Health literacy in Peru: Transcultural adaptation and validation of a brief instrument adapted from HLS-EU-Q47

Abstract

Objective

To analyze the psychometric properties of the abbreviated version of HLS-EU-Q47 in the Peruvian population.

Design

Cross-sectional analytical design.

Setting

Two capital cities in the northern region of Peru.

Participants

1196 male and female residents aged 18–59.

Interventions

The translation and back-translation processes into the local language of the instrument under study were carried out. It was subsequently tested with a focus group, after which the final version was finalized.

Main outcome

Both exploratory and confirmatory construct validation of an abbreviated version of the HLS-EU-Q47 were conducted.

Results

The abbreviated version showed adequate internal consistency (McDonald's Omega) and a well-defined three-factor structure. Exploratory factor analysis yielded 12 items grouped into functional literacy, critical literacy, and health promotion/action, explaining 68% of the total variance, with strong factor loadings (0.58–0.95) and adequate communalities (0.58–0.82). Confirmatory factor analysis supported the three-factor correlated model showing adequate goodness-of-fit indices and confirming the structural validity of the instrument (Comparative Fit Index, CFI = 0.970; Tucker–Lewis Index, TLI = 0.961; Standardized Root Mean Square Residual, SRMR = 0.056); although Root Mean Square Error of Approximation, RMSEA was elevated (0.096), consistent with parsimonious models. Additionally, factorial invariance across sex and age groups was confirmed (ΔCFI ≤ 0.001; ΔRMSEA ≤ 0.006), supporting the stability and comparability of the instrument.

Conclusion

The short version of 12 items of the HLS-EU is a valid and reliable tool for the Peruvian population and is theoretically more powerful for integrating the functional and critical levels of health literacy.

Resumen

Objetivo

Analizar las propiedades psicométricas de la versión abreviada del HLS-EU-Q47 en la población peruana.

Diseño

Diseño analítico transversal.

Lugar

Dos capitales de la región norte del Perú.

Participantes

1196 residentes, hombres y mujeres, de 18 a 59 años.

Intervenciones

Se realizaron los procesos de traducción y retrotraducción al idioma local del instrumento en estudio. Posteriormente, se probó con un grupo focal, tras lo cual se finalizó la versión definitiva.

Mediciones principales

Se llevó a cabo la validación de constructo exploratoria y confirmatoria de la versión abreviada del HLS-EU-Q47.

Principales resultados

La versión abreviada mostró una consistencia interna adecuada (McDonald's Omega) y una estructura bien definida de tres factores. El análisis factorial exploratorio arrojó 12 ítems agrupados en alfabetización funcional, alfabetización crítica y promoción/acción de la salud, que explican el 68% de la varianza total, con fuertes cargas factoriales (0,58–0,95) y comunalidades adecuadas (0,58–0,82). El análisis factorial confirmatorio apoyó el modelo correlacionado de tres factores, mostrando índices de bondad de ajuste adecuados y confirmando la validez estructural del instrumento (Índice de Ajuste Comparativo, CFI = 0,970; Índice de Tucker-Lewis, TLI = 0,961; Residuo Cuadrático Medio Estandarizado, SRMR = 0,056); aunque el Error Cuadrático Medio de Aproximación, RMSEA fue elevado (0,096), que es consistente con modelos parsimoniosos. Además, se confirmó la invarianza factorial entre grupos de sexo y edad (ΔCFI≤0,001; ΔRMSEA≤0,006), lo que apoya la estabilidad y comparabilidad del instrumento.

Conclusión

La versión corta de 12 ítems del HLS-EU es una herramienta válida y confiable para la población peruana y, teóricamente, más eficaz para integrar los niveles funcional y crítico de alfabetización en salud.

