Keywords

1 Introduction

Health promotion is fundamental in the drive to reduce the growing chronic disease burden across the globe (WHO, 2008). The prevention of chronic diseases requires behaviour modifications. Health promotion interventions aim to engage and empower individuals and communities to choose healthy behaviours and make changes that reduce the risk of developing chronic diseases and other morbidities. “SMART Eating” is an example of such a health promotion intervention aimed at optimizing the consumption of fat, sugar, salt and fruits and vegetables among urban North Indian adults from diverse socio-economic backgrounds, i.e. low-income group (LIG), middle-income group (MIG) and high-income group (HIG) (Kaur et al., 2018). SMART stands for Small, Measurable and Achievable dietary changes by Reducing fat, sugar and salt consumption and Trying different vegetables and fruits.

Health promotion interventions like “SMART Eating”, being contextual and community-based, are often complex (Tremblay & Richard, 2014). Considering the complexity of this intervention trial, following the pragmatic worldview (Creswell, 2014), a mixed methods approach applying both deductive and inductive logics combining both quantitative and qualitative approaches was used to better understand unhealthy dietary behaviours. The intervention design was based on a healthy and productive combination of theory-based (top-down) approach and active involvement of end-user communities (Kaur et al., 2020a).

The research objectives were identified and refined based on our personal experiences, interactions and discussions with the community, the literature reviews and discussions with public health and nutrition experts. Based on the work experience in this community setting, we realized that the high prevalence of chronic diseases is one of the major health issues. Discussions with the community people to explore their understanding of the causes of chronic diseases revealed that the majority were not even aware of the link between unhealthy diet and chronic diseases and, even more, what constitutes a healthy diet as a preventive measure. These discussions motivated us to work on exploring innovative ways to deal with unhealthy dietary behaviour – a major risk factor for chronic diseases. This chapter describes the epistemological analysis of the research framework, paradigms, approaches and methods used in the planning, implementation and evaluation of the “SMART Eating” health promotion intervention that used a cluster randomized trial design.

2 The “SMART Eating” Research Framework

Evidence of theoretical use for developing methodology and deciding on analyses can guide navigating the intervention process. The transtheoretical model , the theory of planned behaviour and the health belief model are the common behaviour change theories attempting to understand dietary behaviours at the individual level (Davis et al., 2015). These have been used less in collectivistic cultures like India, where dietary decision-making occurs at the family level (Daivadanam et al., 2015). Furthermore, there is seldom a “one-size-fits-all” solution to address an issue adequately. No single theory can account for all the complexities of dietary behaviour change as health behaviour, culture and context differ widely (Darnton, 2008).

A potential solution could be the use of planning models such as the PRECEDE–PROCEED model (PPM) (Glanz et al., 2008) or the intervention mapping (IM) approach (Bartholomew et al., 1998), both of which provide specific guidance for systematically developing theory-based interventions through the selection of appropriate micro-, meso- and macro-level behaviour change theories. The PRECEDE–PROCEED model is one of the most comprehensive and most used approaches in ecological and ethical health promotion practice (Porter, 2015) and has been found useful to researchers in conducting health behaviour change trials (Aldiabat, 2013). Both the PPM and IM share similar characteristics in applying a social-ecological framework and behaviour theories for intervention building, emphasizing multilevel multiple interventions and involving stakeholders and the existing resources in intervention development (Yoo & Kim, 2010). However, IM employs PRECEDE assessments based on the PPM and calls for logic models, while the PPM itself is an example of a logic model , in that it links the causal assessment and the intervention planning and evaluation to one overarching planning framework (Glanz et al., 2008).

Therefore, we chose to use the PPM to plan the “SMART Eating” trial. The model was a road map for the entire research work and specific theories and models (the social-ecological model (SEM), the transtheoretical model , the attitude–social influence–self-efficacy model (ASE) and the UK Medical Research Council (MRC)’s framework) directed towards the goal through different phases of the project. Hence, the research framework for “SMART Eating” was guided by a combination of compatible theories, models and frameworks (Fig. 30.1).

