PSYCHOMETRIC VALIDITY OF THE HEALTH LITERACY QUESTIONNAIRE (HLQ) FOR BRAZILIAN PEOPLE WITH MENTAL DISORDERSABSTRACT
Objective: to analyze evidence of validity of the Brazilian version of the Health Literacy Questionnaire for use with people with mental disorders.
Method: a cross-sectional psychometric study with individuals undergoing treatment at a Psychosocial Care Center. Data collection took place between April and October 2023. The Sociodemographic and Health Conditions Questionnaire, Health Literacy Questionnaire, Health Literacy Scale-14, and Medical Outcomes Study Social Support Scale were applied. Reliability was assessed using McDonald’s omega and Cronbach’s alpha while the model was tested by adjusted chi-square and confirmatory factor analysis.
Results: a total of 444 people participated. The Health Literacy Questionnaire showed good internal consistency, with omega of parts 1 and 2, respectively, of 0.898 and 0.797, and alpha of 0.893 and 0.788, maintaining the 44-question model. Convergent, divergent and discriminant validity between scales were evidenced.
Conclusion: the Health Literacy Questionnaire demonstrated valid and reliable evidence for assessing health literacy in people with mental disorders. The nine-scale model was statistically adjusted.
DESCRIPTORS:
Health literacy. Mental disorders. Validation study. Psychometrics. Factor analysis; statistical
RESUMO
Objetivo: analisar as evidências de validade da versão brasileira do Health Literacy Questionnaire para aplicação em pessoas com transtornos mentais.
Método: estudo psicométrico, transversal, com pessoas em acompanhamento em Centro de Atenção Psicossocial. Coleta de dados ocorreu entre abril e outubro de 2023. Foram aplicados Questionário Sociodemográfico e de Condições de Saúde, Health Literacy Questionnaire, Health Literacy Scale-14 e Escala de Apoio Social do Medical Outcomes Study. A confiabilidade foi avaliada por Ômega de McDonalds e Alfa de Cronbach, enquanto o modelo foi testado pelo Qui-Quadrado ajustado e análise fatorial confirmatória.
Resultados: participaram 444 pessoas. O Health Literacy Questionnaire apresentou boa consistência interna, Ômega das partes 1 e 2, respectivamente, 0,898 e 0,797, e do alfa 0,893 e 0,788, mantendo o modelo de 44 questões. Validades convergente, divergente e discriminante entre escalas foram evidenciadas.
Conclusão: o Health Literacy Questionnaire demonstrou evidências válidas e confiáveis para avaliar o letramento em saúde em pessoas com transtornos mentais. O modelo com nove escalas mostrou-se estatisticamente ajustado.
DESCRITORES:
Letramento em saúde; Transtornos mentais; Estudo de validação; Psicometria; Análise fatorial
RESUMEN
Objetivo: analizar la evidencia de validez de la versión brasileña del Health Literacy Questionnaire para su aplicación a personas con trastornos mentales.
Método: estudio psicométrico, transversal, con personas en seguimiento en un Centro de Atención Psicosocial. La recolección de datos se realizó entre abril y octubre de 2023. Se aplicaron el Cuestionario de Condiciones Sociodemográficas y de Salud, el health literacy questionnaire, health literacy scale -14 y la medical outcomes study social support scale. La confiabilidad fue evaluada mediante Omega de McDonalds y Alfa de Cronbach, mientras que el modelo fue probado mediante chi-cuadrado ajustado y análisis factorial confirmatorio.
Resultados: participaron 444 personas. El health literacy questionnaire mostró buena consistencia interna, con omega de las partes 1 y 2, respectivamente, de 0,898 y 0,797, y alfa de 0,893 y 0,788, manteniéndose el modelo de 44 preguntas. Se evidenció validez convergente, divergente y discriminante entre escalas.
Conclusión: el health literacy questionnaire demostró evidencia válida y confiable para evaluar la alfabetización sanitaria en personas con trastornos mentales. El modelo con nueve escalas resultó ser estadísticamente ajustado.
DESCRIPTORES:
Alfabetización en salud; Trastornos mentales; Estudio de validación; Psicometría; Análisis factorial
INTRODUCTION
Mental disorders (MDs) are characterized by manifestations that alter an individual’s cognition, thinking, emotions, and behavior. They generally result in distress or impairment in fundamental areas of their psychological and biological functioning, and interfere with individuals’ lives in their personal, family, work, and social contexts1. Worldwide, MDs have been MOSt common diseases. It is estimated that one in eight people in the world present diagnostic criteria for MDs and behavioral disorders, which accounts for approximately one billion people2. When analyzing the Brazilian scenario, 23 million people are estimated and, of these, 5 million meet the criteria for moderate to severe disease, which corresponds to 3 % of the population with severe and persistent MDs3.
It is important to consider subjects based on their potential, awareness, freedom and responsibility, treating them as co-responsible for treatment, not only as a consumer of care and medication4. Users’ autonomy and empowerment in making health-related decisions depend on the ability to learn and use information received or provided in the care process5. Using information to enable subjects to make decisions regarding their care depends on individuals’ level of health literacy (HL)5.