Introduction

Health literacy is a key factor in improving population health indicators and reducing inequalities. The comprehensive definition of health literacy has been developed by Sorensen, who explains that it aims to imply knowledge, motivation, and skills to access, understand, evaluate, and apply information so that people can make judgments and decisions in terms of medical care, disease prevention, health promotion, and maintaining quality of life.¹ Adequate health literacy supports favorable health environments, efficient implementation of public policies aligned with health programs, and effective efforts to promote prevention, higher-quality health care, and lower costs for the population.¹ Likewise, health professionals must understand patients’ health literacy from the perspectives of their knowledge, motivation, and skills before designing and implementing health interventions to promote more effective health literacy.²

Misunderstandings about disease transmission, treatment, and prevention are often based on information from unofficial sources that constitute clear cases of disinformation. These contents can incite the population to intensify exaggerated actions or even to react to the disease with erroneous attitudes and behaviors, such as a lack of trust in health services, self-medication, and the use of ineffective prevention methods.³ In addition, these knowledge, beliefs, perceptions and thoughts are immersed in the adoption of behaviors, habits and behaviors of each person, therefore, the adoption of a new health behavior depends on the perception of risk of each person who, due to the influence of age, education and place of residence, can adopt a new behavior and acquire new knowledge.⁴ Additionally, these difficulties in understanding and using health information are not homogeneous across countries or populations, as they are influenced by sociocultural, educational, and structural factors specific to each context.

In Latin America, and particularly in Peru, a persistent infodemic has emerged since the COVID-19 pandemic, significantly impacting public health decision-making. This overexposure to misinformation or unverified information has influenced people's attitudes and behaviors, generating distrust in health services and prevention methods. Despite strategies implemented to control misinformation and strengthen health communication, its effects and impact remain perceptible in various social contexts. In this scenario, it is essential to promote the use of tools, methods, and interventions aimed at strengthening health literacy, in order to foster informed decisions and healthy behaviors among the population.⁵

Several instruments have been developed to measure health literacy in community and clinical settings, such as the Test of Functional Health Literacy in Adults (TOFHLA), which evaluates the ability to understand health information,⁶ and the Rapid Estimation of Literacy in Adult Medicine (REALM), which evaluates the ability to read health terms,⁷ the scale (eHEALS) that evaluates digital literacy and the skills to find, evaluate and apply knowledge.⁸ However, comprehensive tools for measuring health literacy across settings and populations are scarce. The European Health Literacy Survey (HLS-EU-Q47) allows evaluating three health domains (medical care, disease prevention and health promotion) and includes four information processing subdomains (find, understand, judge and apply) and can be applied to different contexts, populations and communities,⁹ being used in different countries and in different populations (adolescents, adults and older adults).^10–15 Several studies have shown that health literacy instruments require cross-cultural adaptation and psychometric validation processes specific to each national context, due to cultural, linguistic, and structural differences in health systems.^10,11 Research conducted in Europe, Asia, and Latin America has reported variations in the factor structure and behavior of the items of the HLS-EU-Q47 and its abbreviated versions, demonstrating that their psychometric properties cannot be assumed to be equivalent across populations.^9,16 In this sense, validation in the Peruvian context is necessary to ensure an adequate measurement of health literacy, considering the educational, sociocultural, and information and health service access inequalities present in the country.

Having a validated instrument will allow us to accurately assess health literacy levels in the Peruvian population, identify relevant gaps, and guide the design of interventions and public policies based on evidence, thereby improving health decision-making and reducing inequalities in access to and use of health information.

In this context, this research aims to analyze the psychometric properties of an abbreviated version of the HLS-EU-Q47, translated and adapted into Spanish, in the Peruvian population.

Methodology

Design

A quantitative instrumental study with a cross-sectional design was conducted.

Participants

The research was conducted between July and October 2025, with the participation of adult residents of two major cities in Peru (Trujillo and Cajamarca). The sample was selected using simple random sampling, with parameters set at a 95% confidence level, a 4% margin of error, and maximum variability (p = 0.50). The sample size for Trujillo was 601. The same parameters were applied to Cajamarca, resulting in a final sample size of 599. Data collection took place in homes in different districts of both cities. The sample size was estimated at 1200 residents, taking into account the possibility of attrition. Inclusion criteria were adults between 18 and 90 years of age, residents of the cities of Trujillo and Cajamarca, of both sexes, who agreed to participate in the study and signed the informed consent form. Exclusion criteria included individuals with cognitive impairment, difficulties understanding the research objectives, severe medical conditions preventing participation, incomplete questionnaires, and persons who were not residents of the selected cities or were under 18 years of age. During data collection, it was observed that 4 participants did not fully complete their questionnaires and were therefore excluded from the study. Thus, the final sample size was 1196.

