Introduction

The relationship between dietary habits and NCDs has been widely proven1,2. Global Burden of Disease studies have shown that unhealthy diets are one of the leading factors contributing to the incidence and mortality of NCDs. Specifically, in 2017, almost 11 million deaths and 255 million disability-adjusted life years (DALYs) were attributable to dietary risk factors worldwide3. In China, the number of deaths and DALYs due to dietary risk factors related to chronic diseases reached 2.493 million and 549.95 million person-years, respectively, in 2016, posing a significant burden on public health and the socio-economy4,5.

Numerous factors, including demographic, socio-economic, environmental, and socio-cultural aspects, influence dietary quality6. Among these, nutrition knowledge (NK) plays a central role in guiding healthy eating behaviors7,8. Substantial evidence indicates that nutrition education can effectively improve individuals’ dietary practices and nutritional status9,10,11. China’s top-level designs, such as the "Healthy China 2030" Initiative and "Healthy China Action (2019–2030)," also emphasize improving residents’ nutrition knowledge as a key strategy12. However, significant deficiencies exist in current nutrition education programs targeting adults in rural China. While the Essential Public Health Services program includes the promotion of a “healthy diet”13, the absence of concrete operational guidance means that outcomes depend heavily on the capabilities of individual grassroots health workers. This fails to ensure systematic delivery, results in low population coverage, and limits the initiative’s effectiveness for adults.

Specifically, existing nutrition education models face multiple challenges in bridging the knowledge-behavior gap, making effective implementation difficult in resource-limited rural settings. First, the health effects of diet often manifest decades later, and the lack of immediate feedback discourages sustained behavioral change among adults. Second, the knowledge system for healthy eating is fragmented and complex, easily leading to information overload among the public. Third, most educational efforts are short-term interventions, whereas knowledge internalization requires sustained, repeated input. Fourth, interventions predominantly focus on children and adolescents, systematically neglecting the working-age adult population (18–64 years), who are a key demographic for NCDs prevention. Finally, face-to-face, small-scale education models struggle to balance coverage with cost-effectiveness14.

Concurrently, grassroots health workers, as the primary educators, often receive insufficient nutrition training15, resulting in generally low nutrition literacy16,17, which constrains their guidance capacity. In rural areas, factors such as low income, low educational attainment, and weak preventive healthcare infrastructure18 further complicate the delivery of effective nutrition education, contributing to the rising incidence of diet-related chronic diseases19. Research conducted in similar settings demonstrates that improving NK is a key mediator for enhancing dietary behaviors20,21, highlighting the urgent need for innovative educational models.

Therefore, to address these gaps, this study implemented a multi-channel nutrition intervention designed for resource-limited settings in rural Chengdu. This study has the following characteristics: first, it precisely targets the systematically neglected rural working-age adult population; second, it explores low-cost, sustainable communication channels through community participation to reduce reliance on specialized human resources and achieve broad coverage; finally, it aims to develop a replicable “intervention menu” through systematic evaluation. The objectives of this study are: (1) to improve residents’ NK levels; (2) to develop a practical “intervention menu”; and (3) to assess the coverage, feasibility, and cost of different channels to determine priority strategies.

Methods

Setting

The study was conducted in rural Pengzhou, a county-level city under Chengdu’s administration. Of the six initially selected villages (QY [population (pop.) 2,575], ZW [pop. 2,589], JTC [pop. 8,730], MMH [pop. 2,247], LY [pop. 4,297], and SLS [pop. 2,129]), MMH was excluded due to withdrawal of designated implementation staff, resulting in five completed study sites. The included villages represented both plains (QY, ZW, JTC) and hilly terrain (LY, SLS) (see Supplementary Fig. S1 for locations).

Study design

The study employed a five-phase sequential design (Fig. 1): (1) Baseline survey: A cross-sectional NK survey was conducted to establish pre-intervention levels; (2) Intervention development: Focus group discussion was held to identify optimal delivery methods and tailored education materials were designed. (3) Intervention implementation: A 4-week nutrition education intervention was delivered across all target villages. (4) Post-intervention survey: Second cross-sectional survey with new participants to prevent recall bias. (5) Outcomes evaluation: A comprehensive evaluation assessed intervention effectiveness and method prioritization.