HL is a multidimensional construct that encompasses functional, communicative and critical HL, which represents the progressive capacity that people have to understand health issues and exercise greater control over them. This multidimensional approach allows individuals to navigate the health system, having the ability to critically assess health information and identify strengths and weaknesses in population samples6.
Tools have been created to assess HL in different global contexts7. Among them, we have the Health Literacy Questionnaire (HLQ), used to understand the strengths and weaknesses of HL in different socioeconomic and ethnic scenarios8. HLQ was developed with a multidimensional profile to enable the exploration of several dimensions of the construct and to assess the population’s HL conditions, in addition to assisting managers in their decisions to qualify public health policies and, thus, strengthen the health network8. The Brazilian version of HLQ was called HLQ-Br (2021)5.
In international literature, few studies were found in which people with MDs were assessed. Among them, a study carried out an assessment of multidimensional HL, with people with MDs9. Another study, in Australia, obtained a satisfactory assessment regarding its applicability10. A study in Quebec demonstrated some weak points11. International studies on functional literacy (FL) with people with MDs showed discrepancies between the populations studied. Most studies identified that part of the population with MDs has low levels of FL12-13.
In the Brazilian context, a study was found that assessed the FL of people and adherence to treatment with antidepressants, assessing unidimensionally14. Another study assessed the association between sociodemographic variables and the HL levels of people with MD from HLQ-Br15.
Considering the knowledge gap and the importance of assessing multidimensional HL, it is necessary to expand research on HL assessment in the general population and, specifically, in people with MDs, since, to date, there are no validated instruments in Brazil for use in this segment. From this perspective, the use of HLQ-Br and its psychometric validity for this population segment is justified.
Based on this context, this study aimed to analyze evidence of validity of HLQ-Br for application to people with MDs.
METHOD
This is a methodological study with a cross-sectional design, carried out in a Psychosocial Care Center (In Portuguese, Centro de Atenção Psicossocial - CAPS) in a municipality with 83,947 inhabitants16, in the northwest region of Rio Grande do Sul, Brazil. The study included users aged 18 or over and diagnosed with severe, chronic and persistent MDs, recorded in medical records, in a stable clinical condition at the time of collection, previously assessed by health professionals from the team. Those who, in addition to MDs, had a diagnosis of intellectual or mental disability or retardation recorded in medical records, and who had a judicial interdiction, were excluded. The exclusion included diagnoses classified in ICD-10 codes F70 to F79.
Data collection was carried out between April and October 2023 by scientific initiation scholarship holders, previously trained by the first author, through reading and explanations of the items of each instrument, and solving doubts to standardize the collections. For sample calculation, we used sample size estimates of 90 % and an assumed margin of error of 9.5, considering a weighted population size of 720 people, with a minimum sample size of 432 respondents. A total of 512 people who met the inclusion criteria were invited to participate and, of these, 68 refused, resulting in 444, with a response rate of 87 %.
People were invited to participate in the study when they entered the unit. Upon acceptance, the Informed Consent Form (ICF) was signed in two copies of equal content. Then, the instruments were applied, ensuring participant privacy. The instruments used were Sociodemographic Questionnaire (sex, age group, race, color, marital status, cohabitation, personal education, maternal and paternal education, family income) and Health Conditions (time of follow-up in the service, diagnosis), prepared by the researchers, and the Health Literacy Questionnaire (HLQ)5,8, the Health Literacy Scale-14 (HLS-14)17-18 and the Medical Outcomes Study Social Support Scale (MOS-SSS)19-20.
HLQ is a multidimensional instrument that allows exploring several dimensions of the HL construct and assessing the Brazilian population’s HL conditions. It comprises 44 items, distributed across nine scales: 1. Feeling understood and supported by healthcare providers (4 items); 2. Having sufficient information to manage my health (4 items); 3. Actively managing my health (5 items); 4. Social support for health (5 items); 5. Assessment of health information (5 items); 6. Ability to interact actively with health professionals (5 items); 7. Navigating the healthcare system (6 items); 8. Ability to find good health information (5 items); 9. Understanding health information well enough to know what to do (5 items). HLQ does not provide an overall score for the questionnaire, but scores for each of the nine scales separately. The questionnaire assesses responses using a four-point Likert scale (strongly disagree, disagree, agree, and strongly agree) for the first five domains, referred to as Part 1. The last four scales are five-point scales whose items address the difficulty of performing a task (ranging from cannot do to always easy), and are referred to as Part 2. The scores indicate each person’s strengths and needs in relation to their HL5,8.
For convergent analysis, we used HLS-14, a multidimensional instrument for measuring HL. It consists of 14 questions, each assessed on a five-point Likert scale. The instrument assesses three distinct dimensions of literacy: the functional dimension (LF) and the communicative dimension (CoL), each with five items, and the critical dimension (CrL), with four items17-18.
For divergent analysis, MOS-SSS was used, a multidimensional instrument that assesses the social support (SS) perceived by users. It consists of 19 questions, distributed in five dimensions: tangible support; affectionate support; emotional/informational support, and positive social interaction. Participants answer each question based on the initial instruction “How often is each of the following kinds of support available to you if you need it?”, selecting one of the five options on a five-point Likert scale: 0 (none of the time) to 4 (all of the time). It is assumed that higher total scores indicate greater perception of the type of support19-20.
