Development and validation of a rapid health literacy assessment tool for Chinese primary and secondary school students

Abstract

Health literacy (HL) plays a vital role in shaping health behaviors and long-term well-being. Existing HL assessment tools for children and adolescents remain fragmented, with limited theoretical foundations, limited age adaptability, and insufficient psychometric validation. A scientifically developed, developmentally appropriate, and psychometrically strong measurement tool specifically designed for younger populations can support HL monitoring, health promotion, and educational evaluation. This study conceptualized HL for children and adolescents and developed the Chinese Rapid Health Literacy Questionnaire for Primary and Secondary Students (CRHLQ-PS). Using an exploratory sequential design, we combined policy and literature reviews, expert interviews, and Delphi consensus to develop an indicator system comprising seven domains—hygiene, nutrition, physical activity, safety, preventive health, psychosocial adaptation, and digital health—with 20 indicators organized across four dimensions: knowledge, skills, motivation, and participation. Items were generated for each indicator. The CRHLQ-PS includes four age-specific sub-questionnaires for Grades 1–2 (39 items), 3–4 (40 items), 5–6 (52 items), and junior high school (58 items). A total of 7113 students from four regions in China participated in psychometric testing. Reliability and validity were evaluated using the Rasch Partial Credit Model. Delphi findings indicated high expert authority and moderate consensus. Rasch analysis supported strong psychometric performance, with good model-data fit, high item reliability, and acceptable person reliability. The scale exhibited a predominantly unidimensional structure and appropriate targeting overall, although it appeared relatively easy for the current sample. The CRHLQ-PS provides a rapid, theory-informed, and developmentally appropriate instrument for assessing HL among school-aged children and adolescents. Future research should broaden item difficulty and undertake cross-cultural validation to enhance its global applicability and comparability.

Contribution to Health Promotion.

• Establish a comprehensive HL indicator framework for children and adolescents encompassing seven domains, providing a more operationalized and multidimensional conceptualization of HL.

• Employ an exploratory sequential design that integrates Delphi consensus and Rasch modeling, offering a rigorous methodological reference for instrument development.

• The CRHLQ can support HL monitoring, evaluation of health promotion program implementation, and school health promotion assessments, with further refinement and cross-cultural validation planned in subsequent research.

Introduction

As emphasized in the Global Strategy for Women’s, Children’s and Adolescents’ Health (2016–30) issued by the World Health Organization (WHO), children and adolescents should not only survive but also thrive and ultimately transform to realize their full potential (World Health Organization 2015). Strengthening the health of this population supports progress toward Universal Health Coverage and sustainable social development. Despite notable global progress, children and adolescents continue to face diverse and complex health challenges worldwide. The WHO has launched multiple initiatives that position health literacy (HL) as a key enabling factor within a whole-of-society approach (World Health Organization 2017).

HL was first introduced by Simonds in 1974. Since then, HL has evolved into a multidimensional construct, widely defined as “the integration of skills, knowledge, and motivational drivers that enable individuals to access, understand, appraise, and apply health information for decision-making in healthcare, disease prevention, and health promotion contexts” (Sørensen et al. 2012, Liu et al. 2020). The U.S. Department of Health and Human Services conceptualized HL as a dual construct encompassing both personal and organizational dimensions, highlighting its dynamic nature and the need for alignment with societal systems to achieve health equity (U.S. Department of Health and Human Services 2020). Childhood and adolescence are critical developmental stages for shaping health behaviors, with promoting HL during these periods contributing significantly to lifelong well-being (Fleary et al. 2018). Previous studies have proposed that children’s and adolescents’ HL is a multidimensional construct, encompassing functional, interactive, critical, and media literacy. Its development and application are influenced by age-related cognitive and socio-emotional characteristics, as well as contextual factors such as school, family, peers, and the broader community environment (Massey et al. 2012, Joseph and Fleary 2021).

Globally, improving HL in schools has become a strategic public health priority (Okan et al. 2023), accompanied by increasing research on the development and validation of HL assessment tools. For children and adolescents, domain-specific assessment tools have been developed to assess areas such as nutritional and food literacy (Truman et al. 2017, Kelly and Nash 2021), physical literacy (Edwards et al. 2017, Barnett et al. 2023, Grauduszus et al. 2023), and mental HL (Mansfield et al. 2020, Simkiss et al. 2021, Riebschleger et al. 2022). There is a growing variety of HL instruments created for younger populations (Guo et al. 2018, Okan et al. 2018). The Health Literacy for School-Aged Children (HLSAC) (Paakkari et al. 2016) measures HL in school contexts, while REALM-Teen (Davis et al. 2006) and TOFHLA (Chisolm and Buchanan 2007) assess functional literacy in health care settings. The QUIGK-K (Teufl 2020) offers a performance-based approach for children aged 8–10 years that captures the processes of obtaining, understanding, appraising, and applying health information. An ideal instrument for children and adolescents should adopt an integrated, multidimensional perspective that reflects how HL is expressed through everyday learning, communication, and health-promoting behaviors (Van Boxtel et al. 2024).

