 | 2.43 ± 0.65 | 2.44 ± 0.67 | 2.45 ± 0.68 |
| OC 4 | 2.58 ± 0.60 | 2.63 ± 0.59 | 2.52 ± 0.61 | 2.58 ± 0.60 | 2.59 ± 0.60 | 2.58 ± 0.59 | 2.58 ± 0.63 |
| OC 5 | 2.52 ± 0.60 | 2.57 ± 0.60 | 2.47 ± 0.59 | 2.53 ± 0.60 | 2.52 ± 0.58 | 2.50 ± 0.58 | 2.56 ± 0.62 |
| OC 6 | 2.75 ± 0.59 | 2.74 ± 0.59 | 2.76 ± 0.58 | 2.76 ± 0.60 | 2.74 ± 0.56 | 2.74 ± 0.56 | 2.77 ± 0.63 |
Table 2 presents the structural validity for the three competitive models specified a priori. The three-factor model (Model 1) fit the data better than the null model (Model 0). However, Model 2, which contained three additional components (esteem, job security and job promotion), demonstrated improved goodness-of-fit of the dimensional structure of the ERI theory compared with Model 1.
| Model 0 | Model 1 | Model 2 | ||||||
|---|---|---|---|---|---|---|---|---|
| Item | Scale | Item | Scale | Component | Item | |||
| ERI 1 | (0.54) | Effort (0.82) | ERI 1 | (0.66) | Effort (0.82) | ERI 1 | (0.66) | |
| ERI 2 | (0.50) | ERI 2 | (0.61) | ERI 2 | (0.61) | |||
| ERI 3 | (0.57) | ERI 3 | (0.64) | ERI 3 | (0.64) | |||
| ERI 4 | (0.04) | Reward (−0.44) | ERI 4 | (0.43) | Reward (−0.43) | Esteem (0.70) | ERI 4 | (0.46) |
| ERI 8 | (0.12) | ERI 8 | (0.62) | ERI 8 | (0.68) | |||
| ERI 6 | (−0.17) | ERI 6 | (0.22) | Job security (0.68) | ERI 6 | (0.42) | ||
| ERI 7 | (−0.13) | ERI 7 | (0.36) | ERI 7 | (0.75) | |||
| ERI 5 | (0.10) | ERI 5 | (0.54) | Job promotion (0.86) | ERI 5 | (0.54) | ||
| ERI 9 | (0.12) | ERI 9 | (0.76) | ERI 9 | (0.77) | |||
| ERI 10 | (0.03) | ERI 10 | (0.68) | ERI 10 | (0.68) | |||
| OC 1 | (0.40) | Overcommitment (0.75) | OC 1 | (0.40) | Overcommitment (0.76) | OC 1 | (0.40) | |
| OC 2 | (0.61) | OC 2 | (0.67) | OC 2 | (0.67) | |||
| OC 3 | (0.59) | OC 3 | (0.51) | OC 3 | (0.51) | |||
| OC 4 | (0.54) | OC 4 | (0.58) | OC 4 | (0.58) | |||
| OC 5 | (0.64) | OC 5 | (0.68) | OC 5 | (0.68) | |||
| OC 6 | (0.47) | OC 6 | (0.51) | OC 6 | (0.51) | |||
| χ2 | 3,281.04 | χ2 | 891.75 | χ2 | 739.54 | |||
| GFI | 0.76 | GFI | 0.94 | GFI | 0.95 | |||
| AGFI | 0.69 | AGFI | 0.92 | AGFI | 0.93 | |||
| RMSEA | 0.13 | RMSEA | 0.06 | RMSEA | 0.06 | |||
| CFI | 0.50 | CFI | 0.87 | CFI | 0.90 | |||
| CAIC | 3,073.04 | CAIC | 695.75 | CAIC | 555.55 | |||
Model 0: a one-factor model in which all items load on a single factor. Model 1: a model with three first-order factors: effort, reward and overcommitment. Model 2: a second-order model representing the theoretical ERI model. AGFI, adjusted goodness-of-fit index; CAIC, consistent Akaike information criterion; CFI, comparative fit index; GFI, goodness-of-fit index; RMSEA, root mean square error of approximation.
