Psychometric properties of the Korean version of the Health Literacy on Social Determinants of Health Questionnaire (K-HL-SDHQ)

The association between the social determinant of health (SDH) and sustainable development goals, has directed attention toward the influence of SDH. However, there is a lack of evidence regarding the instruments used to assess SDH. Thus, this study was conducted to assess the validity and reliability of the Korean Version of the Health Literacy on Social Determinants of Health Questionnaire (K-HL-SDHQ). A total of 660 workers in Korea participated in an online survey. The K-HL-SDHQ measures four dimensions (Access, Understand, Appraise, and Apply) with 33 items. The HL-SDHQ was translated into Korean using the forward-back translation method. To test the validity and reliability of the Korean translated HL-SDHQ, item analysis for the 33 items was conducted. Internal consistency was examined using Cronbach’s α, an exploratory factor analysis, and a confirmatory factor analysis. The scale-level content validity index (S-CVI)/universal agreement of this study was .12 and S-CVI/average was .83 (item-CVI range = .50–1.00). The goodness of fit determined through a confirmatory factor analysis of the four dimensions was acceptable (χ2 (489) = 1475.054, p < .001, RMSEA = .06, CFI = .87, TLI = .85). The K-HL-SDHQ also demonstrated satisfactory internal consistency reliability (Cronbach’s α = .92). The findings indicate that the K-HL-SDHQ is a valid and reliable tool that can be used to assess the SDH of workers in Korea. It is suggested that this tool can be applied through repeated research with workers and non-workers for health promotion, and to enhance researchers’ understanding of the different levels of the HL-SDHQ.


Introduction
Globally, health inequities due to the influence of social determinants of health (SDH) are a crucial issue warranting urgent actions to provide interventions for healthy living and wellbeing for all. SDH have gained recognition since the proposition of sustainable development a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Study design
This descriptive study was designed to ensure that the translation of the HL-SDHQ [13] into Korean was methodologically sound for use for the Korean population. The HL-SDHQ was originally developed by Matsumoto and Nakayama [13], two Japanese scholars who measured health literacy regarding SDH among workers. The original tool was found to be valid and reliable.

Participants
A convenience sample of 660 volunteers who registered to participate in an online panel in Korea was selected. Because most of the participants in the original study were workers, we also limited the sample to adults aged 20-64 years who were currently employed. Using a nonproportional quota sampling method, the sample was selected across regions (metropolitan areas and non-metropolitan), age groups, and gender. A previous study [15] suggested that the instrument related to health literacy does not appropriately identify the SDH of the elderly population in terms of their ability to understand and apply knowledge. Therefore, this present study excluded people over the age of 65 years. For the factor analysis, the sample size met the recommendation with a ratio of at least 20 cases per item [16].

Instrument
The HL-SDHQ was developed in Japan in 2017 by Matsumoto and Nakayama, and its validity and reliability with Japanese workers were examined [13]. The HL-SDHQ includes 33 items in 4 dimensions: Access (n = 7), Understand (n = 7), Appraise (n = 7), and Apply (n = 12). Each item is rated on a 4-point Likert scale ranging from "1 = very difficult" to "4 = very easy." In addition, to help assess participants' level of comprehension, an additional response category of "don't know/not applicable" was included and treated as a missing value. It was imputed by the average of each item following the guideline of the initial instrument developers. Each participant's score was summed and divided by the total number of items. Higher scores indicated higher HL-SDH. In addition, demographic and socioeconomic characteristics of participants including gender, age, living area, size of residential area, education level, marital status, living alone, and religion were also assessed.

Translation procedure
The HL-SDHQ, was originally written in Japanese, and did not examine the psychometric property tests of the English version of HL-SDHQ at the time of this study. Thus we were advised to translate the Japanese version to the Korean version by the original instrument developer. The Committee approach translation method is recommended to be appropriate in cases where linguistic and cultural differences are distinct [17]. In this study, however, the forward-back translation technique [18] was used because it was not available to have committee members who are have Japanese and Korean language expertise, in-depth knowledge of the social determinant health field, and translation process.
In the initial step, a bilingual translator, fluent in both Korean and Japanese and had an understanding of the cultural differences the two countries translated the Japanese version into Korean and then, another Korean-Japanese bilingual translator who did not know the original version, translated the Korean version back into Japanese. The exact meaning of the words and phrases are partly determined by cultural factors, so we instructed the translators to consider the different cultural and institutional contexts between the two countries. Next, an authorized Japanese translator compared the back translation with the original version, reviewed the differences and discrepancy in content, and confirmed the linguistic and conceptual equivalence of the Korean and Japanese versions. The Korean version was finally reviewed and reconfirmed for content clarity and cultural relevance by the first and second authors based on the English version reported by the instrument developer. Finally, a doctoral student majoring in Korean linguistics, reviewed the Korean version for the norms of the language and confirmed a final version. This version was named the Korean Health Literacy-Social Determinants of Health Questionnaire (K-HL-SDHQ).
The final version of the K-HL-SDHQ was tested with ten workers of different ages to ensure that the items were easily understandable, appropriate, and culturally sensitive. The Korean version questionnaire was emailed, and feedback was received by email or in person. One individual stated that the response categories ranging from "very easy" to "very difficult" were not familiar. However, most participants completed the questionnaire in 10-15 minutes without any assistance.

