Abstract
Objectives:
Health literacy refers to the ability of individuals to gain access to, use, and understand health information and services in order to maintain a good health. The assessment of health literacy profiles in a population is potentially crucial to respond to health needs. The Health Literacy Questionnaire explores nine dimensions of health literacy and has been shown to display robust psychometric properties. The aim was to test the validity of the multidimensional Health Literacy Questionnaire and to describe the health literacy profiles in a French population at risk of cardiovascular disease.
Methods:
Data were collected using self-administered questionnaires from 175 participants attending health education and support programmes in local associations of patients in Paris. Analysis included scale reliability, confirmatory factor analysis, and health literacy profiles via descriptive statistics.
Results:
In confirmatory factor analysis, the nine-factor structure was close to the original Health Literacy Questionnaire. A nine-factor confirmatory factor analysis model was fitted to the 44 items with no cross-loadings or correlated residuals allowed. Given the restricted nature of the model, the fit was satisfactory: χ2WLSMV(866 df) = 1383.81, p = 0.0000, comparative fit index = 0.925, Tucker–Lewis index = 0.918, root mean square error of approximation = 0.058, weighted root mean square residual = 1.175. Composite reliability ranged from 0.77 to 0.91. Among the 9 scales of the Health Literacy Questionnaire, the highest scores were found for scale 1 ‘Feeling understood and supported by healthcare professionals’ and scale 9 ‘Understand health information enough to know what to do’ and the lowest for scale 2 ‘Having sufficient information to manage my health’ and scale 7 ‘Navigating the healthcare system’.
Conclusion:
The French version of the Health Literacy Questionnaire was shown to be psychometrically robust with good reliability. In the context of France, the 9 scales of Health Literacy Questionnaire allow a thorough assessment of health literacy strengths and weaknesses to respond to health literacy needs and improve the accessibility of health information and services.
Introduction
According to the World Health Organization, health literacy refers to ‘the cognitive and social skills which determine the motivation and ability of individuals to gain access to, understand and use information in ways which promote and maintain good health’. 1 Health literacy has been mainly assessed through functional tests of reading ability, understanding, and/or numeracy and has been linked with various health outcomes: poorer health status, increased mortality, increased hospital admissions, poorer medication adherence, risk factors for poor health, and increased healthcare costs. 2 Assessing the health literacy of individuals within a community, region or country can provide insight into the problems faced by individuals when trying to access health services and manage disease. 3 Early measures of health literacy include the Rapid Estimate of Adult Literacy in Medicine (REALM), 4 the Test of Functional Health Literacy in Adults (TOFHLA), 5 and the Newest Vital Sign. 6 These tools have important limitations: narrow range of health literacy dimensions, and insufficient power to show unbiased differences across groups.7,8 Recent tools have been developed which attempt to embrace the full breadth of dimensions embodied in the concept of health literacy. 9 The multidimensional Health Literacy Questionnaire (HLQ) has been developed on the theoretical basis of the World Health Organization (WHO) definition of health literacy, and its construction followed a grounded validity-driven approach. 10 The conceptual development of constructs and the cognitive testing of items benefitted from extensive patient engagement. 11 The HLQ explores health literacy through nine separate conceptually distinct scales whose psychometric properties have been showed to be robust in several languages.11–15
In France, while a few research teams are exploring health literacy in diabetes, cancer, rheumatic diseases, and its relations to health inequities, health and care pathways, and health education and prevention,16–24 to date no health literacy instrument has been developed or adapted in France with comparable properties as for the HLQ, to allow assessment in communities and across specific populations with the aim of adaptation of services and structures to the health needs of people struggling with disease management or prevention. The present work reports the French translation of the HLQ, the testing of its psychometric properties, and its usefulness in a population of individuals at risk for cardiovascular diseases attending counselling support and education in three community settings in France.