Fig.1
figure 1

Flow chart of the nutrition intervention project in Pengzhou.

Full size image

Nk survey

Sample size calculation

The required sample size was calculated as 280 using the formula n = 2[(uα+uβ)s/δ]2 (α = 0.05, 1-β = 0.90, s = 13.722, δ = 2.622). To improve the stability and reliability of the results, the actual sample size was increased by 20%, resulting in a final sample size of 336.

Sampling method and participant selection

This study employed two fully independent cross-sectional samples at baseline and post-intervention phases. To ensure sample balance and operational feasibility, an equal number of participants (56 per village) were planned for recruitment across all villages. Households were randomly selected using computer-generated random numbers from village residency registries, with one permanent resident aged 18–64 years surveyed per selected household. Inclusion criteria consisted of: no intellectual disability and voluntary provision of informed consent. Exclusion criteria included: age outside target range, significant cognitive impairment, or refusal to participate. The sampling flowchart is presented in Fig. 2.

Fig.2
figure 2

Pengzhou nutrition intervention project: populations sampling protocol.

Full size image

Questionnaire design

The questionnaire comprised two sections: (1) demographic information (gender, age, education level, occupation type, and health status), and (2) nutrition knowledge (covering four dimensions: Dietary Recommendations, Food Characteristics, Diet-related Diseases, and Food Selection). Items were adapted from the "Nutrition and Health Knowledge Questionnaire for Chinese Adults Aged 18–64" developed by the China Center for Disease Control and Prevention (CDC). The original questionnaire achieved expert consensus through two rounds of Delphi consultation (11 experts; authority coefficient = 0.900), with moderate agreement (Kendall’s W = 0.324, p = 0.001)23. In this survey, the nutrition knowledge module demonstrated good reliability (Cronbach’s α = 0.73, meeting the > 0.70 threshold) and validity (KMO = 0.681; Bartlett’s χ2 = 4071.416, p < 0.001). Exploratory factor analysis identified 16 factors (cumulative variance explained = 63.80%; all factor loadings > 0.40). Post-intervention, additional items assessing NK sources were added following expert review.

NK scoring

Each original 4-option multiple-choice question was decomposed into four binary true/false items for analytical transformation. Together with the additional single-choice questions, the total number of questions amounted to 50, with each correct answer earning 2 points, summing up to a total score of 100 points. The four dimensions of NK content had maximum scores of 38, 26, 20, and 16 points respectively. Higher scores represent higher knowledge levels.

Survey implementation

Data collection used a hybrid approach: (1) Centralized surveys: Participants were invited in small batches (≤ 15 persons/batch) to designated venues (village clinics/committee meeting rooms) by village health workers. (2) Household visits: Conducted for individuals unable to attend centralized sessions. To accommodate varying literacy levels, dual administration modes were implemented: self-administered questionnaires completed under researcher supervision (maintaining ≥ 1 m interpersonal distancing to prevent answer sharing) and immediate quality checks by trained surveyors, supplemented by standardized face-to-face interviews for illiterate or visually impaired participants. All collected questionnaires underwent rigorous two-stage quality control involving fieldworker self-review followed by team leader random audits (20% verification rate), ultimately yielding final valid samples from 5 villages (baseline n = 275; post-intervention n = 280) with completely independent sampling between phases per repeated cross-sectional design.

Focus group discussion and intervention material development

Participant recruitment

We employed purposive sampling to recruit six village health coordinators (one per target village), consisting of three primary healthcare providers from rural clinics and three village committee members responsible for health work. Each coordinator was authorized to invite one local resident, resulting in a final focus group of ten participants (six coordinators and four residents aged 27–58 years). This structure ensured representation of both institutional health perspectives and community voices across all study sites.

Focus group discussion content

In-depth discussions were held on the current status of rural science popularization and the potential intervention methods to be adopted in this study. The outcomes of these discussions will serve as a direct guide for the design of intervention materials and the determination of intervention approaches.

