Cross-cultural adaptation and psychometric properties of the short version of COPSOQ II-Brazil

ABSTRACT OBJECTIVES Translate and culturally adapt the short version of Copenhagen Psychosocial Questionnaire II (COPSOQ II) into Brazilian Portuguese (COPSOQ II-Br) and evaluate its psychometric properties. METHODS Translation and cultural adaptation followed the standardized guidelines. Structural validity was assessed using exploratory factorial analysis. Test-retest reliability was evaluated by intraclass correlation coefficient (ICC3,1) and internal consistency by Cronbach’s alpha. Floor and ceiling effect was considered acceptable if less than 15% of participants reported the lowest or highest scores. Measurement error was assessed by standard error of measurement (SEM), while construct validity was tested by correlating the COPSOQ II-Br, the Job Content Questionnaire and the Nordic Musculoskeletal Questionnaire. RESULTS The study evaluated a total of 211 civil servants and service providers in the test and 157 in the retest. After cross-cultural adaptation, the COPSOQ II-Br structure comprised seven domains and 11 dimensions. Most dimensions showed acceptable floor and ceiling effects, excepting “Work family conflicts” (floor effect of 26.1%), and “Meaning and commitment” and “Job satisfaction,” with ceiling floor of 27.5% and 22.3%, respectively. Cronbach’s alpha values reached the recommended levels (varied between 0.70 and 0.87). Test-retest reliability indicated that all dimensions had ICC between 0.71 and 0.81. SEM ranged from 0.6 to 2.2 and the construct validity showed good results with the tested instruments (significant positive and negative correlations). CONCLUSIONS All psychometric properties of the short version COPSOQ II-Br are suitable for use in Brazil. The instrument is thus validated and can be used by occupational health and human resources professionals to evaluate psychosocial working conditions.


INTRODUCTION
providers in the areas of education, health, administration and human resources, general services (cleaners and conservation), maintenance and repair (building assistance and zookeepers). The inclusion criteria were: workers between 18 and 65 years of age 14,16 , with at least six months on the job and a work routine of at least 20 hours a week (4 hours/day, 5 days a week). All participants were informed about the study and those who agreed to participate signed an Informed Consent Form (ICF).
A total of 785 workers were invited to participate in the study. Of the 228 who agreed to participate, 211 were included in the study (Figure 1).

COPSOQ II
COPSOQ comprises seven conceptual models to identify psychosocial risk factors, and all versions have dimensions that measure exposure indicators (risks), health effects, satisfaction and stress 5,15,16 .
COPSOQ II has three versions: long (41 dimensions and 127 questions), medium (28 dimensions and 87 questions) and short (23 dimensions and 40 questions) 16 . In the short version, questions are answered according to a 5-level Likert scale, with question 1B being the only one with an inverted value. Score is calculated by adding up the responses of each dimension, except for the domain offensive behaviour 16 . According to the number of questions in each dimension, the score can vary from 0 to 3, 0 to 4, 0 to 6 or 0 to 8 points. For each dimension, the values obtained can be classified as favorable (green), require attention (yellow) and unfavorable (red) 5 . Another classification method involves calculating the average value of each dimension.

Job Content Questionnaire (JCQ)
JCQ is a standardized instrument proposed to measure the dimensions of the demand-control-social support model 22 . The short and adapted version into Brazilian Test: n = 211 Retest: n = 157 Did not meet the inclusion criteria: Less than 6 months on the job (n = 17) Accepted to participate in the study: n = 228 Portuguese consists of 17 questions covering psychological demand, control, and social support, evaluated by a Likert scale. Score is obtained by adding the questions that make up the dimension, with demand varying between 5 and 20 points, control between 6 and 24 points, and social support between 6 and 24 points 22 .

Nordic Musculoskeletal Questionnaire (NMQ)
NMQ is used to identify the presence of musculoskeletal symptoms in the past seven days and 12 months, seeking health care and having functional limitations due to these symptoms 23 . Participants report the presence of musculoskeletal symptoms such as pain, numbness or discomfort in the neck, lower back, shoulders and other areas in the past 12 months and seven days, on a dichotomous scale (presence or absence) 23 .