Instrument

The health literacy instrument was created by Sorensen et al.⁹ This survey was validated across Europe in various countries and languages (Austria, Bulgaria, Germany, Greece, Ireland, the Netherlands, Poland, and Spain) and subsequently unified into a final English version.^10,14,15 The instrument comprises 47 items and assesses four dimensions of information processing (access, understanding, evaluation, and application of health information for decision-making), organized into three health domains: health care, disease prevention, and health promotion. Responses are formatted as a 4-point Likert scale (1 = very difficult, 2 = difficult, 3 = easy, and 4 = very easy) and an option 5 = I do not know is chosen by the interviewer.

Procedures

Direct translation and back translation: The translation and back-translation processes were conducted by two bilingual professionals. Both translators had specialized training in official and academic translation, more than ten years of professional experience, and previous participation in research and cross-cultural adaptation of instruments. During this stage, two versions of the questionnaire were obtained.

Translation synthesis: The sequential debate between the translators and the team of experts resulted in an agreement that produced the initial Spanish edition of the questionnaire, taking into account Peru's cultural aspects. There were no significant semantic or conceptual changes due to back translation; however, during adaptation, the term “prospectus” was changed to “bull” in item 6 because it is an uncommon term not used by the Peruvian population. In addition, in item 08, the word “pharmaceutical” was removed, as in Peru, it is prohibited for a pharmacist to issue indications or instructions to patients or the public. Such changes were made in accordance with expert recommendations, as reported in the Original study.¹⁷

Consolidation by a committee of experts: A consensus was reached among participants in the previous phases, all of whom were professionals with doctoral degrees in various areas (biostatistics, psychometrics, and public health), and a final document, including a Spanish translation, was produced, forming the final version.

Pre-test: In a pilot study, the questionnaire's legibility was evaluated with 20 participants who met the sample requirements. The study's subject was explained to the participants, and, under supervision, the questionnaires were completed using the self-report method. The process took between 15 and 20 min each. Finally, a group discussion was held with all participants to ensure linguistic and cultural relevance.

Subsequently, the questionnaire was distributed to residents of two cities who met the predefined inclusion criteria. With this data, the internal validity of the Spanish version of the questionnaire was established.

Data analysis

The data collected through the forms were entered into an SAV file, considering the template information and the respective codes for the response options of the categorical items. The reading was performed using the software chosen for the statistical analysis, and the cleaning process involved eliminating cases with at least one missing value in the main variable.

The descriptive results were used to characterize participants based on the global dataset of 1196 cases; descriptive statistics for each health literacy item were also calculated. For exploratory and confirmatory factor analysis, two data subsets were created, each with 598 cases.

The evaluation criteria based on the descriptive results were: Asymmetry and Kurtosis between −2 and +2, and a corrected homogeneity index greater than 0.30.^17,18 The exploratory factor analysis was conducted on the first subset, for which a polychoric correlation matrix was calculated because the items were ordinal. The Optimal coordinates and Parallel analysis methods determined 8 factors through the n factors function. However, there were problems of high complexity, cross-loadings with values greater than 0.3, cross-loadings differences less than 0.2, some factors with a low proportion of explained variance, and correlations between factors that suggest excessive fragmentation. The exploratory factor analysis was carried out using the minimum residual extraction (MINRES) and oblique rotation methods recommended for Likert scales.^19,20 Items were evaluated according to established psychometric standards commonly recommended in exploratory factor analysis literature, including factor loadings ≥ 0.30, communalities ≥ 0.40, and factorial complexity ≤ 2 for the initial analyses, while stricter criteria (loadings and communalities ≥ 0.50) were applied for the short version development.^21,22 These thresholds were used to ensure conceptual coherence and adequate psychometric performance of the retained items.

This situation led the researchers to conduct confirmatory factor analyses directly on the second data set, based on the theoretical models proposed by previous research. A unifactorial model with 3 factors, a 4-factor model, and a hierarchical model with 4 first-order factors were tested; the bifactorial model was not estimated due to data problems. The fit indices considered for confirmatory factor analysis were the Comparative Fit Index (CFI) and Tucker–Lewis Index (TLI), with values ≥ 0.95 indicating good model fit; the Root Mean Square Error of Approximation (RMSEA), with values ≤ 0.06 indicating low approximation error; and the Standardized Root Mean Square Residual (SRMR), with values ≤ 0.08 reflecting adequate residual fit.²³ As adjustment indices were not reported within the parameters, it was decided to evaluate the statistical indicators in exploratory and confirmatory factor analyses, as well as the relevance of each item within its factor, in line with theoretical adjustment. It was chosen to consider 4 items per factor for the short version. All analyses were performed using R version 4.5.1 and RStudio IDE version 2025.09.1+401.

