Abstracts

INTRODUCTION

The demand-control model states that workers exposed to high levels of job strain – a combination of high psychological demands and low job decision latitude – have an increased risk of ill health compared to those exposed to low job strain.¹⁶16  Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York: Basic Books; 1990. Social support at work was added later as a dimension that would potentially buffer the effect of high strain on stress-related illness.¹³13  Johnson JV, Hall EM. Job strain, work place social support, and cardiovascular disease: a cross-sectional study of a random sample of the Swedish working population. Am J Public Health. 1988;78(10):1336-42. DOI:10.2105/AJPH.78.10.1336
https://doi.org/10.2105/AJPH.78.10.1336... The model has been tested mostly in relation to cardiovascular diseases and related risk factors.³3  Backé EM, Seidler A, Latza U, Rossnagel K, Schumann B. The role of psychosocial stress at work for the development of cardiovascular diseases: a systematic review. Int Arch Occup Environ Health. 2012;85(1):67-79. DOI:10.1007/s00420-011-0643-6
https://doi.org/10.1007/s00420-011-0643-... ^,3232  Toivanen S. Social determinants of stroke as related to stress at work among working women: a literature review. Stroke Res Treat. 2012;2012:873678. DOI:10.1155/2012/873678
https://doi.org/10.1155/2012/873678...

The Job Content Questionnaire (JCQ) is a 49-item instrument, originally proposed to measure job strain dimensions.¹⁴14  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(4):322-55. DOI:10.1037/1076-8998.3.4.322
https://doi.org/10.1037/1076-8998.3.4.32... Abridged versions were proposed for multidimensional studies, such as the Swedish demand-control-support and the Whitehall Job Characteristics questionnaires.¹⁹19  Landsbergis P, Theorell T, Schwartz J, Greiner BA, Krause N. Measurement of psychosocial workplace exposure variables. Occup Med. 2000;15(1):163-88. The demand-control-support questionnaire (DCSQ) includes 17 items grouped into three main dimensions: psychological demands (work fast, work intensely, work effort, enough time, and conflicting demands), decision latitude – subdivided into skill discretion (learning new things, skill level, take initiative, repetitive work) and decision authority (how to do the work and what to do at work) – and social support at work (pleasant atmosphere, relationships at work, coworkers’ support, helpful coworkers, relationship with supervisors and friendly workers). According to the psychometric literature, dimension represents the latent variable (not observed), which summarizes the original set of observed variables (e.g., items).⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).^,1111  Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL. SEM: confirmatory factor analysis. In: Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL, editors. Multivariate data analysis. 6.ed. Upper Saddle River: Pearson Prentice Hall; 2006. p.770-842.^,3030  Streiner DL, Norman GR. Health measurement scales: a practical guide to their development and use. 4.ed. Oxford: Oxford University Press; 2008.

