Individual health literacy has been defined as “the degree to which people are able to access, understand, appraise and communicate information to engage with the demands of different health contexts to promote and maintain health across the life-course”.1 Limited health literacy has been linked to various adverse outcomes, including higher mortality, and is more common among the elderly, immigrants, and those with lower levels of education.2–5

In the past three decades, numerous instruments have been developed to screen for limited individual health literacy in research or clinical settings.6 The most widely used include the 66-item Rapid Estimate of Adult Literacy in Medicine7 (REALM), and the full and short versions of the Test of Functional Health Literacy in Adults (TOFHLA8 and STOFHLA9).3 Most of them were originally developed in English and are being adapted to other languages and populations.6 The REALM10 is a 125-item instrument developed as a fast screening tool to identify patients with limited abilities to read common medical and lay terms for body parts and illnesses. It presents words in ascending order of difficulty and is based on the idea that patients having trouble reading and pronouncing words probably will have issues with reading comprehension. The most commonly used is the reduced 66-item version7 that is frequently used to estimate patient reading levels (converting raw scores into grade equivalents) and tailor communication with patients accordingly.

In languages with very high letter to sound (phoneme-grapheme) correspondence, such as Spanish and Portuguese, the adaptation of health literacy assessment instruments based on word recognition and pronunciation, such as the REALM, is hindered by their inability to discriminate between health literacy and ability to read.11,12 The Short Assessment of Health Literacy for Spanish-speaking Adults (SAHLSA)13 was designed to overcome this issue by incorporating word comprehension. It has been adapted to Portuguese and validated in the Brazilian population as the Short Assessment of Health Literacy for Portuguese-speaking Adults (SAHLPA).14 These instruments have been proposed as less intimidating alternatives to assess health literacy in a clinical setting, and particularly helpful in assessing health literacy in the population groups most vulnerable to limited health literacy.13

In Portugal, limited health literacy has been estimated to affect between 4915 and 73%16 of the population. There is a lack of health literacy instruments designed specifically for low-literate populations that can be used to tailor health education interventions, as well as to study the impact of this social determinant of health.17 Because of its characteristics, brevity and ease of administration, we aimed to culturally adapt and validate SAHLPA in the Portuguese population.

Characteristics of the sample are summarized in Table 2. Women made up the majority of respondents in all validation groups except for the group of engineering researchers (p <0.001). Engineering researchers and laypersons from the general population were older (p <0.001).

The scree plot curve inflected at the first component, revealing a single dimension of the instrument (Fig. 1). This dimension explained 44,4% of the total variance. The global fit of the underlying model was good (CFI = 0.97 and RMSEA = 0.037; 90%CI: 0.030-0.043). Two items (SAHLPA 5 e 7) were removed because of a ceiling effect and 15 items because they had factor loadings below 0.5 (Table 1). The final version contained 33 items (SAHLPA-33) (see Appendix online to this article). SAHLPA-33 showed an adequate degree of reliability, with a Cronbach's alpha of 0.73 (95%CI: 0.68-0.78).

Figure 1.

Scree plot of eigenvalues after exploratory factor analysis.

(0.03MB).

The distributions of SAHLPA-33 scores were left skewed and with positive kurtosis in all validation groups, but with different distribution shapes (Fig. 2). The scores ranged from 24 to 33 in the general population subsample, 31 to 33 for researchers and 32 to 33 for physicians.

Figure 2.

SAHLPA-33 score distribution by validation group.

(0.12MB).

There was a statistically significant difference in mean ranks of SAHLPA-33 scores between the four validation groups (p <0.001), with the group of laypersons from the general population exhibiting a lower mean rank of scores than the other groups (all p <0.001), and the group of engineering researchers showing a lower rank of scores than physicians (p = 0.042).