Ethical aspects

This research was approved by the Ethics Committee of the Health Sciences School of the Peruvian Union University, with the certificate number code 2025-CEB-FCS-UPEU No. 179. Respondents’ participation was through a consent form. Participation was anonymous, voluntary, and risk-free for participants. Ethical principles were respected, as well as national and international ethical standards in accordance with the Declaration of Helsinki (2000).

Results

Table 1 presents the sociodemographic characteristics of the 1196 participants. The majority of participants had a single marital status (56.7%) and presented mainly secondary education (52.2%), followed by technical or university training (22.8%). Regarding occupation, 33.9% were employed, while 28.1% were unemployed. Almost half of the participants reported not receiving family income (47.0%). Own housing was the most frequent (57.6%), and the origin was distributed evenly between Trujillo (50.3%) and Cajamarca (49.7%). The average age was 29.53 ± 10.49 years, and the average number of people living with them was 3.02 ± 1.96.

Table 1.

Sociodemographic characteristics of the participants.

Variables and categories	n	%
Sex
Male	554	46.3
Female	642	53.7
Marital status
Single	678	56.7
Married	480	40.1
Widowed	17	1.4
Divorced	15	1.3
Cohabitant	6	0.5
Educational level
Illiterate	7	0.6
Elementary incompleted	130	10.9
Elementary completed	163	13.6
Secondary incompleted	251	21
Secondary completed	373	31.2
Technical	136	11.4
University	136	11.4
Occupation
Retired	90	7.5
Housewife	209	17.5
Works	405	33.9
Unemployed	336	28.1
Student	156	13.0
Family income
No income	562	47.0
Less than 500	289	24.2
From 501 to 1000	161	13.5
More than 1000	184	15.4
Type of housing
Own	689	57.6
Rented	484	40.5
Other (Sublet)	21	1.8
Usufruct	1	0.1
Temporary accommodation	1	0.1
Origin
Trujillo	601	50.3
Cajamarca	595	49.7

	X (S)	Me (RI)
Age	29.53 (10.49)	27 (15)
Number of people you live with	3.02 (1.96)	3 (2)

Note: n = number of participants, X = Mean, S = standard deviation, Me = Median, RI = interquartile range.

Table 2 summarizes the sequential process of exploratory factor analysis and item purification. The initial eight-factor solution (AFE1) showed high factor complexity, cross-loadings, and overlap among dimensions, limiting its interpretability. The three-factor theoretical model (AFE2), although conceptually plausible, presented high interfactorial correlations and conceptual overlap. The four-factor model (AFE3) was selected as the reference, enabling a systematic process for eliminating items that did not meet the minimum psychometric criteria.

Table 2.

Exploratory factor analyses in the iterative refinement process.

Stage	Factor	Items	Criteria	Removed	Justification
AFE1	8	47	Empirical exploration of dimensionality	No deletion	High factor complexity, cross-loadings, and overlap among factors.
AFE2	3	47	Contrast of the three-factor theoretical model	No deletion	High interfactorial correlations and conceptual overlap between factors limit discrimination between dimensions.
AFE3	4	47	Loads ≥ .30, difference between loads ≥ .20, commonality ≥ .40, complexity ≤ 2, theoretical coherence	p1, p3, p16–p23, p26, p29–p31, p33, p37–p38, p40–p42	Reference model by AFC. Items that do not meet strict criteria are discarded to begin an iterative elimination process.
AFE4	4	27	Loads ≥ .50, difference between loads ≥ .20, commonality ≥ .50, complexity ≤ 2	Fourth dimension, but not the item	A single item represented the fourth factor, and it had a higher load on another factor.
AFE5	3	27	Loads ≥ .50, difference between loads ≥ .20, commonality ≥ .50, complexity ≤ 2	p2, p15, p27, p28	Items removed were those that did not meet at least two criteria.
AFE6	3	23	Loads ≥ .50, difference between loads ≥ .20, commonality ≥ .50, complexity ≤ 2, intra-factorial theoretical coherence, conceptual representativeness, and equilibrium (4 items per dimension)	p4, p8, p13, p24, p25, p32, p34, p35, p36, p39, p45	The final selection of items was based on theoretical criteria, prioritizing conceptual coherence and construct representativeness, resulting in a 12-item short version.