Validation studies have been conducted using different versions of the JCQ;²2  Araújo TM, Karasek R. Validity and reliability of the Job Content Questionnaire in formal and informal jobs in Brazil. Scand J Work Environ Health Suppl. 2008;(6):52-9.^,55  Edimansyah BA, Rusli BN, Naing L, Mazalisah M. Reliability and construct validity of the Malay version of the Job Content Questionnaire (JCQ). Southeast Asian J Trop Med Public Health. 2006;37(2):412-6.^,66  Escribà-Aguir V, Más Pons R, Flores Reus E. Validación del “Job Content Questionnaire” en personal de enfermeria hospitalario. Gac Sanit. 2001;15(2):142-9. DOI:10.1016/S0213-9111(01)71533-6^,1717  Kawakami N, Fujigaki Y. Reliability and validity of the Japanese version of Job Content Questionnaire: replication and extension in computer company employees. Ind Health. 1996;34(4):295-306. DOI:10.2486/indhealth.34.29510
https://doi.org/10.2486/indhealth.34.295... ^,2020  Li J, Yang W, Liu P, Xu Z, Cho SI. Psychometric evaluation of the Chinese (mainland) version of Job Content Questionnaire: a study in university hospitals. Ind Health. 2004;42(2):260-7. DOI:10.2486/indhealth.42.260
https://doi.org/10.2486/indhealth.42.260... ^,2525  Niedhammer I. Psychometric properties of the French version Karasek Job Content Questionnaire: a study of the scales of decision latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health. 2002;75(3):129-144. DOI:10.1007/s004200100270
https://doi.org/10.1007/s004200100270... ^,2626  Sale JE, Kerr MS. The psychometric properties of Karasek’s demand and control scales within a single sector: data from a large teaching hospital. Int Arch Occup Environ Health. 2002;75(3):145-52. DOI:10.1007/s004200100289
https://doi.org/10.1007/s004200100289... ^,2828  Schreurs PJG, Taris TW. Construct validity of the demand-control model: a double cross-validation approach. Work Stress. 1998;12(1):66-84. DOI:10.1080/02678379808256849
https://doi.org/10.1080/0267837980825684... five studies using the DCSQ were identified:¹⁰10  Griep RH, Rotenberg L, Vasconcellos AGG, Landsbergis P, Comaru CM, Alves MGM. The psychometric properties of demand-control and effort-reward imbalance scales among Brazilian nurses. Int Arch Occup Environ Health. 2009;82(10):1163-72. DOI:10.1007/s00420-009-0460-3
https://doi.org/10.1007/s00420-009-0460-... ^,1212  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... ^,2222  Mase J, Ota A, Inoue K, Iida T, Tsutsumi A, Yatsuya H, et al. Reliability and validity of the Japanese translated version of the Swedish Demand-Control-Support Questionnaire. Ind Health. 2012;50(6):467-75. DOI:10.2486/indhealth.MS1282
https://doi.org/10.2486/indhealth.MS1282... ^,2727  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921... ^,3131  Theorell T. The demand-control-support model for studying health in relation to the work environment: an interactive model. In: Orth-Gomér K, Schneiderman N, editors. Behavioral medicine approaches to cardiovascular disease prevention. Mahwah: Lawrence Erlbaum Associates; 1996, p.69-85. one was restricted to an internal consistency analysis³¹31  Theorell T. The demand-control-support model for studying health in relation to the work environment: an interactive model. In: Orth-Gomér K, Schneiderman N, editors. Behavioral medicine approaches to cardiovascular disease prevention. Mahwah: Lawrence Erlbaum Associates; 1996, p.69-85. and four evaluated dimensional validity using factor analysis.¹⁰10  Griep RH, Rotenberg L, Vasconcellos AGG, Landsbergis P, Comaru CM, Alves MGM. The psychometric properties of demand-control and effort-reward imbalance scales among Brazilian nurses. Int Arch Occup Environ Health. 2009;82(10):1163-72. DOI:10.1007/s00420-009-0460-3
https://doi.org/10.1007/s00420-009-0460-... ^,1212  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... ^,2222  Mase J, Ota A, Inoue K, Iida T, Tsutsumi A, Yatsuya H, et al. Reliability and validity of the Japanese translated version of the Swedish Demand-Control-Support Questionnaire. Ind Health. 2012;50(6):467-75. DOI:10.2486/indhealth.MS1282
https://doi.org/10.2486/indhealth.MS1282... ^,2727  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921...

Mase et al²²22  Mase J, Ota A, Inoue K, Iida T, Tsutsumi A, Yatsuya H, et al. Reliability and validity of the Japanese translated version of the Swedish Demand-Control-Support Questionnaire. Ind Health. 2012;50(6):467-75. DOI:10.2486/indhealth.MS1282
https://doi.org/10.2486/indhealth.MS1282... (2012) evaluated the Japanese version of DCSQ applied to 212 workers from 19 day nurseries and nursery schools. The exploratory factor analysis showed a three-factor solution, which did not corroborate the hypothesized theoretical model since psychological demands and skill discretion were combined to form a single dimension. Sanne et al²⁷27  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921... (2005) conducted an exploratory factor analysis on data obtained from Norwegian workers and found a three-factor solution, namely, psychological demands, decision latitude, and social support at work. The authors found that for white-collar men and blue-collar women, decision latitude was split into skill discretion and decision authority in a four-factor solution.

The best fit model of a Brazilian version of DCSQ¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... assessed through confirmatory factor analysis (CFA) was achieved when psychological demands, skill discretion and decision authority were considered in a three-factor solution, after the exclusion of social support at work. Moreover, a correlated error measurement between the items work fast and work intensely (psychological demands) was observed and the item repetitive work evidenced low and non-statistical significant loadings on all dimensions. Another Brazilian study¹⁰10  Griep RH, Rotenberg L, Vasconcellos AGG, Landsbergis P, Comaru CM, Alves MGM. The psychometric properties of demand-control and effort-reward imbalance scales among Brazilian nurses. Int Arch Occup Environ Health. 2009;82(10):1163-72. DOI:10.1007/s00420-009-0460-3
https://doi.org/10.1007/s00420-009-0460-... also found correlated residuals (measurement errors) when fitting a one-dimensional confirmatory model on psychological demands. Unfortunately, the second order solution modeled for decision latitude could not be estimated.

