Participants were recruited from among those registered with a Japanese Internet research company. The research company had approximately 2.5 million voluntarily registered participants. We wanted to collect data from a minimum of 1000 men and women aged 20 to 69 years. On March 11, 2013, potential respondents (n = 6613) were randomly invited via e-mail to participate in a cross-sectional Web-based anonymous health literacy questionnaire.
When determining who to invite to participate, we tried to match participants’ gender, age group and region (we divided the country into eight regions) to the results of the 2010 Japanese census . We accepted emailed responses from potential participants until we reach the targeted number in gender, age group and region. In all, we collected data from 1054 Japanese adults. Participants voluntarily signed an online informed consent form, approved by our institutional review board. The study received prior approval from the Research Ethics Committee of St. Luke’s International University, Japan.
Japanese version of the HLS-EU-Q47
To help validate the Japanese translation of the HLS-EU-Q47, we employed several experts in the field of translation: two Japanese translators (to translate the original English version of the survey into Japanese); three native English translators (to translate the new Japanese version of the HLS-EU-Q47 back into English); and a final translator who compared and checked the resulting Japanese- and English-language versions. For further validation, we hired a professional medical interpreter and a non-professional adult who was bilingual in English and Japanese to compare and check the surveys. We checked the translation process and took into consideration the opinions of these last two individuals, and then compiled the final Japanese version questionnaire (see Additional file 1).
The survey’s answer categories were all phrased similarly to “On a scale from very easy to very difficult, how easy would you say it is to understand why you need health screenings?” and were ranked on a four-point Likert-type scale (1 = very difficult, 2 = fairly difficult, 3 = fairly easy, 4 = very easy). In the original orally administered English-language questionnaire, a “don’t know” answer option was not provided, and was only used when stated spontaneously. However, as this survey was a self-administered questionnaire, we included the response “don’t know” to further help assess participants’ level of health literacy. The initial survey also included an item about health promotion efforts in the workplace. In our questionnaire, we added the response “don’t know/not applicable”; this response was coded as a missing value.
Health literacy indices are constructed as a general health literacy index (GEN-HL) comprising all items. This mechanism provides a general overview as well as three sub-indices: health care health literacy index (HC-HL), disease prevention health literacy index (DP-HL) and health promotion health literacy index (HP-HL).
Index scores, as with the original scale, were standardized on a metric between 0 and 50, using the formula: (MEAN-1) × (50/3) . MEAN is the mean of all item responses for each participant. Index scores were computed only for respondents who had rated (validly answered) at least 80 % of the items associated with all indices (n = 927).
We defined four levels of health literacy within these indices: 0–25 for “inadequate”, >25–33 for “problematic”, >33–42 for “sufficient” and >42–50 for “excellent”.
Confirmation of scale validity
We used two scales to confirm scale validity. The First was the Communicative and Critical Health Literacy (CCHL) scale, comprising three items for communicative health literacy (items i–iii) and two items for critical health literacy (items iv–v) . These items asked in Japanese whether participants would be able to (i) collect health-related information from various sources, (ii) extract the information they wanted, (iii) understand and communicate the obtained information, (iv) consider the credibility of the information and (v) make decisions based on the information, specifically in the context of health-related issues. Each item was rated on a five-point scale, ranging from 1, “strongly disagree”, to 5, “strongly agree”. The second was an assessment of eHealth literacy  using the Japanese version of eHEALS (J-eHEALS). This survey uses a five-point Likert scale (ranging from 1, strongly disagree, to 5, strongly agree; score range, 8–40) to measure perceived eHealth literacy.
