India has implemented several programs to tackle anemia, the first of which is the National Nutritional Anemia Control Program, implemented in 1970, to promote regular consumption of iron rich foods and to provide iron and folic acid (IFA) supplements to pregnant and breastfeeding women ²¹ . Since that time, although many nutritional programs exist in India, only few focus on anemia control.

The emergence of two programs followed in 2000, one is the Weekly Iron and Folic Acid Supplementation (WIFS) program, which provides weekly IFA supplements to adolescents through The Adolescent Girls Anemia Control Program ^{22, 23} . Following this, the 12 by 12 initiative was launched in 2007 with the goal of making sure every child in India achieves a healthy hemoglobin level within 12 years ²³ .

The Ministry of Health and Family Welfare also developed the National Iron+ Initiative Program (NIPI) to address supplementation interventions for pregnant and lactating women and supplementary nutrition for pregnant, lactating women and adolescent girls to increase protein intake ²⁴ . In addition to the existing programs, several researchers have designed and conducted randomized controlled trials in India to reduce anemia in WRA ^{25– 28} .

Despite universal distribution and receipt, the intake of the IFA supplement remains low. Data from the 2015–2016 cycle of the India Demographic Health Data show that 91 percent of pregnant mothers reported that they received IFA supplements during their last pregnancy but only 37 percent consumed them for more than 100 days during pregnancy. Clearly, there are significant demand side barriers that contribute to low uptake and compliance.

Several studies, including recent investigations in Kenya and Nepal have indicated that knowledge barriers exist, with women reporting that they did not receive adequate information from their providers when they distributed their IFA tablets ^{29, 30} . A summary of four qualitative studies in India showed that in each study, many women did not know the importance of consuming iron supplements or the link between anemia and iron supplements ³¹ . Another recent study in India, reported that pregnant women were unaware of the association between anemia and maternal outcomes ³² .

Perceived and real side effects have also been implicated by numerous studies as being critical barriers. Specifically, adolescents and WRA reported that side effects include gastric problems, stomach pain, weight gain and nausea ^{33, 34} . Additionally, pregnant women in India report beliefs that taking too much iron may cause too much blood or a large baby, making labor more challenging ^{32, 35} .

The WHO has expressed concern that the current dose of 100mg IFA supplement for women in India is higher than their recommendations and may therefore itself be a barrier to use ¹⁰ . And despite the demonstrated acceptability of weekly or twice weekly supplementation in smaller trials and state-level pilot programs in India, the 2012 national scale-up of weekly IFA supplementation with 100mg iron targeting school-age youth has met resistance due to concerns regarding the gastrointestinal side effects ³⁶ .

Additional barriers in reducing anemia include food insecurity, which may result in a poor-quality diet. A 2002 qualitative study conducted in developing countries reported that dietary restrictions could vary across gender as a woman reported, “eating last” or eating “whatever is leftover” as reasons for inadequate diet ³⁷ .

WRA do not make decisions about their health in isolation; social networks can influence behaviors and family members play a crucial role in supporting the use of IFA tablets in women ³⁸ . Pregnant women whose spouses actively participated in antenatal visits were more likely to demonstrate a significantly higher adherence to IFA supplements than women whose husbands were not active throughout antenatal visits ³⁹ . Mothers and mothers-in-law can also influence the intake and adherence of IFA in pregnant women, even more than their partners as demonstrated by a formative research study among pregnant women in Bangladesh. However, when it came to money matters, husbands were involved and considered the most important decision-makers ⁴⁰ . Financial decision-making and control within a family may also have implications on a woman’s ability to obtain IFA tablets.

A growing body of literature points to social norms as a critical barrier to IFA uptake ^{41, 42} . Social norms are mores or rules of behavior that are considered acceptable in a group or society ⁴³ . Evidence from a meta-synthesis of qualitative research on the social determinants of iron supplementation among WRA conducted in 17 countries identified social norms as one of the primary factors limiting uptake of iron supplements, particularly during pregnancy ¹⁰ . However, to our knowledge, no studies have used a social norms approach to reduce anemia ³⁶ .

Social norms are different from laws in that they are negotiated through social interactions, whereas laws are codified. Traditions, like social norms are also socially negotiated, but they are more stable. The primary difference is that norms are more dynamic, and shaped and understood through communication processes ⁴⁴ and can thereby be shaped to promote positive healthy behaviors ⁴⁵ .

