The current study aimed to develop iWorkHealth, a valid and reliable self-administered instrument which identifies workplace psychosocial risk factors in Singapore.
The survey was conducted among 2718 employees who were primarily salaried workers and working in five companies from the healthcare, banking and finance, and legal sectors in Singapore. Factor extraction and item reduction were conducted using exploratory factor analysis (EFA) and Mokken scale analysis (MSA). Construct validity, internal consistency and convergent validity of the final scale were confirmed using confirmatory factor analysis (CFA), Cronbach's alpha and Pearson correlation coefficients, respectively. Multiple Indicators Multiple Causes model was used to detect Differential Item Functioning (DIF).
EFA and MSA identified a five-factor solution (job demand, job control, employee and management engagement, supervisor support and colleague support) for the 27 items iWorkHealth instrument. CFA demonstrated that the five-factor model fitted the data with high internal consistency (Cronbach's alpha ranged from 0.79 to 0.92). The convergent validity was shown through significant association with existing scales—high job demand was significantly associated with high burnout and depression, while high job control, employee and management engagement, supervisor support and coworker support were significantly associated with low burnout and depression. Ten items were detected with significant DIF, but impact was minimal on the associations between socio-demographics factors and iWorkHealth subscales.
The findings provided evidence that the iWorkHealth instrument which comprises 27 items in five domains of psychosocial risk at the workplace is a reliable and valid instrument that could be used to measure and compare the level of psychosocial risk factors across companies and industries in Singapore.
Citation: Abdin E, Subramaniam M, Chan A, Chen J-A, Chong CL, Wang C, et al. (2019) iWorkHealth: An instrument to identify workplace psychosocial risk factors for a multi-ethnic Asian working population. PLoS ONE 14(8): e0220566. https://doi.org/10.1371/journal.pone.0220566
Editor: Mohamad Alameddine, Mohammed bin Rashid University of Medicine and Health Sciences, UNITED ARAB EMIRATES
Received: February 5, 2019; Accepted: July 18, 2019; Published: August 7, 2019
Copyright: © 2019 Abdin et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data underlying the results of this study are available upon request due to ethical restrictions imposed by the SingHealth Centralised Institutional Review Board, in Singapore. All study subjects were informed that only non-identifiable and aggregate data will be used in subsequent reports and scientific publications, and communications to companies and stakeholders. Hence, individual data is not available for public online access. Readers may write to the Workplace Safety and Health Institute, Ministry of Manpower at email@example.com (https://www.wsh-institute.sg/) for requests on aggregated data.
Funding: Author CLC received salary support from Clanworks, Singapore. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: Author CLC received salary support from Clanworks, Singapore. Clanworks, Singapore had no conflict of interest relating to employment, consultancy, patents, products in development, or marketed products, etc. with the current study. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
The increasing presence of workplace psychosocial risk factors such as poor organizational climate, introduction of new technologies, renewed business models, social relationship and leadership, work-family conflict, high work pressure, and job insecurity has resulted in significant negative impact on employees' psychological wellbeing, physical health and safety, as well as organizational performance-related outcomes such as absenteeism, job dissatisfaction and loss of productivity [1, 2]. The International Labor Organization  defines psychosocial risk at work as interactions between the individual and a range of workplace factors including job design, management, and the organizational environment that have the potential to have a hazardous influence on employee’s health. A hazard occurs when one or more of these factors have a detrimental effect on a worker’s wellbeing, resulting in poor health outcomes such as exhaustion, anxiety or depression . To manage any of these psychosocial risk factors at work, it is crucial to first determine their existence and prevalence followed by identifying the specific groups at risk .
Over the past decades, various instruments have been developed that measure psychosocial risk factors at the workplace including the Job Content Questionnaire , the Effort-Reward Imbalance Questionnaire  and the Copenhagen Psychosocial Questionnaire . In other countries, especially those in Asia, there are cultural differences and concepts of psychosocial risk factors may be perceived differently and even uniquely depending on the population . Hence, it is important to adopt some of these key instruments and use their items to specifically develop an instrument that is relevant for use in a culturally diverse population that recognizes how within that population, the various cultural subgroups may think, feel and act differently on various issues at work .