Consistently with the DCSQ findings,¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... ^,2727  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921... some JCQ studies concluded that Decision Latitude should be better operationalized as two separate dimensions.¹⁷17  Kawakami N, Fujigaki Y. Reliability and validity of the Japanese version of Job Content Questionnaire: replication and extension in computer company employees. Ind Health. 1996;34(4):295-306. DOI:10.2486/indhealth.34.29510
https://doi.org/10.2486/indhealth.34.295... ^,2626  Sale JE, Kerr MS. The psychometric properties of Karasek’s demand and control scales within a single sector: data from a large teaching hospital. Int Arch Occup Environ Health. 2002;75(3):145-52. DOI:10.1007/s004200100289
https://doi.org/10.1007/s004200100289... ^,2828  Schreurs PJG, Taris TW. Construct validity of the demand-control model: a double cross-validation approach. Work Stress. 1998;12(1):66-84. DOI:10.1080/02678379808256849
https://doi.org/10.1080/0267837980825684... Thus, using decision latitude score as a whole, when combined with demands for the Karasek’s quadrant terms, may be unable to distinguish risk factors and outcomes related to each (sub) dimension. Additionally, repetitive work has consistently appeared as the most troublesome item.⁵5  Edimansyah BA, Rusli BN, Naing L, Mazalisah M. Reliability and construct validity of the Malay version of the Job Content Questionnaire (JCQ). Southeast Asian J Trop Med Public Health. 2006;37(2):412-6.^,1414  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(4):322-55. DOI:10.1037/1076-8998.3.4.322
https://doi.org/10.1037/1076-8998.3.4.32... ^,2020  Li J, Yang W, Liu P, Xu Z, Cho SI. Psychometric evaluation of the Chinese (mainland) version of Job Content Questionnaire: a study in university hospitals. Ind Health. 2004;42(2):260-7. DOI:10.2486/indhealth.42.260
https://doi.org/10.2486/indhealth.42.260... ^,2626  Sale JE, Kerr MS. The psychometric properties of Karasek’s demand and control scales within a single sector: data from a large teaching hospital. Int Arch Occup Environ Health. 2002;75(3):145-52. DOI:10.1007/s004200100289
https://doi.org/10.1007/s004200100289...

According to the Consensus-based standards for the selection of health status measurement instruments – the COSMIN checklist –, cross-cultural validity should be assessed in addition to structural validity and hypotheses testing as part of the construct validity domain.²³23  Mokkink LB, Terwee CB, Patrick DL, Alonso J, Stratford PW, Knol DL, et al. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res. 2010;19(4):539-49. DOI:10.1007/s11136-010-9606-8
https://doi.org/10.1007/s11136-010-9606-... CFA is one of the recommended statistical techniques to evaluate how equivalent the performance of items and dimensions are, in both the original and culturally adapted version of the instrument.²³23  Mokkink LB, Terwee CB, Patrick DL, Alonso J, Stratford PW, Knol DL, et al. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res. 2010;19(4):539-49. DOI:10.1007/s11136-010-9606-8
https://doi.org/10.1007/s11136-010-9606-... However, rarely, the DCSQ studies applied the CFA approach to a heterogeneous sample, which enables to evaluate correlated items and factorial convergent and discriminant validity – crucial information to allow scale improvements.

Thus, the purpose of this study is to further explore the DCSQ cross-cultural validity using the CFA approach in two culturally distinct contexts: Sweden, a high income country where the questionnaire was originally developed; and Brazil, a middle income country where work opportunities and job related social policies and rights are more unevenly distributed.

METHODS

Health workers were considered for this study because they represent a relatively high status occupational class, albeit with some internal sociocultural heterogeneity. Brazilian and Swedish databases covering the same time period, of similar sample size and including relevant variables for the present analysis were available.

Brazilian data (BR) originated from a cross-sectional multidimensional study (2004-2005) investigating all workers in a public health research institute in Rio de Janeiro city, Southeastern Brazil. Of 497 estimated active workers, 424 agreed to participate and 399 (80.3%) completed the self-administered DCSQ at the workplace. Workers less than 18 years of age, and those either on a leave of absence or retired, or who had worked for less than two years, were excluded.

Swedish data (SW) were drawn from the Swedish Longitudinal Occupational Survey of Health (SLOSH), a nationally representative cohort of the Swedish working population, in a follow-up of the Swedish Work Environment Survey, conducted by Statistic Sweden in 2003. For the present study, we analyzed data from the first follow-up wave carried out in 2006. The self-administered questionnaires were sent to the participants by mail. We selected workers with occupational characteristics similar to those of the Brazilian sample and included workers from specialized health centers, general hospitals, laboratories, ambulance crews, and veterinary staff. Of the 391 health workers who were followed up in 2006, 362 (92.6%) completed the DCSQ.

The cross-cultural adaptation process of the DCSQ Brazilian version was described elsewhere.¹1  Alves MGM, Chor D, Faerstein E, Lopes CS, Werneck GL. Short version of the “job stress scale”: a Portuguese-language adaptation. Rev Saude Publica. 2004;38(2):164-71. DOI:10.1590/S0034-89102004000200003
https://doi.org/10.1590/S0034-8910200400... Briefly, conceptual, item, semantic and operational equivalences were evaluated. Measurement equivalence was restricted to test-retest reliability and internal consistency.