Material design

Based on the results of the focus group discussion and expert opinions, various forms of nutrition education materials were designed and produced.

Intervention implementation

Between June 26 to July 23, 2023, using the designed promotional materials and the determined promotional frequency, a 4-week education intervention was completed in five rural villages (JTC, QY, SLS, ZW, and LY). Full methodological details are provided in Table 2.

Evaluation of intervention methods

Eighteen community-based implementers (3–4 per village) assessed intervention methods through a standardized matrix scoring system across four domains: workload, implementation difficulty, sustainability, and recommended priority, with each domain rated on a 10-point scale (1 = least favorable, 10 = most favorable).

The project team prioritized interventions through a comprehensive evaluation integrating participant-reported NK sources, implementers’ method evaluation scores, and actual intervention costs.

Statistical analysis

To ensure comparability between survey phases, post-stratified random subsampling (R 4.4.2; seed = 2023) standardized post-intervention participants to 55 per village (n = 275), matching the baseline sample size. Full reproducibility documentation is provided in Supplementary Code S2. Sensitivity analyses validated the robustness of standardization (Supplementary Material S3).

Data processing included categorical organization of demographic variables: participant age was stratified into 18–34, 35–49, and 50–64 years; occupations were classified as either health/wellness workers (healthcare/food/education/wellness) or other; and health status was dichotomized based on presence/absence of physician-diagnosed chronic diseases (hypertension, diabetes, hyperlipidemia, coronary heart disease, or stroke).

All data were double-entered in EpiData 3.1 with consistency checks. Categorical variables were presented as counts (percentages) with χ2 tests for comparisons. Continuous data exhibited non-normal distribution based on Shapiro–Wilk testing (all p < 0.05) and Q-Q plot evaluation, and were therefore reported as medians with interquartile ranges (IQR; Q1-Q3) and full ranges. Case weighting was applied to adjust for inter-village population heterogeneity. Statistical comparisons used Mann–Whitney U test for two-group comparisons and Kruskal–Wallis H test for multi-group analyses, with post hoc pairwise Mann–Whitney tests (Bonferroni-adjusted) when statistically significant. The two-sided α level of 0.05 defined statistical significance.

Results

Focus group results and intervention strategy development

Focus group discussions with village health coordinators and villagers revealed critical insights into current NK education implementation and existing promotional approaches in rural communities (Table 1). Participants consistently noted that conventional methods, including lectures, printed leaflets/pamphlets, and posters, suffered from low engagement due to unappealing content formats. Through discussions, participants and the project team collaboratively identified six enhanced delivery methods with higher anticipated adoption rates, specifying implementation modalities for each: (1) village loudspeaker broadcasts, (2) WeChat group video sharing, (3) redesigned NCDs-focused leaflets, (4) informational posters, (5) online quizzes with prizes, and (6) Children-Teaching-Families (CTF) Program (Table 2).

Table 1 Current status of nutrition promotion in rural communities: Key findings from focus group discussions.
Full size table
Table 2 Six nutrition intervention methods identified through focus group discussions.
Full size table

Development of intervention materials

The project team developed six tailored health communication tools: 1. Audio recording: Ten nutrition education audio clips (3–4 min each) were developed based on the Dietary Guidelines for Chinese Residents24, featuring lively and conversational presentations to improve engagement. 2. Video: Colorful and lively PowerPoint presentations were produced and converted into videos accompanied by audio. 3. NCDs-focused leaflets: Colorful, large-print leaflets (A4 size) were designed for diabetes, hypertension, hyperlipidemia and cancer prevention, featuring bolded text on high-contrast colored backgrounds to maximize visibility and reader engagement. 4. Posters: Two types of posters were created, depicting the Chinese Dietary Guidelines Tower and food labels. 5.Question bank: A 100-item NK question bank was deployed on WenJuanXing®, where each participant was randomly assigned 10 questions with immediate feedback providing correct answers and explanations. 6. Children’s engagement materials: The intervention incorporated a color-coded NK-sharing logbook for children to document family education activities, along with a pre-validated science booklet from project resources for independent learning.

