After obtaining formal authorization to use and adapt the original CCI, the research team followed the standardized cross-cultural adaptation model proposed by Sperber et al [33]. The process was conducted in 4 stages. First, 2 native simplified Chinese translators with expertise in health communication independently produced forward translations of the instrument (T1 and T2). Then, these were compared and reconciled by a senior bilingual translator with a medical background, who synthesized them into a single version (T3). To verify the linguistic equivalence, T3 was subsequently back-translated into English (T4) by an independent professional bilingual translator who had no prior knowledge of the tool and was not involved in the forward translation process, thereby minimizing potential bias and ensuring semantic fidelity. Finally, to further evaluate translation quality, we applied the linguistic equivalence assessment method recommended by Sperber et al [33]. This method involved comparing the original English version with the back-translated version across 2 dimensions. The first dimension is comparability of language, which refers to the similarity of words, phrases, and sentence structures. The second dimension is similarity of interpretability, which refers to the extent to which different wordings convey equivalent meaning. A linguist with a doctoral degree conducted this assessment using 4-point ordinal scales (1=extremely comparable or similar; 2=comparable or similar; 3=not comparable or similar; 4=not at all comparable or similar). Items rated as 3 or 4 on either dimension were reviewed by an expert panel consisting of one health education specialist and 2 nurses, all of whom had formal training in health communication and substantial experience in patient-facing health education and counseling. Revisions were made through iterative panel discussions until consensus was reached. When the initial agreement was unattainable, additional rounds of discussion were used to determine the most contextually and semantically appropriate wording. The finalized items were then incorporated into the final C-CCI. The overall process is summarized in Figure 1.

Composite reliability (CR) and average variance extracted (AVE) were calculated to evaluate convergent validity [42]. Convergent validity was considered strong when CR exceeded 0.70 and AVE was greater than 0.50, and acceptable when AVE ranged between 0.36 and 0.50 [43]. Discriminant validity was assessed by comparing the square root of the AVE for each construct with the inter-construct correlations. When the square root of AVE for a construct was greater than its correlations with other constructs, discriminant validity was deemed satisfactory [42].

To assess the reliability of the C-CCI, Fleiss kappa (κ) test was applied to evaluate inter-rater agreement, and Cronbach α coefficient was computed to examine internal consistency. A κ value between 0.61 and 0.80 was interpreted as indicating substantial agreement [44], whereas a Cronbach α coefficient greater than 0.70 was deemed indicative of acceptable internal consistency [45].

A comprehensive search of the official websites of all 31 provincial CDCs in Mainland China was conducted, focusing on diabetes as the evaluation theme given its global rise, major public health burden, and reliance on health education for prevention and management [46,47]. The search was completed on May 18, 2025.

The keyword “diabetes” was used in the internal search engines of the provincial CDC websites to ensure accuracy and consistency. Consistent with the “least effort” principle, whereby the public typically focuses on top-ranked results [48], the highest-ranked diabetes-related health article from each provincial CDC was selected for analysis. Exclusion criteria included government reports, bidding information, and policy or regulatory documents exceeding 1000 words. The Uniform Resource Locators of the selected articles were compiled into an Excel database for systematic evaluation. The Jilin Provincial CDC website did not provide relevant content and was therefore excluded. In total, 30 provincial-level CDC websites were included, covering regions with diverse economic development levels and cultural backgrounds, thereby ensuring a nationally representative sample.

