The Brazilian Portuguese version of the revised Maastricht Upper Extremity Questionnaire (MUEQ-Br revised): translation, cross-cultural adaptation, reliability, and structural validation

This study was conducted from December 2012 to April 2013. Questionnaires were sent to 627 computer-office workers at the University of São Paulo (USP), Ribeirão Preto campus. USP is a public institution, and 51% of the office workers are female. In addition, 24% have a higher-education degree, and 76% have a technical formation or high school degree.

For this project, the computer-office-worker sample consisted of men and women employees who were between the ages of 18 and 60 years, had been in the same job position for at least 12 months, and used a computer for a minimum of four hours each work day [ 6].

This phase included 55 individuals (mean age: 33.56 years, SD: 7.93, 95% CI: 31.48–35.64, 41 women, 14 men) who had worked with a computer at their jobs for an average of 10.40 years (95% CI: 8.31–12.49), been at the same job for 7.08 years (95% CI: 5.23–8.93), and worked an average of 7.21 hours/day using a computer (95% CI: 6.87–7.56). The pre-final version included a pre-test with 15 volunteers (mean age: 31.54 years, SD: 5.08, 95% CI: 28.97–34.11 years, 12 women, and 3 men) who had worked with a computer for an average of 10.27 years (CI 95%: 6.50–14.04).

The reproducibility test was conducted with 50 workers, with an average age of 36.04 years (SD: 8.46, 95% CI: 33.70–38.38) who had worked using a computer at their jobs for a mean of 12.86 years (95% CI: 10.71–15.01); 22 were women and 28 were men.

Of the 627 questionnaires handed out, only 386 (participant mean age: 37.44 years, SD: 9.38, 95% CI: 36.50–38.38) were included in the confirmatory factor analysis (CFA) and internal consistency phases. Thus, we obtained a response rate of 62%, and the mean time taken to complete the questionnaire was 14.67 minutes (95% CI: 13.88–15.46). Of the participants, 216 were women and 170 were men (see Table  1), and they had used a computer at their jobs for 13.52 years (95% CI: 12.38–14.36).

The exclusion criteria were: illiteracy or functional illiteracy, visual impairment (not corrected with glasses/contact lenses), and hearing impairment (not corrected with a device). This project was reviewed and approved by the Human Research Ethics Committee of the University Hospital at Ribeirão Preto Medical School, University of São Paulo (Process HCRP N° 10299/2012). All participants signed an informed consent form.

The MUEQ addresses the occurrence, nature, and possible work-related physical and psychological factors associated with CANS among computer users. The MUEQ is the result of the combination of previous tools; questions related to psychosocial factors are based on the Job Content Questionnaire, and the questions related to physical factors at work are based on the Dutch Musculoskeletal Questionnaire [ 21, 22].

The original MUEQ consisted of 59 questions, and individuals were allowed 15 minutes to complete it (see Additional file 1). The first part of the MUEQ assesses socio-demographic characteristics. The MUEQ consists of seven basic domains: (l) work station (seven questions), (2) body posture during work (11 questions), (3) job control (nine questions), (4) job demands (seven questions), (5) break time (eight questions), (6) work environment (nine questions), and (7) social support (seven questions).

The work-station domain has seven questions, with a maximal score of 7 points and two response options (“No” = 1 and “Yes” = 0). For all the other domains, the response options were “Always” (3 points), “Often” (2 points), “Sometimes” (1 point), “Seldom” (0 points), and “Never” (0 points). The body-posture-during-work domain ranges from 0 to 33 points, the job-control domain ranges from 0 to 27 points, the job-demands domain ranges from 0 to 21 points, the break-time domain ranges from 0 to 24 points, the work-environment domain ranges from 0 to 27 points, and the social-support domain ranges from 0 to 24 points.

Complaint items that assess the frequency and clinical features of neck and upper-arm complaints are included in the final portion of the MUEQ. These items can be used to characterize work conditions and workers’ clinical features, but the scores for the complaint items are not included in the total sum.

Before initiating this study, we obtained written permission from the author of the original MUEQ, Prof. Shahla Eltayeb. The process was performed in three phases: English-to-Portuguese translation, synthesis of the translation (committee translation agreement), back-translation, a consensus committee for the pre-final version, and a field test of the pre-final version as recommended by Beaton et al. [ 23] (see Figure  1).

Once the English version was obtained from the author [ 2], it was translated into Portuguese. This version was translated into Brazilian Portuguese by two translators who were fluent in both languages and whose native language was Portuguese; one was a lay translator and the other was an expert translator. Both Brazilian Portuguese versions were compared and synthesized by a committee of translators that included 11 members who were researchers, experts on ergonomics, and civilians [ 23]. This process resulted in the synthesized version T12 (see Figure  1).

Following this phase, two new translators (one native English-speaker and one Brazilian fluent in English) were invited to translate the T12 version into English (i.e., back-translation; see Figure  1). These two translators did not have experience in the health field and were “blind” to all the previous versions of the questionnaire [ 23]. Beyond the English to Portuguese translation, a Dutch to Portuguese translation was also performed. One lay translator, whose native language was Brazilian Portuguese, who was fluent in Dutch had previously translated the original MUEQ into Dutch (see Figure  1). Only back translation from Portuguese (T12) to Dutch (T3) was performed because it was not possible to find another Brazilian native who was fluent in Dutch to perform the initial translation.