Cross-cultural adaptation of COPSOQ II-Br
After obtaining the approval for translation, the process of translation and cross-cultural adaptation followed 5 stages 21 : 1. Translation from English into Brazilian Portuguese (a layman translator and an expert translator, both bilingual and Brazilian Portuguese native speaker, worked independently); 2. Synthesis of translation (synthesis of the translations made by two experts in the field of ergonomics, discussions of differences and consensus among translators); 3. Back-translation (made by two English native speaker specialist translators without prior access to COPSOQ content); 4. Meeting with a committee of specialists for the pre-final version (professionals trained in ergonomics, language professionals, researchers specialized in cross-cultural adaptation and translators); 5. Pre-test (2 pre-tests, where participants were interviewed and the understanding of the questionnaire items assessed).
The final version of the COPSOQ II-Br short version was thus developed and cross-culturally adapted into Brazilian Portuguese. After this 5-step process, we finally tested the psychometric properties of the instrument.

Procedures
Data collection included in-person and online (invitation via link on social and institutional media) questionnaires 14 . In-person data collection took place both in the process of cross-culturally adaptating the instrument into Brazilian Portuguese (pre-test) and in the evaluation of psychometric properties. Online data collection was conducted only for the measurement properties. We chose one or the other according to the participants' schooling level, e.g., cleaners, building assistance and zookeepers were unfamiliar with answering online forms. First, the participants received the ICF and those who consented were included in the study. Then, after a explanation about the questions and guarantee of confidentiality, a researcher sent the questionnaires to the participants, or devilered them in person. The questionnaire is self applicable. Doubts about the questionnaire were solved in the workplace and via e-mail (online format).
To assess test-retest reliability, we perfomed two data collection with an interval of two weeks between measurements to ensure that no changes had occurred 24 . Participants received e-mails reminding them of their participation in the study, with a five-day deadline set for the retest response.
To assess construct validity, we applied the NMQ in the testing phase and the JCQ in the retest, so that the evaluations would not become extensive. The participants completed the questionnaires without additional hours of work or sallary reduction.

Structural validity
Structural validity concerns the degree to which the scores of an instrument adequately reflect the dimensionality of the construct to be measured 20 . Exploratory factor analysis determines whether items compose one or more dimensions and should be carried out when internal consistency is adequate or in the impossibility of conducting confirmatory analysis 24 . The present study carried out a exploratory factor analysis, considering the same sample as that used to assess internal consistency, i. e., a minimum of 100 participants 20,24 .
Principal Component Analysis (PCA) with a final solution based on the Varimax orthogonal rotation was used to evaluate the internal structure of the measurement. The relevance of performing a factor analysis was evaluated by the sample adequacy ratio Kaiser-Meyer-Olkin (KMO) 13,25 . Retained factors were derived according to the magnitude of the eigenvalues (greater than 1) and the proportion of extracted variance. Scree plot was used to determine the number of factors to be extracted. When items had low loads (less than 0.4) in more than one component (complex structure), they were removed from the analysis 25 . A forced analysis with one or two factors was performed when results indicated more than two factors 13,25 .

Internal consistency
Internal consistency assess the degree of interrelation between items, i. e., whether the items are measuring the same concept 20 . We considered a minimum sample of 100 participants to ensure the stability of the variance-covariance matrix and evaluate the internal consistency 24 . Cronbach's alpha was calculated separately for each dimension using baseline data (n = 211). Scores between 0.70-0.95 indicate acceptable internal consistency 24 .

Floor and Ceiling Effect
Floor and ceiling effects reflect problems with content validity and can limit detecting changes over time 24 . These were considered acceptable if less than 15% of participants reported the lowest or highest scores 24 . Distribution of the dimensions is presented in mean, standard deviation (SD), median and interquartile range (IQR).

Test-retest reliability
Reliability is the degree to which a measurement is free of measurement error, and if the scores are identical in repeated measurements over time (test-retest) for the same participants 20 . Studies recommend a minimum sample of 50 participants to test the reliability measure 24 . Test-retest reliability was assessed using the intraclass correlation coefficient (ICC (3,1) ), with ICC values greater than 0.70 24 indicating reliability.

Measurement error
Measurement error means the systematic and random error of a participant's score that is not assigned to real changes in the construct being measured 20 , which requires a minimum sample of 50 participants 24 . This psychometric property was estimated by calculating the standard error of measurement (SEM) at baseline, using the following formula: SEM = SD*√(Cronbach's alpha-1), where SD is the standard deviation of all the participants' baseline scores 26 . Minimal detectable change (MDC) with 90% confidence was calculated by the following formula: MDC = 1,64*SD*√ (2 (1-R)), where R is the test-retest ICC value and SD the standard deviation of all the participants' baseline scores. A change equal to or greater than the MDC, before and after an intervention, indicates that there has been a real change at the individual level 26 .