Data collection
The study was approved by the Institutional Review Board of the affiliated institution prior to initiating the study (IRB No. Y-2018-0001). Data were collected from voluntarily registered participants through an internet research panel institute, from May 30 to June 7, 2018. Before starting the online survey, the purpose of the research, anonymity and confidentiality were explained on the first screen. It was designed to begin the survey by clicking the "Next" button, indicating agreeing to participate voluntarily in the survey. When the questionnaire was completed, participants were given a token of appreciation for responding to the online survey.

Statistical analysis
The data were analyzed using SPSS 23 and AMOS 23. The participants' demographic characteristics were expressed as numbers, means, standard deviations (SD), and percentages. For item analysis, the mean, standard deviation, and item-total correlation (ITC) were used. The content validity of the K-HL-SDHQ was tested by calculating an item-level content validity index (I-CVIs) and a scale-level content validity index (S-CVI) according to Polit and Beck's recommendation [19]. In order to measure the content validity of the K-HL-SDHQ items, we invited 10 nursing professors who have teaching or research experience pertaining to SDH. Each item was evaluated using a 4-point Likert scale (1 = highly irrelevant, 2 = slightly irrelevant, 3 = relevant, 4 = highly relevant) to calculate the CVI and K � (modified kappa). The K � was calculated using the formula suggested by Polit et al. [20] to estimate the I-CVIs. Finally, S-CVI/universal agreement (UA) and S-CVI/average (Ave) were calculated to derive the S-CVI. The criteria used for evaluating each index are as follows: I-CVI � .78, S-CVI/UA and S-CVI/Ave � .80 [19].
To assess construct validity, an exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were conducted. EFA was performed to support construct validity by examining the characteristics of the factors and concept correspondence of each item [21]. EFA was conducted after following the Kaiser-Mayer-Olkin (KMO) measure and Bartlett's test of sphericity using the principal component method from Oblimin, an oblique rotation. After that, CFA was performed to assess model fit, convergent validity, and discriminant validity. CFA involved root mean square error of approximation (RMSEA), comparative fit index (CFI), Tucker-Lewis Index (TLI), and chi-squared test. The RMSEA is a measure of the average residual variance and covariance; good models have RMSEA values that are equal to or less than .08. The CFI is an index that falls between 0 and 1, with values greater than .90 considered to be indicators of good fitting models. Convergent validity indicates if the latent variables are consistently measured well. To present convergent validity, a model should have a factor loading value (λ) � 0.5, significance p < .05, average variance extracted (AVE) � 0.5, and construct reliability (CR) � 0.7 [22]. Discriminant validity was assessed when the AVE was more than the squared correlation coefficients [23]. To test the reliability of the Korean translated version, internal consistency using Cronbach's α for the items in each factor and for all items as a whole was assessed.

General characteristics of the sample
The mean age of the participants was 44.20 years with a standard deviation of 13.08 years. About 82.0% were college graduates, 56.5% were married, 20.5% were living alone, and 50.5% had no religion. Regarding the type of occupation, 39.1% of the participants were engaged in office work, 14.5% in service, and 13.9% in manual labor (Table 1).

Item analysis
An analysis of the 33 items of the K-HL-SDHQ showed that the mean of the total item scores was 1.92-3.05. Item 12: "Understand that work that is not stable becomes a huge stress" scored the highest. Item 27: "Approach the manager or the employer regarding rewards that do not match efforts done at work" scored the lowest. ITC values ranged from .316 to .570, which is considered acceptable. As a result, all 33 items were used in this study [24] ( Table 2).

Content validity
Ten experts evaluated content validity for the 33 items. I-CVI values ranged from 0.50 to 1.00, and 13 items that occupied the largest proportion of total items had an I-CVI of .90, followed by .80 for 8 items, and .70 for 7 items. Twenty-five items had I-CVI values � .78 and K � >.74, which indicated excellent validity as interpreted by Halek et al. [25]. Seven items had I-CVI values < .78 and � .60 and K � � .74, which was regarded as indicating good validity. Only one item was found to have I-CVI values < .60 and �.40 and K � � .59, indicating fair validity. I-CVI for 7 items was .70 was .50 for 1 item in this study. The values satisfied the standards which suggested that I-CVI values less .50 are unacceptable [20]. Therefore, all the 33 items were used in this study. The S-CVI/Ave was .83, making the content validity acceptable [19] ( Table 3).

