Upscaling Participatory Action and Videos for Agriculture and Nutrition (UPAVAN) trial comparing three variants of a nutrition-sensitive agricultural extension intervention to improve maternal and child nutritional outcomes in rural Odisha, India: study protocol for a cluster randomised controlled trial

Estimates suggest that implementing ten nutrition-specific interventions, addressing the immediate determinants of undernutrition, at 90% coverage could reduce deaths by 15% and stunting prevalence by 20% among children under 5 years old [3]. Accelerating reductions in undernutrition will entail coupling nutrition-specific interventions with nutrition-sensitive programs that address the underlying causes of undernutrition in other sectors, such as agriculture, education, and social welfare [7].

Agriculture can affect maternal and child nutrition through several interrelated pathways [8]. For example, agriculture is a source of income and food, and it has effects on food prices [9]. Women’s participation and decision-making in agriculture also affect their, and their children’s, nutritional outcomes [7, 9]. Between 2001 and 2013, more than a dozen literature reviews have sought to understand whether agricultural programs improve nutrition outcomes [7, 10‐15]. These reviews highlighted several critical methodological and evidence gaps, calling for more rigorous, theory-driven evaluations of nutrition-sensitive agriculture (NSA) intervention impacts on nutrition outcomes. The findings of the most recent systematic review [16] testify to a swift response to this call in recent years. Their review finds that NSA interventions show positive impacts on maternal dietary diversity, micronutrient intakes, child wasting and maternal underweight, and reduction in anaemia among mothers and children, but not child stunting. However, there remains a paucity of studies assessing the impact on maternal nutrition outcomes (only two studies in the review assessed maternal anthropometry [17, 18]) and the cost-effectiveness of NSA interventions.

The review of Ruel et al. [16] shows encouraging progress, with the addition of recent studies on nutritional impacts of diverse agriculture interventions such as biofortification, homestead food production, livestock, nutrition-sensitive value chains, and irrigation. This review and several recent global reports [19, 20] highlight the critical need to continue to develop, improve, and assess innovative NSA approaches in this Decade of Action on Nutrition.

Agricultural extension services aim to foster smallholder farmers’ access to technical knowledge and information, improve their skills through training and demonstration, and facilitate access to credit and markets [21‐23]. Nowadays, these services are also expected to play a broader role in improving rural livelihoods and food security and promoting farming techniques that are not harmful to the environment; these services could also provide a platform to improve nutritional outcomes [22]. Most agricultural extension services involve a worker going from door to door in villages and interacting with a select number of people, usually men with larger farms [22, 23]. Studies have shown that many farmers have limited access to extension services and, even when they do, farmers may not adopt extension workers’ advice if these agents do not possess location-specific, relevant knowledge [24]. Extension workers visit infrequently or erratically, and their information does not reach farmers with the lowest yields, many of whom are women and poor [24, 25]. Consequently, agricultural extension programs have been inefficient with mixed impacts on agricultural productivity [23, 26]. Recognising this, agricultural extension systems in India and worldwide are adopting pluralistic, decentralised, participatory, gender-sensitive, demand-driven approaches to improve coverage and relevance [27, 28].

A promising innovation is the use of digital platforms that can strengthen the delivery and uptake of extension services in agriculture, even in the poorest settings [29‐32]. One such approach features low-cost participatory videos with women’s groups and farmers’ groups. In a pilot trial involving 16 villages, this approach increased the adoption of promoted agricultural practices sevenfold compared with the traditional extension approach [33]. On a cost-per-adoption basis, the method was ten times more effective per dollar spent than a classical extension system [33]. A trial in the Indian state of Bihar found that this intervention resulted in a 21% increase in agricultural productivity compared with the control areas [34]. By integrating nutrition-sensitive video content to improve yields of nutrient-rich foods, increase incomes, reduce women’s workloads, and promote more participation of women in agricultural decision-making, this video approach could leverage these agricultural improvements to improve dietary quality and nutritional status. However, to our knowledge, no studies have examined whether this approach, if made nutrition-sensitive, can improve nutrition outcomes for mothers and children.

As with agriculture, experimentation with digital platforms for health- and nutrition-related behaviour change has gained momentum worldwide. Published literature on the impact of digital interventions to improve nutrition is only just emerging, and the overall quality of evidence remains low [35]. Participatory interventions integrating mobile phone messages have been shown to improve infant and young child feeding practices in two trials in Nigeria and China [36, 37]. Building on this success and on the success of low-cost videos for agriculture, videos with nutrition-specific messages could provide viewers with knowledge and opportunities to adopt new nutrition practices. However, we did not identify any studies that evaluated the impact of any other digital technologies, such as videos shown in women’s groups, on undernutrition outcomes.