Nine graduate nursing students and 3 senior nurses with more than 5 years of endocrinology nursing experience were invited as evaluators. Prior to the formal evaluation, all evaluators completed a 3-week systematic training that covered the theoretical foundation of the tool, scoring criteria, and application methods. Training included multiple discussions and simulated scoring sessions to enhance the consistency and understanding of the tool’s dimensions. After the training, twelve raters independently evaluated health education materials collected from 30 provinces, producing a total of 360 data entries. To ensure stable and reliable results, factor analysis requires an adequate sample size. Recommendations vary, with common rules of thumb suggesting a minimum of 100 to 200 participants, or a subject-to-item ratio of at least 5:1 or 10:1 [49]. CFA, in particular, requires larger samples to achieve stable parameter estimates and sufficient statistical power [50]. Simulation-based studies using Monte Carlo methods indicate that sample size requirements for CFA depend more on model complexity, factor loadings, and indicator structure than on fixed numerical thresholds [50]. Under typical psychometric conditions, about 200 participants generally provide sufficient power for models of moderate complexity [50]. Accordingly, this study used a 2-step validation process using 2 independent samples: used 2 independent samples: 150 cases for EFA and 210 for CFA, the latter yielding more than 0.80 power for detecting RMSEA deviations at 0.08 [50], ensuring model stability and reliable fit evaluation.

All statistical analyses were conducted using IBM SPSS Statistics (version 27.0; IBM Corp) and IBM SPSS Amos (version 28.0; IBM Corp). Based on the study design, consistency tests, reliability analyses, EFA, and CFA were performed to assess the scale’s reliability and validity. All analyses were 2-tailed, and a P value <.05 was considered statistically significant.

This study was approved by the Ethics Committee of Hangzhou Normal University (Approval No. 2025063). All participants provided informed consent before participating. All rater data were anonymized and deidentified prior to analysis, and the evaluated health education materials were obtained from publicly accessible sources. All evaluation data were coded, securely stored, and used solely for research purposes. The study was conducted in accordance with relevant institutional guidelines and the Declaration of Helsinki. All participants involved in this study were provided a small token of appreciation in the form of a gift worth approximately 30 RMB (US $4.13).

Comparison of the original and back-translated versions showed strong conceptual and semantic equivalence (Table 1). Across all items, the mean score for comparability of language was 1.308 (SD 0.63), and the mean score for SI was 1.077 (SD 0.28), indicating a high level of linguistic and interpretive consistency between the 2 versions. Overall variation across items was minimal, with the exception of item 3. While the phrase “policy change” can be directly translated, the research team cautioned that, within the context of public health in Mainland China, it may be misconstrued as implying radical “reform.” To avoid this potential misinterpretation, the phrase was revised to “alignment with national health policies,” thereby softening the connotation of drastic change and emphasizing policy continuity and optimization.

To further enhance comprehensibility and applicability in the Mainland simplified Chinese context, culturally specific expressions were localized while maintaining consistency with the original scoring framework. For example, in the “NOTE” section of the scale, “Facebook posts,” “Twitter messages,” “scripts for podcasts” and “call center responses” were replaced with “WeChat official account articles,” “Weibo posts” “short video scripts” and “customer service responses” respectively. These adaptations enhanced cultural relevance and user understanding.

The KMO value was 0.769, and Bartlett test of sphericity was significant (χ²₆₆=1034.141; P<.001), indicating the data’s suitability for factor analysis. Four factors with eigenvalues greater than 1 were extracted, accounting for 78.7% of the total variance, consistent with the factor structure of the original scale. The scree plot also supported a 4-factor solution, with a clear inflection after the fourth factor (Figure 2).

Item loadings (Table 3) were generally high and well defined, and all items showed satisfactory communalities, ranging from 0.475 to 0.903 [52]. Although item 4 had a slightly lower loading (0.484), it exceeded the 0.40 retention threshold and demonstrated adequate communality and semantic alignment within the “Main Message and Call to Action” dimension; thus, it was retained. Overall, the EFA indicated a clear and stable factor structure with well-defined dimensions, providing strong support for the construct validity of the C-CCI.