Using health literacy as a binary variable, our data revealed evidence of an association between limited health literacy and validation group (Fisher's test p <0.001; p for trend of the original hypothesis <0.001). In regression analyses, limited health literacy was less common with increasing age, although not significantly (Table 3). There was a negative association between limited health literacy and education attainment (p for trend <0.001). The strongest association was observed for people with education attainment above the twelfth grade; they were significantly less likely to have limited health literacy when compared to people with education attainment below the ninth grade (sex and age-adjusted OR = 0.05; 95%CI: 0.02-0.15).

When considering only the subsample of laypersons, results were similar: there was a significant negative association between limited health literacy and education attainment (p for trend = 0.001) and no significant association with age, although the direction of the association was the same.

We adapted a brief and simple health literacy instrument to European Portuguese, and showed that it was valid and fairly reliable in the Portuguese population. Regarding construct validity, health literacy was significantly associated with health occupation and higher education attainment.

Our results revealed an evident left skew and positive kurtosis in the SAHLPA scores. This asymmetry in scores distribution was more pronounced in our study than in the Brazilian one, which may be explained by the use of a more diverse and literate sample: the average score of the 50-item SAHLPA in our sample was 6 points higher than that found in the Brazilian study, even when considering only the laypersons subsample (43.8; standard deviation [SD] = 4.4 vs. 37.7; SD = 9.0), and 9 points higher when considering the whole sample (46.7; SD = 3.8).14 Furthermore, validation of SAHLPA in the Brazilian population was restricted to patients over 60 years old (mean 74.4 years) and a quarter of the sample (25.7%) had less than 4 years of schooling.14 In contrast, our sample included participants between 18 and 86 years (median 38.5 years), and only 14.8% had less than 4 years of schooling. Hence, our study design oversampled people with higher health literacy, pushing scores to the upper end of the scale.

Our findings show that SAHLPA-33 is fairly reliable. The lower internal consistency (Cronbach's alpha = 0.73), when compared to that of the Brazilian SAHLPA-18 version (Cronbach's alpha = 0.90), could be explained by the lower variability in score distributions, that is known to underestimate the reliability.20

Although two different screening instruments previously validated in the Portuguese population were available,12,16 we decided not to test concurrent validity, because neither of them is considered a gold standard in health literacy assessment. Instead, our strategy relied on examining known-groups validity, that is, administering the instrument to different groups that logically should have different levels of the construct to confirm whether the hypothesized difference was reflected in the scores of the groups.21 Thus, we assumed health literacy would decrease across groups with progressively lower familiarity obtaining and processing health information, in the following order: physicians, health researchers, engineering researchers, laypersons from the general population. Although our data showed a significant trend (p <0.001), the instrument was better at discriminating people in the lower range of the health literacy spectrum, as it was designed to do.10,13

Less educated people tended to have lower health literacy, in accordance with results from previous studies.22

We were not able to find a significant association between limited health literacy and age. The magnitude and direction of this association appears to vary according to the type of assessment instrument used. A recent systematic review found that limited health literacy, when assessed using instruments based on medical vocabulary, such as the REALM (the precursor of SAHLPA), only weakly associates with older age.23 Instruments based on reading comprehension, reasoning, and numeracy skills, such as the NVS or the TOFHLA, in contrast, usually reveal positive associations between limited health literacy and age. The authors argue that crystallized cognitive abilities, such as the ones involved in word recognition and pronunciation, are not affected by aging-related decline, as opposed to those requiring fluid cognitive abilities, more related to reasoning and problem solving. In addition, it is also plausible that as people age and become more exposed to healthcare, their medical vocabulary increases, altering the traditional direction of the association between limited health literacy and age.24 Arguably, our study was underpowered to detect this association.

