The study was conducted at the Kersa Demographic Surveillance and Health Research Centre (KDS-HRC) of Haramaya University, east Ethiopia, from July to August 2011. In the Demographic Surveillance Site (DSS), 48192 adults and 7198 children aged under- five years lived in 10256 households during the study period.
The DSS which had three climatic zones, namely low land, midland and high land, was divided into two semi-urban and ten rural ‘kebeles’ (the smallest administrative units in Ethiopia). In Kersa district, there were no hospital and ambulance services, the nearest hospital being 50 km from the DSS. There were three health centers and ten community health posts within the geographic coverage of the DSS. Each community health post had two health extension or community health workers who provided basic primary health care services. The health care coverage of the district was 80% in 2010 [35, 36].
Agriculture is the main means of livelihood for the study population. Crop production is basically on an annual basis, except in few locations where it is biannual. Sorghum and maize are the common grains cultivated, while potato and other vegetables are scarcely produced. Subsistent crops are often planted during the wet season (June-September) and harvested during the dry season (October–January). Khat, a stimulant plant with amphetamine like effects, is predominantly produced as a cash crop. Polygamous marital relationship are common with no profound cultural taboos related to eating habits.
A community-based analytical cross-sectional study was conducted on mothers of children under-two years of age (mother–child pair). The mothers and their children were initially selected at random from the DSS sampling frame. In addition to the random selection of the study participants, samples were proportionally allocated to the 12 ‘kebels’ of the DSS based on the total number of their households. If more than one under-five child lived in the selected household, one child was selected by a lottery method. The sample size was estimated using two approaches based on the objectives of the study. To estimate the prevalence of non-EBF in infants aged under six months a single population proportion formula was used with the following assumptions: expected prevalence of non-exclusive breastfeeding 51% [21, 22], a 5% type I error, margin of error 5% (desired precision between sample and population parameter), and a 20% contingency for the non response. Accordingly, a sample size of 461 mother–child pair was calculated. On the other hand, to assess the predictors of non-exclusive breastfeeding, the sample size was estimated using Epi Info Version 3.5.1 with the following assumptions: a 5% type I error, power of the study 90%, control to case ratio 4:1 to detect the odds ratio of 2.0 among the cases which was estimated from the other study , being illiterate (main exposure) 66.1% among controls , and a 20% contingency for the non-response. Thus, the minimum sample size required for the study was 881 (220 mothers of non-exclusively breastfed children and 661 mothers of exclusively breastfed children). In this study, the latter sample size was considered to increase the power of the study. However, during the baseline survey for the longitudinal study, which was designed for a different study, the total number of mothers of children aged 6 to 23 months was found to be 860 (243 non-exclusively breastfed children and 617 exclusively breastfed children). To address other exposure variables and increase the power of the study, we included all the cases that were identified beyond the minimum requirement in the final analysis.
Data on socio-demographic and obstetric factors and infant feeding practices were collected using a structured and pre-tested interview questionnaire. The questionnaire was adapted from the Ethiopian Health and Demographic Survey (EDHS), WHO, and LINKAGE project questionnaires used to assess IYCF in developing countries. It was initially prepared in English and then translated into the local language, Afan Oromo, by fluent speakers of both languages, and it was translated back into English to check its consistency. A one week intensive training was given to 24 data collectors and 4 supervisors who were selected from KDS-HRC and the surrounding community. They were informed about the objective and relevance of the study, the confidentiality of the information, the respondent’s right, the questionnaire, the pre-test, the informed consent, and techniques to conduct the interview.
In this study, the outcome variable was the status of exclusive breastfeeding to infants under six months of age following the birth of the index child. EBF was understood as feeding only breast milk (including milk expressed or from a wet nurse) without anything else for the first six months of life with the exception of oral rehydration salt (ORS), drops, and syrups (vitamins, minerals, medicines) for therapeutic purposes . Non-EBF was coded as “1” while EBF was coded as “0” for regression analysis.
The independent variables were age, education, residence, the marital status of the mother, the household socio-economic position and food security status, maternal access to a health facility, knowledge about IYCF, the practice of colostrum feeding, and the child’s sex.
The age of the mothers which was categorized into “15-34”and “35-49”years, was coded “0” and “1”, respectively. Illiterate mothers were coded “0” while the literates coded “1”. The “semi–urban” and “rural” residences of mothers were also coded “1” and “0”, respectively. Regarding their marital status, “others” was coded “0” and married “1”.