Construct validity
Regarding the EFA of the 33 items, the KMO was high at .93 [26]. Bartlett's test of sphericity showed statistically significant results for EFA (KMO = .93, χ 2 = 7703.79, p < .001). The EFA was conducted using a principal component method to minimize data loss during factor extraction. Oblique rotation was performed by considering characteristics of HL-SDHQ that could be correlated with each item. Four fixed factors were extracted each with eigenvalues > 1.00. Items considered for these four factors were those with a loading value greater than 0.2 [27]. Factor details are shown in Table 4.
The factor loading was .553-.780 for factor 1, .388-.715 for factor 2, .256-.623 for factor 3, and .237-.730 for factor 4, which satisfied the standard [27]. The factor loadings were consistent with the corresponding domains in the original version. The explanatory power was 46.36% of the total variance, 27.76% for factor 1, 10.03% for factor 2, 4.58% for factor 3, and 3.99% for factor 4. These extracted factors were verified and named similar to that of the original version (Understand, Apply, Appraise, and Access) ( Table 4). Based on the EFA results, 4 components were extracted. Then, CFA was conducted using maximum likelihood to examine model fit. The findings showed that the model fit was good: χ 2 (489) = 1475.054, p < .001, CFI = .87, TLI = .85, RMSEA = .06 [22] (Table 4).
To test construct validity, convergent validity and discriminant validity were examined, which showed a low correlation and independence between sub-factors. First, convergent validity was .412-.764. When analyzed using AVE, the convergent validity was between .41-.55, and CR was between .82-.91, which partially met the standard of convergent validity. Second, discriminant validity was assessed which showed low correlation and independence between sub-factors. AVE was smaller than the squared correlation between Factors 1 and 2 and Factors 1 and 3, and so the criterion for acceptable discriminant validity was partially satisfied [23].

Reliability
Cronbach's α, which was .92 in Matsumoto and Nakayama's HL-SDHQ [13], was confirmed as .92 in the present study as well. Four subdomains of Understand, Apply, Appraise, and Access showed Cronbach's α values of .86, .86, .77, and .75, respectively (Table 4). Considering the large difference in the education level of the participants, the internal validity was examined by the education level. The findings showed minuscule differences; Cronbach's α was .92 in those with college education and .91 in those with less education, indicating that the instrument is equally appropriate for both groups.