Construct validity
Construct validity reflects the degree to which an instrument score is consistent with hypotheses, based on the assumption that the instrument measures the construct under assessment 20 . This psychometric property was tested between COPSOQ II-Br and JCQ, which also assesses psychosocial aspects, and NMQ, as musculoskeletal symptoms are closely related to psychosocial aspects. Musculoskeletal pain is common in the working population and has been associated with high work demands and low control [27][28][29] .
Spearman's correlation coefficient (r s ) was calculated to assess the construct validity between the JCQ scales 22 and dimensions of the COPSOQ II-Br. The relation between symptoms in the neck and lower back in the past 12 months and seven days using the NMQ 23 and the dimensions of COPSOQ II-Br was assessed by point-biserial correlation coefficient (r pb ). We also tested the Spearman's (r s ) correlation between the number of areas with musculoskeletal symptoms in the past 12 months and seven days and the dimensions of COPSOQ II-Br. Correlation coefficients greater than 0.70 were considered strong, between 0.70 and 0.50 moderate, and less than 0.50 weak 24 .
Regarding the expected relationship between the tested instruments, we formulated the following hypotheses: 1) there will be a significant and positive correlation between the dimension "Demands at work" of COPSOQ II-Br and the "Demand" scale of JCQ; 2) there will be a significant and positive correlation between the dimensions "Influence and development" and "Meaning and commitment" of COPSOQ II-Br and the "Control" scale of JCQ; 3) there will be a significant and positive correlation between the dimensions "Interpersonal relationships" and "Leadership" of COPSOQ II-Br and the "Social support" scale of JCQ; 4) there will be a significant and positive correlation between the dimensions "Demands at work", "Work family conflict" and "Burnout and stress" of COPSOQ II-Br and musculoskeletal symptoms; 5) there will be a significant and negative correlation between the dimensions "Interpersonal relationships", "Job satisfaction" and "General health" of COPSOQ II-Br and musculoskeletal symptoms.
The data were analyzed using SPSS software version 22.0 (SPSS Inc, Chicago, IL, EUA).

Sample Characteristics
Our study sample comprised mostly women (71%), with mean age of 40 years (SD: 11), married/living with a partner, with more than 15 years of study and up to 2 years working in the same company. Table 1 presents the complete sociodemographic characteristics of the sample.

Cross-Cultural Adaptation Process
We kept the format of the original questionnaire, with minor changes. Discrepancies were resolved by consensus between translators and by the expert committee.
We conducted two pre-tests. In the first, a sample of 30 workers from education (50%) and cleaning services (50%), with high and low level of education, respectively, answered the instrument. Workers (57% of the sample) had difficulty in answering question 4A due to semantics. Since the percentage of doubt exceeded 20% 27 , the requestion was reprhased (from "Do you have a high degree of influence in relation to your work?" to "Do you have a high degree of influence in decisions about your work?"). A sample of 30 workers from the education, cleaning and maintenance/repair services, 15 women (50%) and 15 men (50%) with high and low level of education (50% in each group), participated in the second pre-test. At this stage, workers encountered difficulties only with questions 1A, 4A, 10A. As only 7% of workers had doubts on each question, no redesign was necessary 27 . Based on the results of the pre-test, we finally proposed a short version of COPSOQ II culturally adapted to Brazilian Portuguese (COPSOQ II-Br) a .

Structural validity
PCA identified 11 factors, which represent approximately 50% of the variance ( Table 2)   Continue suggested the structure with two factors for "Health and wellness" (KMO = 0.77). For this domain, the "General health" dimension comprised one question and the "Burnout and stress" dimension consisted of four questions. For the "Offensive behaviours" domain, factor analysis obtained one factor.

Internal consistency
For ten of the 11 dimensions of COPSOQ II-Br, the Cronbach's alpha values varied between 0.70 and 0.87; for one dimension, the Cronbach's alpha value was less than 0.70 ("Offensive behaviours", Cronbach's α = 0.54) ( Table 3).