Fig. 30.1
figure 1

Research framework for ‘SMART Eating’ intervention. MRC Medical Research Council, ASE Attitude–Social-influence–Self-efficacy, QUAN Quantitative, QUAL Qualitative

Full size image

The PRECEDE–PROCEED model ’s “Educational Diagnosis” PRECEDE (Predisposing, Reinforcing and Enabling Constructs in Educational Diagnosis and Evaluation) guided the intervention development and the “Ecological Diagnosis” PROCEED (Policy , Regulatory and Organizational Constructs in Educational and Environmental Development) guided intervention alignment, implementation and evaluation (Glanz et al., 2008). Designing effective interventions for achieving the desired dietary behaviour changes requires an in-depth study of people’s behaviours situated in their socio-cultural and interpersonal contexts. The social-ecological model (SEM) is pertinent to understanding behaviour in terms of interactions among various multilevel influences (Bronfenbrenner, 1977). In the first phase (qualitative formative research ), we explored stakeholders’ perspectives to understand the multilevel influences on dietary behaviours, the barriers and facilitators to dietary behaviour change and intervention preferences (Kaur et al., 2020a). The formative research findings helped in understanding the local contexts and tailoring the design and conduct of the “SMART Eating” intervention for enhancing its efficacy (Darlington et al., 2020).

During the next phase, we used the transtheoretical model to identify the stage of readiness for fat, sugar, salt, fruit and vegetable consumption of the participants. However, we realized that applying the stages of change to dietary behaviour poses difficulties because of the complexity of dietary behaviours and the lack of stage stability over time even before intervention (Brug et al., 2004). As dietary behaviours may be different for a different set of diets, people may hold different self-efficacy beliefs and may be in different stages of change for each of the desired behaviours (Adams & White, 2004). The lack of knowledge about the recommended dietary intake among the participants (98%) (baseline survey) could have led to misconceptions about their consumption, and the majority were categorized in the maintenance stage (Lechner et al., 1998; Bogers et al., 2004). Their intention to change based on subjective qualitative assessment (whether they think their consumption is sufficient/low/high? yes/no) of the adequacy of their dietary intake did not match the dietary consumption estimated quantitatively using a food frequency questionnaire (FFQ) (Lechner et al., 1998), which is regarded as the pre-contemplation stage (Greene et al., 1999). Hence, considering the participants' pre-contemplation stage, a common intervention was developed based on the qualitative formative research (Kaur et al. 2020a), the empirical literature on interventions and the dietary guidelines by the National Institute of Nutrition, India (ICMR, 2011).

The UK Medical Research Council (MRC)’s framework for designing and evaluating complex interventions guided the process evaluation (Moore et al., 2015). The attitude–social influence–self-efficacy (ASE) model (Engbers et al., 2006) and the SEM-based formative research findings (Kaur et al., 2020a) informed an a priori identification of the potential mediators, namely, attitude, social influence, self-efficacy, monthly household purchase and consumption, and guided the mediation analyses for understanding the mechanisms of the intervention’s effect on the outcomes. The ASE model , a social cognition model that integrates concepts from the theory of planned behaviour and social cognitive theory, is commonly used to predict and explain dietary behaviours (Sandvik et al., 2007; de Vries et al., 1988).

3 Methods

3.1 Community Involvement

Community involvement is a fundamental principle of health promotion. We involved the stakeholders during all trial phases, i.e. intervention development, implementation and evaluation. Given that an ecological approach is central to the concepts and methods of health promotion, involving participants as co-researchers rather than as just data sources as in conventional biomedical approaches – being our priority – and the lack of prior empirical evidence from India on understanding multilevel influences on multiple dietary behaviours among urban North Indian adults, we conducted formative research to develop a context-specific culturally acceptable intervention. We chose a qualitative approach to explore the diverse experiences and perspectives related to the consumption of fat, sugar, salt, fruits and vegetables (Brink & Wood, 1998). Focus group discussions (FGDs) were conducted with adults. The participants were recruited with the help of influential people from local communities who provided immense support and cooperation. The FGDs were held at venues suggested by the participants as per their convenience. As our department has been working in the study areas to strengthen healthcare services, its strong rapport with the communities over several years has resulted in openness in responses. The key findings are presented in Box 30.1 (Kaur et al., 2020a).

Box 30.1 Multilevel Influences on Dietary Behaviours

  • Individual level: Lack of knowledge about the recommended dietary intake; lack of food measurement skills; low-risk perception for chronic diseases; frequent consumption of packaged food; frequent reporting of eating out; low self-efficacy in adapting to healthy eating related to the inability to influence family cooking habits.

  • Family level: The role of women as the “main chef” of the family, family norms and preferences of family members.

  • Structural level: Societal norms, high cost of fruits and vegetables and use of pesticides/harmful injections on fruits and vegetables.