In following phase, the same committee as above met again (see Figure  1). The committee’s role was to create the pre-final version that would be administered in the field tests [ 24].

The test of the pre-final version consisted of administering the questionnaire to a sample of participants that corresponded to the target population. The purpose of this step was to assess the acceptability of the tool [ 25].

The pre-final version of the questionnaire was first administered to 15 subjects, and the index of doubts and suggestions obtained was lower than (0.07%). Thus, it was not necessary to reformulate the questionnaire [ 23]. However, suggestions that did not change the meaning of the questions were incorporated. The sample size in this phase was 55 workers, which is a good sample size as suggested by the COSMIN protocol [ 26].

The acceptability of the tool was controlled by means of an open-field form so that the workers who answered the pre-final version of the MUEQ-Br could report any doubts, their impressions and incomprehension of each item, answer choices, item headings, instructions, and the tool’s layout.

The volunteers who completed the final version of the MUEQ-Br (Additional file 1) were invited to complete the questionnaire again one week later in order to verify answer reproducibility. The volunteers were asked whether there had been any changes to their work conditions and musculoskeletal complaints, and they were excluded from the reproducibility study if such changes had occurred. The MUEQ was self-administered.

The internal consistency and confirmatory factor analysis of the instrument were assessed. Internal consistency is the degree to which the items in the questionnaire are correlated [ 24, 26].

Factor analysis is an important statistical phase for the validation of questionnaires. Its objective can be to reduce data, assess the structural factors (i.e., the dimensions of the tool), or investigate whether the questionnaire shows the same dimensions across different groups [ 27]. Because the MUEQ factor structure has been previously described [ 11, 18, 19], a better approach is to use confirmatory factor analysis to assess structural validity.

The confirmatory factor analysis was applied to confirm the model adopted in previous publications [ 2, 11, 18, 19]. The sample of workers was the same used to calculate the internal consistency (n = 386) (see Table  1). The sample size used followed the recommendation of the COSMIN manual [ 20, 26], which states that the sample size should be 5–7 times the number of items with a minimum of 100 individuals. Considering that there are 55 questions across the six domains of the MUEQ-Br, a sample based on COSMIN recommendations would need to have at least 385 respondents (see Table  2) to verify the consistency and confirmatory factor analysis of the MUEQ-Br, and the current study had 386.

The statistical analysis for reproducibility was performed using intraclass correlation coefficient (ICC), with levels of classification as described by Fleiss et al. [ 26] and in which ICC <0.40 is considered weak, 0.40 < ICC < 0.7 is considered moderate, and ICC >0.75 is considered excellent. The statistical model specified was a two-way mixed effects model, based on absolute agreement measures [ 28].

Internal consistency was analyzed using Cronbach’s α coefficient, with results between 0.7 and 0.95 [ 24], and an item-total correlation between 0.2 and 0.7 was considered acceptable [ 29]. The Spearman correlation between the standard error of measurement (SEM) of the total score of the questionnaire on test-retest and the time taken(minutes) to complete the questionnaire was calculated. The SEM was calculated using the formula: SEM = SD √1 - ICC, where SD = the standard deviation between scores from the test and retest [ 30]. For comparisons between mean values among the groups, one-way ANOVAs were applied.

Confirmatory factor analysis was used to test the factor structure of the MUEQ-Br in computer workers. The IBM SPSS AMOS (version 22) was used for conducting the confirmatory factor analysis. Maximum likelihood was used to assess the fit of three different models. The goodness-of-fit for each factor structure was evaluated using several descriptive criteria: (1) consistent Akaike information criterion (CAIC), (2) root mean square error of approximation (RMSEA), (3) the comparative fit index (CFI), (4) the goodness-of-fit index (GFI), (5) the non-normed fit index (NNFI) or Tucker Lewis index (TLI), (6) the expected cross-validation index (ECVI), and (7) chi-square (also called the discrepancy function, likelihood ratio chi-square, or chi-square goodness of fit) (CMIN). The CAIC is a goodness-of-fit measure that adjusts the model’s chi-square to penalize for model complexity and sample size [ 31]. Low measures indicate better fit. The RMSEA quantifies the divergence between the data and the proposed model per degree of freedom. Values below 0.08 indicate an adequate fit [ 31]. The CFI, GFI, and NNFI measure how much better the model fits as compared to a baseline model in which the observed items are assumed to be uncorrelated. These indices are relatively independent of sample size. The CFI avoids underestimation of fit in small sample sizes. For the CFI and GFI, values above 0.90 indicate an adequate fit, and values above 0.95 indicate a good to very good fit. The ECVI is a relative measure to compare competing models: The model with the lowest value has the best fit [ 31]. The CMIN/DF (degrees of freedom) should be less than three [ 32]. The magnitudes of factor loadings for each variable were considered when analyzing the items’ contributions to the model. Variables with a factor loading of 0.4 or greater [ 33] were considered representative of the construct being measured in each domain.

All analyses were performed using the Statistical Package for the Social Sciences (SPSS) for Windows and IBM SPSS AMOS, version 22 (IBM, SPSS Inc., Chicago, USA).