Demographic and socioeconomic characteristics
The following demographic and socioeconomic characteristics were analyzed:
Gender (men, women)
Age groups (20–29, 30–39, 40–49, 50–59, 60–69)
Highest level of education (junior high school, high school, 2-year college, college/university, graduate school)
Annual pre-tax household income in millions of yen (<2.5, 2.5–3.5, 3.5–4.5, 4.5–6.0, 6.0–8.5, 8.5–12.5, ≥ 12.5, unknown)
Self-assessed living conditions (very hard, a little hard, common, a little well, very well) 
Occupation (self-employed, managerial and administrative, professional and technical, other regular staff [routine and manual], part-time, homemaker, student, unemployed)
Municipality size (very large [population of 500,000≤], large [100,000–499,999], moderate [<100,000], small)
Self-rated health (SRH) was measured by participants’ responses to the question, “Recently, how would you describe your state of health?” Five response categories were available: five = good, four = fairly good, three = fair, two = fairly poor and one = poor. We used the same question and response that appeared in a survey administered by the Ministry of Health, Labour and Welfare to be able to compare our findings against national Japanese data . Additionally, we included in our survey questions from the Five-Item Mental Health Inventory (MHI-5); this tool comprises the following five questions: “How much of the time during the past month (i) have you been a very nervous person?, (ii) have you felt calm and peaceful?, (iii) have you felt downhearted and blue?, (iv) have you been a happy person?, and (v) have you felt so down in the dumps that nothing could cheer you up?” [18, 19]. For these questions, we used a five-point Likert scale ranging from one (all the time) to five (never). All scores were totaled, with higher scores indicating better mental health.
Reliability and validity
Cronbach’s alphas were calculated to examine internal consistency. For construct validity, confirmatory factor analysis (CFA) was conducted separately for the three domains of health (health care, disease prevention, and health promotion). The number of factors was set to four related to the four information-processing competencies (accessing, understanding, appraising, and applying). In CFA, the Comparative Fit Index (CFI) and the root mean square error of approximation (RMSEA) were used as the model fit indices. A CFI value of .90 or larger is generally considered to indicate acceptable model fit. RMSEA value of less than .05 represents good fit, and a value < .08 is acceptable .
Construct validity was also assessed through the calculation of a Pearson or Spearman’s correlation coefficient between four health literacy indices (GEN-HL, HC-HL, DP-HL, HP-HL) and two other scale of health literacy (CCHL, J-eHEALS) and health status (SRH and MHI-5). When CCHL and J-eHEALS were compared with the HLS-EU-Q47, both scales were narrower concepts; CCHL because it did not contain a functional health literacy measurement and J-eHEALS because it was limited to electronic health information. No strong correlations were expected, but moderate correlations from .4 to .6 were thought likely. In the HLS-EU survey, the correlations between GEN-HL and self-assessed health varied from .15 to .33 across the eight countries , so similar correlations between health literacy indices and health status (measured by SRH and MHI-5) were expected.
As to GEN-HL, we compared the mean and standard deviation (SD) by demographic and socioeconomic characteristics and a multiple linear regression analysis was used to explore the associations between GEN-HL and demographic and socioeconomic characteristics (gender, age group, education, income, living conditions, occupation and municipality size). To assess GEN-HL’s associations with health status, multiple linear regression analyses were used to explore the associations between health status (SRH, MHI-5) and GEN-HL, together with demographic and socioeconomic characteristics.
Comparison between Japan and Europe
The European Health Literacy Survey (HLS-EU) was conducted in 2011 across eight European countries (Austria, Bulgaria, Germany, Greece, Ireland, Netherlands, Poland, and Spain); its findings are openly available online . The accompanying report shows the answer distributions of the survey’s 47 health literacy items, displaying the combined answer categories “very difficult” and “fairly difficult”, the mean percentage of respondents perceiving the items as difficult in the total sample of eight countries.
For the purposes of our research, using the published HLS-EU report as a starting point, we compared the rate of difficulty with health literacy items between Japan and Europe. The reason why we combined “very difficult” and “fairly difficult” in our analysis is that “very” and “fairly” had a subtle difference of nuance in the language of each country, making possible differences in responses. The judgment as to whether something was difficult or easy to understand is, however, rarely problematic in responses between languages. We also compared the mean and standard deviation of the four health literacy indices (GEN-HL, HC-HL, DP-HL, HP-HL) and the distribution of proportions of the categorized index.
Data were analyzed using IBM SPSS Statistics and Amos version 23.0.