Methods
Health Literacy Questionnaire
The HLQ is composed of 44 items measuring 9 scales comprising 4–6 items each (Table 1). Scales 1–5 are scored on a 4-point Likert-type response scale (strongly disagree, disagree, agree, strongly agree) and scales 6–9 are scored on a 5-point Likert-type scale which was improved recently by the authors of the original questionnaire with response options focusing on difficulty (cannot do or always difficult, usually difficult, sometimes difficult, usually easy, always easy). 15
Health Literacy Questionnaire (HLQ) scales with high and low descriptors. 11
French language and cultural adaptation
The French translation of the HLQ followed a thorough procedure provided by the authors of the questionnaire. 25 Based on a detailed item intent document, the HLQ was first translated into French by (a) a professional interpreter (French–English bilingual and native French speaking) and (b) two French–English bilingual speakers (native French speakers) with broad experience in French local contexts in both rural and urban settings. Differences in the three forward translations were discussed by the French research team (X.D., M.B.D, D.B.) and a recommended draft French version was prepared and then blindly back translated by a bilingual English native translator. An HLQ author (R.H.O.) reviewed the final version. Nuances in French translation and the original English intent of each item were then discussed in detail in three consensus conferences with X.D., M.B.D., R.H.O., the translators, three representatives of civil society or local association of patients and one allied health professional with an interest in local cultures. Cultural relevance and measurement equivalence of items were discussed. With this team, and with reference to the detailed written intent of every item, the team was able to assess every nuance of the forward (French) translation from diverse lay and professional perspectives in order to achieve the best possible version for the French setting. Following the exhaustive translation and consensus phases, the final forward translation was applied in the field. No further improvements to the items were deemed necessary.
Setting and participants
Participants were recruited from February to October 2016 among voluntary individuals attending to the counselling and education services of three local associations of patients in Paris and suburbs (‘Maisons DOC: Diabète Obésité Cardiovasculaire’). These associations are dedicated to the health support and health education for persons with diabetes, obesity, or high cardiovascular risk. Participants gave their informed consent and fulfilled anonymously the questionnaire before attending to the education session or counselling encounter: self-administered paper questionnaire, with help for 10 participants (from health professional, n = 8; family, n = 1; other user of Maisons DOC, n = 1). In addition to the HLQ, other variables collected included sociodemographics (age, gender, living alone or not, employment status, educational attainment), and health conditions (diabetes, obesity, cardiovascular disease, and other associated chronic conditions: arthritis, cancer, other). Oral informed consent was obtained from all participants included in the study. In accordance with French regulation, a written consent was not required as the study was observational and strictly anonymised. Data collection procedures complied with the French regulatory rules and were declared to the national French CNIL (Commission Nationale Informatique et Libertés, n°1594321v0).
Sample size
A target number of 200 participants were first determined considering the feasibility of recruitment in the Maisons DOC over 6 months. A minimum sample size of 150 was however considered acceptable, as this study was a validation of a careful translation of an already well-researched questionnaire where all items were expected to load well on their target factors (and thus the communalities of all items would be anticipated to be high) and all target factors were anticipated to be quite strongly determined by between four and six variables. 26
Data analysis
Analyses were conducted with SAS version 9.4, Mplus version 7.4 and STATA version 13.1. Characteristics of participants were summarised by median (interquartile range) or percentages. Missing values and difficulty level by scale were determined by descriptive statistics produced for each item. For scales 1–5 (part 1 of the questionnaire), the difficulty level was calculated as the fraction of disagree/strongly disagree responses as against agree/strongly agree responses. For scales 6–9 (part 2 of the questionnaire), the difficulty level was calculated as the fraction responding cannot do or always difficult/ usually difficult/ sometimes difficult as against usually easy/always easy.