DCSQ dimension scores were calculated by summing up the unweighted item scores by dimension; for psychological demands scores ranged from five to 20, and for decision latitude scores ranged from six to 24. The four-point response options varied from “often” (four points) to “never/almost never” (one point).

Social Support at Work consists of six items answerable by a four-point-Likert scale ranging from “strongly agree” to “strongly disagree” (scoring from six to 24). This dimension is applied to respondents as part of the DCSQ, but is hypothesized to buffer the association between job strain and stress-related disorders.¹³13  Johnson JV, Hall EM. Job strain, work place social support, and cardiovascular disease: a cross-sectional study of a random sample of the Swedish working population. Am J Public Health. 1988;78(10):1336-42. DOI:10.2105/AJPH.78.10.1336
https://doi.org/10.2105/AJPH.78.10.1336...

Sociodemographic characteristics included age, sex and education level (< high school, high school and university degree). Job titles were evaluated according to the major groups of the International Standard Classification of Occupations (ISCO, 1988) and grouped as white-collar (major ISCO groups one to four) and blue-collar workers (major ISCO groups five to nine).

Factor analysis was carried out using Mplus software, version 6, which provides the robust weighted least squares mean and variance-adjusted (WLSMV) estimator, which is suitable for categorical indicators.²⁴24  Muthén LK, Muthén BO. Mplus statistical analysis with latent variables: user´s guide. 5.ed. Los Angeles: Muthén & Muthén; 1998-2007. In agreement with the theoretical model¹³13  Johnson JV, Hall EM. Job strain, work place social support, and cardiovascular disease: a cross-sectional study of a random sample of the Swedish working population. Am J Public Health. 1988;78(10):1336-42. DOI:10.2105/AJPH.78.10.1336
https://doi.org/10.2105/AJPH.78.10.1336... ^,1616  Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York: Basic Books; 1990. and previous study results,¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... we excluded the Social Support at Work dimension from factor analysis of the DCSQ (to be referred to as DCQ hereafter). Since the DCQ is comprised of 11 four-level ordinal items, polychoric correlation matrices were used.⁷7  Finney SJ, DiStefano C. Non-normal and categorical data in structural equation modeling. In: Hancock GR, Mueller RO, editors. Structural equation modeling: a second course. Greenwich: Information Age Publishing; 2006. p.269-314.

First, CFA was performed on each sample to corroborate the factor structure proposed by the original demand-control model. Basically, two models were tested: Model 1 predicted the existence of two correlated latent factors representing psychological demands and decision latitude; Model 2 consisted of three correlated latent factors (due to the subdivision of decision latitude in two factors): psychological demands, skill discretion and decision authority. We evaluated the magnitude of the factor loadings, item residuals (or uniqueness), i.e., the portion of an indicator not explained by the latent factor, and factor correlations.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).

Modification indices were evaluated to inspect potential item cross loadings and item residual (error) correlations. Possible item cross loadings and correlations between item measurement errors involving modification indices values equal to or greater than 10 would be further examined, as well as the magnitude of the corresponding expected parameter changes for freely estimated parameters.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).

Goodness of fit was evaluated using three indices. The root mean square error of approximation (RMSEA) incorporates a penalty function for poor model parsimony, accounting for sample size vis-à-vis the number of estimated parameters. Values under 0.06 suggest an approximate (good) fit, whereas values above 0.10 indicate a poor fit and that the model should be rejected.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences). The Comparative Fit Index (CFI) and the Tucker-Lewis index (TLI) represent incremental fit indices contrasting the hypothesized model to a more restricted nested baseline model, the “null model”. Both range from zero to one, with values > 0.9 considered an adequate fit.¹⁸18  Kline RB. Principles and practice of structural equation modeling. New York: Guilford Press; 2005. (Methodology in the Social Sciences).

Second, exploratory structural equation modeling (ESEM) was performed on each sample to investigate plausible alternative dimensional structures to be further evaluated through the CFA approach. ESEM has been utilized to additionally evaluate whether the same dimensional pattern replicates across the two samples (factorial invariance). Basically, this procedure consisted of fitting separate and grouped exploratory models (one to four factors) within the CFA framework. We fixed factor variances to a value of one and the latent means to zero in the Swedish and Brazilian samples. For the grouped ESEM analysis we also fixed the item residual variances to a value of one in the Swedish sample only. All loadings are freely estimated and geomin oblique rotation is used. ESEM allows implementation of other functions that may otherwise only be accomplished within a CFA framework, such as evaluating error measurement correlation and measurement invariance across groups.²¹21  Marsh HW, Muthén B, Asparouhov T, Lüdtke O, Robitzsch A, Morin AJS, et al. Exploratory structural equation modeling, integrating CFA and EFA: application to students’ evaluations of university teaching. Struct Equ Modeling. 2009;16(3):439-76. DOI:10.1080/10705510903008220
https://doi.org/10.1080/1070551090300822...