Data statistical treatment was performed using the IBM Statistical Package for Social Sciences version 25.0 (IBM SPSS Statistics 25) for Windows, combined with the AMOS module specific for structural equation modeling, to employ Confirmatory Factor Analysis (CFA) through Structural Equation Modeling (SEM). In the model estimated by CFA, the existence of outliers was verified through the squared Mahalanobis distance (D2), where the absence of these cases is assumed for the application of the technique. The assumption regarding normality was investigated, studying the univariate distribution through skewness (Sk) and kurtosis (Ku) as well as multivariate distribution (Mardia’s coefficient for multivariate kurtosis) (ISkI < 4.0 and IKuI < 10)21.
The model estimation considered the variance-covariance matrix and estimation by maximum likelihood. The validity considered the magnitude and statistical significance of the standardized coefficients. The model improvement strategy was adopted, in which correlations between errors in variables or construct were inserted22.
In determining the model quality, the absolute adjustment indices were assessed by chi-square (χ²), Root Mean Square Residual (RMR), Root Mean Square Error of Approximation (RMSEA), Goodness-of Fit Index (GFI) and by the comparative adjustment indices Comparative Fit Index (CFI), Normed Fit Index (NFI), Tucker-Lewis Index (TLI). For CFI, GFI, NFI and TLI, the literature suggests values greater than 0.950, while RMR and RMSEA should be below 0.080 and 0.060, respectively23-24. Unidimensionality was assessed based on standardized residuals related to the indicators of each latent variable, where the construct is considered unidimensional if it presents standardized residuals lower than 2.58 for a significance level of 5 %22.
The scale reliability study was carried out using Cronbach’s alpha (α) and McDonald’s omega (ω) coefficients. The coefficient (α) reflects the degree of covariance between items of a scale, while the estimate using McDonald’s omega (ω) tends to capture inter-item reliability, being less susceptible to bias22,25. Thus, like α, ω ranges from 0 to 1, and values considered acceptable are between 0.70 and 0.95. Values lower than 0.70 may not be sufficient to demonstrate internal consistency of an instrument, and very high values may indicate redundancy between items22,25. Evidence of convergent/divergent validity was analyzed by the linearity relationship between HLQ, HLS and MOS-SSS.
The research obtained prior authorization from the author who translated the instrument into the Brazilian version (HLQ-Br) and from the authors of HLQ by email (hl-info@swin.edu.au).
RESULTS
A total of 444 people participated, of which 75.9 % were female. MOSt represented age group was 55 to 64 years old, with 32.88 %. Moreover, 63.29 % of participants were in the age group of 24 to 54 years old, a period generally considered to be the population’s productive phase; 76.57 % declared themselves to be white; 58.10 % were without a partner; and 83.33 % lived with other people, such as a partner, children or people with other relationships. Regarding education, 31.76 % had completed high school; 53.60 % had incomplete or complete elementary school; and 12.16 % had completed higher education or graduate studies. In addition, 77.48 % of participants had a family income of up to two minimum wages.
Among participants with a medical diagnosis of severe MDs, the following prevalence was observed: depressive disorders (61 %); anxiety disorders (15 %); bipolar mood disorder (10 %); schizophrenia, schizotypal disorders and delusional disorders (11 %); personality and behavioral disorders (1 %); and other disorders (1 %).
As for the validity of HLQ scale structure, the basic assumptions for applying the technique were initially assessed. Evidence of univariate normality was identified, with absolute values of asymmetry and kurtosis estimated below 5 and 9, respectively. Since it presents an asymmetric multivariate distribution, the structural model was estimated using the maximum likelihood method.
The results obtained, based on the weight of the items’ individual loadings, indicate that, in the initial model, 23 items in part 1 saturated with a magnitude greater than 0.400 (p<0.001). The items with the highest weight were Q12 (Ꮧ=0.874), Q8 (Ꮧ=0.788), Q23 (Ꮧ=0.727) and Q19 (Ꮧ=0.756). Items with lower weights, i.e., items with less explanatory power, but no less important, were observed in Q21 (ꮧ=0.402), Q18 (Ꮧ=0.411) and Q13 (Ꮧ=0.422), all belonging to the D3 scale.
Concerning the estimated loadings for items in part 2, question 12 (P2 HLQ 12) saturated with a loading of 0.350, which does not compromise the robustness of the model, since, in most cases, the other items presented loadings higher than 0.500, reaching an explained variance of at least 25 %. Furthermore, regarding the loadings (or weights) of lesser expression, Q8 (Ꮧ=0.530) and Q19 (Ꮧ=0.541) stood out. In relation to the loads with the greatest explanatory power, they occurred in Q15 (Ꮧ=0.780), Q14 (Ꮧ=0.740) and Q13 (Ꮧ=0.690).
Regarding correlations between scales, it is worth noting that, in part 1, most estimates were classified as weak (0.100 < r≤0.300), as they indicate a low relationship or significant independence between scales, i.e., scales in this part of the scale are covering content from different concepts. The highest correlation was evidenced between scales D1 and D2 (r=0.501 - moderate correlation), i.e., among the scales mentioned, it was where the greatest conceptual similarity was identified (similarity defined as moderate). As for the correlation estimates between scales in part 2, correlations of moderate magnitude were observed between D6 and D7 (r=0.549) and D9 (r=0.518). These results indicate a moderate conceptual proximity between scales D6 compared to D7 and D9. However, such estimates were not excessively high (strong correlations) to the point of suggesting a factorial structure with a smaller number of scales.
