Data source
In this paper, we used the data from a cross-sectional survey that was conducted as part of an evaluation of the Maternal, Infant and Young Child Nutrition (MIYCN) Program of BRAC (formerly known as Bangladesh Rural Advancement Committee), an international non-governmental organization based in Bangladesh. Data was collected during April-May 2016.
Study area
The study area comprised nine districts of Bangladesh where BRAC was implementing its MIYCN program (Fig. 1). BRAC has selected these districts considering the availability BRAC programme-delivery infrastructures including the availability of its community health workers who were trained to implement the MIYCN interventions at the community level.
Study population
The study population included children aged 6–59 months and their caregivers. The caregiver was defined as the child’s biological mother or the person who takes care of or looks after and gives the child most meals on most days in the past seven days before the survey. Inclusion criteria allowed selection of the households of caregivers who had at least one child of 6–59 months and where the caregivers had resided in that household for at least one year. We excluded households if the caregiver was unable to attend an interview during the day of the survey due to illness or was unable to give consent to participate in the survey. If the household has more than one eligible child, we randomly selected one child for the survey.
Sample-size and sampling
We calculated sample-size for a district-level estimation; we considered 50% prevalence, the precision of ± 10%, Zα value of 1.96, and a design effect of 2. The use of the standard sample-size calculation formula yielded the minimum sample-size of 192 households per district. Thus, the total sample-size was 1,728.
We followed the two-stage cluster-sampling procedure. In the first stage, systematic random- sampling procedure was applied to select 16 Primary Sampling Units (PSUs) from the complete list of BRAC communities in a district. This procedure helped to ensure the equal chance of being included in the sample, and the resulting sample was close to an even spatial sample of BRAC’s target areas. In the second stage, the survey team ensured the population-size, total approximate households, and the boundaries of the PSU on arrival at the selected PSU and in consultation with and assistance of the local people (Union Parishad Chairman, Member, Counselor, school teacher, elderly person, and the relevant personnel of the locality). A physical map-segment sample approach was exercised to segment the selected community or PSU. The detailed sampling procedures have been reported elsewhere [20].
Outcome variable
Child morbidity status was the outcome variable for this study. We considered children to be experiencing morbidity if their caregivers reported that their child had been sick either due to ailments, such as diarrhea (diarrhea with 3 or more loose or watery, bloody, pussy or mucous stools in a 24 hour period), fever (illness with fever), illness with cough and had difficulty in breathing or fast breathing, difficult or fast breathing with blocked or runny nose in the last 14 days before the survey. In addition to the caregivers recall, we also collected child’s morbidity related information from the doctor’s prescriptions/medicines if the child received any treatment from a doctor during the last episode of illness. If caregivers reported any of the illnesses being present in their child, or if medical records indicated a sickness episode, then we considered them as experiencing morbidity.
Exposure variable and covariates
Household FI status (categorized as food-insecure, food-secure) was considered the main exposure variable in this study. We assessed households FI status based on 9 questions from the Household Food Insecurity Access Scale (HFIAS) developed by the Food and Nutrition Technical Assistance (FANTA) group in collaboration with Tufts University and Cornell University [21]. The response to each question ranges from 0 to 30. We made scoring of these responses as 0 = 0, 1–2 = 1, 3–10 = 2, and 11–30 = 3. The total score ranged from 0 to 27 for 9 questions. We then categorized as score 0–1 = food-secure household and 2–27 = food-insecure household. Other exposure variables included toilet availability, categorized as: Improved toilet (flush or pour flush to a piped sewer system, septic tank, pit latrine, Kumasi Ventilated Improved Pit latrine, pit latrine with slab), unimproved toilet (pit latrine without a slab, hanging latrine or defecate in bush or field). We also combined food security and toilet facilities of households to see the combined effects in regression analysis and we categorized them as: Food-secure and improved toilet, Food-secure and unimproved toilet, Food-insecure and improved toilet, Food-insecure and unimproved toilet.
Other covariates included household-size (categorized as: <5, ≥ 5), number of 6–59 months old children in the household (categorized as: one, two, or more), child’s age (6–23 months, 24–59 months), any children in the household aged 5–14 year who were attending school, caregiver’s age (< 25 years, ≥ 25 years), caregiver’s education (< 5 years, ≥ 5 years), father’s age (categorized as: <30 years, ≥ 30 years), caregiver’s religion (categorized as: Muslim, Hindu/Other religion), caregiver’s occupation (categorized as: other, housewife), wealth index (categorized as: poor, middle, rich), and monthly household income [categorized as: <11000 BDT (Bangladeshi taka), ≥ 11000 BDT (83 BDT = 1 USD)].
Data collection
We measured the level of anxiety and uncertainty of the participants about household food supply, insufficient quality of food, and insufficient food intake by following the HFIAS that comprises a brief survey instrument to assess whether households have experienced problems with accessing food during the last 30 days of survey. The questionnaire used a nine-item household hunger scale questionnaire [(i) worry about food, (ii) unable to eat preferred foods, (iii) eat just a few kinds of foods, (iv) eat foods they really do not want to eat, (v) eat a smaller meal, (vi) eat fewer meals in a day, (vii) no food of any kind in the household, (viii) go to sleep hungry, and (ix) go a whole day and night without eating)].
Before finalizing the questionnaire, a field test was conducted in a real-field setting in the non-survey areas, and the feedback from the field test was incorporated into the final version of the questionnaire. It was then submitted to the Institutional Review Board (IRB) of icddr,b for review and approval. A Standard Operating Procedure (SOP) was developed for the interviewers. This SOP was a guide for the interviewers on how to ask each of the questions to the participants. The electronic data-collection procedures used an Android-based Smartphone program of survey questionnaire. To support the Android operating system, Open Data Kit (ODK) software was used for developing the program. TABs/Smartphones were used and both Bangla and English versions questionnaire were used in the ODK software.
Data analysis
Weighted and cluster (PSU)-adjusted descriptive statistics were estimated and presented in percentages with respective 95% confidence intervals. Bivariate analysis using a chi-square test was performed to measure the association between the outcome variable (morbidity status of the children) and main exposure variables (household FI and toilet facilities). We performed multivariable logistic regression analysis to measure the association between outcome variables and other independent variables. At first, we performed unadjusted logistic regression to find the significant variables for the final multivariable regression model; p-value of < 0.05 was considered for the significance level. Finally, multivariable logistic regression was performed to assess the association of child morbidity with household FI and unimproved toilet facility after adjusting for potential confounders and presented in adjusted odds ratios with a 95% confidence interval. All analyses were performed using statistical software STATA (Version 13).