The participants’ narratives indicated that individuals alone may not be able to adapt healthy dietary behaviours without family support, given that food cooked at home is consumed by all members of the family and family food habits are strongly influenced by the prevalent social norms and cultural practices regarding food consumption. Hence, family or community approach could be more likely to be effective in settings where collectivistic cultures exist. Furthermore, the lack of knowledge about the recommended dietary intake suggested that efforts should be made to translate written documents on these guidelines into practical applications. Given that provision of knowledge alone may be insufficient to motivate people to initiate the desired changes, considering multilevel influences on dietary behaviours (Box 30.1), we decided that, alongside knowledge provision, the intervention should focus on increasing awareness about the benefits of adapting healthy dietary behaviours, improving risk perceptions related to chronic diseases, developing food measurement skills and enhancing self-efficacy. These qualitative findings may be applicable to other similar settings in India (transferability or analytical generalizability). We discuss the involvement of the stakeholders in intervention implementation and evaluation in the forthcoming sections.

3.2 Multi-Channel Communication Approach

Identifying appropriate communication channels is crucial for successful intervention implementation. Face-to-face individual counselling or group education, being effective methods of education, have been often used for promoting healthy dietary behaviours; however, they may be expensive to scale up in resource-constrained settings (Pomerleau et al. 2005). As unhealthy dietary behaviours are prevalent in most populations, nutrition interventions should reach large numbers of people at a low cost. mHealth (mobile health) could be a potential solution to this problem as IT has penetrated across different socio-economic strata. However, in developing countries, mHealth is still at an early stage of development, and the effectiveness of such interventions to improve dietary behaviours has not yet been explored adequately (Marcolino et al., 2018). Furthermore, the exclusive use of a single technology alone may be insufficient to alter complex, multifactorial, multiple dietary risk behaviours (Svetkey et al., 2015).

Therefore, in the qualitative formative research, we also explored the participants’ preferences for the intervention content, messages, channels and modes of communication, duration of the intervention and the target audience for implementing the intervention (Box 30.2). Qualitative information on preferences for the use of IT tools was supplemented by a rapid feasibility survey of the study area and the baseline survey, which revealed extensive use of mobile phones (100%), including smartphones (91%) and the Internet (92%), among all families, including those from the low-income group. Thus, considering the participants’ intervention preferences (Box 30.2), the intervention was implemented for a period of 6 months using a multi-channel communication approach (Figs. 30.2 and 30.3) including information technology (short message service – SMS, email, WhatsApp and a project website) and interpersonal communication along with the provision of innovative educational tools (“SMART Eating” kit) (Kaur et al., 2020b). Based on the participants’ language preferences and literacy levels, the contents were made available in the local languages (Hindi and Punjabi) and in English. Content validation was done by eight experts from related fields. The website design was validated by three experts. The intervention aids were pre-tested, and modifications were made.

Fig. 30.2
figure 2

Description of the ‘SMART Eating’ intervention

Full size image
Fig. 30.3
figure 3

Educational aids. Dining tablemat, kitchen calendar © Kaur et al. CC BY 4.0; project website by authors, in School of Public Health, PGIMER, Chandigarh. (Open from 2016 to 2020)

Full size image

Box 30.2 Participants’ Intervention Preferences: Findings from the Qualitative Formative Research

  • Content of the intervention: Provision of knowledge on the recommended quantity of fat, sugar, salt, fruits and vegetables; how to measure the quantity; benefits of eating the recommended quantity; risks of eating more or less.

  • Methods of communication: Telephone calls, SMS, educational videos, Internet, social media (WhatsApp), mass media (TV and radio) and face-to-face individual or group education. In addition, low-income group participants preferred innovative printed materials (pamphlets/posters/banners).

  • Duration of the intervention: Preferences ranged from 1 month to 1 year, and the majority preferred at least 6 months.

  • Frequency of message delivery: Weekly, fortnightly.

  • Language preferences: Hindi, Punjabi, English.

  • Target audience for intervention implementation: The majority suggested that the person responsible for cooking should be the primary target audience for implementing the intervention. Some participants indicated the need to involve other family members to help the target audience (Kaur et al., 2020b).

Given the importance of the family’s role in dietary behaviours, we used the family champions approach for implementing the intervention, adapted from the health champions approach (Warwick-Booth et al., 2013). One family champion (an adult in the family who usually cooks food) was selected as the target audience from each family to motivate other family members to adapt healthy eating behaviours. Considering the potential challenge that not all family champions will be using IT tools, a co-champion to assist the family champion was identified by the family champion. Guidance was provided on the use of different components of the intervention. The trial had no provision for any kind of monetary incentives for participation in the trial.