Concerning the correlations between scales of parts 1 and 2, higher estimates reached moderate correlations (0.300 < r≤0.600), and these occurred between scale D6 compared to D1 (r=0.461) and D2 (r=0.471). The other correlations reached weak magnitude (r<0.300).
In relation to the factorial structure tested, Figure 1 shows the path diagram identifying the allocation of items in their respective scales.
Regarding the results achieved by the structure of the model with nine scales, quality indicators (or modification indexes) were observed, which make it possible to identify how much the sample was adjusted to the original structure.
Initially assessing the parsimonious fit of the sample [(X2/gl)] to the model, the results indicated minimal discrepancy between observed and estimated covariance matrices, since the fit value was less than 4.00 [(X2/gl)=3.849<4.0]. Thus, sample data, when adjusted to a nine-scale structure, were able to reproduce estimates very close to the real values observed in the scale items. Furthermore, in relation to the sample adjustment to the scale theoretical structure, it was investigated whether the degree of disagreement between the model and sample covariance is within the acceptable adjustment acceptance limits, through RMSEA. According to the result obtained [RMSEA = 0.053 < 0.08], we can believe that the observed discrepancies did not compromise the model credibility.
In the information relating to comparative indexes, where estimates above 0.900 define the model as acceptable, indicating that the estimated model proved to be efficient in its estimates, absolute and relative indicators were considered.
According to the results in Table 1, the absolute errors of GFI [GFI=0.913] and the Adjusted Goodness of Fit Index (AGFI) [AGFI=0.968] were above the acceptable minimum (0.900), indicating that the estimated structural model is satisfactory, where the errors identified, in comparison to the observed values, are not representative.
As for the relative indexes, which show how much better the proposed model fits the data, compared to the base model generated by AMOS (hypothetical model), it was found that estimates for TLI and CFI were above the acceptable cut-off point, with estimates of 0.906 and 0.924, respectively.
NFI assesses the proportion of the difference between the chi-square value of the proposed model in relation to a null model. According to Table 1, NFI was estimated at 0.924 (>0.900), indicating that the model for the sample studied does indeed present representative correlations that allow maintaining the proposed scales.
Based on the quality indicators obtained, it is worth noting that, in the analysis of data adjustment quality to the factorial model, at least three adequacy indexes must have values higher than the minimum required for a good adjustment. Hence, the model structured for the nine scales presented a large number of satisfactory indicators, indicating evidence of high quality of the factorial model tested.
Concwerning the results obtained for the total scale, ω values for parts 1 and 2 were, respectively, 0.898 and 0.797. Similarly, αC values, respectively, 0.893 and 0.788 were considered satisfactory (Table 2). According to reliability estimates, in relation to if-item-deleted, measured by αC, the exclusion of any item did not substantially alter the result of the internal consistency for the total scale.
The highest means were in scales D6 - Ability to actively engage with healthcare providers (3.82), D9 - Understanding health information well enough to know what to do (3.57), D7 - Navigating the healthcare system (3.56) and D8 - Ability to find good health information (3.51), respectively, corresponding to part 2 of the questionnaire. Lower means were evidenced in scales D5 - Appraisal of health information (2.60), D3 - Actively managing my health (2.70), D4 - Social support for health (2.75), D2 - Having sufficient information to manage my health (2.83) and D1 - Feeling understood and supported by healthcare providers (2.97), corresponding to Part 1.
Construct equivalence aims to demonstrate convergent and divergent validity through the relationship of the main scale with other measures. The scale was compared with HLS-14 and MOS-SSS. Regarding the comparisons made, correlations with at least moderate classifications were achieved, where positive coefficients in the comparison with HLS-14 identified convergent validity, while negative correlation coefficients with MOS-SSS indicated divergent validity.
Concerning the estimates achieved in equivalence scales, it was found that, for HLS-14, internal consistency was satisfactory both for the total scale (ω=0.755 and αC=0.738) and FL (ω=0.814 and αC=0.801) and CoL (ω=0.892 and αC=0.803) and CrL (ω=0.720) dimensions. Regarding the mean scores, it was identified that the highest scores occurred in the CoL (3.6±0.6) and CrL (3.6±0.7) dimensions, while the lowest estimate occurred in the FL (2.8±0.9) dimension. As for the estimate for the total scale, the mean reached 3.3 (SD=0.5) points.
In the information from MOS-SSS, there was internal consistency (αC>0.700) in all dimensions. The lowest estimate for McDonald’s omega (ω) was 0.877 and Cronbach’s alpha (α) was 0.864 in the tangible support dimension, while a higher coefficient was observed in the emotional/informational support dimension (ω=0.901).
To study the convergent validity of HLQ, HLS-14 was used as an external variable. It was found that correlations of greater magnitude occurred with scales D9 - Understanding health information well enough to know what to do (r=0.648; p<0.001) and D8 - Ability to find good information about health (r=0.630; p<0.001). These estimates indicated significant, positive and strong correlations (0.600<r ≤0.900), which indicate high scores on HLQ scales, and were related to high scores on the total HLS.