As shown in Figure 3, the CFA results supported the 4-factor structure of the C-CCI. Model fit indices indicated an excellent fit: χ²₄₈=90.95 (χ²/df=1.895), suggesting strong model adequacy; RMSEA=0.065, further supporting the good fit assumption. The CFI value was 0.972, exceeding the conventional cutoff of 0.90 and indicating strong model fit. Other indices also performed well: goodness-of-fit index=0.934, incremental fit index=0.972, Tucker–Lewis index=0.961, and standardized root mean square residual=0.072 (Table 4). All standardized factor loadings were substantial, ranging from 0.70 to 0.95, with each indicator loading strongly on its intended latent construct. No post hoc model respecifications were made based on modification indices, and the final model was retained as theoretically specified. These results demonstrate an excellent overall model fit, with stable and theoretically coherent parameter estimates across the 4-factor structure.

CR values were 0.907, 0.897, 0.914, and 0.897 for Main Message and Call to Action, Behavioral Recommendations, Numbers, and Risk, respectively. The AVE values were 0.831, 0.814, 0.774, and 0.645, demonstrating strong convergent validity. Discriminant validity was assessed using the Fornell–Larcker criterion [43], comparing the square root of each construct’s AVE with its inter-construct correlations (Table 5). In all cases, the square root of AVE exceeded its highest correlation with any other factor. Specifically, the Risk construct exhibited a square root of AVE of 0.803, showed a higher value than its strongest correlation with Behavioral Recommendations (r=0.442), and the same pattern was observed for Main Message and Call to Action (0.911), Behavioral Recommendations (0.902), and Numbers (0.880). Although moderate correlations were observed among some constructs, these findings indicate that the 4 dimensions are empirically distinct while remaining theoretically related.

Expert ratings demonstrated substantial consistency (κ=0.624). The scale demonstrated high internal consistency overall (Cronbach α=0.837), with dimension-specific coefficients of 0.861 (Main Message and Call to Action), 0.894 (Behavioral Recommendations), 0.827 (Numbers), and 0.910 (Risk), confirming strong reliability across dimensions.

Of the 31 provincial-level administrative divisions in Mainland China, 30 CDC websites were included (with a coverage of 96.8%), ensuring strong national representativeness. Among diabetes-related health education materials, one province (Qinghai) provided video content with text captions, 8 used text–image formats, and 21 (70%) provided text-only content. Provinces were categorized into eastern, central, and western economic regions in Mainland China [53]. Mean C-CCI scores did not differ significantly across regions (East: mean 50.06, SD 29.06; Central: mean 55.26, SD 35.52; and West: mean 56.48, SD 26.32). The national mean score was 53.84 (SD 29.74), which fell well below the recommended threshold of 90, indicating considerable room for improvement in health communication quality. No regional differences were found across dimensions except for Behavioral Recommendations, where scores in the western region were significantly higher than those in the eastern and central regions (P=.002). However, the effect size was small (η²=0.034), suggesting that regional grouping accounted for only a minor proportion of the variance in Behavioral Recommendations (Table 6).

To our knowledge, this study represents the first rigorous and systematic cross-cultural adaptation process to translate, culturally adapt, and psychometrically validate the CCI, resulting in the C-CCI. Raters with substantial professional experience underwent structured training encompassing theoretical foundations, scoring standards, and simulated evaluations. This process enhanced scoring consistency and minimized subjective bias, strengthening the methodological rigor of data collection. The C-CCI retained the 4 original dimensions: Main Message and Call to Action, Behavioral Recommendations, Numbers, and Risk, thereby preserving theoretical coherence, a structure that was further supported by EFA and confirmed through CFA. Targeted revisions improved cultural applicability, including revising item 3 from “policy change” to “alignment with national health policies” and deleting item 4 due to the incompatibility of passive voice in simplified Chinese. The deletion of items does not alter the scoring rules and, consequently, has no impact on the scoring threshold of the C-CCI. Following these adjustments, expert content validity scores improved, confirming the scientific soundness of the modifications. Psychometrically, the C-CCI demonstrated robust validity and reliability that not only met but exceeded international benchmarks for comparable Health Information Quality Assessment Tools. Prior reviews have shown that most instruments remain at Grade B due to limited structural validation and modest explained variance [30]. In contrast, the C-CCI achieved a cumulative variance of 78.7% in EFA and excellent model fit in CFA (RMSEA=0.065 and CFI=0.972), exceeding Grade B thresholds and reflecting superior structural coherence. Convergent validity was further supported by high CR and AVE values, while strong internal consistency (Cronbach α>.80) and substantial interrater agreement (κ=0.624) confirmed the instrument’s stability and reproducibility.