This research focuses on Singapore which is an island city-state in Southeast Asia with a multi-ethnic Asian population of approximately 5.64 million people in 2018. The population comprises Chinese (74.3%), Malays (13.4%), Indians (9.0%), and other ethnic groups (3.2%) . It has a predominantly working population where 65.1% of residents aged 15 and over were employed in 2018 . Hence, it is not surprising that work-related stress, psychological, physical and behavioural problems such as depression, high blood pressure, burnout, absenteeism, turnover intention etc., are being recognised as an emerging area of concern in this multi-ethnic Asian working population. However, a valid and reliable instrument to assess psychosocial risk factors in this working population, which is applicable across companies and industries, is currently lacking. Moreover, few studies have considered measurement bias such as differential item functioning (DIF) when developing a psychosocial risk factor instrument . Hence, the current study aimed to develop a self-administered instrument that embraces the multi-faceted workplace psychosocial risk in Singapore.
Material and methods
Study design and participants
This multiphase study involved the development and validation of the instrument in a phased manner. In the development phase, the research team conducted an extensive literature review to identify the plausible factors that influence an employee’s psychosocial risk (Fig 1) and derived a model of psychosocial risk factors and outcomes. Instruments relevant to measuring these psychosocial risk factors were then identified and all the researchers proceeded to select the appropriate scales and items through collaborative discussions for the draft version of the instrument.
Next, the draft instrument was field-tested in focus group discussions (FGDs) where a total of 29 employees belonging to different sectors and educational qualifications participated. The participants were also encouraged to identify factors that they felt contributed to psychosocial risk but had not been included. They were also asked to identify those that they felt were not relevant. In addition, participants also provided input on the phrasing and sentence construction of the questionnaire and the response scales.
In the validation phase, a cross-sectional study was conducted among employees aged 18 years and above who were working in Singapore. Research data were collected using a questionnaire survey from five companies operating in the healthcare, banking and finance and legal sectors. The most common types of jobs among participants were nurses, allied health professionals and lawyers. A total of 2718 employees from the three sectors participated with response rates ranging from 70% (banking industry) to 84% (healthcare industry). However, only 2713 were included in the current analyses after excluding five respondents aged below 20 years. The sample comprised twice as many females as males (69.1% vs 30.9%). The majority were Chinese (65.2%) followed by Malays (15.8%), Indians (11%), and other ethnic groups (8%). Education ranged from Primary school (≤ 6 years of education) to Master/Doctorate, with the majority having a Bachelor degree (43.4%), followed by Secondary education (16.1%), Diploma (15%) and Master/Doctorate degree (14.4%) (Table 1). The ethical approval for the study was obtained from the SingHealth Centralised Institutional Review Board in Singapore. Waiver for informed consent was approved. Implicit consent was obtained from all participants using a participant information sheet which was included with the data collection form. The participant information sheet provided all the study details on the voluntary nature and risk benefits of the study. Contact numbers of Principal Investigator and Co-Investigators were also provided in case the participants had any other related queries. Diligent care was exercised to ensure that no identifiers were collected from the participants.
Demographic data, psychosocial outcomes, and items to assess psychosocial risk factors were included as part of the questionnaire. Demographic data included age, gender, ethnicity, marital status and highest education level attained.
Psychosocial outcomes were measured using the following two scales:
- Copenhagen Burnout Inventory (CBI) . This has three sub-dimensions, namely personal burnout, work-related burnout and client-related burnout. The personal burnout subscale is generic, for all participants to be able to respond. The work-related burnout subscale assumes that respondents are engaged in some kind of paid work and the client-related burnout subscale includes questions with the term client, which can be replaced by terms such as colleague, customer, or patient .
- Patient Health Questionnaire-2 (PHQ-2) . This comprises the first two items of PHQ-9 and looks at the frequency of depressed mood and anhedonia over the past two weeks. Score ranges from 0–6 with higher scores indicative of more severe depressive symptoms.
45 items of psychosocial risk factors were identified from the literature search and FGDs in the development phase, with participants responding using a five-point Likert scale, where ‘1’ represented ‘strongly disagree’ and ‘5’ corresponded to ‘strongly agree’.