Correlations of lower magnitude, but no less important, were identified between the total HLS and scales D2 - Having sufficient information to manage my health (r=0.339; p<0.01), D3 - Actively managing my health (r=0.320; p<0.01), D5 - Appraisal of health information (r=0.540; p<0.001), D6 - Ability to actively engage with healthcare providers (r=0.338; p<0.01) and D7 - Navigating the healthcare system (r=0.375; p<0.01). Regarding these results, the correlations were classified as having moderate magnitude (0.300 < r≤0.600).
It is worth noting the comparison of total HLS with scale D1 - Feeling understood and supported by healthcare providers (r=0.206; p<0.01), where the correlation was weak, and scale D4 - Social support for health (r=0.054; p=0.258), indicating an absence of correlation.
In relation to the estimated correlations between HLQ scales compared to HLS-14 dimensions, D8 and D9 scales stood out, which showed correlations classified as moderate with all HLS-14 dimensions.
The convergent validity between the CoL dimension was evidenced in comparison to the scales as follows: D2 - Having sufficient information to manage my health (r=0.330; p<0.001); D3 - Actively managing my health (r=0.375; p<0.001); D5 - Appraisal of health information (r=0.469; p<0.001); D6 - Ability to actively engage with healthcare providers (r=0.383; p<0.001); and D7 - Navigating the healthcare system (r=0.368; p<0.001).
As for the investigation of convergent validity of FL and CrL dimensions, few correlations reached moderate magnitude. Thus, there is evidence of convergent validity between the total HLS and most of HLQ scales, a characteristic that was also present in the comparison of the CoL dimension with HLQ.
According to the results observed in Table 3, divergent validity was evidenced by the absence of significant correlations as well as by the presence of significant correlations of weak magnitude between the dimensions compared. The only HLQ scale that presented correlation estimates that were significant and of moderate magnitude, with all dimensions of MOS-SSS, was D4 - Social support for health. This fact is justified by the fragility of the support network perceived by service users, with direct implications for improving people’s health conditions.
The discriminant validity of HLQ was investigated in its ability to differentiate independent groups of subjects, who theoretically should present different mean scores, at different levels of education. The scores of the dimensions were compared in independent groups, stratified by level of education. Statistically significant differences were detected on the dimensions presented in Table 4, evidencing discriminant validity. It was identified that mean scores were significantly higher in levels of education with longer study time.
Regarding discriminant validity, the hypothesis that the scale is sensitive to people’s education level in some dimensions was considered. We know that better levels of education contribute to a better understanding of information and enable a better assessment of its quality, in order to contribute to more appropriate decision-making.
DISCUSSION
This is the first study conducted in Brazil that used a multidimensional instrument to assess HL in people with MDs. It proved to be valid and reliable in this population. It maintained the original structure, with 44 questions distributed in nine scales, according to original studies8 and in the Brazilian version5.
HLQ can be used to assess HL in Brazilian individuals with MDs, since the data matrix factorability demonstrated similarity with the original structure8. It presents similarity when compared to validity studies, as it demonstrates high internal consistency and reliability5,8,10,26-29.
The RMSEA 0.053 presented better indices than the original instrument8, but higher than that validated for use in the general Brazilian population5 and that demonstrated in Australia26 and China30. Approximate scores were found in validity studies carried out in Portugal28 and China30.
As in the original instrument and in other validity studies, CFI (0.924) and TLI (0.906) indexes were above the acceptable minimum (0.900)8,27-28, but lower than in other studies5,26,29, which indicate a good fit of the model.
It is worth noting that all dimensions presented McDonald’s omega and Cronbach’s alpha indicators with values above 0.76, similar to the Brazilian validity study5, except in D9 - Understanding health information well enough to know what to do (αC 0.688, ω 0.725), which presents questions related to correctly filling out medical forms, following instructions from health professionals, reading and understanding information on medicine labels and understanding what professionals are asking for, similar to studies carried out in Portugal with diabetic adults28 and in Brazil5.
In this dimension, the factor loading in Q12P2 - Able to read and understand written health information was the lowest (0.35), and it can be understood that, beyond the simple ability to read, understanding health information for decision-making is considered a communicative skill. Other studies presented reduced loadings in questions of D9; however, none showed a low factor loading in this question5,28. Mental health care goes beyond using medication correctly, as it requires cognitive, family, social, environmental and financial skills that are not always under a person’s own management.
Factor loadings above 0.40, but borderline, were evidenced in Q21P1 - There are things that I do regularly to become healthier (0.402), Q18P1 - I set my own goals about health and fitness (0.411) and Q13P1 - Despite other things happening in my life, I make time to stay healthy (0.422), corresponding to D3 - Actively managing my health, and classified according to Nutbean’s scheme as communicative HL6,8. A study carried out with diabetic people in Portugal showed indications of greater fragility in HL, when considering D3, since it refers to the capacity for active self-management of health care28. Considering that HL D3 - Actively managing my health and D4 - Social support for health were 2.70 and 2.75, respectively, which represents moderate HL, we can understand low factor loadings in this D3, since people have greater difficulty in managing their own care and are more dependent on other people, which, in many cases, is fragile or even non-existent.