Third, factor-based convergent and discriminant validity were evaluated using the parameters obtained in the CFA models based on the theoretical proposition and outlined in ESEM analysis. Factor-based convergent validity assesses the degree to which items of the same dimension are correlated and is measured by internal consistency and average variance extracted. Internal consistency was measured by composite reliability, since the traditional Cronbach’s alpha coefficient tends to underestimate reliability except under restricted assumptions of equal common factor loadings and uncorrelated measurement errors. Composite reliability is defined for each factor by the ratio between the true variance (the squared sum of the standardized factor loadings) and the observed variance (the squared sum of the standardized factor loadings plus the sum of their related uniquenesses, ranging from zero to one. Values ≥ 0.70 indicates acceptable internal consistency.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).^,1111  Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL. SEM: confirmatory factor analysis. In: Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL, editors. Multivariate data analysis. 6.ed. Upper Saddle River: Pearson Prentice Hall; 2006. p.770-842.

Average variance extracted assesses the amount of variance captured by a common factor in relation to the amount of variance due to random measurement error.¹¹11  Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL. SEM: confirmatory factor analysis. In: Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL, editors. Multivariate data analysis. 6.ed. Upper Saddle River: Pearson Prentice Hall; 2006. p.770-842. Average variance extracted ranges from 0 to 1 and a factor shows convergent validity when ≥ 0.50, indicating that at least 50.0% of the variance in a measure is due to the hypothesized latent factor. If correlated measurement errors were considered in the CFA model, the sum of nonzero error covariances, multiplied by 2, was added to the denominator for average variance extracted and composite reliability.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences). Ninety five percent confidence intervals (95%CI) for average variance extracted and composite reliability were obtained via a bootstrap method with 1,000 replications.

Factor-based discriminant validity assesses the degree to which different dimensions of the same construct are distinct,¹⁸18  Kline RB. Principles and practice of structural equation modeling. New York: Guilford Press; 2005. (Methodology in the Social Sciences). which is corroborated if the average variance extracted square root of a given factor is above the absolute value of its correlation with any other factor in the model.⁸8  Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39-50. DOI:10.2307/3151312
https://doi.org/10.2307/3151312...

External construct validity (hypotheses testing) assessment was evaluated when inspecting mean-score distribution of demand and control dimensions according to sociodemographic characteristics and social support at work levels. T-test and ANOVA were used to evaluate mean differences across two or more independent groups, respectively. We tested the hypothesis that younger and female workers of lower socioeconomic status who reported lower scores of social support at work would perceive higher levels of psychological demands, and lower levels of skill discretion and decision authority. These analyses were performed in Stata SE, version 10.0.

The Brazilian study was approved by the Research and Ethics Committee of the Instituto de Pesquisa Clínica Evandro Chagas (Protocol 0024.0.009.000-03). The SLOSH was approved by the Regional Research Ethics Board in Stockholm. All participants signed the informed consent form.

RESULTS

The Brazilian sample was younger, had a lower proportion of women, had fewer years of formal education and a lower proportion of blue-collar workers (Table 1). White-collar workers included mostly clerks and professionals, respectively, in the Swedish and Brazilian samples. Health professionals (SW = 60 and BR = 78) were predominantly medical doctors (SW = 27 and BR = 38), nurses (SW = 30 and BR = 12) and psychologists (SW = 5 and BR = 14). Technicians (SW = 132 and BR = 82) included more nurse assistants (SW = 75 and BR = 21) and lifescience technicians (SW = 25 and BR = 43).

The CFA model results are presented in Table 2. Compared to the two-factor model, the three-factor model showed better item loadings and uniquenesses, particularly related to the items postulated as part of the skill discretion dimension and more evidenced in the Swedish sample. The item repetitive work resulted in low loadings and high uniquenesses in both models and samples. The magnitude of factor correlations ranged from 0.10 to 0.44. These models had poor fit for both samples (RMSEA > 0.08), the three-factor model (M2) fit better (CFI/TLI > 0.90) than the two-factor model (M1).