From a methodological perspective, previous research has often relied on classical test theory (CTT) and item response theory (IRT) to validate scale reliability and validity. However, CTT faces several limitations, including the assumption of item equivalence (individual items are treated as equally difficult) and sample dependency (the item and scale statistics apply only to a specific group of respondents). In contrast, IRT links a person’s ability and item characteristics with the likelihood of responses, enabling inferences across different samples of individuals and items (Osterlind et al. 2010, Streiner and Norman 2015). Among IRT models, the Rasch model is one of the most widely used. It assumes that the probability of a person passing an item on a test depends on the respondent’s ability and the item’s difficulty. Rasch analysis can generate linear measures with equal-interval units, called logits, from counts of qualitatively ordered observations (Linacre and Wright 1993), allowing both respondent ability and item difficulty to be quantified on a single, unified scale (Jabrayilov et al. 2016, Woudstra et al. 2019). Furthermore, the Rasch model offers a probabilistic framework for analyzing data–model fit and response patterns, helping guide iterative item refinement and instrument improvement.

Existing HL instruments have several significant limitations. Many instruments lack rigorous psychometric validation and a thorough theoretical foundation, with insufficient evidence of reliability, validity, and construct representativeness (Haun et al. 2014, Hawkins et al. 2018). Most emphasize functional literacy while neglecting interactive and critical dimensions that underpin active health engagement. Digital HL—an increasingly vital component—remains underrepresented in existing tools (Arias López et al. 2023, Ban et al. 2024). Furthermore, many instruments lack explicit conceptual frameworks aligned with integrated HL models (Okan et al. 2015, Bröder et al. 2017). These limitations are particularly pronounced for children and adolescents. Numerous instruments have been adapted directly from adult scales without adequate consideration of developmental differences in cognition, language, and socio-emotional capacities (Caldwell and Killingsworth 2022, Chou et al. 2023), resulting in limited age appropriateness and measurement precision. Furthermore, existing tools frequently lack cultural and ecological awareness, neglecting how family, school, and peer surroundings impact young people’s health behaviors (Van Boxtel et al. 2024). Few studies have included children or adolescents in the design of instruments, which limits their contextual relevance, linguistic clarity, and applicability in real-world settings (Fretian et al. 2020). Conceptually, the field remains fragmented and largely adult-centered. Existing frameworks are predominantly extrapolated from adult HL models that emphasize healthcare and disease prevention, while insufficiently capturing developmental learning and daily health promotion (Ormshaw et al. 2013, Okan et al. 2015, Bröder et al. 2017, Guo et al. 2018). Consequently, these tools fail to effectively reflect the evolving cognitive, social, and digital skills of younger populations. Differences in definitions, domains, and measurement approaches across various studies complicate the comparison and integration of data findings (Urstad et al. 2022).

Developing a scientifically sound, developmentally suitable, and psychometrically reliable HL assessment tool for children and adolescents is crucial. Key challenges involve: (i) developing a comprehensive theoretical framework and defining core constructs to clarify the conceptual foundation; (ii) enhancing age adaptability to align with cognitive and socio-emotional differences at various developmental stages; and (iii) maintaining measurement reliability and validity while enabling quick and practical assessments.

To address these challenges, this study uses the Rasch model to examine item difficulty, measurement accuracy, and psychometric properties. The goal is to develop the Chinese Rapid Health Literacy Questionnaire for Primary and Secondary Students (CRHLQ-PS)—an age-appropriate, contextually relevant tool to assess HL among Chinese children and adolescents aged 6–14. The research questions include: (i) How can the HL framework be refined based on current models to better match the cultural, cognitive, and emotional developmental stages of children at different ages? (ii) Which core domains and competencies should be emphasized, and how can full coverage be ensured without repetition? (iii) Does the questionnaire show reliable and valid results across various age groups?

Methods

Scale development and psychometric testing are essential in health, social, and behavioral sciences. Usually, this process involves three stages: (i) concept modeling and item generation, (ii) questionnaire construction, and (iii) questionnaire evaluation, including testing for dimensionality, reliability, and validity (Boateng et al. 2018). The CRHLQ-PS was developed and validated through an exploratory sequential research design.

Development of the CRHLQ-PS

Step 1: concept modeling

In the concept modeling phase of this study, a systematic literature search was conducted to develop an HL theoretical framework suitable for children and adolescents and to guide the generation of an indicator system. Searches were performed in PubMed, Web of Science, and CNKI covering the period 2000–24, aiming to (i) identify existing HL theoretical models and frameworks, including those for adults and adolescent populations; and (ii) identify existing HL instruments for children and adolescents and their measurement domains. The search terms included “health literacy,” “eHealth literacy,” “health competence,” “health capability,” “health empowerment,” “health promotion,” “health education,” “behavioral model,” “theoretical framework,” “conceptual model,” and “child/adolescent/school-age/youth.”