Discussion
HL-SDHQ was developed originally in Japanese as a tool to measure and access health information regarding social determinants of health as well as to understand, evaluate, and apply this health information to promote health [13]. This study is the first translation after the development of the original instrument; there is no other language version yet. After the translation, the validity and reliability of the Korean version of HL-SDHQ was tested with an appropriate sample size of 660 Korean workers in Korea. The ITC between the items of analysis and total score was over .20; thus, 33 items were accepted and used in this study [13]. The 4 factors had the same theoretical structure in the EFA as they did in the original instrument. By linking SDG to health literacy, greater attention is now being paid to the influence of SDH on population health. However, there is still very little awareness of the ability of SDH. Therefore, it would be worthwhile to first introduce K-HL-SDHQ to SDH researchers and provide them with evidence for the applicability of HL-SDHQ to other countries. The 7 items from factor 1, Access category, in the original tool were extracted as factor 4 in this study, and the final 7 items were composed. Items 27, 28, 32, and 33 (original items-1, 2, 6, and 7) overlap with the Understand category in the original factor 1. This might be because "know", "find", and "understand" were considered as the same concept. The 7 items in factor 2, Understand category, of the original tool were extracted as factor 1 comprising 7 items. Factor 3 consisted of 7 items, and the 7 items in the Appraise category were extracted from the 3 factors of the study tool to comprise the final 7 items. It is assumed that items 15 and 20 included in the Understand category of Factor 1 are also in this study; it is considered bound to both sides because it is assumed to be a premise to understand the judgment. The 12 items in the Apply category of factor 4 in the study tool were extracted as factor 2 and the final 12 items were composed. "22. Cooperate in the creation of a fair society in which to live a healthy life" was included in the third factor Appraise on the assumption that cooperation is possible through the appraise process.   2. Understand that abuse suffered as a child has an impact even when one becomes an adult (9) .762 .764 3. Understand that being isolated from the community and workplace impacts health (10) .734 .742 4. Understand that determining how to proceed working on one's own is related to stress (11) .647 .657 5. Understand that work that is not stable becomes a huge stress (12) .780 .726 6. Understand that widening income disparities dilute the ties between people (13) .753 .682 7. Understand that in a society with a high level of stress, there is a tendency toward dependency on drugs (14) .672 .627 Apply 8. Cooperate in the creation of a fair society in which everyone can live a healthy life (22) .388 .478 .46 .91 9. Involve oneself in politics and public administration to help small children live a healthy life (23) .546 .579 10. Participate in childcare support activities (24) .600 .553 11. Participate in activities to eliminate poverty (25) .646 .620 12. Involve oneself in politics and public administration to protect the health of workers both institutionally and legally (26) .715 .685 13. Approach the manager or the employer regarding rewards that do not match efforts done at work (27) .595 .539 14. Participate in activities to increase employment and vocational training opportunities (28) .700 .637 15. Participate in activities that support an individual, including his or her family, who is in trouble in the community or workplace (29) .715 .665 16. Participate in activities to spread the importance of ties with people for health (30) .680 .602 17. Involve oneself in politics and public administration to make it easier for persons who have used illegal drugs to receive treatment (31) . 28. Find information related to the impact of the daily life of a mother on the growth of the child to be born (2) .494 29. Find someone who is isolated from society and whose health is failing (3) .575 30. Find information on the relation between unemployment and stress (4) .723 31. Find out the support required by someone in trouble in the community or workplace (5) .730 32. Find out smoking is not going to eliminate the cause of stress (6) .504 33. Find information about the relationship between dietary changes and health (7) .590 Eigen value 9.16 3.31 1.51 1.32 (Continued ) CFA confirmed the original 4-factor structure of HL-SDHQ, indicating satisfactory construct validity [28]. The criteria of convergent validity include � .70 CR and � .50 AVE. In this study, CR values were �.70 for all factors, indicating adequate reliability of internal consistency. Except for 2 factors, the AVE extracted was �.50, which means that the amount of variance is a too small to explain the latent construct. However, Fornell and Larcker [22] suggested that the use of CR alone is enough for confirming convergent validity. Thus, this study confirmed convergent validity of K-HL-SDHQ. In this study, the structure extracted was the same as the original tool; however, in the content validity test, the CVI of the 27th item ("Approach the manager or the employer regarding rewards that do not match efforts done at work") was found to be lowered to .50. This may be due to differences in the cultural backgrounds of the experts involved, which may affect the factor structure [29]. In Korean society, it is difficult to question the unreasonableness in the workplace as workers are reluctant to go directly to their employers even if they experience difficulties. Therefore, it is reasonable to consider this as one of the causes of low content validity. However, continued exposure to these organizational cultures increases job stress and is a major contributor to health risk factors such as smoking, drinking, and insomnia [30]. Therefore, as part of SDH, understanding the influence of the aforementioned item on health is an important approach [13,14]. Thus, it may be reasonable to retain the item for the worker population, which in turn suggests further research on the non-worker population.
In evaluating the internal consistency of the instrument, Cronbach's α for all the items was .92, which indicated it to be a highly reliable instrument. Cronbach's α ranged between .75-.86 in the reliability analysis of the four subscales. Therefore, it has proven to be a suitable tool with acceptable internal consistency reliability to measure the level of literacy regarding the determinants of Korean social health. In addition, the K-HL-SDHQ is a valid and reliable tool extracted from 4 factors similar to the original tool at the time of development. The validity and reliability of the K-HL-SDHQ were confirmed in this study. Thus, the results of this study are helpful to understand the level of HL and its social determinants among Korean workers and enable the comparison of the findings with the social HL levels of workers in other countries. Furthermore, the findings of this study can be used as basic data for developing strategies to improve health promotion behavior; more specifically, by focusing on the literacy level among decision makers in the field of social health.
This study has certain limitations. First, the participants were not diverse, as the sample was confined to workers aged between 20 and 64 years. EFA was used to test construct validity but extracted factors by eigenvalue rather than through a theoretical basis. The results are likely to vary according to the characteristics of the collected sample [31]. Therefore, it is necessary to confirm the validity of the tool according to the changes in the target group. It is also necessary to reflect on cultural considerations and the possibility that bias may have been involved in the process of evaluating the validity of the translated tools [32]. In this study, only nursing professors who had experience in research and education of SDH were examined. Thus, it is recommended that future research seek to examine content validity by consulting various experts with different backgrounds.

Conclusions
This study attempted to translate the HL-SDHQ [13] into Korean, which was originally developed in Japanese in 2017, and to test its validity and reliability so that it can be used to assess the literacy levels of workers' SDH. Results of the reliability and validity test of the K-HL-SDHQ revealed that the tool was valid and suitable for measuring HL-SDH among Korean workers. Based on the results in this study, it is possible to conduct comparative studies with other countries with different cultural characteristics. It is also expected that the tool can be used in the screening of literacy levels of Korean SDH and in developing interventions to improve the literacy of Korean SDH. It is suggested that this tool be applied through repeated research on workers and non-workers, so that the applicability and usefulness of the tool can be expanded.
Supporting information S1 Appendix. Korean version of the health literacy on social determinants of health questionnaire. (PDF)