A series of steps were taken to develop and validate the scale. To obtain the underlying factors, a spilt-half test was first done where the 2713 participants were randomly divided into two groups. Exploratory factor analysis (EFA) was used on the first subsample (n = 1356) to extract the factors. The Mokken scale analysis (MSA) was used to examine item performance and to carry out the final item reduction. The derived factors were then applied to the second subsample (n = 1357) and validated using confirmatory factor analysis (CFA). EFA examined the polychoric correlations with weighted least squares with the mean- and variance-adjusted chi-square (WLSMV) estimator. An oblique geomin rotation was then applied to obtain a more discriminating factor structure. Several criteria were used to determine the number of factors in EFA such as eigenvalues (values >1.0), visual sighting of scree plot, identifying pattern of loadings on each factor (i.e. loading > 0.4 or cross-loading items), and robustness of interpretability for each solution. The psychometric properties of each scale and its associated items were further assessed using MSA. Generally, the MSA assessed four fundamental assumptions including unidimensionality, local independence, monotonicity and non-intersection. A set of items was assumed to form a good scale if the Hi coefficient (Loevinger’s scalability coefficient for the item) for each pair of items was greater than or equal to 0.3. The scale was classified based on H value (Loevinger’s scalability coefficient for the scale) which classified H value of 0.3 to <0.4 as indicating a weak scale, 0.4 to <0.5 a medium scale, and 0.5 to 1.0 a strong scale . After determining the factor solution and eliminating items with low factor loadings and scalability index from both EFA and MSA, the construct validity of the final model was confirmed using CFA’s fit indices that included comparative fit index (CFI) >0.95, Tucker-Lewis index (TLI) >0.90, and root mean square error of approximation (RMSEA). The RMSEA value at 0.05 and below and 0.08 and below indicated good and moderate fit [17, 18]. Internal consistency of each subscale was evaluated using Cronbach’s alpha coefficient, where the acceptable level was set at 0.7 . The convergent validity in overall sample and by age, gender, ethnicity and education status was examined with Pearson correlations between the psychosocial risk factors instrument and the external instrument which measured burnout and depression. In order to ensure that the items were not biased due to differences in age groups (young: 20 to 39 years, middle: 40 to 59 years, older: 60 years and above), gender (Male and Female), ethnicity (Chinese, Malay, Indian and Others), marital status (Single, Married, Divorced/Separated), education level (Secondary and below, Vocational/ITE, Diploma, Tertiary including Bachelor, Postgraduate Diploma, Master and Doctorate Degree) and sector (Healthcare, Legal and Banking); DIF tests were incorporated in the analyses. According to Zumboo  the instrument may exhibit DIF when the respondents of different groups endorse an item unequally given the same latent trait that the item intends to measure. Previous studies have suggested that if an instrument exhibits DIF it might lead to different results , and therefore, statistical adjustments such as a Multiple Indicators Multiple Causes (MIMIC) modeling should be used to address DIF and reduce its impact on group comparisons [21, 22]. Hence, MIMIC modelling was implemented to explore associations between socio-demographic factors and iWorkHealth subscales and address significant DIF in the regression model. Statistical significance was set at p value < 0.05.
Factor extraction and item reduction
The factors identified through the literature review and FGDs included workload, job demand, job control, job meaning, job reward, organizational culture, colleague support, and supervisor support. A series of EFAs were conducted to determine the dimensionality of the instrument. The plot of eigenvalues for the initial 45-item indicated that five, six and seven factor solution were plausible. Upon examination of each of the rotated solutions including pattern of factor loadings i.e. cross-loading and loadings > 0.40, 6 items were removed and the six-factor solution with 39-item was found to be optimal. Item performance and item reductions were further tested using MSA. 12 items with poor item fit (Hi less than 0.3) were deleted. The range of item scalability index value for the remaining 27 items ranged from 0.40 to 0.78. Re-examination of factor structure using 27 items concluded that the five-factor solution was optimal with higher factor loadings within each factor. Based on the content of the remaining items in each factor, the five factors were re-named: job demand, job control, employee and management engagement, supervisor support and colleague support. The H value for their subscales ranged from 0.462 (job demand) to 0.756 (colleague support) indicating medium to strong subscales according to Loevinger’s scalability classification criteria (Table 2).