For children and adolescents, models place greater emphasis on cognitive, behavioral, emotional, and social competencies. Regarding measurement tools, child and adolescent HL scales include HLSAC, HLAT-8/16, eHEALS, newest vital sign, and the Adolescent Health Literacy Questionnaire, with core dimensions covering information access, understanding, appraisal, application, decision-making, communication, self-management, and digital literacy. Guided by Bandura’s Social Cognitive Theory and Self-Determination Theory (SDT) for motivation, and Zarcadoolas’ multidimensional framework (Zarcadoolas et al. 2005) for participation and empowerment, we synthesized overlapping constructs to define four interrelated dimensions: (i) Knowledge: understanding fundamental health concepts; (ii) Skills: the ability to access, appraise, and apply health information; (iii) Motivation: intrinsic drivers that promote engagement in health-enhancing behaviors; and (iv) Participation: active involvement in both individual and collective health-related decision-making (Fig. 1). These dimensions capture the essential components of HL and serve as the theoretical basis for the assessment tool.

Figure 1.

Dimensions of the HL conceptual model.

Step 2: indicator system construction

Considering the significant differences in cognitive skills, language development, and independent behaviors across various educational stages, we integrated developmental adaptation into the indicator system. The target population was segmented by educational level: early primary (grades 1–2, ages 6–8), middle primary (grades 3–4, ages 9–10), upper primary (grades 5–6, ages 11–12), and junior high (ages 13–15). In terms of content, the domain indicators encompass context-specific literacies validated in prior research, such as ehealth literacy, food literacy, physical literacy, and mental HL, offering a comprehensive evaluation of children’s and adolescents’ skills and needs in multiple health areas. The procedure for constructing the HL indicator system is shown in Fig. 2.

Figure 2.

Diagram for the procedures followed to develop the HL indicator system.

Phase 1: policy review

Firstly, we collected and analyzed policy documents relevant to HL from Chinese government’ websites, including the State Council of the People's Republic of China (SC), the National Health Commission of the People's Republic of China (NHC), and the Ministry of Education of the People's Republic of China (MOE) (Supplementary Table S1). Key documents, including Health Literacy of Chinese Citizens—Basic Knowledge and Skills (National Health Commission of the People's Republic of China 2008) and Healthy China 2030 (The State Council of the People’s Republic of China 2016), emphasized the four core competencies: knowledge, skills, motivation, and participation. Developmental benchmarks from the Standards for Health Education guided the design of age-specific items, while the Ecological Environment and Health Literacy (2018) guided contextually grounded domains, such as environmental health.

Phase 2: expert interview

Secondly, we conducted semistructured interviews and panel discussions with 13 multidisciplinary experts to explore key themes, including: (i) the conceptual boundaries of children’s and adolescents’ HL; (ii) the challenges in promoting students’ health behaviors and lifestyles; (iii) current modalities for accessing health information (e.g. digital platforms, school-based resources); (iv) students’ perceived needs and barriers in obtaining, interpreting, evaluating, and applying health information; and (v) Practical strategies to enhance accessibility and usability of health information. All experts had expertise in health promotion, behavioral science, and children's health, with a minimum of 10 years of professional experience.

Phase 3: panel discussion

Thirdly, we held a series of structured panel discussions and brainstorming sessions to synthesize expert insights into specific health domain indicators. Eight domain indicators were identified through multiple brainstorming sessions and group discussions, covering key aspect indicators such as eye, oral, and hand hygiene; food safety and nutrition; exercise, sedentary behavior and sleep; injury prevention and natural disaster evacuation; nutrition-related, infectious, chronic and endemic diseases; sociocultural adaptation, behavior regulation and psychological support; digital health access, evaluation and application; and pollution exposure prevention and sustainable resource utilization. When choosing each indicator, we carefully considered the real-life circumstances and behavioral abilities of children and adolescents. For example, topics such as daily hygiene, nutrition, physical activity, and access to digital health information were included because they are either directly within students’ control or achievable with support. For indicators that may seem beyond children’s independent abilities, we emphasize that they should be handled in educational or supportive settings, rather than expecting children to manage them entirely on their own.

Finally, after phases 1, 2, and 3. The initial Chinese Health Literacy Indicator System for primary and secondary students (CHLIS-PS) was developed.