NK survey

Following the withdrawal of MMH Village from the intervention, analyses were restricted to the five remaining villages.

Demographic equivalence of independent samples

Comparison of baseline and post-intervention independent samples showed balanced distributions across all measured demographics: gender (p = 0.932), age groups (p = 0.273), education (p = 0.311), occupation (p = 0.617), and chronic disease status (p = 0.335) (all p > 0.05; Table 3). This demographic equivalence between the two cross-sectional surveys supports the validity of pre-post comparisons.

Table 3 Demographic characteristics of independent samples in baseline and post-intervention surveys.
Full size table

NK scores: baseline vs. post-intervention samples

The results showed that the median NK score of villagers increased from 60.0 in the baseline sample to 70.0 in the post-intervention sample. Median NK scores significantly improved across all demographics, including gender, age group, village, education level, occupation, and chronic disease status (all p < 0.05) (Table 4).

Table 4 Nutrition knowledge (NK) scores: Baseline vs. post-intervention samples.
Full size table

When comparing age groups, the 35–49 and 50–64 age groups scored significantly lower than the 18–34 group (median: 58.0/56.0 vs. 64.0, p < 0.001) at baseline. Post-intervention, all age groups reached comparable median scores (70.0), with no residual inter-group differences (p = 0.779).

Significant inter-village disparities were observed at baseline (p < 0.001). Specifically, SLS village demonstrated significantly lower scores than both JTC and QY villages (median: 58.0 vs 62.0/62.0). The post-intervention sample SLS and ZW villages showed the most pronounced improvements (median increases: 14 and 11 points). After Bonferroni adjustment, village differences became non-significant (p > 0.05).

Regarding education, villagers with primary/junior high education had significantly lower baseline scores (both median = 56.0) vs. higher education groups (64.0–68.0, p < 0.001). The post-intervention sample demonstrated gains inversely related to baseline education: + 16 (primary), + 14 (junior high), + 4 (high school/college), compressing knowledge gaps. Within the post-intervention sample, although overall differences between education-level groups remained significant (p = 0.039), all possible pairwise contrasts became non-significant after Bonferroni adjustment (adjusted p > 0.05).

Occupational analysis revealed significantly higher baseline NK scores among health/wellness workers (p < 0.001), but these inter-group disparities were resolved in the post-intervention sample (p = 0.670). Chronic disease patients initially scored higher, but patients showed smaller gains than the non-patients(+ 8 vs + 12 points), eliminating baseline disparities (p = 0.581).

Domain-specific NK scores: baseline vs. post-intervention samples

Between-survey comparisons revealed significant improvements across all four domains (all p < 0.05). The median score for the Diet-Related Diseases domain showed the largest increase (4 points, 33.3%), while the Dietary Recommendations and Food Selection domains each increased by 2 points (8.3% and 20.0%, respectively). Although the median score for the Food Characteristics domain remained unchanged, the score distribution showed significant upward shift (p < 0.05), indicating generalized improvement (Table 5). Most notably, the lowest score in the Dietary Recommendations domain rose dramatically from 2 to 10 points.

Table 5 Domain-specific NK scores:Baseline vs. post-intervention samples.
Full size table

Among the 50 questions, 30 showed significant improvements in correct rates after the intervention. Notably, the improvement rates for questions regarding "Maximum daily intake of added sugars" "Excessive salt intake can increase the risk of developing gastric cancer" and "Consuming excessive amounts of sugary drinks can increase the risk of developing dental caries" showed particularly marked improvement (> 50% increase in correct responses).

Sources of villagers’ NK

Survey results indicated that the primary source of NK for villagers was WeChat group video sharing (70.0%), followed by loudspeaker broadcasts (43.2%) and posters (43.2%). Nutrition quizzes and leaflets accounted for 30.8% and 32.5%, respectively, while the CTF program was the least common source, accounting for only 5.4%. Approximately 43.6% of participants reported using three or more sources to obtain NK.