For scale reliability, Cronbach’s coefficient α was calculated, but completed by unbiased estimates of composite reliability. 27 Since the HLQ scales were stated a priori, confirmatory factor analysis (CFA) was used to test factor structure. A one-factor CFA model was fitted to the data for each scale. 11 The response options were scored as ordinal variables. Mplus provided the weighted least squares mean and variance adjusted (WLSMV) estimator, the unstandardised and standardised factor loadings as estimate to the variance in the measured variable explained by the latent variable (R2), and the associated standard errors, together with fit statistics (χ2, CFI – comparative fit index; TLI – Tucker–Lewis index; RMSEA – root mean square error of approximation; WRMR – weighted root mean square residual). Threshold values for the tests of ‘close fit’ used were CFI > 0.95; TLI > 0.95; RMSEA < 0.06; WRMR < 1.0. RMSEA < 0.08 was taken to indicate a ‘reasonable’ fit. A full nine-factor CFA model with no correlated residuals or cross-loadings was fitted to the data to investigate discriminant validity.
Means differences on the HLQ scores across a range of sociodemographic variables were determined using one-way analysis of variance (ANOVA). Effect size (ES) with 95% confidence interval (CI) for standardised differences in means between sociodemographic characteristics were detected using Cohen’s d. ES was considered ‘small’ when >0.20–0.50, ‘medium’ when 0.50–0.80, and ‘large’ >0.80. α risk was set to 1.67% using the Bonferroni correction (5%/3) for multiple comparison.
Results
Sociodemographic characteristics of the sample
A total of 198 participants were recruited, 6 declined, 17 could not complete the questionnaire (French language difficulties, n = 15; had no time to fill the questionnaire, n = 2). Finally, a total of 175 questionnaires were collected. Responses to the HLQ items were high: there was only one missing data point (item 9.1). Table 2 depicts the sociodemographic and health features of the participants. The median age was 66 years (extremes 18–87). In this sample, there were more women (76.6%) than men. This sample had a high level of education with 61.1% who had a university level. More than half of the people lived alone (50.3%) and were retired (56.6%). Metabolic, musculoskeletal, and cardiovascular diseases, depression or anxiety, cancer and asthma were the most frequent reported disorders.
Characteristics of participants.
IQR: interquartile range.
Median (IQR).
More than one possible.
Difficulty level
Difficulty level of items for each of the 9 scales of the translated HLQ is shown in Table 3. For scales 1–5 with 4-point response options (strongly disagree to strongly agree), scale 1 ‘Feeling understood and supported by healthcare providers’ showed the lowest difficulty level (average item difficulty 0.14), and scale 2 ‘Having sufficient information to manage my health’ showed the highest difficulty level (average item difficulty 0.39).
Difficulty level of the translated Health Literacy Questionnaire (HLQ) in a French population at risk for metabolic or cardiovascular disease.
CI: confidence interval.
Difficulty level was calculated as the proportion responding disagree and strongly disagree as against agree or strongly agree.
Difficulty level was calculated as the proportion responding cannot do, very difficult or quite difficult as against quite easy and very easy.
For scales 6–9 with 5-point response options (cannot do or always difficult to always easy), scale 9 ‘Understanding health information well enough to know what to do’ showed the lowest difficulty level (average item difficulty 0.41), and scale 7 ‘Navigating the healthcare system’ the highest level of difficulty (average item difficulty 0.62). In Part 1, the easiest item was on scale 3, ‘Actively managing my health’ (item 3.4 ‘I set my own goals about health and fitness’ (0.12)), and the hardest item on scale 2, ‘Having sufficient information to manage my health’ (item 2.3 ‘I am sure I have all the information I need to manage my health effectively’ (0.51)). In Part 2, the easiest item was on scale 9, ‘Understand health information well enough to know what to do’ (item 9.5 ‘Understand what healthcare providers are asking you to do’ (0.32)) and the hardest item in scale 7, ‘Navigating the healthcare system’ (item 7.1 ‘Find the right healthcare’ (0.69)).