Phase 4: Delphi method

Finally, following established procedures of previous studies (Bale et al. 2018), the Delphi method was employed to refine the CHLIS-PS. The 13 experts who participated in the Delphi consultation were the same multidisciplinary experts who participated in the phase 2 interviews. All experts participated in an online consultation using a structured questionnaire and proposed any additional amendments to the system. Throughout the amendment process, we stressed that experts should not only evaluate the scientific rigor and feasibility of the initial indicators but also offer revision suggestions based on the needs and capacities of students at different grade levels. Based on the experts’ judgments regarding age-appropriate suitability, we stratified or adjusted specific indicators to ensure that the final indicator system fully aligns with the principle of developmental adaptation. Subsequently, experts were asked to evaluate each indicator for importance and suitability (Holey et al. 2007). Expert reliability was assessed through: (i) response rate; (ii) authority coefficient (Cr = [Ca + Cs]/2), where Ca (0.3–0.1) reflects judgment basis and Cs (0.2–1.0) indicates their familiarity with the topic, providing a quantitative measure of each expert’s credibility in the Delphi process (Hasson et al. 2000); (iii) consensus consistency (Kendall's W, 0–1); (iv) opinion convergence. Open-ended feedback was used to inform iterative revision.

Step 3: item generation

Using the constructed indicator system, we developed an item pool through three brainstorming sessions, complemented by items independently created by an expert. Each session, facilitated by the researcher, lasted 2–4 hours. All experts involved were either health education practitioners or health promotion specialists. Our goal was to generate at least four items per domain indicator to cover the related aspect indicators thoroughly. This method aligns with previous research findings that three to four items per scale are sufficient to achieve adequate internal consistency reliability (Harvey et al. 1985). Within the same aspect indicator, the number of items, their wording, and content were suitably adapted according to students’ knowledge, cognitive skills, and reading comprehension across different grade levels. Items assessing motivation were developed in accordance with SDT principles, covering introjected, identified, and integrated regulation, as well as intrinsic motivation.

Regarding question formats, lower-grade sub-questionnaires used a 3-point Likert scale, single-choice, and judgment questions, often with pictorial representations. Higher-grade sub-questionnaires adopted a 5-point Likert scale to better capture students’ motivations and engagement in health behaviors and included single-choice, true/false, and calculation questions. Key priorities during item creation encompassed: (i) ensuring balanced coverage of each HL subdomain, (ii) maintaining developmental appropriateness, and (iii) keeping items concise to secure feasibility. Consensus was reached through discussion.

Participants

A stratified cluster random sampling method was used to conduct the validation survey on-site from March to June 2023 across various regions of China, including Luzhou (western Sichuan Province), Zhengzhou (central Henan Province), Taiyuan (central Shanxi Province), and Beijing (north). In each region, one primary school and one secondary school were selected, and multiple classes from different grades within each school were randomly chosen to ensure the sample's representativeness. The questionnaire was developed using the digital platform “Yiqixiu” and distributed via a QR code for online completion through WeChat, allowing each account to submit only once. To ensure data completeness, the system required all items to be answered before submission; as a result, all returned questionnaires were complete and valid. The questionnaire distribution and collection were managed by trained teachers, who shared the QR code with parents, explained the survey's purpose and importance, and emphasized the importance of protecting personal data. They provided clear instructions for completing the questionnaire without offering any leading or suggestive advice. After securing informed consent, primary and secondary school students independently completed the questionnaire on their mobile devices—the final valid sample comprised 2788 secondary school students.

Ethical approval was granted by the Institutional Review Board of the Chinese Academy of Medical Sciences & Peking Union Medical College, which supervises research involving human participants. Informed consent was obtained from the parents of all participants after providing a detailed explanation of the survey’s content and goals.

Statistical analysis

The Rasch analysis was performed using Winsteps 3.72.3. Each sub-questionnaire was analyzed separately. The Rasch model can be extended to handle polytomous items with ordered response options. Two common extensions are the Partial Credit Model (PCM) (Masters 1982) and the Rating Scale Model (Andrich 1978). Since each sub-questionnaire contained both dichotomous and polytomous items, the PCM was chosen. This model allows each item to have its own set of step calibrations, offering greater flexibility for estimating item difficulty and person ability, primarily when response categories differ among items.

Individual item and person fit were analyzed using Infit (weighted) and Outfit (unweighted) statistics to assess whether the data fit the model’s expectations. For each of these fit statistics, Winsteps provides the Mean Square (MNSQ) and Z-Standardized Scores (ZSTD) (Boone et al. 2014). For individual fit, the MNSQ values assess the consistency of an individual’s response pattern with the model’s predictions. Unusual response patterns may affect item fit and can be caused by cognitive deficits in respondents. Thus, deleting such responders can improve the tool’s internal construct validity. For items fit, MNSQ values reflect the extent to which each item contributes to defining a common measurement construct, with an ideal value around 1.0. Values exceeding 1.5 may reflect unpredictable or inconsistent response behavior, while values below 0.5 may indicate overly deterministic or repetitive responses. Values exceeding 1.5 may indicate a lack of structural homogeneity with other items, and values below 0.5 may suggest item redundancy. However, the cutoff values tend to vary depending on the purpose for which the ratings are used (Karabatsos 1997). If the MNSQ value is beyond the acceptable range, ZSTD must be checked. ZSTD values of 2.0 or more indicate statistically significant model misfit (Linacre 2012). If misfitting items or individuals are found, an iterative process proposed may be used to address these issues (Linacre 2010). Infit is sensitive to responses on items near the person’s ability level, whereas outfit is sensitive to responses on items that are relatively distant from the person’s ability level (Piquero et al. 2001). Point–measure correlations, PTMEA CORR, were assessed. A positive PTMEA CORR indicates that respondents with higher ability are more likely to answer an item correctly, reflecting good item discrimination and consistency with the latent trait.