3.3 The Cluster Randomized Controlled Trial Design

The randomized controlled trial (RCT) design is a widely accepted “gold standard” experimental design for evaluating health interventions (The Lancet, 2019). We opted for a two-arm cluster randomized controlled trial design to test the proposed intervention’s effectiveness, as, in an individual-level RCT, there is a threat of contamination (Campbell et al., 2012). Compared with individually randomized trials, cluster RCTs are more complex to design. They require a greater number of participants for equal statistical power and an estimate of between-cluster variance or intraclass correlation (often not available from the published literature) and entail more complex analysis due to clustered observations – all of which are important for internal validity in pragmatic trials measuring effectiveness (Kerry & Bland 1998a, b). Cluster RCTs involve randomization of clusters (groups of individuals) to intervention conditions, and participants’ recruitment often takes place post randomization. Thus, the lack of blinding to allocation status to those recruiting the participants can result in poor internal validity. However, due to the nature of interventions, blinding is not always feasible, especially in behavioural interventions (Campbell et al., 2012), like “SMART Eating”. However, the random selection of the families and individuals from clusters can minimize that bias.

Based on the type of housing, a total of 12 clusters (i.e. 4 each from the LIG, MIG and HIG) were recruited before randomization. Similar socio-economic group clusters were paired based on the clusters’ geographical distance to form six pairs to avoid a possible spillover effect. Then, from each of the six pairs, clusters were randomly allocated to the intervention and comparison arms in a 1:1 ratio using computer-generated simple randomization by a researcher not involved in the study. While accounting for clustering in sample size estimations, across 12 clusters, a total of 732 families were recruited with equal numbers of families from each cluster using systematic random sampling. One adult per family was randomly selected as an index case to measure the change in the outcomes.

Given that most intervention studies target high-risk groups or vulnerable populations, we aimed to focus on the general population in a real-world setting as unhealthy dietary behaviours and related chronic conditions are common. Furthermore, we believe that, although chronic diseases are relatively well managed within the government healthcare systems, government healthcare services lack the resources to focus on health promotion or to facilitate behaviour changes among otherwise healthy people. For example, for a focus group participant, a low-salt diet meant not adding extra salt. Additionally, we believe that reinforcement is an essential aspect of health promotion. Hence, to enable all people to achieve their fullest health potential – “equity in health”, a basic principle of health promotion (Grabowski et al., 2017) and the mission of the International Union for Health Promotion and Education (IUHPE) – we included all participants, except pregnant women and those with critical health problems.

Furthermore, to ensure equal representation of all socio-economic groups in the project, the sample was stratified by the type of housing as assigned by the Chandigarh Administration – i.e. low-income group (LIG), middle-income group (MIG) and high-income group (HIG). The type of housing thus served as a proxy for socio-economic status (SES). Formative research indicated that participants from all SESs perceived themselves at greater risk of nutrition-related diseases and all, even those from the LIG, were supportive of the dietary change intervention. Therefore, the LIG was also included in the project, despite the fact that affordability could be a barrier to “SMART Eating” in this group (John & Ziebland, 2004). In recognition of the high price of fruits and vegetables, we emphasized the use of seasonal fruits and vegetables available at low price and explicitly pointed out that money spent on junk foods could be used to purchase fruits and vegetables.

3.4 Mixed Methods Design

Mixed methods research (MMR) is becoming increasingly used and recognized as a major research paradigm , along with quantitative and qualitative approaches (Johnson et al., 2007). Mixed methods approach based on the pragmatic worldview (using pluralistic approaches based on the complexity of the problem) involves a combination or integration of quantitative (post-positivism philosophy) and qualitative (constructivist worldview) approaches (Creswell, 2014). It provides a more comprehensive understanding of the problem under study rather than using either of these (Creswell & Plano Clark, 2007). MMR is unique and superior to the single approach as it improves inference quality (internal validity in QUAN and credibility in QUAL) and inference transferability (external validity in QUAN and transferability in QUAL) (Tashakkori & Teddlie, 2003).

Mixed methods research is increasingly used in the health promotion field (Nutbeam, 1998), and the experts emphasize that the use of mixed methods is most appropriate for evaluating complex health promotion interventions (Pommier et al., 2010). As health promotion is concerned with producing generalizable knowledge and transforming reality, qualitative and quantitative methods should be viewed as complementary in health promotion research (Rootman et al., 2001). Choosing the most appropriate design for a particular study, i.e. whether purely mixed with equal status to both QUAN and QUAL or dominant design, is pivotal to MMR (Almalki, 2016). We used an embedded mixed methods research design, which sees one method of inquiry being used in a supportive secondary role to another primary method of inquiry, which enables making sense of the study in its entirety. This method is used in quantitative experimental designs (Creswell, 2014). Our approach is a quantitative dominant mixed methods research, with the QUAN component in the dominant role (cluster RCT) and the QUAL component in the supporting role in the overall design (formative research and process evaluation) (Johnson et al., 2007). The intervention was informed by the qualitative formative research (qual data), the effectiveness of the intervention was evaluated using a cluster RCT design (QUAN data) and the participants’ perceptions regarding the effect of the intervention (process evaluation) were explored using post-intervention extreme case interviews (qual data).