Psychometrics properties
The psychometric properties of the HLQ are described in Table 4. The composite reliability was ⩾0.8 for all scales except for scale 9: ‘Understand health information well enough to know what to do’ (0.77). The scales with the highest composite reliability were scale 1: ‘Feeling understood and supported by healthcare providers’ (0.91), and scale 4: ‘Social support for health’ (0.90). Cronbach’s alpha ranged from 0.70 for scale 9 to 0.86 for scale 6.
Psychometric properties of the HLQ, French version, in a population sample at risk for cardiovascular disease, n = 175.
HLQ: Health Literacy Questionnaire; CI: confidence interval; CFI: comparative fit index; TLI: Tucker–Lewis index; RMSEA: root mean square error of approximation; WRMR: weighted root mean square residual.
The model fit for all scales was generally good except for scales 9. ‘Understand health information well enough to know what to do’. For this scale, CFI and TLI were low (<0.95), and RMSEA was high (>0.08). For each scale, factor loadings were satisfactory to high, with 40 out of the 44 items showing factor loadings of ⩾0.60 (ranging from 0.60 to 0.99). Four items were low: item 2.1. ‘I feel I have good information about health’ (0.55), item 3.4. ‘I set my own goals about health and fitness’ (0.57), item 5.4. ‘I know how to find out if the health information I receive is right for me or not’ (0.48), and item 9.2. ‘Accurately follow the instructions from healthcare providers’ (0.41).
A nine-factor CFA model was fitted to the 44 items with no cross-loadings or correlated residuals allowed. Given the very restricted nature of the model, the fit was quite satisfactory: χ2WLSMV(866 df) = 1383.81, p = 0.0000, CFI = 0.925, TLI = 0.918, RMSEA = 0.058, and WRMR = 1.175. While the CFI and TLI are lower than the pre-specified cut-off and the WRMR is higher, this is not surprising given the large numbers of parameters in the model set precisely to 0.0. The ranges of the factor loadings in this model were as follows: scale 1. ‘Feeling understood and supported by healthcare providers’: 0.74–0.99; scale 2. ‘Having sufficient information to manage my health’: 0.55–0.88; scale 3. ‘Actively managing my health’: 0.57–0.83; scale 4. ‘Social support for health’: 0.60–0.91; scale 5. ‘Appraisal of health information’: 0.48–0.76; scale 6. ‘Ability to actively engage with healthcare providers’: 0.69–0.85; scale 7. ‘Navigating the healthcare system’: 0.66–0.82; scale 8. ‘Ability to find good health information’: 0.69–0.83; scale 9. ‘Understand health information to know what to do’: 0.41–0.77.
Table 5 shows the inter-factor correlations. There was a clear discrimination between the scales in Part 1 with disagree/agree response options (range of 0.33–0.67), and a clear discrimination between the scales in Part 1 and the scales in Part 2 with cannot do/very easy response options (range of 0.35–0.70). However, discrimination was less marked between the scales in Part 2 (range of 0.69–0.86). Inter-factor correlations were >0.80 for scales 6 and 7 (0.81), 6 and 9 (0.83), 7 and 9 (0.83), 8 and 9 (0.86).
Inter-factor correlations in the nine-factor model.
Health literacy profiles of the sample
Distribution of the mean scores for the 9 scales of the HLQ is shown in Table 6. For the first five scales, the highest score was found for scale 1 ‘Feeling understood and supported by healthcare professionals’, and the lowest for scales 2 ‘Having sufficient information to manage my health’ and 4 ‘Social support for health’. For the remaining scales (6–9), the highest score was seen for scale 9 ‘Understand health information enough to know what to do’ and the lowest score for scale 7 ‘Navigating the healthcare system’.
Health Literacy Questionnaire scores for the overall sample. French population at risk for metabolic and cardiovascular disease (n = 175).
HLQ: Health Literacy Questionnaire; SD: standard deviation; CI: confidence interval.
One missing data in this scale.
