STUDY SITE AND POPULATION We conducted a qualitative study in two villages of Melandah Upazila (sub-district) of Jamalpur districts in northern Bangladesh. Melandaha Upazila was selected as more than 60% of the population are considered poor, with nearly half of the population (47%) extremely poor according to the 2010 Bangladesh Poverty Map.21 In this impoverished rural region less than one in ten household heads had completed primary education. The feasibility of recruiting an adequate number of pregnant and lactating women was carefully considered. We visited the local BRAC office, the largest NGO in Bangladesh, to seek assistance in selecting and accessing the communities. We visited the community clinics, as advised by BRAC staff, to obtain a list of the pregnant and lactating women in selected villages. We collected data in two phases –from January to February 2015, and from May to June 2015. PARTICIPANTS A total of 36 pregnant and lactating women were purposively selected for in-depth interviews (IDI) using sampling techniques adopted from ProPAN (Process for the Promotion of Child Feeding - a tool to improve infant and young child feeding) methodology. 22 The field research team consulted the community health workers to obtain a list of eligible pregnant and lactating women currently receiving antenatal care and postnatal care services. We selected women who were pregnant at the time of data collection, had delivered a live baby in the past two years, or had a child 0-23 months of age. A total of 36 IDIs were conducted in two phases (16 pregnant, 18 lactating women, and 2 pregnant and lactating women). ETHICAL APPROVAL We obtained ethical approval from the Human Research Ethics Committee, University of Sydney (Ref.: 2014/874). In Bangladesh, the study was approved by the Ethical Review Committee of the International Centre for Diarrhoeal Disease Research, Bangladesh (Ref.: PR-14101/2014). Informed written and oral consents were obtained from respondents at each stage of the study. DATA COLLECTION METHODS We recruited two experienced qualitative researchers, who were then trained in data collection tools by an investigator (WK) (a nutritionist and skilled qualitative investigator). Data was collected using in-depth interviews, focus group discussion, market surveys, and food attribute exercises using methods and tools adapted from ProPAN – an approach to capture infant feeding practices through the development of a key food list and food attribute exercises.22 KEY FOOD LIST DEVELOPMENT PROCESS We developed a key food list based on the results of in-depth interviews on dietary assessment, and from focus group discussions and market surveys. The key food list, which included 25-30 foods, was developed based on five criteria: 1) foods frequently consumed by the women; 2) sources of important nutrients, especially protein, iron, and calcium; 3) availability and affordability in the local markets 4) homestead produce or readily available free foraged foods and 5) seasonal availability. Out of 36 in-depth interviews, we conducted 20 interviews to explore the usual food intake and dietary patterns of pregnant and lactating women. For the structured face to-face interviews, we used an open-ended format to generate data about the foods consumed by the women in the previous day (24-h dietary recall) and in the last seven days (7-day recall). The respondents were also asked about their dietary history and the daily variations in diets until a full picture of food variety was obtained, and enabled us to understand their daily dietary patterns. Based on the results on the 24-hour and 7-day recall and discussion on dietary history, a list of foods was compiled from the various food groups usually consumed by the women (eg, cereals, lentils/legumes, roots and tubers, vegetables, fruits, fish, meat, egg, milk and milk products, oils). Foods mentioned more often in the 24-hour recalls and dietary history were considered as the most frequently consumed foods. Similarly, foods that appeared less often in the 24-hour dietary recalls but appeared in the 7-day dietary recalls and sometimes mentioned in the dietary history were considered as the less frequently consumed foods. Foods that were consumed a few times in a month/six months/year and hardly mentioned in the dietary history were considered as the rarely consumed foods. Two focus group discussions (FGD) were conducted with the husbands of the women. The FGDs engendered information about the availability of foods in the local market, food costs, seasonal availability, homestead production and access to free foraged foods. This information helped identify a list of locally available and affordable, seasonal foods. The foods listed were grouped according to their source of origin and nutritive values, for example, animal source foods (egg, chicken, beef, small fish, large fish, and dried fish and milk/dairy products), green leafy vegetables, other vegetables, fruits, nuts/seeds, and beans. Animal source foods were also selected as rich sources of protein and minerals such as iron, zinc, and calcium, while vegetables and fruits were regarded as the source of vitamins such as Vitamins A, C, B, and folate. We reviewed the “Food Composition Table for Bangladesh”23 to select foods rich in protein, iron and calcium and other vitamins (Vitamin C, and folate) (Table 1). Hence, based on the information, we developed an initial list of key foods to use for the market survey. For the market survey, we selected and visited two local markets where the household members of respondents usually purchased foods. In these markets, local farmers gathered every morning (and once a week in a larger scale) to sell their own produce including vegetables, fruits, fishes, eggs and chickens in small quantities, although there were some small permanent grocery shops. We interviewed the mobile sellers and grocers to collect information on available foods, the retail price per unit, and seasonal availability. Although the market survey was conducted in winter, the sellers and the grocers were asked about the foods available in the market in the other seasons as well. The survey was important to capture the seasonal variations of foods, especially for fruits and vegetables, as the seasonal foods were available at low-cost in the local market. The key food list was developed to use for the food attribute exercises in the next step of the study. FOOD ATTRIBUTE EXERCISES We asked women about their perceptions, attitudes and practices on the consumption of a variety of foods from the list of key foods. The interviewers named these foods and asked for information on whether they were consumed or not, the reasons for consuming/not consuming, the food preparation process and the necessary conditions for consumption if advised to consume the food. DATA ANALYSIS Interviews were audio-recorded, transcribed in Bangla, then translated into English. The quality of translations was discussed within the team. An inductive coding method was used, themes were generated from the data pertinent to the research questions.24 Initially, two transcripts were coded and discussed to finalize the codes. According to the final thematic codes, all transcripts were manually coded and analysed using inductive thematic approach.25 Through iterative discussions among the research team, the data were analyzed to find variations, similarities, and emerging trends in themes generated through the coding process.