CFA of the five-factor model using the final 27 items forming the iWorkHealth instrument resulted in acceptable fit (RMSEA = 0.08, TLI = 0.94, CFI = 0.95). The Cronbach’s α coefficient for the job demand, job control, employee and management engagement, supervisor support and colleague support were 0.79, 0.83, 0.86, 0.92, and 0.90, respectively. We concluded that the internal consistency of the iWorkHealth was good. The convergent validity of the instrument with other existing scales was also examined. Prior to the convergent validity analyses, the five subscale scores were obtained by summing the chosen response from relevant items within each factor. Higher scores indicate greater psychosocial risks experienced by employees. High job demand score was found to be significantly and positively associated with high burnout and depression as measured by the CBI and the PHQ-2 scores, while high job control, employee and management engagement, supervisor support and coworker support were significantly and negatively associated with burnout and depression. We also found that the convergent validity was acceptable across subgroups by age, gender, ethnicity and educational status (Table 3).
Association between socio-demographic factors and iWorkHealth subscales
After controlling for all sociodemographic factors in multivariate regression models, we identified 31 significant associations between sociodemographic factors and iWorkHealth subscales (Table 4: Model 1). Older age group was significantly associated with lower job demand and higher job control, co-worker support and employee and management engagement while the middle age group was significantly associated with higher job control and employee and management engagement than the younger age group. Females were significantly associated with lower job control, supervisor and co-workers support and employee and management engagement than males. Indians were significantly associated with lower job demand and higher job control, supervisor support and employee and management engagement than Chinese. Those with secondary education and below and vocational/ITE qualifications were significantly associated with lower job demand and co-worker support while those with diploma were significantly associated with higher employee and management engagement than those with a university degree. Those who were married were significantly associated with higher job control and employee and management engagement than those who were single. Those working in banking sector were significantly associated with lower job demand and higher supervisor and co-worker support while those working in legal sector were significantly associated with lower job control and higher employee and management engagement than those in the healthcare sector.
Differential item functioning
We found that 10 items from the job demand and job control dimensions had significant DIF in relation to age, ethnicity, education level and sector (Table 5). For example, item 10 “my work in meaningful” and item 11 “my work is important” exhibited DIF in relation to sector i.e., those in the legal sector tend to have lower probability of endorsement of item 10 and item 11 than those in the healthcare sector.
This paper describes the development of a psychometrically sound, robust and easily administrable iWorkHealth instrument which considered all the culturally relevant domains of psychosocial risks in a multi-ethnic Asian working population in Singapore. The iWorkHealth identified five key dimensions of psychosocial risk factors in Singapore. Results from the EFA and MSA showed that they were associated with dimensions of job demand, job control, employee and management engagement, supervisor support and colleague support. Some of these dimensions were quite similar with those reported in the literature [6, 7, 23], such as job demand, job control, supervisor support and co-worker support while others such as employee and management engagement emerged as salient in the local population. Our results showed that the job demand dimension was associated with 5 items measuring high workload—so much work to do, feeling tired after work/shift-work, work that is emotionally demanding, and experience contradictory demands. The job control dimension was clearly associated with 7 items measuring whether the job met their expectation, skills and expertise, whether they had enough information to get the job done, whether they perceived their job as meaningful and important as well as feeling that enough help and equipment was available, and whether they felt motivated and involved in their work. Apart from that, the analyses clearly showed that the instrument was able to differentiate items that measured support from co-workers (3 items) and supervisor (5 items) and demonstrated that the two dimensions are distinct. The employee and management engagement dimension was clearly associated with 7 items measuring satisfaction with the pay and benefits received and opportunities for promotion, whether rewards were given fairly, whether company treated employees well, whether employees received the respect and prestige they deserved, whether company considered employee welfare much more important than operations/sales and profit and whether company managed change well.
