Proposed by Ajzen (1991), TPB is widely used to predict and explain the general behaviour of individuals and their behavioural decision-making processes. This widely used psychological theory suggests that behavioural intention is the most reliable predictor of an individual’s actual behaviour and is determined by three psychological factors: attitude, subjective norms, and perceived behavioural control (Ajzen, 1991). When an individual has more positive attitudes towards performing a certain behaviour, perceive more social pressure to perform the behaviour, and feel empowered to perform the behaviour, then the individual will have a higher intention to perform the behaviour, and consequently, the actual behaviour will more likely occur. The TPB is generally useful for predicting behavioural intention and has been widely recognised as a conceptual model for capturing behaviour. However, the TPB is not perfect, and the main criticism is that it focuses on rational reasoning and does not thoroughly explain behaviour (Visschers et al., 2016). Nejad et al. (2004) pointed out that other specific factors should be introduced to supplement and improve research on individual behavioural decision-making in specific situations to further improve the effectiveness and applicability of the TPB. Many studies have attempted to improve the explanatory power of TPB models by adding other components as predictors of behaviour (Tang et al., 2023).

The TPB has generally been used to predict consumers’ behavioural intentions and has been widely used in related studies on health, pro-environment, and sustainable consumption behaviours (McEachan et al., 2011; Shukri et al., 2016; Tang et al., 2023). Some scholars have used the TPB as a theoretical basis to predict and explain individual intentions regarding healthy or sustainable diets (Visschers et al., 2016; Biasini et al., 2021; Jha et al., 2023). Many of these previous studies have focused on healthy eating intentions, and the few have focused on sustainable eating intentions to reduce food waste. Despite differences in perspectives between the literature and our study, the former provided inspiration and reference for our work. For example, Ploll and Stern (2020) applied TPB analysis to reveal the significant relation between subjective norms, attitudes, behavioural intentions, and vegetarian and vegan behaviours. Sogari et al. (2023) provided new insights into how TPB constructs (attitudes and subjective norms) significantly predict the intention to adopt a healthy diet. Kramer et al. (2023) supported the validity of the TPB in explaining healthy eating and cooking behaviour, suggesting that interventions targeting TPB constructs (attitude, subjective norms, perceived behavioural control, and intention) may help create behaviour change in specific populations. These studies suggested that attitude, subjective norms, and perceived behavioural control are important predictors of consumers’ dietary adoption intention and that the likelihood of triggering an individual’s dietary behaviour practise depends on their dietary intention (Kramer et al., 2023). However, the influence of various aspects (attitude, subjective norms, and perceived behavioural control) on consumers’ intention to adopt dietary patterns may vary with cultural their background. As Hwang et al. (2003) pointed out, different cultural contexts may produce completely different consumer behaviours.

Based on the existing measurement items (Supplementary Table S1 lists the reference sources), we designed structured questionnaires to collect data and analyse the proposed hypotheses (see as Figure 1). The full questionnaire is provided in the Supplementary materials. The three-part questionnaire investigated the participants’ demographic characteristics; past eating behaviour (PEB), health consciousness (HC), and environmental concern (EC); and perceived value (PV), attitude (AT), perceived behavioural control (PBC), subjective norms (SN), and intention to adopt (IA) SHDP. The questionnaire also included a clear definition of Sustainable Healthy Diets, which was carefully explained to respondents by trained investigators during the survey. The items in the questionnaire, which addressed the eight core variables in the analytical framework of our study, were measured on a five-point Likert scale, with respondents answering how much they agreed with specific statements, ranging from ‘strongly disagree’ = 1 point to ‘strongly agree’ = 5 points. All measurement items were derived from well-established scales in existing studies and adapted appropriately to the context of SHDP adoption by Chinese consumers. The initial questionnaire was first test-filled by experts and investigators. Based on the feedback, we modified the semantic ambiguities and inaccuracies of expression in the questionnaire. Next, we conducted a pre-survey amongst 50 consumers in Wuxi. Questionnaire topics were adjusted, added, or subtracted after two rounds of testing, to form a final questionnaire for data collection.