3.5 Measurements

Measurement of outcomes needs to be accurate to check whether an intervention is effective or not. An accurate estimation of diet requires a complete and unbiased assessment of foods consumed and a comprehensive database of the nutrient content of foods. We used food frequency questionnaires (FFQs) to capture information on 113 food items that had been validated in an urban North Indian setting (Mahajan et al., 2013). The reported intakes were converted into food and nutrient data using Indian food composition tables. However, errors inherent in subjective dietary assessment may lead to biases towards or away from the null hypothesis (Naska et al., 2017). FFQs, being highly comprehensive, are prone to overestimation compared to other methods (Steinemann et al., 2017). On the other hand, recall bias in self-reported dietary intake could underestimate the actual intake. Similarly, the possibility of social desirability bias could lead to overestimation of the effect size. Hence, objective assessment, where possible, needs to complement self-reported measures.

Nutritional biomarkers provide an objective assessment of dietary outcomes; however, they are often expensive (Bingham, 2002). Although 24-h urinary sodium excretion, a nutritional biochemical biomarker, is the gold standard for estimating salt intake, spot urine samples can also be used to estimate the mean change in the population salt intake (Petersen et al., 2017b). The equations to estimate 24-h salt from spot urine samples have been previously tested in different Indian populations (Petersen et al., 2017a). Hence, in addition to subjective assessment based on FFQs, we chose spot urine samples to estimate salt intake objectively. Changes in the body mass index, blood pressure, haemoglobin, fasting plasma glucose and serum lipid levels were the other anthropometric, physiological and biochemical biomarkers (Dragsted et al., 2017). The significant net changes observed in some of these biomarkers indicated that the aforementioned biases in self-reported measures did not bias the study results.

3.6 Data Analysis Methods

Besides robust measurements, careful selection of robust analysis methods for both quantitative and qualitative data analyses is imperative for unbiased estimates of the intervention’s effect and correct interpretation and generalizability of the findings. Randomized trials present special requirements for analysis, including the use of the intention-to-treat (ITT) principle, accounting for unobserved causes of the outcome and multiple comparisons and understanding the causal mechanisms. The loss to follow-up is often hard to avoid in randomized trials. We used ITT analysis by the inclusion of all participants in data analysis to reduce selection bias. ITT is strongly recommended in RCTs to minimize type-I error and produce an unbiased estimate of the treatment effect (Campbell et al., 2012). It ensures comparability between groups and maintains sample size as the original randomization, and the number of participants remains unchanged.

The difference-in-differences (DiD) method was used to determine the net mean change in the outcomes in the intervention group relative to the comparison group. This approach removes the biases from comparisons over time in the intervention, which could be due to other unobserved causes of the outcome (Zhou et al., 2016). The analyses were adjusted to account for the clustering in the data using multilevel mixed effects linear regression models to reduce type-I error (Hayes & Moulton, 2017).

Furthermore, multiple comparisons are common in experimental studies due to logistic issues. The statistical inference in experimental research is always drawn from the statistical testing of hypotheses, in which an acceptable cut-off of probability (0.05 or 0.01) is used for decision-making. However, the probability of committing false statistical inferences would considerably increase when more than one hypothesis is simultaneously tested, i.e. in multiple comparisons (Smith et al., 1987). For multiple confirmatory hypotheses testing, there needs to be a proper adjustment to the alpha level to preserve the overall family-wise type-I error rate, which is mandatory. To this end, Holm’s adjustment was made for the four primary outcomes, i.e. fat, sugar, salt and fruit and vegetable intake (Chen et al., 2017). However, several statistical tests conducted for secondary outcomes and subgroups as exploratory analyses could serve as a guide to developing confirmatory hypothesis in future studies.