Exploratory models evaluated within a CFA framework (ESEM) were suitable for both samples and confirmed the three-factor structure clearly related to psychological demands, skill discretion and decision authority (Table 3). Consistently with the CFA models, the item repetitive work resulted in low loadings (< 0.25) and high uniqueness (> 0.90) in both samples (data not shown) and was therefore deleted from the revised ESEM model. Factor loadings arising from the revised ESEM model were suitable for both samples, with the item conflicting demands on psychological demands showing the lowest of all loadings (SW = 0.44 and BR = 0.52). The item initiative loaded not only on skill discretion (SW = 0.52 and BR = 0.54), but also on decision authority (0.38) in the Swedish sample and on psychological demands (0.33) in the Brazilian sample. In the Swedish sample, inspection of the modification indices (MI) suggested that freely estimating the item residual correlation between work fast and work intensely (psychological demands) would decrease the Chi-square of the model by 28.94, with an expected parameter change of 0.52. This item measurement correlation was confirmed in the separate Swedish and Brazilian ESEM models (SW = 0.403 and BR = 0.362) and improved the fit indices: RMSEA = 0.064 (SW) and 0.040 (BR), CFI = 0.985 (SW) and 0.995 (BR) and TLI = 0.960 (SW) and 0.987 (BR). The two-group model considering this item measurement correlation provided a good fit to the data, approximately the same fit as those obtained with separate ESEM models (RMSEA = 0.053, CFI = 0.991, TLI = 0.975). These results support the factorial invariance of the DCQ, which means that the same ESEM model was able to fit data from the Brazilian and Swedish samples.

Since these results had been observed in previous CFA-type analyses on the Brazilian sample,¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... we tested and confirmed an alternate model including the above mentioned item residual correlation in both samples (M3, Table 2). As in the previous model the item initiative cross-loaded on decision authority and psychological demands in the Swedish and Brazilian samples, respectively. The magnitude of factor correlations ranged from 0.18 to 0.34 and from 0.12 to 0.40 for the Swedish and Brazilian samples, respectively. Goodness-of-fit indices were suitable for both samples (RMSEA < 0.07, CFI and TLI > 0.90). Modification indices suggested that the item learning new things has a negative cross loading on psychological demands in the Brazilian sample.

Although almost all composite reliability coefficients were suitable, the average variance extracted was not (< 0.50), particularly with regard to the Skill Discretion dimension in both samples (Table 4). The square root of average variance extracted of all dimensions (M2 > 0.58 and M3 > 0.65) were above the absolute value of all CFA-based factor correlations, which ranged for the Swedish and Brazilian samples respectively, from 0.30 to 0.34 and from 0.13 to 0.44 for the 3-factor model (M2 = psychological demands ⇔ skill discretion ⇔ decision authority) and from 0.09 to 0.40 and from 0.13 to 0.47 when cross loadings and error measurement correlations were added to alternative models M3 (Table 2).

Table 5 describes the mean score distribution of the demand control components according to sociodemographic variables and Social Support at Work. Since the factor analysis showed that the item repetitive work did not belong to the DCQ, it was not computed for the skill discretion score. Workers aged over 30 presented higher scores on psychological demands and decision authority dimensions only in the Brazilian sample. Workers with university degrees had higher scores on job strain components and those with high levels (≥ 21) of social support at work had lower scores for psychological demands and higher scores for decision authority (p < 0.001). When compared to blue-collar workers, white-collar workers had higher scores on skill discretion in the Swedish (M_ISCO1-4 = 10.7, SD_ISCO1-4 = 1.1 and M_ISCO5-9 = 10.1, SD_ISCO5-9 = 1.6; F = 15.8, p < 0.001) and Brazilian samples (M_ISCO1-4 = 10.9, SD_ISCO1-4 = 1.4 and M_ISCO5-9 = 9.2, SD_ISCO5-9 = 2.0; F = 54.2, p < 0.001) (data not shown).

DISCUSSION

Our study corroborated the structural validity of the DCQ, when applied to Swedish and Brazilian hospital workers. The three-factor solution represented by the psychological demands, skill discretion, and decision authority dimensions was also generated in previous studies on Brazilian restaurant workers¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... and on Norwegian male workers in high-status occupations.²⁷27  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921... The same dimensional structure was also identified in at least five studies that used the larger JCQ.¹⁴14  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(4):322-55. DOI:10.1037/1076-8998.3.4.322
https://doi.org/10.1037/1076-8998.3.4.32... ^,1717  Kawakami N, Fujigaki Y. Reliability and validity of the Japanese version of Job Content Questionnaire: replication and extension in computer company employees. Ind Health. 1996;34(4):295-306. DOI:10.2486/indhealth.34.29510
https://doi.org/10.2486/indhealth.34.295... ^,2525  Niedhammer I. Psychometric properties of the French version Karasek Job Content Questionnaire: a study of the scales of decision latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health. 2002;75(3):129-144. DOI:10.1007/s004200100270
https://doi.org/10.1007/s004200100270... ^,2828  Schreurs PJG, Taris TW. Construct validity of the demand-control model: a double cross-validation approach. Work Stress. 1998;12(1):66-84. DOI:10.1080/02678379808256849
https://doi.org/10.1080/0267837980825684... ^,2929  Smith CS, Tisak J, Hahn SE, Schmieder RA. The measurement of job control. J Organ Behav. 1997;18(3):225-37. DOI:10.1002/(SICI)1099-1379(199705)18:3<225::AID-JOB797>3.0.CO;2-E
https://doi.org/10.1002/(SICI)1099-1379(...