In later stages, applying more stringent criteria (AFE4 and AFE5) revealed that one dimension was represented by a single item, prompting the selection of a three-factor structure. Finally, the AFE6 integrated statistical and theoretical criteria, prioritizing conceptual coherence and dimensional balance, yielding a 12-item short version of the instrument distributed equally across three dimensions.

Table 3 shows the results of the exploratory factor analysis of the short version of the instrument, composed of 12 items organized into three correlated factors, consistent with the theoretical dimensions of health literacy. Exploratory factor analysis (EFA) of the short version of the instrument (12 items) demonstrated excellent methodological suitability. Global indicators confirmed the factorability of the matrix, highlighting an optimal Kaiser–Meyer–Olkin coefficient (KMO = 0.85) and a highly significant Bartlett's test of sphericity (X² = 3550.66, p < .001). Additionally, the sampling adequacy values per item were high (MSA: .81–.91). This parsimonious three-dimensional solution explained a substantial 68% of the total variance, with balanced contributions among the dimensions (MR2 = 24%, MR1 = 23%, MR3 = 21%). The internal structure showed a symmetrical distribution of four items per factor with unambiguous and high factor loadings. Factor MR1 (Functional Health Literacy) grouped items p5, p6, p7, and p14 (λ: .66–.85); Factor MR2 (Critical Literacy) comprised items p9, p10, p11, and p12 (λ: .77–.84); and Factor MR3 (Health Promotion and Action) incorporated items p43, p44, p46, and p47 (λ: .58–.95). Communalities were fully adequate (h2: .58–.82), and linear complexity indices remained very low (1–1.4), guaranteeing a clean, well-defined factor architecture free of spurious empirical overlaps or cross-loadings.

Table 3.

Exploratory factor analysis of the short version of the instrument.

	MR2	MR1	MR3	h2	u2	Com	MSA
p5	−0.02	0.80	0.10	0.72	0.28	1	0.81
p6	−0.06	0.85	0.06	0.73	0.27	1	0.84
p7	0.04	0.83	−0.07	0.66	0.34	1	0.88
p9	0.84	0.05	−0.05	0.71	0.29	1	0.82
p10	0.80	0.09	−0.05	0.68	0.32	1	0.87
p11	0.81	0.04	0.00	0.69	0.31	1	0.84
p12	0.77	−0.14	0.19	0.64	0.36	1.2	0.87
p14	0.20	0.66	0.03	0.61	0.39	1.2	0.90
p43	0.28	0.04	0.58	0.58	0.42	1.4	0.91
p44	0.01	0.12	0.78	0.72	0.28	1	0.86
p46	0.00	0.14	0.69	0.60	0.40	1.1	0.83
p47	−0.03	−0.05	0.95	0.82	0.18	1	0.84
Ratio of variance	0.24	0.23	0.21
Cumulative variance	0.24	0.47	0.68

Note: KMO = 0.85, Barttlet's test of sphericity (X² = 3550.66, p0.001, df = 66).

Table 4 summarizes the fit indices of the confirmatory factorial models evaluated for both the full version of the instrument and the short version. Confirmatory factor analysis (CFA) was used to assess the scale's fit through a competitive strategy. For the full version, the unidimensional (CFI = .864, TLI = .851, RMSEA = .093, SRMR = .092), three-factor, and four-factor models were discarded due to poor metric properties. Although the hierarchical model of this macro version initially showed acceptable overall indices (CFI = .972, TLI = .971, RMSEA = .062), it was psychometrically invalidated due to a critically low omega coefficient for the overall factor, confirming the absence of a dominant higher-order latent construct. In contrast, the 12-item short version demonstrated clear statistical superiority. The three-factor correlated model for the short version demonstrated the best empirical and incremental fit to the data (CFI = .970, TLI = .961, SRMR = .056). The marginal increase in the approximation error (RMSEA = .096) is mathematically expected and acceptable behavior in models with high parsimony and few degrees of freedom. Finally, the hierarchical alternative for the short version was rejected due to the degradation of its incremental indices and the instability of its overall factor, confirming that the optimal structure is that of correlated factors.

Table 4.

Confirmatory factor analysis in the complete and short versions of the instrument.

