The practice of exclusive breastfeeding for research purposes is either defined based on detailed history of feeding in the past 24-hours preceding a survey of infants (0–6 months old) or based on a recall period of feeding history of an infant (0–6 months old) since birth. The study data was extracted from the 2008 Ghana Demographic and Health Survey (GDHS); specifically, the individual (women) data set. To validate the data, the rate of exclusive breastfeeding was recalculated using the ‘24-hours’ definition, as used in the 2008 GDHS. The ‘24-hours’ period of exclusive breastfeeding is defined by WHO as a mother or caregiver giving nothing else but breast milk in the last 24 hours preceding an interview or a survey [8].
The GDHS are national-level population and health surveys conducted in Ghana as part of the global Demographic and Health Survey programme. They are conducted every five years. The first survey was conducted in 1988. All the surveys are implemented by the Ghana Statistical Service (GSS) in collaboration with the Ghana Health Service to obtain detailed information including but not limited to maternal and child health [4].
The 2008 GDHS used a two-stage sample design to allow for separate estimates of key indicators for each of the 10 regions in Ghana, as well as for urban and rural areas separately. The first stage involved selecting sample clusters from the 2000 Population and Housing Census. The master sampling frame consisted of 412 clusters which were selected using systematic sampling with probability proportional to size. The second stage of the design involved the systematic sampling of 30 households listed in each cluster. Women age 15–49 and men age 15–59 were eligible to be interviewed in the selected households. A total of 4,916 (out of 5,096) women and 4,568 (out of 4,769) men were interviewed yielding response rates of 97 percent and 96 percent respectively.
In the study, the dependent variable was exclusive breastfeeding. The following socio-demographic variables were used: mother’s age, marital status, education, region, mode of delivery, antenatal visits, place of delivery, infant’s size at birth, and infant’s sex. Some of the variables were recoded while others were adopted as reported in the 2008 GDHS. Ungrouped ages of mothers from 15 to 49 years old were coded as: 15–19, 20–35, or 36–49. Marital status of mothers was recoded as married, or not married. Level of education was also grouped into three categories (no education, primary, secondary/higher). In terms of region, all the ten regions in the country were considered. Antenatal visits by a mother were reorganized into three groups (0, 1–3, 4 or more visits). The generic modes of delivery (normal or caesarean) were considered. In relation to place of delivery, options were regrouped into three categories namely home, government health facility, and private health facility. An infant’s size at birth was considered as large, average, or small, and an infant’s sex was grouped as reported in the 2008 GDHS - male, or female.
Univariate analysis was used to present variables using descriptive analytical methods (frequencies and percentages). Bivariate analysis was also applied on the data. This involved cross-tabulations and results were presented in the form of proportions. Unadjusted odds ratio (UOR) of variables and their confidence intervals (CI) ware used to determine the presence of association between predictor variables and exclusive breastfeeding. Further, binary logistic regression was applied to examine the likely predictors of exclusive breastfeeding in the country. This method was deemed appropriate since binary logistic regression is a type of predictive model that can be used when the target or dependent variable is a categorical variable with two categories. With the present study, the dependent variable was categorized as either a mother was exclusively breastfeeding (code as 1), or not exclusively breastfeeding (coded as 0). All co-variates were included in generating regression model and the results were presented using adjusted odds ratio, and confidence intervals. In this work, 95% CI was used to test for statistical significance if the lower CI does not overlap the null value (Odds Ratio = 1). Stata version 11 was used to analyze the data.
Approval/Authorization to use the data was given by Measure DHS (Demographic and Health Survey), downloaded from DHS Data Archive. ICF Macro, 11785 Beltsville Drive, Calverton, MD 20705. http://www.measuredhs.com. On June 29, 2011.