Q8P2 - Get to see the healthcare providers I need to and Q19P2 - Work out what is the best care for you, items from D7 - Navigating the healthcare system, presented lower factor loadings, 0.530 and 0.541, respectively. Considering that the study participants were followed up by a public health service and that, due to the organization of the Health System itself, people generally do not have the possibility of choosing the health professional or service that will be responsible for their care. Generally, professionals make the referrals and municipal and/or state regulatory teams define the service/professional to which they will be referred, making it impossible for the person to have control over their care in this dimension. Although in different questions, when we refer to this dimension, a similar situation has already been evidenced28.
It is worth mentioning that the results found differ from the Brazilian translation and validity study5, as well as from the original8, where only Q16P1 - I know how to find out if the health information I receive is correct or not presented a low factor loading. The factor loadings as well as Cronbach’s alpha were different from the adapted and validated versions of the Brazilian HLQ and HLQ, which is acceptable, considering cultural particularities and clinical context of individuals.
HLS-14 and HLQ scales have convergence, even moderate in some dimensions, according to the results. We did not find any studies that had used both instruments to analyze this correlation. Therefore, it was not possible to compare the results found. When compared with MOS-SSS for divergent validity analysis, all presented divergence, except with D4 - Social support for health, which can be justified by the results found in the dimension itself, which expresses lower levels of SS that people with MDs present.
Discriminant validity was demonstrated when comparing education with some dimensions of the instrument, reinforcing the reliability and validity of this instrument as well as an international study that assessed the strengths and weaknesses of HL8. Other validity studies assessed the discriminant validity between the dimensions themselves5,8.
The fact that the interviews were conducted in person is considered a strength, since it minimized doubts regarding the question content and the fact that it was not conducted due to low literacy levels. At the same time, it is a limitation, since it may have influenced participants’ responses, especially in questions related to feeling understood and supported by healthcare providers. Furthermore, the fact that the study was conducted in only one health service and with a convenience sample is a limitation. New studies in different specialized mental health services and with a probabilistic sample and with self-administration of the questionnaire may contribute to the generalization of results.
CONCLUSION
The study showed that the data produced are valid and reliable for assessing HL in MD individuals. It presented RMSEA of 0.053, CFI of 0.924 and TLI of 0.906, in addition to Cronbach’s alpha and McDonald’s omega indicators with values above 0.76, except in D9 - Understanding health information well enough to know what to do (αC=0.688, ω=0.725). It maintained the structure of 44 questions, distributed in two parts and nine dimensions, enabling the assessment of multidimensional HL - functional, communicative and critical.
It showed convergent validity when compared with HLS-14, divergent validity with MOS-SSS, and discriminant validity when correlating HLQ dimensions with participants’ educational levels. We suggest that HLQ be used in other studies with people with MDs, enabling the assessment of the strengths and weaknesses of HL in all dimensions of the instrument. We emphasized the importance of using valid and reliable instruments to assess the HL in people with MDs, seeking to improve access, understanding, and use of information in health decision-making for this population with greater equity. In addition to this, it enables professionals and organizations to adapt their practice to contribute to qualifying mental health care.
Having a valid and reliable instrument to assess HL in people with MDs allows for a more accurate analysis of mental health practices, both by the health team and by the service. This resource contributes to establishing priorities in actions that foster and facilitate more assertive communication, strengthening the positive aspects of HL in this population and promoting alternatives to overcome the weaknesses identified in this context.
REFERENCES
-
1. World Health Organization (WHO). Health promotion glossary of terms 2021 [Internet]. 2021 [cited 2023 Oct 21]. Available from: https://iris.who.int/bitstream/handle/10665/350161/9789240038349-eng.pdf
» https://iris.who.int/bitstream/handle/10665/350161/9789240038349-eng.pdf -
2. World Health Organization (WHO). World mental health report: Transforming mental health for all [Internet]. 2022 [cited 2023 Oct 21]. Available from: https://www.who.int/publications/i/item/9789240049338
» https://www.who.int/publications/i/item/9789240049338 -
3. Lopes C de S. Como está a saúde mental dos brasileiros? A importância das coortes de nascimento para melhor compreensão do problema. Cad Saúde Pública [Internet]. 2020 [cited 2023 Oct 21];36(2):e00005020. Available from: https://doi.org/10.1590/0102-311X00005020 doi: 10.1590/0102-311X00005020
» https://doi.org/10.1590/0102-311X00005020» https://doi.org/10.1590/0102-311X00005020 -
4. Mendonça SM. Dignidade e autonomia do paciente com transtornos mentais. Rev Bioét [Internet]. 2019 [cited 2023 Oct 22];27(1):46-52. Available from: https://doi.org/10.1590/1983-80422019271285