Evaluation and prioritization of intervention methods

Eighteen community-based implementers evaluated six intervention methods across four domains—workload, implementation difficulty, sustainability, and recommended priority—using a 10-point scale. Based on priority scores, loudspeaker broadcasts ranked highest with 8.5 points, followed by WeChat group video dissemination with 7 points. The remaining methods—printed leaflets, posters, online quizzes with prizes, and the CTF program—all received comparable scores of 6 points (Table 6).

Table 6 Intervention method evaluations by community-based implementers (10-point scale).
Full size table

The project team evaluated funding requirements, identifying the online quizzes with prizes and the CTF program as the most costly interventions due to recurring expenses, whereas posters and leaflets incurred moderate production costs. In contrast, loudspeaker broadcasts and WeChat video sharing required only initial development costs, establishing them as the most cost-effective options.

Based on comprehensive evaluation, the interventions were prioritized as follows: (1) loudspeaker broadcasts (where feasible), (2) WeChat group video sharing,(3) posters, (4) NCDs-focused leaflets, (5) CTF program, and (6) online quizzes with prizes.

Discussion

This study demonstrates that a comprehensive nutrition education intervention designed for rural areas can effectively improve residents’ NK, with a significant increase of 10 points in the median post-intervention score. More importantly, the research revealed substantial variation in intervention effectiveness across different demographic subgroups. Consistent with extensive existing literature, baseline data indicated significant demographic disparities in NK levels, such as higher scores among females compared to males25,26,27, younger individuals compared to older adults28,29,30, those with higher education levels compared to those with lower education25,26,31,32, and participants involved in health/wellness occupations33. However, post-intervention, NK scores increased significantly across all subgroups, demonstrating a clear "catch-up effect": subgroups with lower baseline levels, such as males, older adults, individuals with lower education, those not working in health-related fields, and non-chronic disease patients, showed particularly marked improvement. This suggests the intervention holds special value for reducing knowledge disparities between social strata and promoting health equity, as most pre-existing inter-group differences became statistically insignificant after the intervention. This finding aligns with and extends the work of Asakura et al.34, providing further evidence that well-designed nutrition education programs can not only enhance overall knowledge levels but also effectively reduce knowledge inequality.

Furthermore, the intervention’s effectiveness varied across different knowledge domains, with the most significant improvement observed in knowledge related to diet-related diseases, where the median score increased by 33.3% from baseline. Particularly noteworthy is that correct response rates for key knowledge points directly related to daily dietary behaviors, such as "the association between high salt/sugar intake and specific disease risks," increased by over 50%. These knowledge areas are crucial for individual health decision-making, and their significant improvement lays a vital cognitive foundation for promoting subsequent behavioral changes (e.g., reducing cooking salt use, limiting sugary beverage intake). It also indicates that targeted, practical nutrition knowledge is more readily accepted and internalized by rural residents.

This study identified WeChat group video sharing and loudspeaker broadcasts as the most cost-effective nutrition education methods for rural Chinese populations, each demonstrating distinct advantages and implementation contexts.

WeChat, developed by the Chinese technology conglomerate Tencent, is the nation’s dominant multipurpose platform integrating instant messaging, social networking, and mobile payment services. With 1.38 billion monthly active users (Tencent Q3 2024 financial report) and penetration exceeding 95% among Chinese smartphone owners, it has been extensively utilized for health interventions. For instance, its application has been shown to produce clinically meaningful outcomes in chronic disease management, such as significantly improving blood pressure and lipid profiles in patients with coronary artery disease35. Furthermore, a randomized controlled trial demonstrated that a health education program delivered via a WeChat public account significantly improved infectious disease-specific health literacy among the general population, proving to be a highly feasible and acceptable model36. To explore this potential in rural nutrition education, our study capitalized on pre-existing village WeChat groups—a multi-user chat platform supporting multimedia exchange—which were established during COVID-19 containment with nearly all households enrolled, to deliver nutrition education videos. The video format facilitated intuitive delivery of nutrition knowledge through visual demonstration, simultaneously engaging both village elders and migrant workers. Anecdotal reports and existing literature37 suggested that video content is often better received than text-based materials due to its accessible format, which reduces cultural and literacy barriers. However, our field observations revealed two major implementation constraints that limit its effectiveness: First, the sheer volume of group messages often caused important content to be overlooked, requiring systematic prompting by village doctors. Second, creating well-received video content proved challenging, requiring dialect adaptation and culturally tailored examples that demanded substantial staff training and resource allocation.