Some patterns were found according to sociodemographic status of participants (Table 7). People with higher education (University vs Secondary school or less) were higher for scale 8 ‘Ability to find good health information’ (ES 0.41 (0.10–0.71); p = 0.01). People living alone had lower scores than those who were living with somebody for scale 4 ‘Social support for health’ (ES 0.52 (0.22; 0.82); p = 0.001). Weaker ES was found for age (<65 vs ⩾65 years) in scale 7 ‘Navigating the healthcare system’ (ES 0.30 (–0.01; 0.60); p = 0.056), and for employment status (Unemployed vs Employed) in scale 5 ‘Appraisal of health information’ (ES 0.32 (–0.01; 0.65); p = 0.059) and scale 7 ‘Navigating the healthcare system’ (ES −0.32 (–0.65; 0.01) p = 0.06). Employed people tended to have lower scores for 5 ‘Appraisal of health information’, but higher scores for 7 ‘Navigating the healthcare system’.
Relationships between sociodemographic variables and HLQ scores. . French version, in a population sample at risk for cardiovascular disease, n = 175.
HLQ: Health Literacy Questionnaire; SD: standard deviation; ANOVA: analysis of variance; ES: effect size; CI: confidence interval.
Welch’s ANOVA.
Self-reported chronic disease.
Discussion
The French version of the HLQ was translated and adapted from the English-language HLQ following a highly rigorous process. Filling of the self-administered questionnaires by participants was found to be easy, taking 10–15 min, and with a very few refusals. The HLQ has strong psychometric properties in this population sample from Paris and surrounding suburbs among people with elevated cardiovascular risk. In a highly restrictive model, CFA showed a nine-factor structure close to the original English language HLQ providing robust evidence of construct validity. All scales also demonstrated satisfactory reliability, with composite reliability >0.80, except for scale 9 (0.77). The HLQ had satisfactory psychometric properties, allowing the fine grained measurement of nine distinct domains of health literacy. 11 Other translated versions of the HLQ have been published with psychometric data analysis. The Danish, German, and Slovak versions of the HLQ demonstrated comparable strong properties.13–15
To our knowledge the present study is the first in France exploring health literacy using an instrument based on a modern conceptualisation of health literacy and including the concepts of access, use, understanding and appraisal, not only of health information, but also of health services, in order to gather information to inform responses to the needs of individuals, communities, regions or countries. In this sample of people with obesity, diabetes, or cardiovascular risk who were undertaking education sessions, health literacy levels were overall quite low. In particular, three constructs of health literacy appeared at stake: ‘Having sufficient information to manage my health’, ‘Social support for health’, and ‘Navigating the healthcare system’. On the other hand, higher levels of HL were found for ‘Feeling understood and supported by healthcare professionals’ and ‘Understand health information well enough to know what to do’. Older (>65 years) and unemployed or retired participants displayed higher difficulties in navigating the healthcare system. In the context of France, it could be important to look at the availability and accessibility of health information and the difficulty of understanding and becoming orientated in the healthcare system, while people appear to be generally confident with their primary care practitioner. 28 Our sample represented a quite selected population with diabetes or other cardiovascular risk since self-selected to attend the educational activities of the association, with potential demand on health system and sufficient delivery of information. 29 On another hand, the population studied here was characterised by a high proportion of potentially vulnerable women, retired or unemployed, living alone albeit with a relatively high level of education. Social support also appears to be too a crucial component of HL. Health inequalities have been shown to coexist with lower health literacy assessed via the HLQ in Australia, 30 Slovakia, 15 and Denmark. 14 In the European Health Literacy Survey (HLS-EU) survey, which only characterised populations into a limited number of categories of health literacy in eight European countries, health literacy levels varied markedly between countries, but limited health literacy was more frequent in vulnerable subgroups (social or economic deprivation, education or old age). 31