The offline field survey was conducted in April 2023 across the administrative districts of Wuxi City, China. Wuxi is known as the ‘land of fish and rice’ in China, with abundant produce and a favourable ecological environment. The economy of Wuxi is well developed, with its GDP per capita ranking first in the country for three consecutive years from 2019 to 2022. Residents of the city have strong purchasing power and a relatively high level of scientific literacy in food safety and dietary health. The Jiangnan diet, an healthy dietary pattern similar to the Mediterranean diet, is popular in Wuxi. We used non-probability sampling, which is suitable for studies that aim to test hypotheses on the relation between specific variables and behaviour (Leary, 2004). This method also offers benefits in terms of cost, convenience, and time (de Medeiros et al., 2016; Fiandari et al., 2019). The survey was conducted mainly in large supermarkets, farmers’ markets, and neighborhoods with high customer flow, targeting Wuxi residents aged 18 years and older. Trained investigators provided clear explanations and definitions of relevant terms and concepts in the process of the one-to-one questionnaire survey. A total of 422 questionnaires were distributed and 402 valid questionnaires were collected, with a valid sample ratio of 95.3%. The studies involving human participants were reviewed and approved by Jiangnan University. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Structural equation modelling (SEM) allows for the simultaneous analysis of many relations and is one of the most commonly used methods for interpreting behaviour and survey data (Secer et al., 2023). It can assess measurement errors, calculate latent structures, and estimate and evaluate complex (Stein et al., 2012) or multifaceted models (MacKenzie, 2001) from observed variables. We applied the SEM method of maximum likelihood estimation to test the research path hypothesis and mediation effect. SPSS (version 25.0) and AMOS (version 26.0) were used for SEM analysis. The SEM algorithm involved two stages: validating the measurement model and testing the structural model.

Table 1 summarises the respondent profiles. Compared with the statistics published by the Wuxi Municipal People’s Government, our sample was similar to the Wuxi population in terms of sex, age, and income; however, the proportion of respondents with higher education was higher than that of Wuxi as a whole. Overall, the survey sample was acceptably representative. The mean values of each variable are shown in Supplementary Table S1, in which the mean value of consumers’ IA (3.69) implied a relatively strong intention to embrace SHDP.

First, we tested the scale using the Kaiser–Meyer–Olkin (KMO) sample fit test and Bartlett’s sphericity test. The KMO value (0.930, more than 0.7) and Bartlett’s sphericity test result (χ² = 6191.881, p < 0.001) indicated that the data were suitable for factor analysis. We then conducted a confirmatory factor analysis to analyse the reliability, convergent validity, and discriminant validity of the scales. As shown in Table 2, the Cronbach’s alpha coefficients for each latent variable were above the critical value of 0.7 (Taber, 2018), and the factor loadings were more than 0.6 (Chin et al., 1997), indicating an acceptable level of reliability. The average variance extracted (AVE) value for each latent variable was greater than the critical value of 0.5 (Bagozzi and Yi, 1988), and the composite reliability (CR) values were all greater than the critical value of 0.7 (Fornell and Wernerfelt, 1987), indicating high convergent validity amongst the variables. In addition, the square root of the AVE for each latent variable (Table 3) was greater than the correlation coefficients of the rows and columns located below the diagonal, suggesting acceptable discriminant validity between the variables (Hair et al., 2019).

Next, to avoid the impact of the non-normal distribution of data on SEM hypothesis testing (Li and Shao, 2023), we used SPSS 25.0 to test the distribution of the sample data. The results (Supplementary Table S1) showed that the absolute value of the skewness coefficient of the sample data ranged from 0.008 to 0.900, none of which was over 3, and the absolute value of the kurtosis coefficient ranged from 0.019 to 0.772, none of which was over 8, indicating that the sample data conformed to a normal distribution and could be used in the SEM for hypothesis testing (Teisl et al., 2009).

Given that the data were obtained from respondents’ self-reported questionnaires, we needed to test for common method bias (CMB). Therefore, we conducted an exploratory factor analysis of all variables through Harman’s single-factor test (Harman, 1976): CMB is confirmed if the first factor accounts for >50% of the variance amongst the variables (Gao et al., 2022). Our test showed that the eigenvalue variance percentage of the factor was 35.087%, much lower than the threshold of 50%, suggesting that CMB was not an issue.