Loudspeaker systems proved highly effective as an adaptive rural communication tool, receiving the highest priority score (8.5/10) from implementers. Their success was attributed to three key factors: (1) programmable automation enabling scheduled broadcasts without staff supervision, significantly reducing workload; (2) passive knowledge delivery overcoming literacy barriers through audio messages; and (3) existing infrastructure utilization, with 64.7% of Chengdu’s villages equipped with broadcast capabilities by 2023. However, geographical limitations persisted, as mountainous terrain reduced coverage—only 30% of villagers in hilly areas reported message reception versus 43.2% in flat terrains. This technology’s effectiveness was particularly notable for elderly populations with limited digital access. This finding is corroborated by international evidence: efficacy in malaria prevention was demonstrated in Myanmar38, high community acceptability was confirmed in Vietnam39, and a successful HIV/AIDS intervention in Thailand highlighted its potential to drive behavior change40. By applying this approach to nutrition education in rural China, our study not only demonstrates its high feasibility but also identifies a critical terrain-related limitation, thereby extending the evidence base for this low-cost intervention.​

Traditional print materials (posters and leaflets) demonstrated moderate effectiveness after redesign, achieving 32.5–43.2% recall rates. This approach aligns with the suitability framework proposed by Finnie et al., which emphasizes moving beyond simple reading level to consider critical factors such as content, graphics, and layout for effective communication41. Specifically, the modifications prioritized content simplification, with each material focusing exclusively on one core topic—for instance, providing dietary guidance for a single prevalent chronic disease. Other key refinements included using enlarged A4 formats with high-contrast typography and distilled key messages, directly addressing the framework’s emphasis on layout/typography and reduced literacy demand. Despite these improvements, focus groups noted persistent engagement challenges, as these materials still struggled to capture attention in information-saturated environments—a limitation that underscores the framework’s often-overlooked dimension of learning stimulation in static print media.

The online quizzes with prizes were effective in boosting public engagement, which is consistent with the known effect of small monetary incentives. However, a critical assessment of sustainability reveals a key limitation. Our study demonstrated that this approach incurred substantial monetary costs (averaging ￥15–20 per rewarded participant), which severely undermined its cost-effectiveness. Therefore, while the method is efficacious for achieving short-term engagement during specific campaigns (e.g., annual health observances), the high recurring costs render it unsuitable for sustainable, routine nutrition education programs.​

The CTF (Child-Teaching-Family) program reached only 4.2% of households, primarily limited by its dependency on having school-aged children enrolled in participating schools. Although valuable for child nutrition education (book distribution cost: ¥8–12/child), its narrow demographic penetration reduced overall intervention efficiency.

It is important to emphasize that our study does not dismiss traditional methods but rather constructs a “communication toolkit” containing various tools with clear application scenarios and a recommended sequence, offering a concrete plan for China’s "Rural Revitalization Action." For instance, posters and leaflets remain indispensable supplements due to their high accessibility42; prize-based quizzes can rapidly boost participation during specific campaigns43; while the CTF program focuses on child education for long-term impact. This comprehensive strategy is consistent with the insights of Morrow, Singh et al., who demonstrated the superiority of integrating multiple methods over relying on a single approach44,45.