Lastly, the policy context can also act as an enabler or as an obstacle for the spread of evidence-informed IFA policy innovations. The adoption of new evidence and its consideration by domestic policy actors and stakeholders remains a complex challenge for the translation of empirical evidence into policy implementation. The evidence-to-action gap is “stickier” in cases of multi- and trans-disciplinary research programs and in research programs, such as this, that deal with complex, inter-sectorial issues ^{46, 47} . These challenges have been studied from various disciplinary perspectives including diffusion of innovation theory ^{48, 49} comparative policy and political science studies ^{50, 51} , knowledge utilization theory and evaluation influence studies ^{52– 54} , and the science of team science ⁵² , among others. What has received less attention, and what comprises a significant focus of this project, are the socio-normative factors that guide policymaking. Two major knowledge gaps are of particular relevance to this study. The first is the need to better characterize the channels and mechanisms through which potential changes in population-level social norms around anemia and IFA use percolate upwards to redesign existing national and state-level policies. The second is the challenge of facilitating the adoption and utilization of the research evidence by policy actors and stakeholders in Odisha leading to system learning and improved performance in IFA policy implementation.

Aims and objectives . By the end of the formative assessment period, we will have designed, adaptively tested, and refined an intervention that can be feasibly and effectively implemented to reduce iron-deficiency anemia among women of reproductive age and pregnant women in Odisha, India. To achieve this goal, the formative assessment has the following aims:

The formative assessment will be conducted from March to December, 2018. After analyzing results, we will hold a convening workshop in July, where we will formulate a first draft of the intervention plan. We will pilot test the various components of the intervention to determine its feasibility and effectiveness. Findings from these pilot evaluations will inform and refine the intervention, which we will finalize by December, 2018.

Figure 1 depicts the formative research workflow.

Odisha is an eastern Indian state on the Bay of Bengal ( Figure 2). According to the 2015–2016 India Demographic Health Survey, 83% of households in Odisha reside in rural areas and 67% of women and 84% of men are literate. The vast majority of household heads are Hindu (95%) and 23% of households belong to a specific tribal culture. The total fertility rate (TFR) in Odisha is 2.1 children per woman. In the last five years, 60% of pregnant women had four or more antenatal care visits and 82% of women delivered in a health facility. About half (51%) of women in Odisha have anemia with a higher rate for women from a tribal culture, those following the Christian faith, and those with no schooling ⁵⁸ .

We will conduct the research as a close partnership between the George Washington University (GW), IPE Global, our implementing partners based in Delhi and DCOR Consulting, a research firm based in Odisha, India. The research team consists of university-based researchers with expertise in qualitative research, social norms and global behavior change interventions. IPE Global and DCOR Consulting also has expertise in nutrition and SHG focused interventions in Odisha. These partnerships will be instrumental to gain access to the community and to understand the local context.

We will collect qualitative data via in-depth interviews, focus groups, Participatory Ethnographic Evaluation and Research (PEER) interviews, and structured observations.

In-depth interviews and focus groups. We will conduct in-depth interviews with key informants including Self-help group (SHG) leaders, health providers, teachers and health officials. We will conduct focus groups with women of reproductive age (including pregnant women), spouses, mothers, and mothers-in-law. Prior to focus group interviews participants will complete a demographic questionnaire including questions about caste. We plan to conduct approximately sixteen focus groups and twenty-four key informant interviews but the final sample size will be determined based on theoretical saturation ⁶⁴ . Table 1 shows the expected number of interviews and focus groups.

Observational Data. Observation can allow the researcher to draw inferences about a phenomenon that they cannot obtain from direct conversation via interviews or focus groups. We will collect observational data from four venues: antenatal care (ANC) clinics, medicine stores/pharmacies, self-help groups, and outdoor food markets. Data collectors will take notes on a standard observation form (See Supplementary File 3) at one of each of the following four venues in each of the four villages (16 observations total):

Medicine store/pharmacy observations will note how well IFA tablets are stocked, where they are stocked within the store, the extent to which they are purchased, their price, and their packaging.

ANC clinic observations will note whether there are existing communications (e.g. health education posters) about anemia and/or IFA supplementation. They will also record notes on the environment including (cleanliness, crowding, wait-time, etc.). Additionally, they will observe the pharmacy within the ANC clinic and note how many IFA tablets are available, cost, packaging and stocking/stock outs. We will not inform clinic staff about the observation prior to arrival to ensure that we observe the facility in its authentic state.