Mediation analysis is a recommended statistical tool for understanding the mechanisms of impact in RCT interventions (Moore et al., 2015). Multiple mediation analyses assessed whether, and to what extent, attitude, social influence, self-efficacy, monthly household purchase and consumption mediated the intervention's effect on dietary behaviour outcomes. As in RCTs, it is the intervention that is randomized, not the mediators; we used the “ANCOVA approach for mediation”, which is strongly recommended to reduce confounding in mediation analysis in RCTs (Landau et al., 2018). Hence, the mediation analyses were adjusted for baseline measures of the mediators and outcomes and the baseline covariates.

The focus group data were audio-recorded and transcribed verbatim. The data were analyzed using narrative thematic analysis informed by the framework analysis approach (Ritchie & Lewis, 2003) and techniques derived from the grounded theory approach (Holton, 2010). Guided by the social-ecological model , two analysts, the first and the second author, developed the coding framework based on “a priori” codes derived from the topic guide and the empirical literature and emergent codes. The coding framework was discussed with the third author, an expert in qualitative data analyses with years of qualitative research experience. Differences in coding were resolved by discussions and consensus.

3.7 Comprehensive Evaluation

Guided by the PRECEDE–PROCEED model , we undertook a comprehensive evaluation that was participatory, continuous and theory-driven, as shown in Fig. 30.1. To ensure a comprehensive evaluation, a mixed methods approach was used, which is the recommended design for evaluations within the trials, including cluster RCTs (Grant et al., 2013). The effectiveness evaluation using a cluster RCT showed significant net improvements in all dietary behaviours (i.e. reduction in fat, sugar and salt consumption and an increase in fruit and vegetable intake) in the intervention group relative to the comparison group (Kaur et al., 2020b) – an indication that it is feasible to change multiple risk behaviours. A significant net beneficial effect on body mass index, diastolic blood pressure, fasting plasma glucose, triglycerides and urinary salt excretion observed from the exploratory analysis supported the above-presented findings of the confirmatory hypothesis testing.

Subgroup comparisons showed significant improvements in all dietary behaviours among all socio-economic groups, except for salt consumption in the HIG. Further, the effect of the intervention was significant for fat, sugar and fruit and vegetable intake irrespective of age and gender. However, the effect on salt intake was significant in the younger age group and in women. These findings indicated that for fat, sugar and salt behaviours, the LIG performed better than did the MIG and HIG. However, for fruit and vegetable consumption, the increase observed in the LIG, though significant, was of much lower magnitude compared to those observed in the MIG and HIG. The lesson learned from these findings is that reducing fat, sugar and salt consumption (activities that do not involve money) is feasible even among low-income groups. However, achieving a higher effect on fruit and vegetable behaviour among the LIG would require supportive government policies. Lack of the intervention effect on salt intake in the HIG group warrants exploration of design-suitable interventions for improving salt consumption among this group – another learning.

Besides effect evaluation, process evaluation (Table 30.1) of complex interventions is an indispensable component of the comprehensive evaluation. Guided by the MRC framework (Moore et al., 2015), the process evaluation documented high fidelity and reach (91%) in both the groups, adequate dose for SMS (93%), WhatsApp (89%) and “SMART Eating” kit (100%) with low website usage (50%) in the intervention group and inadequate use of pamphlets (46%) in the comparison group. The participants reported no unintended adverse effects. The mediation analysis adjusted for the covariates indicated that the improvements in attitude, social influence, self-efficacy, household purchase and consumption behaviours mediated the effect of the intervention on dietary behaviour outcomes. These findings suggest that health promotion interventions that target these components will be more likely to be effective in optimizing the intake of fat, sugar, salt and fruits and vegetables. Post-intervention feedback from the participants indicated that the intervention was well-liked by the participants across all socio-economic groups, with some of them demanding programme extension. We found the evidence for adherence to the intervention from the participants’ responses to the social networking app messages, their queries and interactions with the intervention implementer and home observations on the use of the “SMART Eating” kit provided to them.

Table 30.1 Process evaluation components of the “SMART Eating” intervention
Full size table

Besides, it is important to understand that not everyone treated or intervened will change. Deviant cases always lie on a continuum indicating two extremes, i.e. poor performers exhibiting no change in their behaviour and excellent performers. From the extreme case interviews with poor performers, we learned that no perceived effect of the intervention on fruit and vegetable behaviour among the LIG was related to high cost and among the MIG and HIG to poor quality (use of pesticides/harmful hormonal injections for enhancing the production). These findings highlight the need for government policies for promoting healthy dietary behaviours. Similarly, no effect of the intervention on fat, sugar and salt behaviours among deviant cases across all socio-economic groups was found to be related to the lack of family support and perceived adequacy of personal intake. These findings point to the need to explore more effective strategies to enhance family support and improve perceptions about the adequacy of personal intake in future interventions.