The Rasch model assumes that all items measure a single trait, representing a unidimensional tool. Unidimensionality was assessed with PCA of residuals. In WINSTEPS, the first contrast shows residual variance, indicating potential multidimensionality. Usually, unexplained variance below 5%–10% (eigenvalue ≤ 2) suggests unidimensionality; higher values imply a possible secondary dimension (Linacre 2023).

Model reliability values will be calculated. Person reliability measures how consistently respondents’ ability rankings are maintained across different items, similar to Cronbach’s alpha, and ranges from 0 to 1. Higher values imply that respondents’ skill orderings would stay stable if tested with another set of items measuring the same trait. Item reliability indicates the consistency of the item difficulty order within the sample and ranges from 0 to 1, with higher scores indicating greater reliability. Typically, 0.70, 0.80, and 0.90 are considered acceptable, reasonable, and excellent reliability levels, respectively (Bond 2015). The person separation index gauges how well respondents’ abilities are distributed along the measured construct, indicating how many distinct ability levels the questionnaire can differentiate. Similarly, the item separation index measures the spread of item difficulties along the construct, indicating how distinctly different item difficulties are identified within the sample. Both indices are calculated using standard error units and should ideally be at least 2, which means the measure can reliably distinguish respondents, items, or both into at least two separate levels (Bond 2015).

Item difficulties and person abilities were visually depicted on a person–item map (Wright map), with both measured in logits. Person abilities are displayed on the left, and item difficulties on the right, along a shared continuous scale, offering a clear visual of the overlap and distribution of the measurement range. These maps help evaluate different aspects.

Results

Questionnaire measurement

After careful deliberation, “environmental health” and “endemic disease” were excluded, while “sexual and reproductive health” was added to the indicator system. The developed CHLIS-PS is a comprehensive, multidimensional framework that includes 20 aspect indicators across five main domains: hygiene literacy (covering eye, oral, hand, and sexual and reproductive hygiene), nutrition literacy (about food sanitation and diet), physical activity literacy (including exercise, sedentary habits, and sleep), safety HL (focusing on injury prevention and natural disaster evacuation), preventive HL (covering nutrition-related, infectious, and chronic diseases), psychosocial adaptation literacy (addressing sociocultural adaptation, behavior management, and psychological support), and digital HL (related to accessing, evaluating, and using information). Compared with grades 5–6 and junior high students, those in grades 1–2 and 3–4 do not include the “digital HL” domain or the “chronic disease” and “sexual and reproductive hygiene” indicators. The core components of knowledge, motivation, skills, and participation form the foundation of the entire system (see Fig. 3).

Figure 3.

Final version of the Chinese Health Literacy Indicator System for primary and secondary students.

The Delphi process yielded an authority coefficient (Cr) of 0.88, exceeding the 0.70 threshold for expert reliability, indicating high credibility. The Kendall’s W was 0.36 (P < 0.01), showing moderate, significant consensus. Importance and appropriateness scores ranged from 4.5 to 5.0, with CVs below 0.20, showing strong agreement. Physical activity Literacy, psychosocial adaptation Literacy, and preventive HL scored ≥4.9 with CVs ≤0.06, indicating they are highly essential and well-defined (see Table 1).

Table 1.

The importance and appropriateness ratings in the Delphi method.