Using diabetes health communication materials from 30 provincial CDC websites in Mainland China, including text-only, text-image, and video formats, this study systematically evaluated the real-world applicability of the C-CCI. The tool showed a clear structure and specific scoring dimensions, and raters achieved strong inter-rater reliability after brief training, demonstrating strong practicality and replicability in applied settings. The C-CCI effectively identified weaknesses in health communication materials across 4 dimensions, demonstrating strong problem detection capacity and intervention guidance value. Its stable performance across different regions and formats (text only, text-image, and video) further supports its adaptability to diverse contexts.

Importantly, the C-CCI aligns well with China’s evolving digital communication ecosystem, in which health information is increasingly disseminated through official WeChat (Tencent) accounts, TikTok (ByteDance) videos, and other interactive platforms. Although these channels differ in format and engagement mechanisms, they share core demands for message clarity, concise information hierarchy, and actionable behavioral guidance, principles directly reflected in the C-CCI framework. From an implementation perspective, provincial CDCs may use the tool as a formative evaluation instrument during content development and internal review, using it to guide message structuring and action-oriented framing prior to dissemination. This approach enables consistent quality assurance across regions while maintaining flexibility in adapting to platform-specific communication styles.

Beyond the Mainland China context, the C-CCI’s methodological foundation may also hold relevance for other Chinese-speaking regions that share similar linguistic and readability challenges, mirroring Mainland China communication patterns noted in previous studies [54-58]. However, its broader application should be guided by further adaptation studies that consider local policy environments and platform ecosystems, ensuring evidence-based contextualization rather than direct transfer.

The CCI has been adapted into multiple languages and contexts worldwide. For instance, the Brazilian Portuguese version demonstrated applicability among primary health care professionals through field validation [28], while the Thai version emphasized dissemination of the theoretical framework and facilitated adoption in local health education [29]. These studies collectively established the feasibility of cross-cultural adaptation but were limited in the scope of psychometric testing and in the range of evaluated materials. Compared with prior validations, this study advances the methodological frontier of CCI research through multiformat empirical testing across 30 provincial CDC websites. By systematically evaluating diabetes-related materials in text-only, text-image, and video formats, the study assessed whether the measurement properties of the C-CCI remained stable across media that differ substantially in information organization. This format inclusivity, supported by consistent reliability across modalities, broadens the empirical use of the CCI framework beyond single-medium applications.

In addition, the study incorporated a more comprehensive psychometric evaluation strategy, combining exploratory and confirmatory factor analyses with rigorous validity and reliability testing. The convergence of large-scale institutional sampling and cross-format validation provides a robust methodological blueprint for future adaptations in diverse health communication settings. This study also contributes a transferable model for standardizing and assessing health information quality across varied media environments.

Using the C-CCI, this study systematically assessed the quality of diabetes health communication in mainland China, identifying overall quality levels, regional variation, and areas for improvement. Median scores across the 4 dimensions were approximately half of the maximum, revealing substantial room for enhancement. The national mean score of 53.84 (SD 29.74) fell below the recommended threshold of 90 [10], underscoring deficiencies in information completeness, structural coherence, and balance across dimensions.