To test for multicollinearity amongst the latent variables, we conducted a variance inflation factor (VIF) test. The results showed that the VIF of each variable in the model ranged from 1.622 to 3.465 (see as Supplementary Table S2), much lower than the threshold of 10 for multicollinearity amongst latent variables (Timm, 2002). Therefore, the latent variables in our model had no obvious multicollinearity problem.

We constructed a structural equation model based on the theoretical model. Before testing the hypothesised relations, we compared the extended TPB model with the original TPB model. Table 4 presents the various fit indices of the model. According to previous scholars, the chi-squared degrees of freedom ratio (χ²/d f ) of the model should be between 1 and 3; the root mean square error of approximation (RMSEA) and standardised root mean square residual (SRMR) should be less than 0.08 (Hu and Bentler, 1998); and the normed fit index (NFI) should be greater than 0.8 (Chakraborty et al., 2008). According to the index values of χ²/d f , RMSEA, and SRMR (Table 4), the extended TPB model had a better fit than the original TPB model. Moreover, the explanatory power of the extended theoretical model for consumer IA SHDP (R²_IA = 0.756) was better than that of the original TPB model (R²_IA = 0.640). Thus, our extended theoretical model had a better explanatory power than the original TPB model.

As shown in Figure 2, the assessment results of the structural model demonstrated that AT significantly and positively affected IA (β = 0.347, p < 0.001), confirming H1. SN did not significantly influence IA (β = 0.076, p = 0.186); thus, H2 was rejected. PBC positively influenced IA (β = 0.154, p < 0.05), supporting H3. PEB positively influenced IA (β = 0.293, p < 0.001); thus, hypothesis H4 was supported. HC positively influenced IA (β = 0.270, p < 0.001), supporting H5. EC did not significantly influence IA (β = 0.040, p = 0.417); thus, hypothesis H6 was rejected. PV significantly and positively affected AT (β = 0.610, p < 0.001); thus, hypothesis H7a was supported. Interestingly, PV did not have a direct effect on IA; thus, H7b was rejected. As shown in Figure 2, IA explained 75.6% of the total variance of all its antecedent variables, indicating that our hypothesised model had high explanatory validity for consumers’ intention to adopt SHDP.

Using AMOS 26.0 software, we analysed the mediating effects included in the model. We used the bias-corrected nonparametric percentile bootstrap method, with the number of samples set to 2,000. We thus determined the direct, indirect, and total effects of the PV, AT, and IA. The results of the bootstrap test, shown in Table 5, demonstrated that consumer PV had no direct effect on IA; however, the indirect and total effects were significant. The confidence intervals for the indirect effect of the path PV → AT→AI did not contain zero, indicating that PV had an indirect, fully mediated effect on IA through AT. Therefore, H7c was supported.

This study extended the TPB model to explore the drivers that promote the intention to adopt SHDP of Chinese consumers. The results showed that attitudes played an important role in driving consumers’ intention to adopt SHDP—it increased significantly when the consumers had a positive attitude towards SHDP. This finding coincided with previous conclusions on the positive correlation between attitude and behavioural intention with respect to food purchasing and healthy eating intention (Al-Swidi et al., 2014; Sogari et al., 2023). Our study confirmed that consumers’ perceived value affected their attitudes and had a significant indirect effect on their intention to adopt SHDP through attitude. Similarly, Albertsen et al. (2020) reported that consumer acceptance of food innovations may be related to perceived value. Indeed, SHDP is a novel dietary pattern for Chinese consumers, and its adoption implies food innovation. To the best of our knowledge, in the field of research on sustainable healthy diet consumption behaviour, our study provides the first empirical evidence for the indirect incentive role of consumers’ perceived value on the adoption of SHDP in the Chinese sociocultural context.

We also confirmed the importance of perceived behavioural control as a factor directly related to consumers’ intention to adopt SHDP. This is consistent with the findings of previous studies on Chinese consumers’ food purchase intention (Li and Jaharuddin, 2020; Qi and Ploeger, 2021). Perceived behavioural control reflected consumers’ subjective feelings regarding the availability, identification, and convenience of sustainable healthy diets. When consumers subjectively perceived sustainable healthy diets as more accessible and easier to identify, then they would have fewer obstacles to adjusting their dietary patterns towards healthy and sustainable changes. As such, their intention to adopt SHDP would naturally increase.

