PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • The current study was approved by the ‘Institutional Committee for Ethics and Review of Research’ of the Indian Institute of Health Management Research (www.iihmr.org), Jaipur, India. Informed consent (including signature or left thumb impression of the respondent) was obtained from each agreeing participant before the interview. CARE India, a non-governmental organization, in association with the State Government of Bihar and under financial patronage of the Bill and Melinda Gates Foundation, initiated a multifaceted project named Integrated Family Health Initiative (IFHI) in 2011 with the principal objective of reducing the mortality and malnutrition among the infants and women of reproductive age in Bihar, a resource-poor state. [27] As part of the pilot phase of IFHI, multiple population based cross-sectional surveys were conducted to ascertain various health and developmental indicators in the state. In total, five rounds of such survey, using Lot Quality Assurance Sampling (LQAS) technique (a small sample survey design based on binomial distribution), [28–31] were conducted in eight randomly selected districts (from total 38) of Bihar. The sampling ‘lots’ in this survey were the blocks or sub-districts, which were the unit of programmatic intervention. In brief, using a multi-stage sampling strategy, 19 Anganwadi Center (AWC)–village level institutions providing basic health care services—areas were selected through probability proportional to size (PPS) sampling from each of the 137 blocks belonging to the eight randomly selected districts. From each selected AWC area, four eligible households were identified through systematic random sampling. The systematic random sampling of household involved the following steps: The data collectors obtained an approximate number of households catered by an AWC from the AWC Household Register (or as per the information provided by an Anganwadi worker, where registers were not available)Using a random number table, a random starting point—between 1 and the total number of households—was decided. This was considered the number of index household.Counting from the location of AWC and following a ‘Right-hand rule’, the index household was reached. Starting from the index household, the data collectors visited every 5th household and inquired if there was an ‘eligible mother’ in that household.Eligible mothers were those that had a child belonging to any of the following age categories: 0–2, 3–5, 6–8, or 9–11 completed months.The data collectors continued moving in a circular manner, following the ‘Right-hand rule’, till four interviews were conducted with eligible mothers (one from each age category).Five iterations of above sampling strategy took place between October 2011 (LQAS Round I) and October 2014 (LQAS Round V). For each iteration of LQAS, altogether 2603 (137×19) AWCs were selected from the eight study districts and from the catchment areas of these AWCs, four households with children belonging to each of the four age sub-groups were identified. Consenting and eligible mothers responded to a detailed one-to-one interview conducted by trained study staff. In the current analysis, we used information on infants belonging to three age groups—0–2, 3–5 and 6–8 completed months. The data obtained during Round-II to Round-V of LQAS survey were used. Information collected during Round-I of LQAS were excluded because of methodological shortcomings and changes in questionnaire pattern between the first and the subsequent rounds of survey. Our objective was to ensure adequate sample size for evaluation of various health indicators in each block (district sub-divisions). According to LQAS methodology, the suggested minimum sample size for each decision-making unit is 19 (resulting in <10% α and β errors). [32] Thus, based on the LQAS decision-making rule, 19 respondents per block for every age-group was selected. During the interview, mothers of 0–2 and 3–5 completed months old infants were asked, inter alia, about the edible (or drinkable) items given to their child during the previous 24 hours (previous day’s morning to current day’s morning). Children of the mothers who reported giving only breastmilk were considered to be breastfed exclusively and vice versa. [33] Children who received other food or drinks including tea or other water-like substitutes (except for ORS or medicines) were regarded as non-exclusively breastfed. Thus, for less than 6 month old children, EBF status was determined by a surrogate measure based on 24-hour recall. Regarding feeding practice, if the mother was not available for some reason and some other lactating woman provided breastfeeding, it was still considered as breastfeeding. For the children aged 6–8 completed months, the mothers were inquired about the age at which weaning was done i.e. the age from which they started giving liquid/semi-solid food (other than breastmilk) to their child. Children who were reportedly weaned at or after sixth month were assumed to be exclusively breastfed. In order to assess the effect of seasonal variation on breastfeeding among children aged 0–2 and 3–5 completed months, we first sought out the month during which the interview was conducted. Based on the prevailing weather pattern in Bihar, we classified the interviews conducted during November to February as those conducted in ‘winter’ season, April to August as ‘summer’ and rest of the months as ‘autumn/spring’. For 6–8 month old children, we tried to determine the approximate calendar months during which an infant was supposed to be exclusively breastfed (as per WHO recommendation). In order to do this, we determined the months that comprised the six month period since their birth, counting from the month following the birth month of the child. For example, if a child was born in May, we conjectured that he/she was supposed to be exclusively breastfed from June to November. If three or more months of a child’s EBF period fell during the winter season (November to February), then his/her nursing season was categorized as ‘winter’, otherwise the nursing period was classified as ‘non-winter’. As discussed previously, FLWs provide counselling on various maternal and child health issues including EBF and complementary feeding. As part of LQAS survey, mothers of children aged 0–5 months were asked if they received advice from FLWs regarding recommended EBF period. The mothers who reported receiving such advice following birth of the concerned child, were considered exposed to EBF counselling and vice versa. The mothers of older infants (6–8 completed months) were not inquired about EBF advice but they were asked about counselling on complementary feeding. The mothers who reported receiving advice on starting semi-solid food along with breast milk after their child reached a certain age (6 months) were classified as exposed to complementary feeding advice. We included several demographic and socio-economic parameters in our analyses to control for potential confounding. These include gender of the child, caste, religion, economic status of the household (asset index), and parental education level. Caste-wise, the families were dichotomized into marginalized caste [scheduled castes (SC) / scheduled tribes (ST) / other backward castes (OBC)] and other/general caste. Religion categories were Hindu, Muslim, and other. According to the level of education, mothers and fathers were classified into three categories—no formal education, school education up to eighth standard, and school education above eighth standard. Economic status was assessed using an asset index (AI) based on possession of 25 different household items. For calculation of AI, a relative weight was assigned to each of these items (as per NFHS-3) [5] and an aggregated score was generated by adding the weighted score for each item possessed by a household. The cumulative asset scores were then log-transformed to create the AI. Based on the percentile distribution of AI, we then created AI tertiles and classified the families according to the AI tertile they belonged to—low, middle and high wealth. The calculation of asset index followed the methodology described by Kanungo et al. [34] Descriptive analyses were carried out to determine the distribution of socio-demographic and breast feeding related characteristics of the sampled population. For the purpose of descriptive and regression analyses, the datasets containing information on infants aged 0–2 months and 3–5 months were combined to create a concatenated dataset for 0–5 months old children; whereas the dataset on 6–8 month old children was treated as a separate entity. We estimated the proportion of 0–5 month old infants, who were breastfed exclusively during past 24 hours in month-wise age groups, and further stratified the estimated proportions by season and FLW-counselling. We employed the Cochran-Armitage trend test to assess whether age-specific and/or seasonal trends existed in proportion of children receiving EBF. Among the children aged 6 to 8 months, we calculated the proportion of children who reportedly received EBF for the recommended duration—stratified by ‘winter’ and ‘non-winter’ nursing season. Further, we assessed the proportion of 0–5 month old infants who were breastfed exclusively during past 24 hours stratified by FLW-counselling. We also tested if there were any statistical differences in breastfeeding proportions between the children whose mothers received such advice and those who did not. We implemented separate simple (crude) and multiple (adjusted) logistic regression models to identify determinants for the two dependent variables in our analyses—breastfeeding exclusively during past 24 hours (for 0–5 month olds) and EBF during the first 6 months of life (for 6–8 month olds). Further, only for the first dependent variable, in order to explore whether any age-specific differences existed in the association between ‘season of interview’ and breastfeeding exclusively during past 24 hours, six separate age-restricted models (one for each completed month in age—between 0th and 5th month) were employed. In the crude models for the first dependent variable, we separately tested associations with ‘season of interview’ and ‘FLW-provided EBF advice’, whereas ‘nursing season’ and ‘FLW-provided complementary feeding advice’ were the sole predictors in the unadjusted models for the second dependent variable. The adjusted models additionally included potential confounders—gender, religion, caste, economic status (AI tertile) and maternal education. All descriptive (Proc surveymeans and Proc surveyfreq) and associational (Proc surveylogistic) analyses were carried out using the survey data analysis procedures in SAS 9.4 and incorporated information about multi-stage sampling and relevant sampling weights.
rdf:type