Further, an important issue in evaluating behaviour change interventions is the evaluation of the maintenance of behaviour change or positive gains following the intervention, which could be defined as a significant between-group difference for the outcomes at post-intervention and at follow-up (Fjeldsoe et al., 2011). The importance of maintenance of outcomes to inform the translation of evidence-based health behaviour interventions into practice is often overlooked, and only a few interventions include evaluation of the maintenance of dietary behaviour change. Our initial protocol, approved by the institute ethics committee, included a plan to assess the maintenance of behaviour change through between-group differences at the end of the 6-month intervention (post-intervention) and at 12 months (follow-up). However, this project, being a doctoral project (of the first author) with limited time and resources, could not be continued beyond 5 months.

We believe that the positive effect gained would have been maintained over time as our intervention was well received by the participants and used practical strategies, such as emphasis on small gradual changes, enhancing self-efficacy through family support, development of food measurement skills, interactive help to allow interactions with the intervention implementer, timely replies to their queries by the experts and provision of innovative material to meet their preferences, which they would retain for a long time and use to reinforcing desired behaviours in the family – the enabling, empowering and relapse prevention strategies. Given that behaviour change is a challenging and complex process, further work is needed to determine the maintenance of the intervention effect along with exploratory research on understanding the facilitators of and the barriers to long-term maintenance.

4 Implementation Challenges and Remedies

Through formative research , we identified certain issues that might affect implementation and took steps to address them in the intervention (e.g., developing the intervention based on the participants’ preferences, identifying the target audience for intervention implementation at the family level and the provision of interactive help to discuss difficulties in using the intervention). Despite the best efforts and careful implementation planning, we faced some unforeseen challenges. The intervention implementation through different components was easiest, except the use of the social networking app in the beginning.

First, retaining participants in WhatsApp groups was the most challenging task. Although while obtaining informed consent, they were explained that group messaging will be used for WhatsApp messages, many users left the groups immediately saying “we are not familiar with other group members”, “we do not want to be in any group (especially girl co-champions)”, and others left without specifying any reason. Very few participants continued with the group messages. Second, as an alternative to group messaging, an individual message to a larger audience (91% WhatsApp users) was not allowed by the WhatsApp service. The broadcast list, an alternative to these problems (allowing an individual message to 256 recipients in one go with a single click), requires the recipients to save the sender’s contact number in their phone books – again a tedious exercise in a community-based nutrition intervention. Despite these challenges, we could manage these difficulties, and WhatsApp proved to be one of the most useful, successful and interactive methods of education in this project. Therefore, a solution to these problems needs to be explored for efficient use of WhatsApp or similar social media in future public health interventions.

Another unanticipated implementation challenge was the minimal use of the project website. At 1 month of intervention implementation, the project website was reported to be least used for being password-protected, causing difficulty for the participants to login. Therefore, the password was removed, and the visitor count was added as an indicator of the website usage. Although remedial measures improved the website usage from 7% at 1 month to 50% by post-intervention, the website was still the least used component. The majority of the participants in this project reported using the Internet just for using social networking apps. Despite extensive mobile Internet penetration and timely remedial measures, low website usage indicates the need to explore strategies for improving website usage for planning future web-based interventions (Blackford et al., 2017).

5 Epistemological Issues in Health Promotion Intervention Research

Evidence-based health promotion practice needs to focus on the appropriate measurement of effectiveness using theories (Green, 2000). The “SMART Eating” intervention designed based on a theory-based (top-down) approach and qualitative research evidence through community consultations, evaluated using a cluster RCT, was successful in achieving the desired behaviour changes. The intervention implementation based on users’ preferences using the family approach made the intervention acceptable to the users across diverse socio-economic groups. Further, the intervention posed no additional burden on the families as it used the available community resources. It focused on capacity building and facilitating change at the individual and family levels through enabling the home environment.

One of the major epistemological issues in health promotion intervention research is understanding which paradigm or a set of paradigms is the most appropriate for health promotion intervention research. The complexity in health promotion practice needs to be dealt with simple, cost-effective, innovative, culturally and geographically appropriate models, ensuring community participation. Therefore, we used a transdisciplinary approach, which accommodated diverse perspectives and contributed to enhancing the understanding of unhealthy dietary behaviours (Soskolin, 2000), resulting in a common conceptual framework (Albrecht et al., 1998). Experts from different disciplines (sociologists, epidemiologists, an anthropologist, nutrition experts and other community health professionals) agreed on the methodological pluralism required in answering the research questions of the “SMART Eating” project. Hence, the selection of the research approach was based on the pragmatic worldview that allows pluralistic approaches based on the complexity of the health issues (Creswell, 2014). Using a dialectic method to resolve the differences, a mixed methods approach combining both quantitative and qualitative methods to data collection, analysis and evaluation was deemed appropriate. Furthermore, the experts reached a consensus on the need to use multiple compatible theories to practise evidence-based health promotion.

