Study design and setting
A cross sectional comparative study was conducted in Dhanusha district of Nepal, between October 2016 to March 2017.The total area of the districtcovers1,180 square kilometers and Maithili is the commonly spoken language[28]. The south part of the district is attached with the Bihar border of India. The selected Village Development Committees (VDC) covers about 5-30 kilometers East, North and South from headquarter (Janakpur).
Study population, Sample size calculation and sampling procedure
Children aged 6-23 months and their mothers in Dalit and Non-Dalit were the study population who were living in Dhanusha district. The sample size was calculated by using two population proportion groups [29]. The prevalence of minimum dietary diversity of Dalit and Non-Dalit was taken as 20.1% and 31.32% respectively[2]. By taking power 85%, level of significance 5% and non-response rate 5%, the sample size was 599, where 299 from Dalit and 300 from Non-Dalit. Multi-stage simple random sampling was applied in the study. Initially, three out of seven electoral constituencies were selected and three VDCs were chosen from each electoral constituency by randomly. Finally four wards out of nine wards of each VDC selected randomly by using lottery method. Bahedabela, Fulgama, Nagaraen, Dhabauli, Panchaharba, Sonigama, Sinujoda, Sapahi and Shantipur were picked VDCs. List of children aged 6-23 months were obtained from Vitamin A and Immunization registers. Furthermore, the proportional allocation was considered to estimate the number of children to participate in each selected ward. In the case of more than one child, the young child was purposively selected. Moher who had 18 years and above were involved in the study. Two repeated visits were made when mother-child pairs were not available at the time of data collection.
Data collection techniques, tools and procedure
Pre-testing was completed among 10% of the respondents in Bateshwar VDC. The data collection tool was adopted from Nepal Demographic Health Survey[9] and it was modified in the local context, the questionnaire was translated to the native language (Maithili) to maintain uniformity in the responses. Socio-economic and demographic, health, child feeding and anthropometry and information of mother/family were included in the questionnaire. Community Medical Assistants (CMA) were recruited as enumerators for data collection. Training was given to enumerators which contains the questionnaire, ethical consideration, anthropometric measurements and data collection techniques. Face to face interviews was conducted with child mothers by using structured questionnaire and child height was measured by recommended standard measuring instruments.
Data quality and safety
A day training was given to enumerators. Pretested tool was used and each day cross-checked was made for completeness, clarity, consistency and accuracy by researcher. Researcher himself was involved in the monitoring and supervision of enumerators' through the study. Password protected laptop was used for data entry and analyzed by one door system.
Data processing and analysis
Data were edited, coded and entered into the EpiData 3.1 version. The data was exported in SPSS 16.0 version for analysis. For anthropometry analysis, World Health Organization (WHO) anthropometry nutrition survey tool was used. Nutrition-related data were analyzed by using the WHO Anthro plus software 3.2.2 version. The Z-score of height-for-age (HAZ) was calculated, HAZ less than -2 standard deviations (SD) was defined as stunted from the reference population.
In the descriptive analysis, frequency and percentage were used to summarize the study variables. A binary logistic model was used to find out associated factors. P-value of <0.2 in bivariate analysis were entered to the multivariable analysis to control the possible effect of confounders[30]. Hosmer and Lemeshow Chi-square to get whether the model was fit. Enter method was used to see the independent association between dependent and independent variables. The adjusted odds ratio (AOR) with 95% confidence interval was used to notify the strength of association, and at 95% CI was used to declare the statistical significance in the multivariate analysis.
Ethical consideration
The research proposal was approved by the Institutional Review Committee of Institute of Medicine, Tribhuvan University, Maharajgunj; each respondent was informed on the objective, method and benefit of the study and written informed consent was taken before taking the interview. Using identifiable number, the confidentiality was maintained.