Data collectors will observe SHGs and take notes about the nature of their interactions with women in the community. They will take note of how well they function, what they discuss; the kind of activities that they undertake and any challenges that they face.

Data collectors will observe outdoor food markets, as diet is an important factor to consider while studying iron-deficiency anemia. We will take photos of all food items in the market; list them out and inquire about the cost of each item.

Perceptual Mapping. We will conduct perceptual mapping exercises to obtain a visual model that depicts how women and their influencers (their spouses/partners, their mothers-in-law, and others in their social network) think about IFA supplements, anemia, and other related factors, including physical and mental fatigue, and diet (see Supplementary File 4). Perceptual mapping is a technique used to elicit a mental picture held in common by members of a group. Perceptual mapping helps us understand how people construe various objects, both in terms of what meaning they give to them and how they construe the objects in relation to one another ^{65, 66} . Perceptual mapping is done in three steps – attribute elicitation, scoring, and mapping.

Step 1: Attribute Elicitation. The purpose of this step is to extract the meaningful attributes with which people construe specific objects. For example, if people are asked to list a significant attribute pertaining to “IFA tablets,” some may point to “awful taste” as the primary attribute. For others, IFA tables may represent the idea of “medication for a better tomorrow” or “healthy” as the primary attribute. Knowing these key attributes will help us craft meaningful messages during the intervention phase. For example, we may be tempted to develop messages about IFA uptake on the assumption that it represents “no anemia” in the minds of participants. It may well be, however, that the primary attribute pertaining to taking IFA supplements for certain women may center around issues of forgetfulness or something whose impact is not easily visible. If this is the case, then our messages about IFA supplementation would be far less effective than those that tackle the issue from the perspective of linking it with more immediate positive outcomes.

For attribute elicitation, we will provide two images (on two cards), one being a reference object, of particular objects and ask people to tell us how similar or different the two are from each other (on any dimension that the participant uses). Subsequently, the reference object will be kept the same, but another second object will be shown and the same comparison will be solicited. This process will be done for all permutations of all object pairs. We anticipate having 12 objects, for a total of 55 comparisons. The objects we have currently include: IFA pill; clinic; medicine store/pharmacy; SHG; traditional healer; physicians; fatigue; prenatal care; green vegetables; meat; money; and nausea/diarrhea.

Step 2: Scoring. A handful of attributes are thus identified for each primary behavior of interest. Respondents are then asked to rate the importance of each attribute for each behavior or object.

Step 3: Mapping. Step 2 results in a p (number of attributes) x q (number of behaviors or objects) matrix that is then used to model the relative distances between behaviors or objects on a p-dimensional plane, thus showing the relative distances among behaviors or objects. Two behaviors or objects close to each other on a particular attribute signify their conceptual proximity. This mapping provides an understanding about how the primary behaviors or objects of interest are understood by the audience; it also specifies the relative distances among objects.

Three perceptual mapping exercises will be conducted: one among women of reproductive age (including pregnant women), one among men (who constitute the support group for the women), and one among mothers/mothers-in-law and sisters or sisters-in-law. Approximately 30 individuals per group will participate in the mapping exercise, for a total of 90 individuals.

Rapid PEER. We will also conduct Rapid Participatory Ethnographic Evaluation and Research interviews (Rapid PEER), which will enable us to gain local insights into the beliefs and behaviors of beneficiaries, in the full context of their lived experience. This unique ‘insider perspective’ will ensure intervention is truly designed with the user in mind. We will train ‘ordinary’ members of the target group, in this case pregnant women and women of reproductive age, their spouses, and their mothers/mothers-in-law, after which they will serve as Peer Interviewers ^{67, 68} .

To execute the Rapid PEER, we will train Peer Interviewers to carry out in-depth conversational interviews designed to obtain targeted information from others in their own social group. We will decide on five key questions that form the structure of the conversational interview process. The Peer Interviewers themselves will suggest how to form the questions from formal Odia language into a more conversational format. Non-literate Peer Interviewers will draw pictures to represent each of the five questions. We will ask the same questions at each site to at least two interviewees from the PEER interviewers social group. The informal wording and pictures to remember and ask the question may vary to suit the preference of the Peer Interviewers.