Model	X^2	Df	CMIN	p	CFI	TLI	SRMR	RMSEA	ωh
T_One-dimensional	6324.77	1034	6.12	<0.001	0.852	0.845	0.092	0.093
T_M3F	5223.55	1031	5.07	<0.001	0.883	0.877	0.085	0.083
T_M4F	5066.43	1028	4.93	<0.001	0.887	0.881	0.083	0.081
T_Hierarchical	2423.11	1030	2.35	<0.001	0.972	0.971	0.070	0.062	0.19
Short version	329.86	51	6.47	<0.001	0.970	0.961	0.056	0.096
Hierarchical short version	329.86	51	6.47	<0.001	0.922	0.899	0.056	0.096	0.22

Note: X² = Chi-square, Df = degrees of freedom, CMIN = X2/Df, p = p-value, CFI = Comparative Fit Index, TLI = Tucker–Lewis Index, SRMR = Standardized Root Mean Square Residual, RMSEA = Root Mean Square Error of Approximation, ω_h = Hierarchical Omega.

The confirmatory factor model of the short version of the instrument (Fig. 1) demonstrated robust structural validity and theoretical coherence through a parsimonious three-dimensional structure (four items per dimension) with high standardized factor loadings (λ = .77–.87). The most relevant aspect is that these high loadings coexist with moderate interfactorial correlations (r = .49–.62), demonstrating that the dimensions share conceptual variance without compromising their empirical differentiation. Furthermore, multigroup analyses robustly demonstrated compliance with configural (Table 5), metric-scalar (ordinal), and strict invariance based on sex and age group (≤25 vs. >25 years). Since progressive restrictions on loadings, thresholds, and imposition of equality on residuals maintained an optimal model fit without generating statistically relevant variations in incremental indices (CFI = −0.001, RMSEA = −0.006), it is scientifically understood that the construct is measured equivalently between groups, allowing valid comparisons of latent scores.

Figure 1.

Confirmatory model of the short version of the instrument (n = 1196).

Table 5.

Factorial invariance by sex and age of the short scale of health literacy.

Model	X^2	df	CFI	TLI	RMSEA	SRMR	ΔCFI	ΔRMSEA	ΔSRMR
Sex
Configural	475.36	102	0.992	0.990	0.078	0.058
Metric-Scalar Ordinal	509.14	123	0.992	0.991	0.073	0.059	0.000	0.006	0.000
Strict	509.14	123	0.992	0.991	0.073	0.059	0.000	0.000	0.000
Age
Configural	600.34	102	0.991	0.989	0.090	0.066
Metric-Scalar Ordinal	707.25	123	0.990	0.989	0.089	0.066	0.001	0.001	0.000
Strict	707.25	123	0.990	0.989	0.089	0.066	0.000	0.000	0.000

Note: X² = Chi-square, Df = degrees of freedom, CFI = Comparative Fit Index, TLI = Tucker–Lewis Index, SRMR = Standardized Root Mean Square Residual, RMSEA = Root Mean Square Error of Approximation, Δ = difference.

Multigroup factor invariance analyses of the short version of the instrument evidenced configural, metric-scalar (ordinal), and strict invariance by both sex and age group (≤25 vs. >25 years). In both cases, the configural model presented an adequate fit, confirming the equivalence of the three-factor structure between groups. The progressive restriction of the factor loadings and thresholds did not generate relevant changes in the adjustment indices (sex: ΔCFI = 0.000, ΔRMSEA = −0.006, ΔSRMR = 0.000; age: ΔCFI = −0.001, ΔRMSEA = −0.001, ΔSRMR = 0.000), fulfilling the established criteria for the evaluation of invariance in models with ordinal variables (Table 5). Finally, the additional imposition of equality on waste maintained the model's fit, supporting strict invariance. Together, the results indicate that the short version of the instrument measures the construct equivalently between sexes and age groups, allowing valid comparisons of latent scores.