In our study, the item repetitive work did not belong to any of the factors in the Brazilian or Swedish samples. The same result was observed among Brazilian restaurant workers,¹²12  Hökerberg YH, Aguiar OB, Reichenheim M, Faerstein E, Valente JG, Fonseca MJ, et al. Dimensional structure of the demand control support questionnaire: a Brazilian context. Int Arch Occup Environ Health. 2010;83(4):407-16. DOI:10.1007/s00420-009-0488-4
https://doi.org/10.1007/s00420-009-0488-... although not in a large sample of Norwegian workers.²⁷27  Sanne B, Torp S, Mykletun A, Dahl AA. The Swedish Demand-Control-Support Questionnaire (DCSQ): factor structure, item analyses, and internal consistency in a large population. Scand J Public Health. 2005;33(3):166-74. DOI:10.1080/14034940410019217
https://doi.org/10.1080/1403494041001921... Even though this finding has been observed in several JCQ studies,⁶6  Escribà-Aguir V, Más Pons R, Flores Reus E. Validación del “Job Content Questionnaire” en personal de enfermeria hospitalario. Gac Sanit. 2001;15(2):142-9. DOI:10.1016/S0213-9111(01)71533-6^,1414  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(4):322-55. DOI:10.1037/1076-8998.3.4.322
https://doi.org/10.1037/1076-8998.3.4.32... ^,2525  Niedhammer I. Psychometric properties of the French version Karasek Job Content Questionnaire: a study of the scales of decision latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health. 2002;75(3):129-144. DOI:10.1007/s004200100270
https://doi.org/10.1007/s004200100270... ^,2929  Smith CS, Tisak J, Hahn SE, Schmieder RA. The measurement of job control. J Organ Behav. 1997;18(3):225-37. DOI:10.1002/(SICI)1099-1379(199705)18:3<225::AID-JOB797>3.0.CO;2-E
https://doi.org/10.1002/(SICI)1099-1379(... results may not be comparable, since the length and the items’ response pattern are different from DCQ.⁹9  Fransson EI, Nyberg ST, Heikkilä K, Alfredsson L, Bacquer D, Batty GD, et al. Comparison of alternative versions of the job demand-control scales in 17 European cohort studies: the IPD-Work consortium. BMC Public Health. 2012;12:62. DOI:10.1186/1471-2458-12-62
https://doi.org/10.1186/1471-2458-12-62... ^,1515  Karasek R, Choi B, Ostergren PO, Ferrario M, De Smet P. Testing two methods to create comparable scale scores between the Job Content Questionnaire (JCQ) and JCQ-like questionnaires in the European JACE Study. Int J Behav Med. 2007;14(4):189-201. DOI:10.1080/10705500701638336
https://doi.org/10.1080/1070550070163833...

Error measurement correlation between work fast and work intensely was confirmed for both samples, which may be interpreted as reflecting some redundancy in these items. This finding has been detected among Brazilian restaurant workers,²⁵25  Niedhammer I. Psychometric properties of the French version Karasek Job Content Questionnaire: a study of the scales of decision latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health. 2002;75(3):129-144. DOI:10.1007/s004200100270
https://doi.org/10.1007/s004200100270... nursing staff¹⁰10  Griep RH, Rotenberg L, Vasconcellos AGG, Landsbergis P, Comaru CM, Alves MGM. The psychometric properties of demand-control and effort-reward imbalance scales among Brazilian nurses. Int Arch Occup Environ Health. 2009;82(10):1163-72. DOI:10.1007/s00420-009-0460-3
https://doi.org/10.1007/s00420-009-0460-... and in a study of Canadian hospital workers using the JCQ.²⁶26  Sale JE, Kerr MS. The psychometric properties of Karasek’s demand and control scales within a single sector: data from a large teaching hospital. Int Arch Occup Environ Health. 2002;75(3):145-52. DOI:10.1007/s004200100289
https://doi.org/10.1007/s004200100289...

Since the findings showed that the item initiative not only loaded on Skill Discretion, as expected, but also on the decision authority and psychological demands factors in the Swedish and Brazilian samples, respectively, we reviewed the item contents across the versions, particularly with regard to semantics and wording. We identified an item wording difference between the Swedish and the English versions (“Does your work require creativity?” and “Does your job require you to take the initiative?”, respectively) that might have affected the comparison between the Swedish and the Brazilian versions, since the latter was adapted directly from the English version.¹1  Alves MGM, Chor D, Faerstein E, Lopes CS, Werneck GL. Short version of the “job stress scale”: a Portuguese-language adaptation. Rev Saude Publica. 2004;38(2):164-71. DOI:10.1590/S0034-89102004000200003
https://doi.org/10.1590/S0034-8910200400... Further studies using a qualitative approach could help clarifying the cultural differences regarding the item on ‘initiative’. Nevertheless, the evidence that the item consistently cross-loads suggests its unsuitability in both cultures.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).