Difficulty levels among the 9 scales showed that the hardest items are in the second part of the questionnaire (scales 6–9). This distribution is quite different from the original English-language HLQ, 11 with difficulty levels in the items of part 2 ranging 8%–42% compared to 32%–69% for the French HLQ. The French results, as regards to part 2, are closer to the one observed in the Danish HLQ (12%–51%). 14 This might reflect greater challenges for French people in these domains, as suggested by lower scores compared to the Australian studies. 30 However, two other factors may be at play. The first is that the response options are slightly different to those in the original validation study as the version we used included the concept of frequency to make response options better reflect the real experience of people in the community. Indeed, in an Australian study of the measurement properties of the HLQ among older adults after a fall, 32 inspection of the category probability curves of the first version of the HLQ revealed that some participants struggled with understanding what might be the best response option for them between ‘very difficult’ and ‘quite difficult’. A team from Slovakia systematically compared the two versions, from both the difficulty perspective and the psychometrics, and found the new version to be superior as the analyses demonstrated clear response option choices and reduced ceiling effects across items. 15
A second potential confounder of the item and scale difficulty data is cultural variation across European cultures. While the somewhat onerous translation process (9 h of consensus discussion) with nuanced guidance of what every item is supposed to mean (and not mean) from the lay respondent’s perspective (i.e. the HLQ item intents), and also with detailed explication of minor idiomatic expressions or jargon in the English version, it is still possible that French people (or any other culture) see issues related to healthcare differently than other European or Australian cultures. France’s general attitude may be quite different from that of other countries in terms of the ease with which someone should be able to navigate the system, the level of support they should have, or the amount of information they should have, and therefore rate these as unsatisfactory across the relevant HLQ items (irrespective of whether they are adequate or not). Sociocultural issues can generate bias when comparing countries and diverse populations – akin to the ecological fallacy. 33 Health literacy involves to the ability of individuals to implement health-relevant practices, understand their meaning, and adapt these practices taking into account the strengths and limitations of environments and contexts in which they are actualised. 34 Health literacy of a person or a community is the result of the balance between, on one hand, the individual skills, taking into account the multiple constraints that a person face, including social positioning, gender, culture, and health services, and on another hand the relationships with health professionals, services, and system.34–37 Given these risks of such bias, it is recommended that comparing health literacy scores between cultures should not be undertaken prior to generation of robust evidence that unbiased estimates can be generated. This has only been done to a limited extent within one diverse setting in Australia. 12
This study has some limitations. First, the relatively small sample studied here, with a majority of retired individuals and women with relatively high proportion of social isolation cannot fully reflect the French population. Second, the questionnaire was self-administered by attendees of education sessions or during an encounter with a health educator; therefore, the sample may be biased towards people already productively involved in their own care and preventive activities. Future research should consider engaging populations in other settings such as primary care, the general community and people from diverse migrant or other marginised groups.
In conclusion, the findings from a sample of people at high risk for cardiovascular disease show the potential usefulness for the assessment of HL needs in individuals, settings, and has the potential to provide specific guidance for public health actions in France and French speaking countries. Our study provides a French validated version of the multidimensional Health Literacy Questionnaire. It has a good internal consistency in all 9 scales and is reliable and psychometrically robust in a highly restrictive nine-factor model.
Footnotes
Acknowledgements
The authors thank Audrey Sauger (interpreter: forward translation, consensus conference), Didier Balcou and Yves Derrien (forward translation, consensus conference), Tibor Katelbach (interpreter, backward translation, consensus conference), Gerardo Bautista and Véronique La Hausse (consensus conference), for their active participation in the translation and cultural adaptation process. The authors would also thank all the participants and staff of the Maisons DOC, and especially Marie-Laure Le Pommelet and Jacqueline Fillias, for their active involvement in the conduct of the study.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical approval
Ethical approval for this study was obtained from CNIL (Commission Nationale Informatique et Libertés) n°1594321v0.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the IRESP (Institut de Recherche en Santé Publique) (AAP 2012-20), and the National Health and Medical Research Council of Australia Senior Research Fellowship (APP1059122).
Informed consent
Verbal informed consent was obtained from all subjects before the study. In accordance with French regulation, a written consent was not required as the study was observational and strictly anonymized.