The internal reliability of the instrument was supported by Cronbach’s alpha results. The value of the Cronbach’s alpha for each scale was above the cutoff of 0.70 . The validity of the instrument was also strongly supported by convergent validity of the subscales with other existing scales which measured burnout and depression. In this analysis we found that high job demand was strongly and directly associated with burnout and depression. Meanwhile, high job control, supervisor support, co-worker support, employee and management engagement were inversely associated with the burnout and depression scores. According to the Job Demand Control model, high job demand and low job control needed to exist simultaneously in order to produce high job strain  that could predict mental illnesses . In our study, high job demand and low job control were independently associated with the burnout and depression scores. Hence it strongly supported the Job Demand Control model. We incorporated analyses based on key sociodemographic factors and identified 31 significant associations between sociodemographic factors and iWorkHealth subscales. These findings suggest that the instrument seems to have good properties to differentiate the multi-faceted nature of workplace in our sample across age, gender, ethnicity, marital status, education level and sector. However, we identified 10 items from job demand and job control dimensions that exhibited significant DIF in relation to age, ethnicity, education level and sector. Hence, we implemented MIMIC modelling to address DIF and reduce its impact on group comparisons. After inclusion of significant DIF in the multivariate regression models within the MIMIC framework (Table 4: Model 2), we found that only two associations became non-significant. All the other 29 associations remained significant which suggests that the impact of DIF was minimal.
While the current instrument demonstrated superior reliability and validity in Singapore’s working population, some limitations of this study should be addressed. The study was limited to English speaking adults, those who were working in healthcare, banking and finance, and legal sectors. The 5 firms participated in the study voluntarily. Therefore, the findings may not representative of all firms in Singapore. Moreover, other dimensions of psychosocial risk factors in the wider population such as those from construction or teaching sectors may have fallen outside the scope of this study. Our study had used a cross-sectional design, thus preventing us from concluding any causal relationship between the psychosocial risk factors at the workplace and self-perceived burnout and depression. However, findings from other studies using longitudinal research design, have suggested possible relationship between psychosocial risk factors at workplace and self-perceived depression . Another limitation of this study was that all of the study data were self-reported which may have introduced bias. However, self-report is often the only feasible strategy to gather information concerning workers' psychosocial risk factors .
Overall, the findings provided evidence that the iWorkHealth instrument, which comprises 27 items in five domains of workplace psychosocial risk factors, was reliable and valid for providing insights on the prevalence of psychosocial risk factors and their relationships with employee well-being in the workplace. Subsequently it can be used to provide a normative base to identify successful interventions, contributing to the development of best practice, standards and guidelines for the management of psychosocial risk factors in the workplace in Singapore.
The authors gratefully thank the five companies from the healthcare, banking and finance, government and legal industries for volunteering as well as to all their staff who participated voluntarily in this study.
- 1. Commission on Social Determinants of Health. Closing the gap in a generation: Health equity through action on the social determinants of health (Final Report of the Commission on Social Determinants of Health). Geneva: World Health OrganizationExposure to psychosocial risk factors in the context of work: a systematic review. 2008.
- 2. Leka S, Jain A. International Initiatives to Tackle Psychosocial Risks and Promote Mental Health in the Workplace: Is There a Good Balance in Policy and Practice?. In: Shimazu A, Dollard MF, Nordin RB, Oakman J, editors. Psychosocial Factors at Work in the Asia: from theory to practice 2014.
- 3. ILO. Psychosocial factors at work: Recognition and control. Report of the Joint International Labour Office and World Health Organization on Occupational Health, Ninth Session, Geneva, 18–24 September 1984. Occupational Safety and Health Series No. 56. Geneva: ILO.1986.
- 4. Bailey TS, Dollard MF. National Benchmarking and Standards for Psychosocial Factors. In: Dollard MF, Shimazu A, Nordin RB, Brough P, Tuckey MR, editors. Psychosocial Factors at Work in the Asia2014.
- 5. Houtman I, Jettinghoff K, Cedillo L. Raising awareness of stress at work in developing countries: A modern hazard in a traditional working environment: Advice to employers and worker representatives. World Health Organization http://wwwwhoint/occupational_health/publications/raisingawarenessofstresspdf. 2007.