Another important development relevant to efforts to improve agricultural productivity, health, and nutrition is the spread of participatory women’s groups. Participatory interventions with women’s groups using a Participatory Learning and Action (PLA) approach have shown large positive impacts on essential newborn care practices and have reduced neonatal deaths [38]. These interventions have also shown large effects on women’s nutritional knowledge and dietary diversity [39, 40], but limited impacts on maternal and child nutritional status [40]. In PLA groups, women acquire new nutrition knowledge, receive social support from peers and the community, and exercise both problem-solving and decision-making power [41]. Coupled with a nutrition-sensitive, video-based agricultural extension approach, all of these PLA processes could enable women and their families to further capitalise on the nutrition benefits of an NSA intervention.

The challenge of making agriculture work for nutrition is twofold: it involves (1) strengthening agriculture programs in a way that is feasible and can therefore be effectively implemented at scale, and (2) making these programs ‘nutrition-sensitive’. The UPAVAN trial (Upscaling Participatory Action and Videos for Agriculture and Nutrition) aims to address this dual challenge by testing the nutritional and agricultural effects of (1) a participatory, NSA extension platform with women’s groups viewing videos and follow-up visits and (2) two variants of this intervention, integrating videos on nutrition-specific behaviour change and integrating the PLA methodology. Rigorous assessment of these interventions will determine the effects and cost-effectiveness of NSA extension through participatory, low-cost, video-based approaches on maternal and child nutrition and whether integration of nutrition-specific goals and enhanced participatory approaches can increase these effects.

The objectives of the study are to:

The study is located in four administrative blocks (Patna, Keonjhar, Harichandanpur, and Ghatgaon blocks) in Keonjhar District, Odisha, eastern India, as shown in Fig. 1.

Keonjhar has a population of around 1.8 million which is predominantly rural (86% of the population) and agrarian [42]. Scheduled Castes and Scheduled Tribes, the historically disadvantaged groups in India, comprise 57% of Keonjhar’s population (12% and 45% respectively) [42]. About 30% of women in the district are underweight and 40% are anaemic [6]. The prevalences of child (< 5 years of age) stunting and wasting are 45% and 19% respectively, and less than 10% of children 6–23 months of age are fed the minimum acceptable diet [6]. Around three quarters of people in rural areas earn their livelihood through traditional agriculture, selling wood, and collecting non-timber forest products [43].

The interventions are implemented and evaluated by a consortium of seven partners. (1) Digital Green, a global development organisation working in digital participatory extension services, coordinates the implementation of interventions. (2) The Voluntary Association for Rural Reconstruction and Appropriate Technology (VARRAT), an Odisha-based non-governmental organisation, is responsible for implementing the interventions in Keonjhar. (3) John Snow Inc. (JSI) Research and Training Institute, through its Strengthening Partnerships, Results, and Innovations in Nutrition Globally project, provides technical assistance for the development of interventions, and (4) Ekjut, an Indian non-governmental organisation, provides technical assistance on PLA in the AGRI-NUT+PLA arm. (5) The London School of Hygiene & Tropical Medicine (LSHTM) leads all research activities, in collaboration with (6) University College London’s Institute for Global Health and (7) Development Corner Consulting Pvt. Ltd.

The study is a four-arm, cluster randomised controlled trial with 148 clusters randomly allocated to the four trial arms to give 37 clusters per arm (allocation ratio 1:1:1:1). The trial will run for a total of 53 months, from December 2015 to May 2020, including intervention development, setup, baseline and endline surveys, and intervention implementation for 32 months (from March 2017 to October 2019). The impact of the interventions will be assessed by repeat cross-sectional surveys of eligible households at baseline (November 2016 to January 2017) and endline (November 2019 to January 2020).

Randomisation was undertaken remotely by the Clinical Trials Unit at the LSHTM. The unit of randomisation is a cluster, defined as one or two villages and surrounding hamlets to give a minimum population size of 800 per cluster. Clusters were selected from four administrative blocks (Patna, Keonjhar, Harichandanpur, and Ghatgaon blocks) in Keonjhar District (Fig. 1). One hundred forty-eight clusters were randomly allocated to four trial arms, stratified by distance to nearest town (< 10 km or ≥ 10 km) and low (< 30%), medium (30–70%), or high (> 70%) proportion of Scheduled Tribe or Scheduled Caste households in the cluster, giving six strata in total. The randomisation schedule was drawn up once cluster-level consent was obtained and the clusters completed the baseline survey, thus guarding against selection biases at entry of clusters to the trial. Randomisation occurred sequentially in two batches (December 2016 and January 2017) due to the time it took to complete the baseline survey.