Subjective norms did not have any effect on consumers’ intention to adopt SHDP, coinciding with previous findings on the limited ability of subjective norms to predict behaviour in the TPB (Yazdanpanah and Forouzani, 2015). In a meta-analysis of the psychosocial determinants of pro-environmental behaviours, Bamberg and Möser (2007) found no direct link between subjective norms and intention. A possible explanation is that subjective norms are unstable and vary in different contexts. Moreover, different respondents may be influenced by different target groups, and the effectiveness of subjective norms is reduced when respondents do not identify with a defined target group (family, friends; Trafimow and Finlay, 1996).

The expanded TPB we constructed also contained other constructs, such as past eating behaviour, health consciousness, and environmental concern. Consumers’ past eating behaviours significantly influenced adoption intention towards SHDP. That is, the more a consumer’s actual past eating behaviour met the criteria for SHDP, the more acceptable and willing they would be to adopt it. Rivis and Sheeran (2003) suggested that the impact of past eating behaviours depends on whether the behaviours have become habitual, in which the individual becomes guided by external environmental cues and develops an automatic response that is not driven by cognitive factors. Thus, past eating behaviour may positively influence the intention to adopt a dietary pattern, even if the consumer lacks precise knowledge of sustainable healthy diets.

Health consciousness can predict a person’s intention to engage in healthy behaviours (Rahamat et al., 2022). We found that consumers’ health consciousness directly and positively influenced their intention to adopt SHDP. Indeed, amid the increased health consciousness worldwide, health consciousness is not only a driver for promoting healthy food choices amongst consumers but also a major driving factor behind consumers’ adoption of sustainable dietary behaviours (Kareklas et al., 2014).

Meanwhile, we found that environmental concern did not have a direct effect on consumers’ intention to adopt SHDP. Consumers’ values and personal priorities are known to be important factors that influence their behavioural intentions (Arora et al., 2022); egoism-driven health consciousness is more important than altruism-driven environmental concern in the case of dietary consumption. Research on organic food has shown that consumers’ environmental concerns are drivers of green consumer behaviour in China (Ahmed et al., 2021). However, Chinese consumers’ awareness of SHDP is lacking because it is still an emerging concept in China, and most ignore the connection between diet and environmental sustainability. The objects of environmental concern are familiar environmental behaviours, such as reducing the use of plastic bags (Sun et al., 2017) and purchasing recycled products (Zhang and Luo, 2021). This may explain how environmental concern does not drive the intention to adopt SHDP.

We expected our study to contribute to the literature in two ways. First, it expands the scope of research on the TPB in the fields of food and diet consumption. As a social psychological model that can effectively identify and elucidate consumer behaviour and behavioural intention, the TPB has seen applications in the field of food/diet consumption behaviour, but they have focused on organic food purchasing, environmentally friendly food consumption, healthy eating behaviour, and food waste reduction behaviour (Al-Swidi et al., 2014; Visschers et al., 2016; Jha et al., 2023). Meanwhile, our study extends the TPB in terms of health consciousness, environmental concern, and past eating behaviour and applies it to explore the main influencing factors and internal mechanisms of Chinese consumers’ intention to adopt SHDP. Our work fills the research gap in the TPB in the field of sustainable consumption, especially in the sociocultural context of China.

Second, it provides a new perspective for promoting consumer intention to adopt SHDP. Previous studies on individual-level sustainable healthy dietary behaviours are typically qualitative works on cognitions, attitudes, motivations, and barriers; the few quantitative empirical analyses have focused on developed countries in Europe and on the US (Van Loo et al., 2017; Verain et al., 2017; Claessens et al., 2023). By collecting sample data from a Chinese city in Wuxi, our empirical analysis provides evidence of the importance of enhancing consumers’ perceived value in motivating them to adopt SHDP, thus filling the current research gap related to the relation between consumers’ perceived value and their adoption of SHDP, and providing additional evidence for elucidating the psychological mechanisms of consumers’ dietary change, especially in the context of healthy and sustainable diet consumption.