» https://doi.org/10.1590/1983-80422019271285 -
5. Moraes KL, Brasil VV, Mialhe FL, Sampaio HA de C, Sousa ALL, Canhestro MR, et al. Validação do Health Literacy Questionnaire (HLQ) para o português brasileiro. Acta Paul Enferm [Internet]. 2021 [cited 2023 Oct 20];34:eAPE02171. Available from: https://doi.org/10.37689/acta-ape/2021AO02171
» https://doi.org/10.37689/acta-ape/2021AO02171 -
6. Nutbeam D. Health literacy as a public health goal: A challenge for contemporary health education and communication strategies into the 21st century. Health Promot Int [Internet]. 2000 [cited 2023 Oct 19];15(3):259-67. Available from: https://doi.org/10.1093/heapro/15.3.259
» https://doi.org/10.1093/heapro/15.3.259 -
7. Dodson S, Good S, Osborne RH. Health literacy toolkit for lowand middle-income countries: A series of information sheets to empower communities and strengthen health systems [Internet]. New Delhi, (IN): World Health Organization/Regional Office for South-East Asia; 2015 [cited 2023 Oct 18]. Available from: https://iris.who.int/bitstream/handle/10665/205244/B5148.pdf?sequence=1&isAllowed=y
» https://iris.who.int/bitstream/handle/10665/205244/B5148.pdf?sequence=1&isAllowed=y -
8. Osborne RH, Batterham RW, Elsworth GR, Hawkins M, Buchbinder R. The grounded psychometric development and initial validation of the Health Literacy Questionnaire (HLQ). BMC Public Health [Internet]. 2013 [cited 2023 Oct 18];13(1):658. Available from: https://doi.org/10.1186/1471-2458-13-658
» https://doi.org/10.1186/1471-2458-13-658 -
9. Friss K, Aaby A, Lasgaard M, Pedersen MH, Osborne RH, Maindal HT. Low health literacy and mortality in individuals with cardiovascular disease, chronic obstructive pulmonary disease, diabetes, and mental illness: A 6-year population-based follow-up study. Int J Environ Res Public Health [Internet]. 2020 [cited 2023 Oct 30];17(2):9399. Available from: https://doi.org/10.3390/ijerph17249399
» https://doi.org/10.3390/ijerph17249399 -
10. Degan TJ, Kelly PJ, Robinson LD, Deane FP, Wolstencroft, Turut S, et al. Health literacy in people living with mental illness: A latent profile analysis. Psychiatry Res [Internet]. 2019 [cited 2023 Oct 21];280:112499. Available from: https://doi.org/10.1016/j.psychres.2019.1124
» https://doi.org/10.1016/j.psychres.2019.1124 -
11. Fortin MP, Lavoie M, Dufour I, Chouinard MC. La littératie en santé chez les personnes présentant des troubles mentaux graves. Santé Mentale Québec [Internet]. 2020 [cited 2023 Oct 26];45(1):105-26. Available from: https://doi.org/10.7202/1070243ar
» https://doi.org/10.7202/1070243ar -
12. Clausen W, Galloway SW, Britigan DH, Baerentzen MB. Health literacy among people with serious mental illness. Community Ment Health J [Internet]. 2016 [cited 2023 Oct 25];52(4):399-405. Available from: https://doi.org/10.1007/s10597-015-9951-8
» https://doi.org/10.1007/s10597-015-9951-8 -
13. Bacon O, Vandenberg A, May ME. Provider and patient perception of psychiatry patient health literacy. Pharm Pract (Granada) [Internet]. 2017 [cited 2023 Oct 20];15(2):908. Available from: http://scielo.isciii.es/scielo.php?script=sci_arttext&pid=S1885-642X2017000200005&lng=es
» http://scielo.isciii.es/scielo.php?script=sci_arttext&pid=S1885-642X2017000200005&lng=es -
14. Sá AM de, Casagrande V, Chagas EFB, Pinheiro OL. Influence of functional health literacy on adherence to antidepressant treatment. Rev Gaúcha Enferm [Internet]. 2022 [cited 2024 Aug 02];43:e20210299. Available from: https://doi.org/10.1590/1983-1447.2022.20210299.en
» https://doi.org/10.1590/1983-1447.2022.20210299.en -
15. Rehfeld GEE, Oliveira RF, Kunkel GK, Nogaro A, Caçador AZF, Bernat KAC. Association between Sociodemographic Variables and Health Literacy Levels among İndividuals with Mental Disorders in Brazil. J Health Literacy [Internet]. 2024 [cited 2024 Aug 02];9(3):31-44. Available from: https://journals.mums.ac.ir/article_24659_a5cbbc5f7beb115642ced78eb6c80270.pdf
» https://journals.mums.ac.ir/article_24659_a5cbbc5f7beb115642ced78eb6c80270.pdf -
16. Instituto Brasileiro de Geografia e Estatística (IBGE). Censo Demográfico 2023 - Ijuí, RS [Internet]. 2023 [cited 2023 Oct 20]. Available from: https://cidades.ibge.gov.br/brasil/rs/ijui/panorama
» https://cidades.ibge.gov.br/brasil/rs/ijui/panorama -
17. Suka M, Odajima T, Kasai M, Igarashi A, Ishikawa H, Kusama M, et al. The 14-item health literacy scale for Japanese adults (HLS-14). Environ Health Prev Med [Internet]. 2013 [cited 2023 Oct 29];18(5):407-15. Available from: https://doi.org/10.1007/s12199-013-0340-z
» https://doi.org/10.1007/s12199-013-0340-z -
18. Batista MJ, Marques AC, Silva Jr MF, Alencar GP, De Sousa MD. Tradução, adaptação transcultural e avaliação psicométrica da versão em português (brasileiro) do 14-item Health Literacy Scale. Ciênc Saúde Coletiva [Internet]. 2020 [cited 2023 Oct 19];25(7):2847-57. Available from: https://doi.org/10.1590/1413-81232020257.22282018
» https://doi.org/10.1590/1413-81232020257.22282018 -
19. Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci Med [Internet]. 1991 [cited 2023 Oct 20];32(6):705-14. Available from: https://doi.org/10.1016/0277-9536(91)90150-b
» https://doi.org/10.1016/0277-9536(91)90150-b -
20. Griep RH, Chor D, Faerstein E, Werneck GL, Lopes CS. Construct validity of the Medical Outcomes Study's social support scale adapted to Portuguese in the Pró-Saúde Study. Cad Saúde Pública [Internet]. 2005 [cited 2023 Oct 20];21(3):703-14. Available from: https://doi.org/10.1590/S0102-311X2005000300004
» https://doi.org/10.1590/S0102-311X2005000300004 - 21. Kline RB. Principles and Practice of Structural Equation Modeling. New York, NY(US): The Guilford Press; 2012.