Domain indicator	Aspect indicator	Importance	Appropriateness
x ± s, score	CV	x ± s, score	CV
Hygiene literacy	–	4.5 ± 0.8	0.17	4.6 ± 0.7	0.16
Eye health	4.8 ± 0.6	0.13	4.9 ± 0.3	0.05
Oral health	4.8 ± 0.6	0.13	4.8 ± 0.6	0.12
Hand health	4.1 ± 1.4	0.34	4.3 ± 1.2	0.27
Sexual and reproductive hygiene	4.9 ± 0.3	0.06	5.0 ± 0.0	0.00
Nutrition literacy	–	4.8 ± 0.8	0.17	4.6 ± 0.8	0.18
Nutrition	4.9 ± 0.3	0.06	4.8 ± 0.5	0.11
Food sanitation	4.9 ± 0.3	0.06	4.8 ± 0.5	0.11
Physical activity literacy	–	5.0 ± 0.0	0.00	4.9 ± 0.3	0.05
Exercise	4.7 ± 1.1	0.24	5.0 ± 0.0	0.00
Sedentary behavior	4.8 ± 0.4	0.08	4.8 ± 0.5	0.11
Sleep	5.0 ± 0.0	0.00	4.9 ± 0.3	0.05
Preventive HL	–	4.9 ± 0.3	0.06	4.9 ± 0.3	0.05
Nutrition-related disease	4.5 ± 0.8	0.17	4.5 ± 0.8	0.17
Infectious disease	4.9 ± 0.3	0.06	4.9 ± 0.3	0.06
Chronic disease	3.9 ± 1.5	0.38	4.1 ± 1.3	0.32
Safety literacy	–	4.8 ± 0.4	0.08	5.0 ± 0.0	0.00
Accidental injury prevention	5.0 ± 0.0	0.00	5.0 ± 0.0	0.00
Natural disaster evacuation	4.9 ± 0.3	0.06	4.9 ± 0.3	0.06
Psychosocial adaptation literacy	–	5.0 ± 0.0	0.00	4.9 ± 0.3	0.05
Sociocultural adaptation	5.0 ± 0.0	0.00	5.0 ± 0.0	0.00
Behavior regulation	5.0 ± 0.0	0.00	4.8 ± 0.4	0·08
Psychological assistance	4.8 ± 0.6	0.13	4.5 ± 0.8	0.17
Digital HL	–	4.8 ± 0.4	0.08	4.7 ± 0.6	0.13
Access	4.7 ± 0.6	0.13	4.8 ± 0.6	0.13
Evaluation	4.5 ± 1.2	0.27	4.6 ± 0.8	0.17
Application	4.5 ± 1.2	0.27	4.5 ± 0.8	0.17

The CRHLQ-PS comprises four sub-questionnaires tailored to specific grades: grades 1–2 (CRHLQ-12), grades 3–4 (CRHLQ-34), grades 5–6 (CRHLQ-56), and junior high school (CRHLQ-Junior). In total, CRHLQ-12 includes 39 items, CRHLQ-34 contains 40 items, CRHLQ-56 includes 52 items, and CRHLQ-Junior includes 58 items. The number of items for each indicator is shown in Supplementary Table S2. Item types and formats differ across grade levels, with detailed specifications in Supplementary Table S3. All sub-questionnaires were scored on a 100-point scale, with item weights based on Delphi consultation results. HL levels were classified using percentage-of-maximum scores: high HL (≥90%), moderate HL (80%–89%), and low HL (<80%). These criteria were derived from commonly applied educational assessment frameworks. Detailed classification criteria are provided in Supplementary Material S2, and the full CRHLQ-PS item set is presented in Supplementary Material S3.

Sample characteristic

7113 students participated across four grade groups: grades 1–2 had 1070 students (563 males, 507 females) in Luzhou (792), Taiyuan (116), Beijing (142), Zhengzhou (20); grades 3–4 had 1190 students (636 males, 554 females) in Luzhou (962), Taiyuan (78), Beijing (115), Zhengzhou (35); grades 5–6 included 1065 students (551 males, 514 females) in Luzhou (646), Taiyuan (162), Beijing (218), and Zhengzhou (39). Additionally, 2788 junior high students (1354 males, 1434 females) were surveyed, with 1875 in Luzhou, 650 in Taiyuan, and 263 in Beijing.

Rasch analysis

Item and person fit

Across all sub-questionnaires, participants generally showed acceptable model fit, with average INFIT and OUTFIT MNSQ values close to 1.0 (see Table 2). A total of 50 (4.7%), 19 (1.6%), 26 (2.4%), and 26 (0.9%) participants exhibited significant misfit on the CRHLQ-12, CRHLQ-34, CRHLQ-56, and CRHLQ-junior sub-questionnaires, respectively. To maintain a reliable model-data fit, those with significant misfit were excluded from further analyses.

Table 2.

Person and item fit summary statistics for the CRHLQ-PS.

Sub-questionnaire	Number	Measure	SE	INFIT	OUTFIT
MNSQ	ZSTD	MNSQ	ZSTD
CRHLQ-12	Person (n = 1070)	2.82	0.49	1.01	0.00	0.91	0.10
Item (n = 39)	0.00	0.14	1.00	0.30	0.87	−0.20
CRHLQ-34	Person (n = 1190)	1.81	0.39	0.99	0.00	1.03	0.00
Item (n = 40)	0.00	0.09	0.99	0.40	0.97	0.10
CRHLQ-56	Person (n = 1065)	1.31	0.34	0.98	−0.01	1.04	−0.01
Item (n = 52)	0.00	0.10	0.99	0.10	0.98	0.10
CRHLQ-junior	Person (n = 2788)	0.99	0.18	0.96	−0.30	1.10	−0.10
Item (n = 58)	0.00	0.05	0.98	0.10	1.09	1.80

For all sub-questionnaires, item fit statistics indicated adequate model-data fit, with average INFIT and OUTFIT MNSQ values ranging from 0.87 to 1.09. Most items were within the acceptable range (0.5–1.5), indicating they reliably measured a common underlying construct (see Table 2). Detailed item fit statistics for each sub-questionnaire are available in Supplementary Material S4.