Our findings also provide important insights for the development of dietary change interventions and offer valuable implications for policymakers. First, interventions should be designed and implemented to target attitude and awareness, as attitude and health consciousness are the most direct drivers motivating consumers to adopt SHDP. Given the close relation between attitude and perceived value, interventions should encourage consumers to develop favourable perceived values to boost positive attitudes towards SHDP. As mentioned by Verain et al. (2017) and Jalil et al. (2020), raising awareness and shaping attitude through education and information campaigns to encourage voluntary dietary behaviour change amongst consumers are the most commonly used soft policy interventions aimed at facilitating a shift towards sustainable healthy diets. Educating consumers about the impact of their dietary choices on personal health and the environment may increase their awareness, interest, and recognition of the value of sustainable healthy diets (Van Loo et al., 2017), thereby improving attitudes and facilitating dietary shifts. The accumulation of information increases the overall social awareness, gradually motivates consumers to take action (Cummings and Proctor, 2014), and encourages educational campaigns to raise social awareness (Willett et al., 2019). We recommend that governments and public welfare organizations strengthen dietary education and information popularisation. In particular, education campaigns should inform consumers on not only the ‘what’ (promoting the concept of sustainable healthy diet consumption and providing scientific knowledge about sustainable healthy diets) but also the ‘why’ (emphasising the egoistic and altruistic values of sustainable healthy diets, focusing on the health benefits and complementing it with scientific information on the environmental impacts of diets) and the ‘how’ (providing consumers with personalised advice and guidance, such as healthy and environmentally friendly recipes and easy-to-follow cooking tips).

Given that individuals with little knowledge or interest in sustainable healthy diets may be less likely to respond immediately to awareness-raising interventions (Jalil et al., 2020), policymakers should also focus on the impact of perceived behavioural control and past eating behaviour on consumer adoption of SHDP. We recommend ‘nudging’ interventions that push consumers towards sustainable healthy diets but do not rely on education to raise awareness. In nudging, ‘nudges’ use mild, more implicit intervention strategies that can easily and even unconsciously influence people’s choices and behaviour in a desired direction (Thaler and Sunstein, 2008). Interventions designed on behavioural nudges (product accessibility, default options, priming), such as meat-free days in public canteens and positive positioning of sustainable healthy foods in retail settings, can lead consumers to voluntarily switch to SHDP in a gentle way in the immediate choice context (Junghans et al., 2015; Van Loo et al., 2017; Pelle G Hansen et al., 2021). These types of nudging interventions based on behavioural insights may help repeatedly trigger consumers’ healthy sustainable eating behaviour to change their future dietary choices.

The integrated implementation of multiple policies and interventions will create more opportunities for dietary change in the population. Therefore, in addition to soft policy interventions, hard policy interventions, including laws, fiscal measures, subsidies, penalties, and other economic and structural measures (Willett et al., 2019), are critical for building a supportive environment for sustainable healthy diets. For example, incentives or subsidies should be provided to encourage the production and marketing of sustainable healthy foods, promote the sustainable development of the food industry, and support food production practises that reduce GHG emissions. In addition, promoting sustainable healthy diets at the population level requires multiple actors to work together, which in turn requires strengthened governance and partnerships to address the challenge of changing diets in an effective cross-sectoral manner (De Schutter et al., 2020). Interventions to promote mass sustainable healthy diets must benefit people of low socioeconomic status and be developed in conjunction with the social safety nets necessary to effectively address poverty and inequality (De Schutter et al., 2020).

There are inevitably some limitations in interpreting our study, and these limitations provide opportunities for further research. First, we investigated the drivers of consumers’ intention to adopt sustainable healthy dietary patterns, as exemplified by residents of Wuxi City in China. However, it must be acknowledged that other factors, including consumer self-regulation and the social environment, were not covered due to constraints in the research scope and methodology. Second, our study did not address consumers’ actual food choices. Future studies could combine participation measurements and actual food choice data, such as those obtained through choice experiments or dietary intake assessments, to further investigate the relation between participants and sustainable healthy food choices, and to more objectively estimate the health and sustainability attributes of sustainable healthy diets. Finally, as with most research in the field, our study relied on self-reported measures that, whilst providing valuable insights, may introduce some biases and limitations, such as the possibility that respondents may be influenced by social expectations and thus deviate from actual behaviour (Fisher, 1993). Additional experimental and observational studies are required to overcome these limitations.