Validity and reliability of the Greek translation of the Job Satisfaction Survey (JSS)

Participants

The current study was conducted at the Therapy Center for Dependent Individuals (KETHEA), which is the largest rehabilitation and social reintegration network for drug addicts and their families in Greece, in the spring of 2015. Employees of all categories (administrative staff, therapy – prevention staff, education – research staff, part-time trainers and other staff) comprised the sample. Questionnaires were distributed to 341 employees and were completed by 239.

Instrument

The Job Satisfaction Survey (JSS) has 36 items with nine subscales to assess employee attitudes about their job and its different aspects. Each subscale is assessed with four items, while a total score is computed from all items. Each item is ranked on a 6-point Likert scale ranging from “strongly disagree” to “strongly agree”. The 36 items are written in both directions, so about half of them must be reverse scored. The nine subscales are Pay, Promotion, Supervision, Fringe Benefits (Monetary and nonmonetary fringe benefits), Contingent Rewards (performance-based rewards), Operating Procedures (Operating policies and required rules), Coworkers, Nature of Work, and Communication. Although the JSS was originally developed for use in human service organizations, it is applicable to a wide range of organization types in both public and private sector [11].

Scores on each of the nine-facet subscales, which in turn are based on 4 items each, can range from 4 to 24, and scores for total job satisfaction, which are based on the sum of all 36 items, can range from 36 to 216. JSS has 19 negatively worded items, which must be reversed. A score of 6 representing strongest agreement with a negatively worded item is considered equivalent to a score of 1 representing strongest disagreement on a positively worded item [11].

As far as scoring interpretation is concerned, two approaches, normative and absolute, can be used. The normative approach would compare the target sample to the norms for the sample, which are limited in three ways: first, there is a small number of occupations and organizations represented; second, the norms are not from representative samples; and third, the norms are mainly from North America—Canada and the U.S., which means that these norms are not representative of other countries that are culturally dissimilar to North America. According to the absolute approach, scores with a mean item response (after reverse scoring) of 4 or more represent satisfaction, mean responses of 3 or less represent dissatisfaction, whereas mean scores between 3 and 4, ambivalence. So, as far as the summed scores are concerned, for the 4-item subscales, scores of 4 to 12 represent dissatisfaction, 16 to 24 represent satisfaction, and those between 12 and 16 represent ambivalence. For the 36-item total, the ranges are 36 to 108 for dissatisfaction, 144 to 216 for satisfaction, and between 108 and 144 for ambivalence. If some items are missing, an adjustment must be done, otherwise the score will be too low. The best procedure is to compute the mean score per item for the individual, and substitute that mean for the missing items. An alternative but less accurate procedure is a middle response substitution for each of the missing items, where either 3 and 4 could be used [17].

Translation procedure

The forward–backward translation, which is the most commonly applied translation process for questionnaires or inventories, was performed [18]. In the first step of the procedure, the original English of the JSS was translated into Greek language by two experienced translators. The assessment of forward translation drafts was performed by two other researchers who were asked to review each translated item independently and choose the most adequate in terms of clarity, common language and cultural diversity.

The second step included retranslation of the agreed Greek text to English language by a researcher who had not previously seen the original version. The backward translation was compared with the original version of the survey, and judgments about the inaccuracies were made by two other researchers. The resulting differences were finally checked by another scientist who made the necessary adjustments.

The final version of the questionnaire was given to 12 volunteer participants (5 male and 7 female) for pilot testing. Each participant was given a brief introduction and requested to complete the Greek version of the questionnaire independently after which each one was interviewed about its clarity and understandability. The participants confirmed that the Greek version of the JSS was coherent and easy to fill in. The Greek and the English versions of the JSS are shown in the Additional file 1.

Statistical analysis

Descriptive statistics were used. Quantitative variables were expressed as mean values (SD) and qualitative as absolute and relative frequencies. A confirmatory factor analysis (CFA) with maximum likelihood procedure was conducted in order to test how well the dimensions of the JSS fit the data. The variance of the latent constructs was fixed at one during parameter estimation and the factors were allowed to be correlated. The fit of the CFA model was assessed using the chi square (χ2), the comparative fit index (CFI), the goodness of fit index (GFI) and the root mean square error of approximation (RMSEA) [19]. For the CFI and GFI indices, values close to or greater than 0.95 are taken to reflect a good fit to the data [20]. RMSEA values of less than 0.05 indicate a good fit and values as high as 0.08 indicate a reasonable fit [20]. Pearson coefficients were used to explore intercorrelations among subscales. Reliability analysis included Cronbach’s Alpha for internal consistency and Guttman Split-Half coefficient. Statistically significant level was set at .05 and the analysis was conducted using SPSS and AMOS (SPSS, Chicago, IL, USA) Statistical Software.

Sample characteristics

Descriptive statistics

Statistical analysis of validity

The CFA assessed the fit of the nine-factor structure and the model fitted the data well as defined from the RMSEA, CFI and GFI values that were equal to 0.055, 0.951 and 0.946, respectively. None of the item cross loadings exceeded the item loadings on the intended latent construct. Factor loadings were high and ranged from 0.61 to 0.90 indicating a strong association between the latent factors and their respective items.

Statistical analysis of reliability

Limitations

Like any other research in social sciences, the current study also has certain limitations. First, the sample was relatively small. Second, the data was collected only from employees working in drug addiction treatment; therefore, it can’t be generalized for all employees providing healthcare services. Third, almost half of KETHEA employees took part in the study, as they were the only ones available. Therefore, the extent to which results may be generalized is limited and should be treated cautiously. Forth, we did not estimate the convergent validity by comparing JSS with a similar tool and the discriminant validity by comparing JSS with a tool designed to measure a different concept. Finally, the use of the test retested the reliability method to examine the degree to which the results are consistent over time, as it was not feasible in the current field of research.