- 22. Hair J, Black W, Badin B. Análise multivariada de dados. São Paulo, SP(US): Bookman; 2009.
- 23. Byrne B. Structural equation modeling with AMOS: Basic concepts, applications, and programming. Trenton, NJ(US): Lawrence Erlbaum; 2010.
- 24. Brown TA. Confirmatory factor analysis for applied research. New York, NY(US): Guilford Press; 2012.
- 25. McDonald RP. Test theory: A unified treatment. Mahwah, NJ(US): Lawrence Erlbaum Associates Inc.; 1999.
-
26. Saleem A, Steadman KJ, Osborne RH, Caze AL. Translating and validating the Health Literacy Questionnaire into Urdu: A robust nine-dimension confirmatory factor model. Health Promot Int [Internet]. 2021 [cited 2023 Oct 24];36(5):1219-30. Available from: https://doi.org/10.1093/heapro/daaa149
» https://doi.org/10.1093/heapro/daaa149 -
27. Wahl AK, Hermansen A, Osborne RH, Larsen MH. A validation study of the Norwegian version of the Health Literacy Questionnaire: A robust nine-dimension factor model. Scand J Public Health [Internet]. 2021 [cited 2023 Oct 26];49(4):471-8. Available from: https://doi.org/10.1177/1403494820926428
» https://doi.org/10.1177/1403494820926428 -
28. Do Ó DN, Goes AR, Elsworth G, Raposo JF, Loureiro I, Osborne RH. Adaptação cultural e teste de validade da versão portuguesa do Health Literacy Questionnaire (HLQ). Int J Environ Res Public Health [Internet]. 2022 [cited 2023 Oct 20];19(11):6465. Available from: https://doi.org/10.3390/ijerph19116465
» https://doi.org/10.3390/ijerph19116465 -
29. Rababah JA, Al Hammouri MM, Aldalaykeh M. Validation and measurement invariance of the Arabic Health Literacy Questionnaire. Heliyon [Internet]. 2022 [cited 2024 Aug 02];8(4):e09301. Available from: https://doi.org/10.1016/j.heliyon.2022.e09301
» https://doi.org/10.1016/j.heliyon.2022.e09301 -
30. Huang Y, Ruan T, Yi Q, Wang T, Guo Z. The health literacy questionnaire among the aged in Changsha, China: Confirmatory factor analysis. BMC Public Health [Internet]. 2019 [cited 2024 Jul 02];19(1):1220. Available from: https://doi.org/10.1186/s12889-019-7563-x
» https://doi.org/10.1186/s12889-019-7563-x
NOTES
-
ORIGIN OF THE ARTICLE
Article extracted from the dissertation “Evidências de validade da versão brasileira do Health Literacy Questionnaire (Hlq-Br) em pessoas com transtornos mentais”, presented to the Graduate Program in Comprehensive Health Care, Universidade Regional do Noroeste do Estado do Rio Grande do Sul, in 2024.
-
FUNDING INFORMATION
The study had support from the Brazilian National Council for Scientific and Technological Development (In Portuguese, Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPq) Process 301694/2025-7 Productivity Grant to the last author and Scientific Initiation Scholarship Holders - Institutional Scientific Initiation Program (PIBIC) and Graduate Support Program (PROAP), Process 88881.993188/2024-01 and the Research Support Foundation of the State of Rio Grande do Sul (Fapergs), Researcher Gaúcho Program - PqG process 24/2551-0001530-2.
-
APPROVAL OF ETHICS COMMITTEE IN RESEARCH
Approved by the Ethics Committee in Research of the Universidade Regional do Noroeste do Estado do Rio Grande do Sul, Ijuí, Rio Grande do Sul, Brazil, Opinion 5.966.864/2022 and Certificate of Presentation for Ethical Consideration 66679223.7.0000.5350.
-
TRANSLATED BY
Agência Latintrad - Leonardo Parachú
-
DATA AVAILABILITY
All data supporting the results of this study have been published in the article itself.
Edited by
All data supporting the results of this study have been published in the article itself.