We found that high effort, low reward, high effort-reward ratio and high overcommitment were all independently related to both poor physical and poor mental health functioning (Table 3).
| Poor physical health | Poor mental health | ||
|---|---|---|---|
| Effort | Low | 1 | 1 |
| Intermediate | 1.46 (1.11, 1.91)* | 1.51 (1.14, 1.98)* | |
| High | 1.77 (1.31, 2.39)** | 1.49 (1.10, 2.02)* | |
| Reward | High | 1 | 1 |
| Intermediate | 1.89 (1.49, 2.40)*** | 1.84 (1.45, 2.35)** | |
| Low | 2.07 (1.60, 2.68)*** | 2.68 (2.06, 3.48)*** | |
| Effort-Reward ratio | Low | 1 | 1 |
| Intermediate | 1.34 (1.03, 1.73)* | 1.36 (1.05, 1.77)* | |
| High | 1.99 (1.53, 2.59)*** | 2.02 (1.55, 2.62)*** | |
| Overcommitment | Low | 1 | 1 |
| Intermediate | 1.11 (0.86, 1.44) | 1.10 (0.84, 1.43) | |
| High | 1.66 (1.28, 2.17)** | 2.46 (1.88, 3.21)*** |
Logistic regression.
All ERI scales were categorized by tertiles into three groups (high, intermediate and low). Adjustment for age, gender, education and employment status.
This study suggests that our 16-item Chinese version of the ERI questionnaire boasts satisfactory psychometric properties in terms of internal consistency, structural validity and criterion validity. Our findings are in line with those reported in validation studies of an identical 16-item ERI questionnaire in the German and Swedish languages20,21) as well as a similar German 14-item version22).
It should be noted that the Cronbach's alpha coefficients in our study are slightly lower compared with those from previous reports20–22), i.e., 0.74–0.80, 0.79–0.84, and 0.79–0.85 for the three scales of effort, reward and overcommitment, respectively. This deviation might be due to sample characteristics or cultural connotations that differ between Eastern and Western populations, which deserve more attention in further research. Remarkably, the factorial structure of the scales of the short ERI questionnaire of our study is in line with the one identified in previous psychometric reports on the short version of the ERI questionnaire. A good fit of the data with the relatively complex theoretical structure of the ERI model (model 2), where AGFI values range between 0.92 and 0.94, supports the structural validity of this measurement approach. Furthermore, the results demonstrate discriminant validity, as scale values differ according to gender, education and employment status in expected directions20–22). In terms of criterion validity with a health measure, our findings are in accordance with a robust body of scientific evidence, indicating that effort-reward imbalance at work is associated with poor physical and mental health functioning cross-sectionally and prospectively17,22–25). Considering the fact of rapid rise of chronic diseases (such as cardiovascular diseases, diabetes, depression, etc.) in China8,26), stress management interventions in the workplace might be taken into account to improve the psychosocial work environment and workers’ health; for example, measures to reduce reward frustration at work seem to be of primary interest based on the effort-reward imbalance theory27).
Importantly, this study applies a recently proposed 4-point Likert scale response format ranging from “strongly disagree” to “strongly agree,” in contrast to a previously used two-step format11,19–21). In the previous two-step format, information on application of item content to the respondent's own job was combined with the respondent's assessment of the degree of experienced distress. This procedure was shown to increase the risk of misclassification28) and, thus, was replaced by the above-mentioned format, in agreement with the authors of the questionnaire20). Our study confirms and extends previous insights by documenting that the new response format is associated with satisfactory psychometric properties.
Several limitations of this study need to be addressed. First, we were not able to infer causality from the observed associations between ERI components and poor physical or mental health because of the cross-sectional design. Second, detailed information on occupations and industries was not included in this study, and this fact may reduce the generalization of our findings. Third, despite using a random sampling technique, our study sample was restricted to married people from one city, which may not be representative for the whole urban Chinese working population. Nevertheless, no systematic recruitment bias was observed, and the sample size was appropriate for application of the statistical analyses that are needed for psychometric tests.
In conclusion, this Chinese 16-item ERI questionnaire represents a suitable instrument for occupational studies in China. Given the severe situation of China's mental health problems in the workplace contemporarily, the short measure of ERI questionnaire might be helpful for efforts aimed at promoting a healthy work environment.
Conflicts of interest: None declared.