Another important epistemological issue is the external validity, which refers to the extent to which study results can be applied to other settings. The “SMART Eating” experiences showed that evaluated using a cluster RCT, the intervention showed the evidence of effectiveness in improving dietary behaviour among urban Indian adults from diverse socio-economic backgrounds. Hence, the findings of this research could be used to plan and implement interventions in different settings with modifications to account for the contextual factors. The use of IT as a health promotion strategy for future interventions in different settings may prove successful because it removes the limitations of resources and geographical distances.

Further, growing evidence supports the need for interventions to target factors that influence health behaviours at multiple levels (Paskett et al., 2016). However, it is not always feasible to intervene at all levels due to the challenges in designing, implementing and evaluating multilevel interventions, such as the need for teams with diverse expertise, managing complex teams over extended periods and unpredictability of timelines (Sallis, 2018) – another important issue. Using qualitative formative research, we learned that intervening at the individual, interpersonal and structural levels would be appropriate for “SMART Eating”. However, the “SMART eating” intervention was restricted to individual and interpersonal levels. Although, to address structural influences, namely, high cost and poor quality of fruits and vegetables (due to rampant use of pesticides) and media influence (misleading advertisements portraying the usefulness of packaged food often high in fat/sugar/salt), we devised strategies such as an emphasis on substituting snacks with seasonal low-cost fruits and vegetables, altering the home environment (making healthy food more visible) and food label reading; intervening at the structural level was beyond the scope of this project due to logistic reasons, being a time-consuming activity.

However, we acknowledge that intervening at social-structural levels (increasing tax on unhealthy food, access to unprocessed food, making healthy food available at grocery stores or provision of packaged food containing low fat/sugar/salt) would certainly add to the effectiveness of and complement interventions at the individual and interpersonal levels, leading to more effective and sustainable behaviour changes. Hence, through the dissemination of our study findings, we try to influence the “structure” by the “agent” as our participants demanded supportive policies for promoting healthy eating behaviours (fruits and vegetables) and restrictive policies to facilitate healthy food production and wanted their voice to reach the government, as evident from these quotes from our published study (Kaur et al., 2020a):

These days, media is affecting our thinking very badly. Advertisements are showing usefulness of ready-to-eat food products, not even a single advertisement tells how harmful are they?” (MIG); “Fruits are too expensive to be consumed; same is the scenario for vegetables. Government should open such stores where poor people get commodities [fruits and vegetables] at subsidized rates” (LIG); “Government should do something…vegetables are grown with medicines…ghiya [gourd] is grown to 1 kg with injection…this is all wrong…it is happening everywhere across the globe…corrective measures should be taken…” (LIG); “Can you convey our message to the government that food adulteration [with injections] and use of pesticides/fertilizers should be reduced (MIG).

6 Conclusions

The “SMART Eating” project was guided by the PRECEDE–PROCEED – a community-oriented ecological health promotion model , which guided the selection of compatible theories. A transdisciplinary approach was used, which accommodated diverse perspectives and contributed to the holistic understanding of dealing with unhealthy dietary behaviour. Based on the pragmatic worldview, a mixed methods approach was used combining quantitative and qualitative methods applying deductive and inductive logics. An embedded mixed methods design was used with the QUAN component in the dominant role and the QUAL component in the supporting role in the cluster RCT. Community involvement, empowerment and equity were the other distinctive features of this health promotion intervention research.

From the implementation of the “SMART Eating” health promotion intervention, we learned how different worldviews can be integrated into health promotion research . Furthermore, we illustrated that instead of following only one major research tradition (quantitative or qualitative), philosophy or school of thought and working in watertight compartments, health promotion research needs to be open and flexible and different traditions should learn from each other and collaborate to meet the community’s needs while taking contexts into consideration. Such openness and flexibility will help capture the nuances of the research problems and thus design and implement appropriate solutions for complex health problems. From this project, we learned that several ingredients are crucial for the successful implementation of complex health promotion interventions: actively involving stakeholders in the formative research (intervention development), selecting appropriate enabling and empowering intervention strategies based on their needs, intervening at multiple levels of influences and having a complete theory-driven evaluation strategy from the inception to the end.