Overall, this study demonstrates that hybrid digital-traditional approaches offer distinct advantages over conventional strategies: where individual/group education achieves limited coverage due to staffing constraints46, and environmental modifications (e.g., food labeling) require sustained policy support47, our model enables cost-efficient, wide-reaching dissemination through two synergistic mechanisms—(1) utilization of China’s unique digital ecology (near-universal WeChat adoption and revitalized rural broadcast systems), and (2) streamlined health worker engagement requiring only modest training. The empirically derived implementation hierarchy (loudspeaker broadcasts > WeChat group videos > posters > leaflets > CTF program > quizzes) provides a replicable framework for resource-limited settings with comparable organizational capacity.​This finding not only aligns with WHO’s emphasis on integrating context-adapted digital tools into health systems48 but also offers a concrete, empirically validated pathway for its implementation in resource-limited rural settings.​

Limitations

This study has several limitations. First, the relatively brief 4-week intervention period, while demonstrating significant improvements in NK, precludes assessment of long-term knowledge retention and behavioral sustainability. Future research should extend the follow-up period and incorporate dietary behaviors and biomarkers (e.g., blood pressure, blood glucose). Second, measurement constraints include the NK source survey’s inability to differentiate intervention-acquired knowledge from pre-existing or externally obtained information, potentially confounding effect size attribution across delivery methods. Furthermore, while the comprehensive intervention design reflects real-world conditions, it is difficult to precisely quantify the independent contribution of each communication channel. Future studies could use factorial or stepped-wedge trial designs to disentangle these effects. Third, although baseline recruitment reached 275 participants (98.2% of the minimum target n = 280), while post-intervention met the full target, the final analytical survey for both phases was maintained at n = 275. This marginal baseline shortfall may have reduced statistical power for subgroup analyses. Regarding study design, the absence of a control group reflects the pragmatic realities of implementing community-wide interventions where: (1) within-village controls would risk contamination due to universal exposure to public health messaging, and (2) between-village controls were ethically untenable given baseline NK deficits. While this limits causal inference, our effect sizes align with those from controlled trials in similar settings49.

Policy and practical implications

This study yields the following key implications for policy and practice:

First, promote the institutionalization and precision of effective intervention models. It is recommended to integrate the intervention strategy validated by this study—prioritizing "loudspeaker broadcast as the core component, supplemented by WeChat group videos"—into the routine assessment system for primary public health services. Implementation should adhere to a context-specific approach: in areas with robust internet infrastructure, digital channels should be prioritized to leverage their efficiency and broad coverage; in regions with limited digital access, such as poor signal coverage or device scarcity, the offline model of "loudspeaker broadcasts + visits by community health workers" should be strengthened to bridge the digital divide and ensure health equity.

Second, clarify the leading responsibility of local governments in resource integration and sustainable scaling. The sustainability of this intervention relies heavily on the active participation of local governments. County-level governments should take the lead in establishing a cross-departmental coordination mechanism (encompassing health, education, agriculture, publicity, etc.) to integrate nutrition education objectives into various local initiatives, such as the Rural Revitalization Strategy, thereby forming synergistic policy effects.

Third, optimize resource allocation to address sustainability challenges. To tackle the difficulty of creating high-quality educational materials at the grassroots level, it is recommended that provincial-level authorities centrally develop a repository of high-quality, multi-dialect audio and video materials as public goods. Concurrently, stable fiscal budgets should be established to cover the necessary infrastructure and maintenance costs of the intervention, reducing dependence on short-term project funding.

Conclusions

This study confirms that a comprehensive nutrition intervention strategy, designed for resource-limited rural areas, can effectively enhance the nutrition knowledge of adult residents and demonstrates potential in reducing knowledge disparities across different demographic subgroups, thereby promoting health equity.

The core practical contribution of this research lies in empirically identifying a prioritized “intervention menu” for rural nutrition education through comparative analysis. Specifically, the combination of "loudspeaker broadcasts + WeChat group videos" has been established as the preferred strategy for routine nutrition education in resource-constrained settings, owing to its characteristics of wide coverage, low cost, and sustainability.

In summary, the findings of this study provide critical empirical evidence and a feasible pathway for achieving the core objectives outlined in the "Healthy China 2030" initiative—namely, "improving residents’ health literacy" and "narrowing health inequalities." The successful scaling and sustainability of this strategy depend on its systematic integration into local public health systems, supported by stable policies and funding, to ultimately generate lasting improvements in the health of rural residents.