Some limitations are worth pointing out. SAHLPA is based on the REALM, a test popularly used to assess health literacy, but centred on reading skills. In fact, the REALM was not designed to assess health literacy but to estimate patient reading levels.7 Some authors have suggested that REALM scores should be treated as a correlate or predictor of health literacy and not as a measure of health literacy per se, because the instrument lacks coverage on three primary content areas of health literacy: comprehension, numeracy, and information seeking/navigation.25 SAHLSA and SAHLPA on the other hand, are seen as new instruments26 because they include comprehension of written health materials and thus have better content validity than the REALM. In addition, as is the case with other instruments that directly test individual abilities, they also do not take into account the abilities to interact, communicate or apply critical thinking, which are now included in definitions of health literacy.27,28 According to more recent guidelines, for adequate assessment of the structural validity (the degree to which scores are an adequate reflection of the dimensionality of the construct to be measured), the recommended sample size should be of at least five participants per item.29 Although we had only 3.92 participants per item, based on theory behind the instrument development and on the high bi-serial correlations between the items, it is highly unlikely the instrument could be multidimensional and assess other health literacy sub-dimensions in addition to the word comprehension sub-dimension.13,14,30 We did not examine test-retest reliability and future studies using SAHLPA in less literate samples should determine it. They should also help determine an appropriate cut-off to use health literacy as a binary variable. Future studies should also investigate the relationship between health literacy and gender, as there is a known gender gap in information seeking behaviour, i.e., women are more likely to engage in information seeking than men.31

An instrument based on the Brazilian SAHLPA-18 has been recently validated in the Portuguese population, adding five items to the shortened instrument, all of them drug-related.32 It is undetermined if the addition of these items significantly increased the difficulty of the instrument, rendering it less appropriate for less literate samples. SAHLPA-33 on the other hand, when compared with two other health literacy measurement instruments (the Newest Vital Sign16 and METER12) using item response theory has shown to have better discrimination and precision at lower levels of respondent ability (unpublished manuscript). Future studies should compare the factor structure between Brazilian and Portuguese populations, with both confirmatory factor analysis and differential item functioning. This comparison should also be done with the abovementioned Portuguese version.

SAHLPA-33 fills the gap in health literacy assessment instruments for Portuguese low-literate adults. In contrast to instruments based on self-assessment questions and more vulnerable to non-response bias,17 it offers an objective way to assess health literacy in this vulnerable group. National and international policies now recognize health literacy as a crucial determinant of health and are focusing on strategies to improve it.33–35 Although this instrument does not assess the health literacy demands imposed on individuals or the resources available to individuals and communities, i.e. their distributed health literacy,36 health literacy research in Portugal is very recent and brief assessment instruments are still useful to increase awareness and advance the field. We hope that the SAHLPA-33 can help support policy makers and clinicians in providing more effective health education, specifically targeted to low health literacy adults.

We have adapted a brief and simple instrument to assess health literacy in the Portuguese population. Future studies with less literate samples are needed to supplement and improve on this validation, before SAHLPA-33 is used to explore associations with health outcomes and to guide health interventions, especially in less literate populations. A cross-cultural validation should also be performed to allow comparisons between Brazilian and Portuguese samples, using SAHLPA-18 and SAHLPA-33. In addition, we recommend complementing it with instruments covering other dimensions of the health literacy construct: access, communication, and critical appraisal of health information to make decisions.

Enrique Castro-Sánchez.

The corresponding author on behalf of the other authors guarantee the accuracy, transparency and honesty of the data and information contained in the study, that no relevant information has been omitted and that all discrepancies between authors have been adequately resolved and described.

D. Paiva, A. Azevedo and N. Lunet conceived and designed the study. D. Paiva analyzed the data with support of M. Severo. All authors contributed substantially to the interpretation of the data. D. Paiva drafted the manuscript, which was critically revised by all the authors. All authors approved its final version.

This study was funded by FEDER through the Operational Programme Competitiveness and Internationalization and national funding from the Foundation for Science and Technology – FCT (Portuguese Ministry of Science, Technology and Higher Education) to the Unidade de Investigação em Epidemiologia - Instituto de Saúde Pública da Universidade do Porto (EPIUnit) (POCI-01-0145-FEDER-006862; Ref. UID/DTP/04750/2013); and the FCT Investigator contract IF/01674/2015 (to SS).

There are no conflicts of interest.