Validation of the Greek translation of the multicultural quality of life index (MQLI-gr)

The concept quality of life has been widely applied in the field of medicine. Over the years, the phrase quality of life has broadened to include social functioning and well-being and it is now common to find quality of life measures in social and epidemiological studies [9, 10]. Numerous quality of life measures have been translated and validated in the Greek language including: Quality of Life in Epilepsy (QOLIE-31) [11], the Minnesota Life with Heart Failure Questionnaire (MLWHFQ) [12], the Stroke and Aphasia Quality of Life Scale (SAQOL-39 g) [13], the Transfusion Dependent for Thalassemia Quality of Life Questionnaire (TRANQoL) [14], and the Macular Degeneration Quality of Life (MacDQoL) [15]. The aforementioned QoL Greek translations have been used to examine the impact of heart disease, stroke, and other chronic health conditions on quality of life [11‐15]; however, they do not address overall health and social functioning [10], nor are they appropriate for use with the general population.

The Multicultural Quality of Life Index (MQLI) is a broad based scale and was originally developed in response to a dearth of culturally relevant measures to assess quality of life [10, 16-18]. The MQLI is a 10-item, self-report measure with items rated on a scale from one to ten (poor to excellent). It takes about 3 min to complete and is used to assess ten dimensions of life satisfaction: physical well-being; psychological/emotional well-being; self-care and independent functioning; occupational functioning; interpersonal functioning; social emotional support; community and services support; personal fulfillment; spiritual fulfillment; and overall quality of life [10]. The ten items were developed based on thematic analysis of identified dimensions in relevant international literature. In the literature it is described as one of the most comprehensive measures in its scope as it includes inquiries that range from physical well-being to spirituality. The MQLI demonstrates strong test-retest reliability (r = .87) among two cross-ethnic English samples [10] and has been validated in several languages including: Korean, Chinese and Spanish [16, 17, 193, 20]. Translated versions report strong internal consistency and test-retest reliability [e.g., English, α = 0.92, r = .87 [10]; Korean, α = 0.97, r = .85 [16]; Chinese, α = 0.94, r = .80 [17]; Spanish, α = 0.88, r = .94 [20]. The added value of the current study is that the MQLI-Gr, unlike prior QoL Greek translations, includes a broad set of health and social indicators. In addition, to the best of our knowledge, there are no Greek translations of the MQLI that have been validated on a general population of community-dwelling adults. The current study sought to address this gap in the literature.

The current validation study is part of a larger study that aimed to explore quality of life in Greece following the economic crisis [9]. The Institutional Review Board (IRB) approval was obtained from Springfield College. Next, two independent, professional translators - whose mother tongue was Greek-translated the MQLI from English to Greek according to cross-cultural adaptation procedures for self-report measures [21]. Translations were compared for discrepancies and translated back to the original language to ensure content validity. Discrepancies included minor wording clarifications; however, no changes were required in the conceptual content. The instrument was checked for semantic equivalency and a pilot test was conducted (N = 15). Based on pilot test feedback, revisions were made to the MQLI-Gr. It was reported to be easy to use and took about 3 min to complete. The survey was administered in-person and online (Survey Monkey) to Greek adults in a major Greek city. In order to recruit for online participants, a link to the study was posted on social media pages. Participants had to be 18-years or older, of Greek origin, and have lived in Greece for at least 5 yr [9]. To prevent participants from taking the survey twice they were able to access the link only once. That is, as participants completed the survey the system itself would not permit a second try. Recruitment for in-person participants took place in a central area of a large city in Greece. The research team using convenience sampling approached persons and requested their participation in the QoL study. If the participant agreed, the researcher administered the survey. The requirement to take the survey only once was reiterated at the time in-person surveys were administered.

After reviewing the purpose of the study, the risks/benefits of participation, and given assurances of confidentiality, participants provided informed consent. The following instruments were administered: (a) Multicultural Quality of Life Index-Greek version (MQLI-Gr), (b) Depression Anxiety Stress Scales (DASS-42), and (c) Brief Survey Form which included the demographic variables marital status, income, and employment [9].

Depression Anxiety Stress Scale (DASS-42) is a 42-item, self-report scale which provides separate scores for depression, anxiety, and stress. Each scale consists of 14 items rated on a 3-point Likert scale. The maximum score for each of the three subscales is 42 with higher scores indicating more severe symptoms. The higher the score the more severe the affective state [22]. The following scoring was based on reports by the DASS-42 authors and was used to differentiate two groups of respondents. One group reported none to mild symptoms of Depression (0–9), Anxiety (0–7) and Stress (score 0–14), the other group reported severe symptoms of Depression (28+) Anxiety (20+), and Stress (34+). DASS-42 takes about 10 min to complete and demonstrates strong test-retest reliability and construct validity [22‐25]. DASS-42 has been validated in several languages including Greek [25‐27] (Table 1).