Discussion

The present study aimed to analyze the psychometric properties of an abbreviated version of the HLS-EU-Q47, translated and adapted into Spanish, in the Peruvian population. The findings demonstrate that, although the original instrument has a solid theoretical basis, its application in the local context resulted in significant conceptual overlap and redundancy, as observed in other versions.^16,24,25 Redundancy and diffusion across domains are not phenomena exclusive to the Peruvian population; several international studies have documented similar trends across different geographical and cultural contexts.^26–28 These studies indicate that the extended versions of the HLS-EU may generate cognitive fatigue and construct overlap. However, whereas in European populations these limitations are mainly attributed to educational level and the semantic complexity of the items, in developing contexts within the Asia-Pacific region and Latin America they appear to reflect a more integrated and less fragmented conceptualization of health services by users.²⁸

The transition to an abbreviated 12-item version, organized into three correlated factors, not only optimizes application time but also offers a more parsimonious structure that is theoretically consistent with the levels of health literacy proposed by contemporary literature.²⁴ The item purification process, as reflected in the Exploratory Factor Analysis stages (AFE1–AFE6), evidences the intrinsic complexity of measuring health literacy (HL). The initial eight-factor solution (AFE1) exhibited high factor complexity, making it difficult to interpret, a common phenomenon in extensive instruments in which items tend to measure adjacent constructs indiscriminately.¹

In particular, the original three-factor theoretical model (AFE2) showed extremely high interfactorial correlations. This phenomenon suggests a discriminating lack of validity in the borders between the domains of “Health Care”, “Disease Prevention”, and “Health Promotion” becomes diffuse for the respondent.²⁹ This conceptual overlap indicates that the Peruvian population perceives health actions as a continuum of care rather than an isolated or fragmented action. The decision to move toward AFE6, which opted for a 12-item version, allowed these inconsistencies to be resolved, prioritizing the representativeness of fundamental cognitive processes.³⁰ In addition to resolving the inconsistencies arising from incomplete capture of key elements of a concept due to an “underrepresentation of the construct,” as evidenced in other research.^16,31,32

It is critical to attribute these factorial complexities to factors that transcend mere linguistic representativeness. The use of an instrument designed under a Eurocentric paradigm (HLS-EU) typically assumes a highly literate, homogeneous, and accessible health system. In Peru, the absence of intercultural perspectives in the original matrix prevents the capture of constructs specific to the literal understanding of rural or Andean health, where the understanding of illness is linked to community well-being and holistic conceptions. Furthermore, the non-probabilistic nature of the sampling employed could underrepresent populations with language or geographic access barriers, which explains why the original matrix items exhibit high empirical friction in local measurement.⁵

The main contribution of this study is the conceptual reinterpretation of the dimensions. By moving toward the MR1 (Functional Literacy), MR2 (Critical Literacy), and MR3 (Health Promotion and Action) factors, the model aligns with Nutbeam's theoretical framework.³³ The MR1 factor focuses on operational comprehension skills in clinical contexts. These items, which originally evaluated processes of “understanding” in care and prevention, represent the basic level of SA: the ability to process technical information essential for survival in the health system.³⁴ The MR2 factor, this dimension transcends mere understanding and reflects evaluative and judgment processes. This finding aligns with the need to assess the individual's ability to filter information and actively participate in their treatment, a vital component in health systems where patients must navigate multiple opinions and sources of information.

Finally, the MR3 factor maintains the focus on social determinants and collective health. By grouping these items under the concept of “Action”, it is based on the idea that literacy is not only an internal cognitive process, but a social competence that allows the individual to influence their environment (housing, community, policies), which is consistent with the empowering model of Health Literacy and its integrated model.^1,35 In contrast to other structures, the proposed model suggests a more consolidated approach where these domains are perceived as part of an integrated health management strategy, each factor equivalent to a domain of the original instrument¹: MR1(health care and prevention), MR2 (Prevention) and MR3 (Health promotion), each factor integrates the original information processing processes. This suggests that patients could conceptualize health literacy across multiple, albeit interconnected, dimensions.

From the perspective of Internal Structure Validity, the short version of three correlated factors presented satisfactory global adjustment indices. Although the RMSEA was high (.096), this pattern is consistent with parsimonious models with few degrees of freedom.³⁶ According to Kenny et al., in models with few indicators per factor, the RMSEA tends to overestimate error, so its interpretation should be supplemented by other indices, such as the CFI and the SRMR, which, in this study, confirm excellent adequacy.³⁷

A critical finding was the invalidation of the hierarchical model. This result is inconsistent with Finbråten's findings.²⁴ The deterioration of the adjustment indices and, above all, the presence of an Omega coefficient for the low general factor confirm that SA in the Peruvian population should not be treated as a dominant one-dimensional trait, as suggested by other authors.¹⁵ Forcing a single global score by replicating the original European summation is methodologically inappropriate in our context. The independence of the factors reveals that an individual can exhibit optimal levels of clinical understanding (functional literacy) and, simultaneously, a critical deficit in information assessment or community action. Evaluating as a unidimensional attribute would obscure these internal disparities, negating the diagnostic value of the instrument in targeted intervention plans.³⁸

Finally, equality in waste maintained the model's fit, indicating that the construct is measured equivalently across sexes and age groups, allowing valid comparisons. This suggests that the instrument does not exhibit systematic biases, provides precision and reliability, and is suitable for evaluating the population with a solid, stable factorial structure.