Albeit exclusive to the Brazilian sample, the negative cross-loading from learning new things on the psychological demands factor identified through the CFA approach failed to materialize when using an exploratory modeling process. This may be because the CFA’s strict requirement of zero-loadings on nontarget factors ends up forcing a single spurious negative cross-loading to accommodate information that would otherwise be adequately redistributed over several inconspicuous cross-loadings as implied by the more parsimonious ESEM approach,²¹21  Marsh HW, Muthén B, Asparouhov T, Lüdtke O, Robitzsch A, Morin AJS, et al. Exploratory structural equation modeling, integrating CFA and EFA: application to students’ evaluations of university teaching. Struct Equ Modeling. 2009;16(3):439-76. DOI:10.1080/10705510903008220
https://doi.org/10.1080/1070551090300822... probably more adequate for latent psychosocial constructs such as the DCQ.

Patterns of mean score distribution for skill discretion (BR) and decision authority (SW and BR) across education levels, for skill discretion (SW and BR) across occupation groups, and for psychological demands and decision authority across social support levels (SW and BR) are in accordance with the theoretical model¹⁶16  Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York: Basic Books; 1990. and with other JCQ validation studies.¹⁴14  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(4):322-55. DOI:10.1037/1076-8998.3.4.322
https://doi.org/10.1037/1076-8998.3.4.32... ^,2525  Niedhammer I. Psychometric properties of the French version Karasek Job Content Questionnaire: a study of the scales of decision latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health. 2002;75(3):129-144. DOI:10.1007/s004200100270
https://doi.org/10.1007/s004200100270...

Our findings need to be evaluated in light of the strengths and limitations of our study. To our knowledge, this is the first comparative study using ESEM and CFA approaches on the structural and cross-cultural validity of the DCQ in two culturally different societies. Other methodological strengths can also be cited. For one, we have explicitly evaluated item residual correlation in a unified context of exploratory and confirmatory analyses, which heretofore had not been investigated in studies on the demand-control model. Moreover, sample sizes by group were in accordance with current recommendations suggested by Montecarlo simulations of CFA- and ESEM-based multiple group analyses²¹21  Marsh HW, Muthén B, Asparouhov T, Lüdtke O, Robitzsch A, Morin AJS, et al. Exploratory structural equation modeling, integrating CFA and EFA: application to students’ evaluations of university teaching. Struct Equ Modeling. 2009;16(3):439-76. DOI:10.1080/10705510903008220
https://doi.org/10.1080/1070551090300822... and when using the WLSMV estimator for non-normal observational data.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences). In terms of limitations, our samples were restricted to health workers, although there were workers from six of the nine major ISCO groups in both samples.

Our results showed that DCQ has the same dimensional structure in the Swedish and Brazilian hospital workers confirming the cross-cultural validity of this instrument. Consistently with several other DCQ and JCQ studies, the item ‘repetitive work’ should be removed from the demand control construct at least in our studied work environments. As most items were hypothesized as part of the Decision Latitude dimension, this could indicate the need to revise the content of this dimension, perhaps by contemplating the scrutiny and eventual introduction of new and better items.

Our results suggest that the demand-control model may still be valid in contemporary labor contexts. The DCQ should be operationalized as a three-dimensional instead of a two-dimensional scale and without the item assessing repetitive work. Since this study additionally sought to compare the Swedish and Brazilian data and thus enables an evaluation across different domains, our results may also be regarded as indicating the instrument’s psychometric invariance, be it only provisionally and pending more rigorous analyses.⁴4  Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2006. (Methodology in the Social Sciences).^,2121  Marsh HW, Muthén B, Asparouhov T, Lüdtke O, Robitzsch A, Morin AJS, et al. Exploratory structural equation modeling, integrating CFA and EFA: application to students’ evaluations of university teaching. Struct Equ Modeling. 2009;16(3):439-76. DOI:10.1080/10705510903008220
https://doi.org/10.1080/1070551090300822... International comparative studies on the association of job strain and health-related outcomes would benefit from truly comparable scales. This would help resolve one source of inconsistency of empirical results in this research field. Remaining issues – which argue for revising the model – include how items cover the latent trait in terms of intensity and the reproducibility of these results across a wider spectrum of socio-occupational groups.

ACKNOWLEDGMENTS

To Töres Theorell, Olle Lundberg, Constanze Leineweber and Gabriella Olsson for their comments; Cathrin Wiksell and Reidar Österman for their technical assistance; Maria Angelica Borges dos Santos and Leigh Passman for text revision.

REFERENCES

Publication Dates

History