Regional comparisons showed that western provinces scored significantly higher in Behavioral Recommendations than eastern and central provinces. However, the magnitude of this difference was small (η²=0.034), indicating that regional grouping explained only a limited proportion of the variance in Behavioral Recommendations. This pattern may be partly influenced by unmeasured contextual factors (eg, regional economic capacity, resource distribution, and health communication infrastructure), pending future covariate adjustment and mediation analyses. Given the heterogeneous developmental landscape across and within China’s provinces, aggregation into broad east–central–west categories may obscure underlying structural disparities and inflate within-group variance. For example, within the eastern region, provinces, such as Hainan and Hebei, may exhibit comparatively lower development profiles than other eastern provinces, while within the western region, provinces and municipalities, such as Sichuan and Chongqing, may demonstrate development levels comparable to, or even exceeding, those of some central or eastern provinces. Similar internal disparities may also be present within the central region. These findings underscore the importance of interpreting statistically significant regional differences alongside effect size estimates. Hence, future research should adopt more fine-grained, mechanism-sensitive grouping strategies or multilevel analytical frameworks, incorporating appropriate covariate adjustment and mediation analyses to distinguish structural influences from context-specific practices. Based on standardized percentage scores, Risk emerged as the lowest-performing dimension, followed by Main Message and Call to Action, with Behavioral Recommendations and Numbers showing relatively better performance. Accordingly, improvement efforts should be prioritized in alignment with these empirical deficits: (1) enriching risk communication should be the foremost priority. This could involve contextualizing both absolute and relative risk information in line with evidence-based communication principles, supported by standardized templates and iterative user testing; (2) strengthening the clarity and authority of main messages should follow, through improved headline structuring, message salience testing (eg, digital A/B testing), and inclusion of credible source attribution; (3) improving numerical presentation by using absolute frequencies, consistent denominators, and simplified visual aids can promote better comprehension across literacy levels; and (4) structuring Behavioral Recommendations into concise, sequential action steps with links to relevant resources can reinforce practical applicability. These refinements would elevate both the scientific quality and communicative impact of diabetes health materials, thereby increasing the effectiveness of public health interventions.

This study has several limitations. First, the sample was restricted to provincial CDC websites and focused exclusively on diabetes-related materials, without including hospitals, nongovernmental organizations, or social media platforms, which are also important health communication channels. Future research should expand to a broader range of health topics and communication platforms to test the generalizability of the C-CCI. Second, all raters had medical or nursing backgrounds. While this ensured professional rigor, it may have introduced certain biases. Rater training yielded acceptable interrater consistency (κ=0.624), although potential bias from disciplinary homogeneity remains. Future studies could strengthen methodological reflexivity by adopting hybrid rating panels that combine professional experts with lay raters to systematically compare expert-oriented and user-centered assessments. In this approach, lay raters could be purposively sampled to reflect variation in health literacy, education, and regional backgrounds, enabling a more representative appraisal of communication quality under real-world conditions. Such iterative and diversified designs would not only enhance measurement robustness but also model continuous improvement in cross-rater reliability and interpretive balance.

This study completed the cross-cultural translation and psychometric validation of the C-CCI, establishing a reliable and context-sensitive instrument for evaluating the clarity and actionability of health communication in Mainland China. Beyond measurement performance, the findings highlight the C-CCI’s potential as a standardized quality assurance tool to support systematic improvement in public health communication, in line with national efforts to advance health literacy and equitable access to health information [59].

By demonstrating stable performance across text, text-image, and video formats in diabetes-related materials from provincial CDC websites, the C-CCI shows strong potential for routine application in CDC practice in chronic disease communication. As exemplified in diabetes education, this framework can be extended to broader chronic disease portfolios and used to support longitudinal monitoring of health communication quality over time. Its integration into content development and review workflows could facilitate evidence-informed standardization of health materials nationwide, helping to reduce avoidable comprehension gaps and strengthen public engagement. Future research should further validate this approach across disease areas and communication contexts, consolidating the role of the C-CCI within China’s public health system.