The household socio-economic position or wealth index was computed using 28 household related variables. They included the household assets, income, environmental health, water supply and dwelling. Principal component analysis (PCA) was performed to determine the households’ socio-economic position or wealth index. The categorical variables were made dummy before initiating the analysis, but the ordinal ones were ordered from the least important to the most important. Finally, the socio-economic position or wealth status was used in the consecutive analysis by classifying it into five groups: lowest, second (low), middle, fourth (high), and highest [12, 38].
The household food security status was determined by computing nine standard household Food Insecurity Access Scale (HFIAS) questions adapted from the Food and Nutrition Technical Assistance (FANTA) Project of the United States Agency for International Development (USAID) which were designed for this purpose in 2007 . All “Yes” responses were coded “1” and “No” responses were coded “0”, and the responses were summed up to obtain the household food insecurity index. The index which had a high internal consistency (Cronbach’s alpha = 0.90) , was further dichotomized as “food secure” and “food insecure” for a score equal to zero and greater than zero , respectively, and was coded “1” for “food secure” and “0” for “food insecure” for analysis .
The maternal access to health facilities was based on the mother’s and key informants’ estimate of the distance from the nearby health institution. A distance of 10 km radius was considered as access to health service facility and was coded “1” while a distance beyond this range of values was considered as lack of access and coded “0”.
The knowledge of mothers on IYCF was computed based on seven questions which included awareness of the mothers about the timing of breastfeeding initiation after delivery, exclusive breastfeeding, colostrum feeding and its importance, age at complementary feeding, how long breastfeeding should continue, and whether HIV sero-positive mother should breastfeed or not. The mothers who scored above the mean cut-off point were considered to have high knowledge about the practices and coded “1”, whereas those who scored below this cut–off point were considered to have low knowledge and coded “0” [29, 42]. Likewise, last pregnancy was defined as a state of pregnancy for which a mother was pregnant for her index child. The mother’s colostrum feeding practice was categorized as “colostrum discarded” and coded “0”, and colostrum given to child was coded “1”.
The data were double entered using EpiData Version 3.1 by two data clerks and were exported to SPSS Version 16 for analysis. Descriptive statistics was used to summarize the study variables. Summary measures and proportions of the independent variables were computed against non-exclusive breastfeeding. The bivariate analysis was performed on important explanatory variables, and the crude odds ratio was calculated to identify the correlates of non-exclusive breastfeeding and to select candidate variables for the multivariable logistic regression analysis.
The Hosmer-Lemeshow goodness-of-fit and Omnibus tests of model coefficients tests with enter procedure were used to test for model fitness. The continuous variables, such as maternal and child age were tested for normal distribution using some statistical tests, including Kolmogorov-Smirnov and Shapiro-Wilk tests and through visual assessment, using the normal curve with a histogram. The explanatory variables were tested for multicollinearity before entering them into the multivariable model, using the Variance Inflation Factor (VIF) test, the Tolerance test, and values of the standard error.
Crude Odds Ratio (COR) with 95% confidence intervals was estimated to assess the association between each independent variable and the outcome variable, and a p-value was determined. Variables with p-value ≤ 0.2 in the bivariate analysis were considered in the multivariate analysis, along with variables that were well known predictors of non-exclusive breastfeeding, such as maternal age, and child’s sex, regardless of the cut-off point for p-value.
Adjusted Odds Ratio (AOR) with 95% confidence intervals was estimated to assess the strength of the association, and a p-value < 0.05 was used to declare the statistical significance in the multivariable analysis. Variables with p-value < 0.05 in the multivariable logistic regression analysis were considered as significant and independent predictors of non-exclusive breastfeeding.
The study was approved by the Ethical Review Committee of Haramaya University, College of Health and Medical Sciences, Ethiopia. The approval letter was dated 12 July, 2010, and numbered Ref IRERC/100/1/2010. The mothers of the children enrolled in the study were informed about the nature of the study, its objectives, expected outcomes, and benefits and the risks associated with it. Informed verbal and written consent was obtained from the mothers before the interview. Illiterate mothers consented by their thumb print after verbal consent. Privacy and confidentiality were maintained throughout the study.