Point–measure correlations across all sub-questionnaires mostly ranged from 0.10 to 0.50, indicating satisfactory item discrimination and alignment with the underlying latent constructs. For SRHLQ-12, correlations varied from 0.08 to 0.40, with most items showing moderate-to-strong positive relationships; a few items (e.g. B1, B2, B8, and B11) exhibited weaker correlations, likely due to limited variability or ceiling/floor effects. In SRHLQ-34, correlations spanned from 0.11 to 0.42, with most items effectively distinguishing respondents, while a few (e.g. B9, D5) showed lower correlations, possibly reflecting item simplicity or slight misalignment with the construct. SRHLQ-56 correlations ranged from −0.02 to 0.50, with nearly all above 0.15; only C11 showed a slightly negative correlation, suggesting potential misinterpretation or unusual responses. For CRHLQ-junior, correlations ranged from −0.08 to 0.49, with over 90% of items positive. A few items (e.g. B11, C3, C8, C9, C12) had minor negative correlations, possibly due to reverse wording, misunderstanding, or guessing. (See details in Supplementary Material S5).

Unidimensionaliy

For CRHLQ-12, Rasch measures explained 30.2% of the total variance, with the first contrast accounting for 5.0% (eigenvalue = 2.6), indicating a primarily single dimension. For CRHLQ-34, 54.0% of the variance was explained, and the first contrast was 2.9% (eigenvalue = 2.5), supporting unidimensionality. For CRHLQ-56, the Rasch dimension explained 59.7%, while the first contrast contributed 3.0% of unexplained variance, with an eigenvalue of 3.8. Although this proportion is low (<10%), the high eigenvalue suggests a minor secondary structure or local dependence. For CRHLQ-Junior, 64.9% of the variance was explained, and the first residual contrast accounted for 4.9% (eigenvalue = 8.0). Despite this large eigenvalue, the low unexplained variance (<10%) indicates that the primary Rasch dimension captures most of the construct, with possible minor local dependencies.

Reliability

Table 3 shows the person and item reliability and separation for each CRHLQ sub-questionnaire. For CRHLQ-12, person reliability was 0.59 with a separation of 1.20, while item reliability was 0.98 with a separation of 7.72. For CRHLQ-34, person reliability was 0.56 with a separation of 1.13, and item reliability was 1.00 with a separation of 14.43. For cRHLQ-56, the person reliability was 0.54 and the separation was 1.08, with item reliability at 1.00 and separation at 15.43. For CRHLQ-junior, person reliability was 0.76 with a separation of 1.77, and item reliability was 1.00 with a separation of 24.19. Item reliability was high, and the stable difficulty hierarchy covers different levels. Person reliability was lower, ranging from 1.08 to 1.77, indicating a limited ability to distinguish among respondents’ abilities.

Table 3.

Summary of changes in separation, reliability, and variance explained.

Questionnaire	Separation	Reliability	Variance explained (%)
Person	Item	Person	Item
CRHLQ-12	1.20	7.72	0.59	0.98	30.2
CRHLQ-34	1.13	14.43	0.56	1.00	54.0
CRHLQ-56	1.08	15.43	0.54	1.00	59.7
CRHLQ-junior	1.77	24.19	0.76	1.00	64.9

Item difficulty

All four CRHLQ sub-questionnaires generally covered most respondents’ ability levels. For CRHLQ-12, abilities were between +1 and +4 logits (mean = 2.82), with item difficulties from −4 to +3 logits (mean = 0.00), indicating the scale was relatively easy. For CRHLQ-34, abilities ranged from −1 to +5 logits (mean = 1.81), and item difficulties from −3 to +5 logits (mean = 0.00), with fewer high-ability items. For CRHLQ-56, abilities ranged from −2 to +4 logits (mean = 1.31), with item difficulties from −4 to +5 logits (mean = 0.00). For CRHLQ-junior, abilities ranged from −2 to +3 logits (mean = 0.99), with item difficulties from −4 to +4 logits (mean = 0.00). Overall, item distributions target the sample, but the questionnaires are relatively easy (See details in Supplementary Material S4).

Discussion

This study developed and validated a rapid, comprehensive HL questionnaire for children and adolescents, comprising four sub-questionnaires tailored to different grade levels. The development process focused on achieving rapid assessment by optimizing construct validity and item relevance. Item development was guided by core latent HL constructs, balancing representativeness and measurement efficiency while avoiding superficial repetition, all while ensuring reliability and validity. Additionally, item wording was carefully adapted to align with the cognitive levels and language comprehension skills of children and adolescents at various developmental stages, ensuring that the assessment effectively covers the main aspects of HL within a limited time.