Floor and Ceiling Effect
Most dimensions of the COPSOQ II-Br short version presented acceptable floor and ceiling effects, except for the floor effect in the "Work family conflicts" dimension (26.1%) and, for the ceiling effect in the dimensions "Meaning and commitment" (27.5%) and "Job satisfaction" (22.3%) ( Table 3).

Test-retest reliability and measurement error
All 11 dimensions showed acceptable reliability, with ICC (3,1) between 0.71 and 0.81. The SEM ranged from 0.6 to 2.2, being lower for the dimensions "Offensive behaviors" and "Work family conflicts", and higher for the dimensions "Demands at work", "Interpersonal relationships" and "Influence and development".
Minimal detectable change (MDC) ranged from 0.8 to 5.9 indicating, for example, that a change in the "Interpersonal relationships" dimension score must be at least 6 points to represent a real change for the individual (Table 4).

Construct validity
The analysis revealed significant and positive correlations between the dimensions of the COPSOQ II-Br and the JCQ scales. The point-biserial correlation coefficient between the presence of musculoskeletal symptoms and the dimensions of the COPSOQ II-Br were Offensive behaviours 20) Have you been exposed to undesired sexual attention at your workplace in the past 12 months? 0.44 21) Have you been exposed to threats of violence at your workplace in the past 12 months? 0.84 22) Have you been exposed to physical violence at your workplace in the past 12 months? 0.68 23) Have you been exposed to bullying at your workplace in the past 12 months? 0 significantly and positively correlated for three dimensions ("Demands at work", "Work family conflits" and "Burnout and stress"). We also found a significant and negative correlation between the presence of musculoskeletal symptoms and the dimensions "Interpersonal relationships", "Job satisfaction" and "General health" of COPSOQ II-Br. These results confimed all the hypotheses formulated (Table 5).

DISCUSSION
The present study cross-culturally adapted the short version of the COPSOQ II into Brazilian Portuguese and analyzed its psychometric properties.
The exploratory factor analysis revealed a questionnaire structure composed of seven domains and 11 dimensions (40 questions Regarding internal consistency, the results showed that the Cronbach's alpha values, in general, reached those recommended in the literature (above 0.70), excepting the "Offensive behaviours" dimension. These findings are in agreement with the results of other studies [8][9][10][11][12][13][14] .
As for the floor and ceiling effect, our results for the floor effect are similar to those found in the original Danish study 16 and in the study that cross-culturally adaptated the instrument to Portuguese in Portugal 14 , both evaluating the long version of COPSOQ. Both studies found acceptable floor effects (less than 15%) for all dimensions, except for "Work family conflicts". Our values for the ceiling effect, in turn, differ from studies in the literature 14,16 , as two dimensions presented this effect. This finding can be explained by the type of worker included in the sample, since teachers, managers and administrative positions tend to score higher for these dimensions 15 .
The present study assessed construct validity by the correlations between the dimensions of the COPSOQ II-Br and the JCQ scales. These findings are in accordance with the COPSOQ proposal, which addresses, among others, the Karasek theory with equivalent questions in the questionnaire 15,16 .
The association between COPSOQ II-Br and musculoskeletal symptoms presetend adequate results, in agreement with other studies in the literature that noted the association between these psychosocial aspects and musculoskeletal symptoms [27][28][29] . Interestingly, in our study, some of the dimensions showed a negative correlation with the presence of musculoskeletal symptoms, i. e., the greater the presence of these psychosocial aspects in the workplace the lower the proportion of symptoms.
Hauke et al. 27 found that psychosocial working conditions such as low social support, job control, decision-making power, job satisfaction and high job demands, were associated with increased risk of lower back and neck/shoulder pain. A cross-sectional study conducted in Brazil, in turn, showed that low social support, high psychological demand and low job control increased the prevalence of multiregional symptoms 29 . Thus, the results obtained in this study were consistent with the literature findings.
Some limitations of this study concern aspects related to data collection. First, the online questionnaire had a low response rate (15%) when compared to the in-person format (75%). Second, the sample lacked balance regarding sex, which may have influenced the responses. Finally, the study did not include industry and retail workers, sectors that should be investigated in future studies. Therefore, this study was able to culturally-adapt the short version of COPSOQ II into Brazilian Portuguese and test its psychometric measures, whose results were considered suitable for using the instrument to assess the health of the population of Brazilian workers.