The interventions are delivered at the cluster level, and all women in the intervention clusters are eligible to participate (‘intervention participants’). The trial will evaluate impacts on outcomes in ‘trial participants’, who are children 0–23 months of age, their mothers or female primary caregivers where the mother is absent (aged 15 to 49 years), and their households. Trial outcomes are measured using baseline and endline repeat cross-sectional surveys on an average of 32 randomly selected eligible households per cluster. If more than one child aged 0–23 months is present in a randomly selected household, we will randomly select one of them for inclusion.

Exclusion criteria for trial participants are any disability impairing participation in the survey for mothers; any disability affecting weight, standing height, or recumbent length for children; and household members who are not permanent residents. Residents should have lived in the household at least half of the time during the past 12 months (e.g. 3–4 days of each week for 12 months, or 6 full months of the past 12), except for infants younger than 6 months, newly married women, and domestic help, lodgers, and agricultural labourers who are currently in the household and will be staying in the household for longer than this but arrived less than 12 months ago.

The impact of the interventions will be evaluated through repeat cross-sectional household surveys at the same time of year (November to January) at baseline (2016–2017) and endline (2019–2020), in the same intervention and control clusters.

Our process evaluation has three objectives: (1) to assess the fidelity and quality of the interventions’ implementation, (2) to clarify mechanisms behind the intervention’s effect, and (3) to identify contextual factors associated with variation in effects on the primary outcomes.

The economic evaluation will be conducted from a societal perspective, taking into account all costs incurred by the implementing agencies (program costs), community and health care providers, and users or households, and all associated and measurable outcomes or benefits associated with the intervention.

The program costs will be estimated using a combination of activity-based costing and an ingredients approach [64]. Program costs will be collected prospectively from the project accounts of all implementing partners. These costs will be entered into an MS Excel data capture tool designed for this purpose. Costs incurred by government frontline workers and community level institutions such as gram panchayat (local, elected body of representatives) and health facilities will be collected retrospectively using key informant interviews and publicly available salary information. Costs incurred by households affected by the UPAVAN interventions will be collected using household surveys.

All costs will be estimated from an economic perspective, assessing the full impact on all parties of direct and indirect costs, including time costs and donated goods. All costs will be adjusted for inflation using the Indian Consumer Price Index and converted to International Dollars to adjust for purchasing power parity.

The economic evaluation will take the form of a cost-consequence analysis, comparing incremental costs of the interventions and incremental changes in the trial outcomes [65]. Incremental costs and significant outcomes will be listed separately, allowing policymakers to compare the incremental costs with the incremental consequences of the different UPAVAN interventions. Cost-consequence analysis has been recommended for complex interventions, such as UPAVAN, that have multiple health and non-health effects which are difficult to measure in a common unit [65, 66]. Robustness of the analysis results will be assessed using sensitivity analyses. A full protocol will be developed a priori in an economic evaluation plan.

Our study is not powered to test the differences between interventions due to the large sample size required and the resources available for this trial. The intervention components are not additive, and therefore the study will not be able disentangle the effects of individual components within arms. The reason is that we kept the number of meetings (video disseminations or PLA) at two per month to minimise the burden on participants and to standardise the number of meetings per month across all intervention arms. Had we made the intervention strictly additive, the number of meetings would have also been additive, making it difficult to distinguish between the effect of the number of meetings and the meeting content.

By examining the role of innovative agricultural interventions for improving maternal and child nutrition, the trial will inform ways in which nutrition- and agriculture-related services can be strengthened in rural communities. Specifically, our study will provide robust evidence on (1) the extent to which improving agricultural extension through participatory, low-cost, video-based approaches can improve agricultural productivity and diversity; household income and investments in nutrition-related expenditures; women’s decision-making; and reducing women’s workload and thus maternal and child nutrition outcomes; (2) whether integration of explicit nutrition-specific messages enhances the impact of agriculture interventions on maternal and child nutrition; and (3) the optimal level of participatory action required to improve nutrition outcomes. The trial will provide insights on whether women’s groups can provide an effective platform for integrated agriculture and nutrition interventions. Our process and economic evaluations will provide critical information for translational feasibility and scale-up. The lessons from this trial can be readily adopted at scale in UPAVAN partners’ existing and future collaborations with non-governmental organisations and governments across South Asia and Africa.

Ethics approvals and informed consent were obtained in October 2016 and clusters were randomised. The baseline survey was conducted between November 2016 and January 2017, and the interventions started in March 2017. With the endline survey starting in November 2019, the duration of intervention exposure is 32 months. The Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) timeline showing key trial time points is shown in Fig. 6, and the SPIRIT checklist is available in Additional file 3.

Version number 2; 21 December 2017.