Data collection will be carried out over a one-day period, wherein the PEER interviewers will hold conversations with two same-sex friends who fit the inclusion criteria (pregnant women, women in the reproductive age group, their spouses or mothers/mothers-in-law, all living in the community, 15 years of age or older who speak Odia). Within 24 hours, researchers will debrief the PEER interviewers to obtain detailed in-depth information from each PEER interviewer on what they discussed during the interview. During the debriefing process, the research team will probe PEER interviewers for broader contextual information regarding the responses during the interview. A final workshop with PEER interviewers will explore their experiences of assisting with the formative research.

The Rapid PEER process will take 4 days at each site (two villages total):

We will hold an in-person training in Bhubaneswar, Odisha. Researchers from the George Washington University will conduct the training alongside research staff from DCOR Consulting. Following the training, we will pilot test all instruments and research methods over the course of two days in Gandilo Village, Khurda District located about 45 minutes outside of Bhubaneswar. We will translate and review select KII and FGDs to ensure that the participants understand the research questions and interviewers are using the skills discussed during training. After all instruments are revised, trained researchers from DCOR Consulting will collect all data in person.

We will match interviewer/moderator by gender whenever possible to build interviewer rapport ⁶⁹ . We will conduct all interviews in Odia, the local language in locations selected by our local research partners in collaboration with local officials. Each interview and focus group will be audio recorded. DCOR consulting will transcribe the interviews and focus groups from Odia to English.

We designed the data collection instruments ( Supplementary File 1 – Supplementary File 4) based on the TNSB, a review of the literature on barriers and facilitators to IFA use, and feedback from the expert panel on our team. To explore IFA norms in a less personal and threatening way and to reduce social desirability bias within the focus groups, we will use vignettes, short stories about hypothetical characters who live in a rural village in Odisha. Vignettes are a simple way to explore social norms relating to a behavior ⁷⁰ . During the pilot study, we will test and revise all of the instruments as needed to ensure that the questions flow well, are culturally relevant, and capture the relevant constructs. The pilot study will also help to identify regional lexicon, which will inform revisions of the formative data collection instruments.

Within the state of Odisha, we plan to work in Angul district. We chose Angul for several reasons: 1) it is three hours from Bhubaneswar, the capital of Odisha, representing a rural but not isolated area 2) There are many existing SHGs 3) IFA supplements are available (as this is primarily a behavioral/demand side study); and 4) malaria and anemia rates are average compared to the rest of India. Within the district of Angul, we will conduct the formative research in the Athmalik block including four villages, two tribal and two non-tribal, Barham, Bardabar, Dudum and Tapadhol. We will also conduct the randomized control trial and intervention in Angul including but not limited to the villages from the formative research.

Inclusion criteria are that participants must reside in the selected village because many of the questions pertain to norms in their communities. They must also speak either Odia or English, and they must be at least 15 years old.

The FGD sample will include WRA (both pregnant and not pregnant), husbands, and mothers-in-law. We plan to use a random sampling procedure to select participants for FGDs. Local Anganwadi health workers have a household enumeration of everyone who lives in each village. Based on the number of individuals needed for FGDs, against a sampling frame that will consist of the entire village, we will use a proportional skip pattern that begins with a randomly selected initial participant in order to identify households from which to select every succeeding participant for each category (e.g., mothers in law, husbands and WRA).

For the perceptual mapping exercise, we will use the same random sampling method (at the village level) and participants are only eligible to participate in one research method.

Among the key informant interviews, SHG leaders, Odisha Health Officials, teachers, and health providers, we plan to specifically employ critical case or “reputational” sampling where cases will be chosen based on specialized knowledge or expertise. These interviews will be conducted with individuals who have expertise with respect to a specific perspective on this issue ^{71, 72} . Therefore, we will conduct key informant interviews with SHG leaders, ANC clinic providers who serve pregnant and reproductive age women, health officials who work on anemia or IFA distribution/education in Odisha and front line IFA dispensary workers. This sampling strategy will ensure that we interview the most knowledgeable key informants. Our sampling strategy will be refined in consultation with our local partners and Anganwadi health workers, who will recommend key informants who meet these criteria.

We will conduct Rapid PEER interviews in two villages within the selected district. In each site we will purposively choose six PEER interviewers (one each from women of reproductive age (15 – 25 years old), women of reproductive age (26 – 49 years old), pregnant women (15 – 49 years old), younger married men (15 – 25 years old), older married men (26–49 years old) and mothers-in-law). Each interviewer will interview or hold conversations with two interviewees (with same-sex friends who fit the previously discussed inclusion criteria). Within the two villages, we will train 12 PEER interviewers who will conduct 24-interviews.