Implications for public health in Peru

The validation of this 12-item short version offers an efficient tool for rapid screening in saturated health services. By focusing on functional, critical, and action skills, the instrument enables health professionals to identify specific communication barriers. In the Peruvian context, where gaps in education and access to information are marked, having an instrument that distinguishes between “understanding” and “being critical” is essential for designing health promotion interventions that are not merely informative but transformative.³⁹

This three-dimensional, correlated version has direct implications for strengthening primary health care (PHC) and the family health strategy in Peru. Its concise format facilitates its inclusion in routine screenings at primary care facilities, where consultation time is limited. Furthermore, this instrument is relevant in the field of family health, enabling better management of functional and critical literacy within households. Finally, its use within an intercultural approach to health will allow public administrators to recognize that literacy is not limited to reading and interpretation, but encompasses the capacity for empowerment and action required by communities to engage dialogically with the Western medical system, thereby closing health inequity gaps.¹

Despite the statistical strength, the study had some limitations that should be considered when interpreting the findings. First, the study was conducted in two urban Peruvian cities, which may limit the representation of the cultural, linguistic, and territorial diversity of the country. In addition, although the HLS-EU-Q47 provides a comprehensive framework for assessing health literacy, it was originally developed within a European context and may not fully incorporate intercultural perspectives or culturally specific understandings of health and healthcare practices present in Latin American populations. Therefore, future studies should evaluate the instrument in indigenous and rural populations, as well as explore culturally sensitive approaches that integrate local conceptions of health literacy and healthcare decision-making.

Conclusion

In conclusion, the 12-item HLS-EU is a valid and reliable tool for the Peruvian population. Its structure of three correlated factors solves the overlapping problems of the original model and provides an accurate metric based on differentiated competencies. This abbreviated model is not only psychometrically superior in terms of parsimony but is theoretically more powerful for integrating the functional and critical levels of health literacy.

What is already known about this topic

• 1.

  Health literacy is a key determinant of health outcomes and is commonly measured using instruments such as the HLS-EU-Q47.

• 2.

  However, applying this tool across different cultural contexts may introduce bias without proper adaptation.

• 3.

  In urban Peru, there was a lack of validated and concise instruments suitable for accurately assessing health literacy in populations exposed to sociocultural inequalities and heterogeneous access to health information and services.

What this study adds

• 1.

  This study provides a culturally adapted and psychometrically validated 12-item short version of the HLS-EU-Q47 for the Peruvian population.

• 2.

  It demonstrates a clear three-factor structure that improves measurement efficiency while maintaining reliability and validity.

• 3.

  The instrument offers a practical and concise tool for assessing health literacy in primary care and public health settings.

Authors’ contribution

WDB, WJMB, and SHV have designed the study. Analyses were planned by AJCA WJMB, JGSA and conducted by WJMB. Results were interpreted by WDB, AJCA and JGSA. WDB, AJCA and JGSA wrote the manuscript. SHV reviewed the manuscript and supervised the study. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

This research was approved by the ethics committee of the Health Sciences School of the Peruvian Union University, with the certificate number code 2025-CEB-FCS-UPEU No. 179. Respondents’ participation was through a consent form. Participation was anonymous, voluntary, and risk-free for participants. Ethical principles were respected, as well as national and international ethical standards in accordance with the Declaration of Helsinki (2000).

Consent for publication

Not applicable.

Funding

This work was funded by the Universidad Peruana Unión (UPeU) within the framework of the funding in Development and Training in Teacher Research DFID2025-01 approved under resolution No. 159-2025/UPEU-FCS-CF-E.

Conflict of interests

The authors have declared no competing interests.

Availability of data and materials

The data and materials of the current study are available on reasonable request from the corresponding author.