- 6. Karasek RA. Job demands, job decision latitude, and mental strain: implications for job redesign. Admin Sci Q. 1979;24: 285–308.
- 7. Siegrist J. Effort-reward imbalance at work and cardiovascular diseases. International Journal of Occupational Medicine and Environmental Health. 2010;23:279–85. pmid:20934954
- 8. Kristensen TS, Hannerz H, Hogh A, Borg V. The Copenhagen Psychosocial Questionnaire-a tool for the assessment and improvement of the psychosocial work environment. Scand J Work Environ Helath, 2005; 31(6):438–449
- 9. Fishta A, Backe EM. Psychosocial stress at work and cardiovascular diseases: an overview of systematic reviews. Int Arch Occup Environ Health. 2015;88(8):997–1014. pmid:25687981
- 10. Iwata N. Cultural Distinctiveness in Response Bias. In: Dollard MF, Shimazu A, Nordin RB, Brough P, Tuckey MR, editors. Psychosocial Factors at Work in the Asia2014.
- 11. Singapore Department of Statistics. Yearbook of Statistics Singapore, 2018. Ministry of Trade & Industry, Republic of Singapore. 2018.
- 12. Ministry of Manpower. Labour Force in Singapore 2018. Manpower Research and Statistics Department, Ministry of Manpower, Republic of Singapore. 2018.
- 13. Orhede E, Kreiner S. Item bias in indices measuring psychosocial work environment and health. Scan J Work Environ Health. 2000;25(3):263–72.
- 14. Kristensen TS, Borritz M, Villadsen E, K.B.C. The Copenhagen Burnout Inventory-a new tool for the assessment of burnout. Work & Stress, 2005; 19(3):192–207
- 15. Kroenke K, Strine TW, Spitzer RL, Williams JB, Berry JT, Mokdad AH. The PHQ-8 as a measure of current depression in the general population. J Affect Disord. 2009;114(1–3):163–73. pmid:18752852.
- 16. Sijtsma K, Molenaar IW. Introduction to Nonparametric Item Response Theory.: Thousand Oaks, CA:; 2002.
- 17. Hu L, Bentler PM. Cut off criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling. 1999;6:1–55.
- 18. Browne MW, Cudek R. Alternate ways of assessing model fit. In: Bollen KA, Long JS, editors. Testing structural equation models: Newbury Park, CA: Sage.; 1993. p. 136–62.
- 19. Cronbach LJ. Coefficient alpha and internal structure of tests. Psychometrica. 1951;16:297–334.
- 20. Zhumbo BD. A handbook on the theory and methods of differential item functioning: Logistic regresion modelling as a unitary framework for binary and likert type (ordinal) items scores. Ottawa: Directorate of Human Resources Research and Evaluation, Department of National Defense; 1999.
- 21. Yu YF, Yu AP, Ahn JH. Investigating differential item functioning by chronic diseases in the SF-36 health survey: A latent trait analysis using MIMIC models. Med Care. 2007;45:851–9. pmid:17712255
- 22. Abdin E, Subramaniam M, Picco L, Pang S, Vaingankar JA, Shahwan S, et al. The importance of considering differential item functioning in investigating the impact of chronic conditions on health-related quality of life in a multi-ethnic Asian population. Qual Life Res. 2017;26(4):823–34. Epub 2016/09/30. pmid:27679497.
- 23. Karasek R, Theorell T. Healthy work: Stress, productivity and the reconstruction of working life.: New York: Basic Books.; 1990.
- 24. Kline P. Psychometrics and Psycholog: London, Academic Press.; 1979.
- 25. Rugulies R, Bultmann U, Aust B, Burr H. Psychosocial work environment and incidence of severe depressive symptoms: prospective findings from a 5-year follow-up of the Danish work environment cohort study. Am J Epidemiol. 2006;163:877–87. pmid:16571741
- 26. Karasek R, Brisson C, Kawakami N, Houtman I, Bongers P, Amick B. The Job Content Questionnaire (JCQ): an instrument for internationally comparative assessments of psychosocial job characteristics. J Occup Health Psychol. 1998;3: 322–55. pmid:9805280