The Statistical Package for the Social Sciences (SPSS) version 24 and the Mplus program were used for data analysis. The default estimator for CFA in Mplus is WLSMV (diagonally weighted least squares) and is appropriate for ordinal level data. Next, data were cleaned and checked for anomalies and frequencies run to check the accuracy of each variable. To assess internal reliability, internal structure, and construct validity, the following analyses were conducted: Cronbach’s α coefficient, confirmatory factor analysis (CFA), independent t-test, and ANOVA. A confirmatory factor analysis (CFA) [28] was conducted to evaluate a one-factor model (personal satisfaction) [18]. The following four indexes along with cut-off criteria [29] were used to assess model fit: the model Chi-Square [cut-off p > .05]; the comparative fit index (CFI) [cut-off > .90]; the root mean square error of approximation (RMSEA) [cut-off < .05 represents close approximate fit, results between .05 and .08 suggest reasonable error of approximation, and > .10 indicates poor fit] [29]; and the standardized root mean square residual (SRMR) [cut-off < .08].

The MQLI-Gr showed high internal consistency (Cronbach’s alpha = .90). Table 3 presents the results of four fit indexes achieved by the one-factor model tested by means of confirmatory factor analysis (CFA). Results of two indexes did not meet the criteria for good fit: the X² statistic (criteria p > .05, result = p < .001), and the RMSEA statistic (criteria <.08, result = .16). Two fit indexes met the criteria for good model fit: the CFI statistic (criteria >.90, result = .95), and the SRMR statistic (criteria <.08, result = .03). Correlated errors were also run; however, there were no substantive improvements in the model fit indices. Factor loadings on each of the individual items of MGLI-Gr were .56 and above except Community and Support Services (.31) [see Table 4.].

Pearson correlation tests were run on depression, anxiety, stress (DASS-42) and MQLI-Gr. Results showed an inverse relationship with depression (r = −.63, p < 0.001), 95% CI [.59, .67] anxiety (r = −.50, p < 0.001), 95% CI [.44, .55] and stress (r = −.51, p < 0.001), 95% CI [.45, .55] and the higher the score on depression, anxiety and stress, the lower the MQLI-Gr score.

DASS-42 was used to assess discriminant validity. Independent t-tests were run to compare the means of two groups hypothesized to differ in quality of life-those with none-mild symptoms versus those with severe symptoms of depression, anxiety, and stress. For all three subscales, there was a statistically significant difference in MQLI-Gr scores between the two groups, with the mild group scoring higher on MQLI-Gr [indicative of better quality of life] as compared to the severe group. Results are as follows: DM = 3.04, 95% CI [2.6, 3.4], t (102.43) = 14.99, p = 0.001; ANM = 2.34, 95% CI [1.9, 2.7], t (133.56) = 12.27, p = 0.001; SM = 2.45, 95% CI [1.9, 3.0], t (63.98) = 8.882, p = 0.001.

To further assess the ability of MQLI-Gr to discriminate between groups hypothesized to differ on quality of life, marital status, income, and employment were examined [3, 9, 30‐32]. Because the Levine test of homogeneity was violated for each variable (p < .05), the Welch ANOVA with post hoc Games-Howell test was used to interpret the results. The following results were statistically significant at p < .05 level. Results indicated those who reported single as compared to married were more likely to report lower quality of life; those who earned less than €499 were more likely to report lower quality of life; and students-unemployed were more likely to report lower quality of life [see Table 5].

This was the first step in the validation of a quality of life measure for community-dwelling Greek adults. As in any beginning endeavor, there were a number of limitations. As this was a non-representative sample, results should be interpreted cautiously. The use of non-probability, as opposed to probability sampling, was a significant limitation. Persons who self-selected to participate in the quality of life survey may not represent most Greek adults, and thus results are prone to biased estimates. Convenience sampling also prohibits generalizing beyond the sample, and thus results may be limited in overall applicability. Secondly, the use of online and in-person data collection may have caused ‘mode effects’ for which we did not control. As well, some variables such as Income had a high level of ‘missing values’ and we were not able to account for this non-response. Had we incorporated additional ‘missing value’ categories this may have been avoided. Another limitation was in terms of the internal structure of the MQLI-Gr where only two out of four indices indicated good model fit. Therefore, we recommend further work be done to strengthen internal structure--perhaps by continuing to explore the idea of a two-factor model. Finally, with additional tests our validation study of MQLI-Gr would have been strengthened. For example, a test-retest reliability procedure would have provided an indication of stability and reproducibility, and similarly, an assessment of criterion validity would have strengthened the validation of MQLI-Gr.