A key step in the development process involved conceptualizing HL for younger populations. Recently, Seidl developed the first child and adolescent-centered HL model (Seidl et al. 2025). Building on existing theoretical models, we introduced a developmental, multidimensional framework comprising four interconnected dimensions: knowledge, skills, motivation, and participation. This approach views HL as a dynamic competence that develops through learning, social interaction, and empowerment. It goes beyond mere factual knowledge by emphasizing active participation in health decisions. Based on this concept, we developed an HL indicator system that reflects the multidimensional and developmental sides of HL in children and adolescents. To our knowledge, this is the first comprehensive indicator system tailored for primary and secondary students, enabling evaluation of HL across different developmental stages. The CHLIS-PS provides both a theoretically robust and practically useful framework that links developmental theory with empirical assessment.

Compared with widely used international instruments, the CRHLQ-PS targets aspects of school-aged HL that are not fully addressed elsewhere while remaining developmentally oriented and time-efficient. Tools such as s-TOFHLA (Chang et al. 2012) and REALM-teen (Davis et al. 2006) emphasize functional and text comprehension in clinical contexts, and HLS-EU-Q47 (Domanska et al. 2018) was designed for adults. Additionally, the WHO GAMA framework’s 47 adolescent health indicators (World Health Organization 2024) are primarily for system-level monitoring rather than individual measurement. In contrast, the CRHLQ-PS implements a theory-based indicator system for hygiene, nutrition, physical activity, safety, preventive health, mental health, and digital health. The Rasch analysis results provide preliminary psychometric evidence supporting the validity and reliability of the CRHLQ-PS. All sub-questionnaires showed item and person fit indices within acceptable limits, demonstrating good alignment between the model and data. While the first contrast eigenvalues slightly surpassed the recommended threshold, this probably indicates minor multidimensionality, which is expected given the complex and multifaceted nature of HL, especially in early developmental stages (Aryadoust et al. 2021, Mancone et al. 2024). Item reliability was consistently high, whereas person reliability was moderate, suggesting limited discrimination among respondents—possibly due to restricted ability variance, ceiling effects, or relatively homogeneous educational backgrounds. Although these factors might limit the sensitivity to detailed individual differences, the tool still offers reliable results for group comparisons. Future improvements should focus on enhancing measurement accuracy by including a broader range of item difficulties and updating overly simple items to distinguish higher-ability respondents better.

Several limitations should be noted. Although the CRHLQ-PS showed acceptable fit and high item reliability, its modest person reliability and possible multidimensionality suggest that further psychometric development is necessary before confirming its unidimensionality and robustness. Using a participatory design approach might yield useful insights into item clarity, cognitive interpretation, and relevance to context. Future versions should include stakeholder input from students, teachers, and health educators to improve face validity and contextual fit. Additionally, participants were recruited via cluster sampling, which could affect the sample's representativeness. However, this concern is partly alleviated by Rasch analysis, which is less influenced by sample characteristics than CTT.

Conclusion

This study introduced and validated the Chinese Rapid Health Literacy Questionnaire for Primary and Secondary School Students (CRHLQ-PS), a rapid yet comprehensive instrument for assessing HL among children and adolescents. Based on a multidimensional theoretical framework and a structured indicator system, the CRHLQ-PS offers an evidence-based method for examining developmental and contextual factors of HL in school-aged groups. As a newly created instrument, it still requires ongoing refinement. Future studies should focus on testing its stability across larger, more diverse populations, evaluating its ability to predict health-related behaviors and outcomes, and exploring its cross-cultural relevance in various educational and social settings.

Author Contributions

Jiayue Guo (Conceptualization, Formal analysis, Writing—original draft, Writing—review & editing), Yuxing Wang (Formal analysis, Writing—review & editing), Ziqing Zan (Investigation, Methodology, Writing—review & editing), Lizhu Liu (Investigation, Methodology, Writing—review & editing), Mengyu Li (Investigation, Writing—review & editing), and Lili You (Conceptualization, Funding acquisition, Supervision, Writing—review & editing)

Supplementary material

Supplementary material is available at Health Promotion International online.

Conflicts of interest

None declared.

Funding

This study was funded by the Science and Technology Innovation Project in Medicine and Health, Chinese Academy of Medical Sciences (grant no. 2021-I2M-1-046) and the National Natural Science Foundation of China (grant no. 71904205).

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

AI Use Disclosures

The authors affirm that this manuscript was conceived, analyzed, and written by the authors. ChatGPT was solely used for language refinement, grammar correction, and sentence formatting during the drafting process. No AI tools were used to generate or analyze data, and all intellectual contributions, including conceptual framing and writing decisions, remain the authors’ sole responsibility.