Determinants of early initiation of breastfeeding in Ghana: a population-based cross-sectional study using the 2014 Demographic and Health Survey data

We used data from the child recode file of the 2014 version of the Ghana Demographic and Health Survey (GDHS). The GDHS is a nationwide survey which covers all the erstwhile ten regions and is conducted every 5 years [12]. The survey is conducted by the Ghana Statistical Service and the Ghana Health Service with Inner City Fund(ICF) International giving technical support through the MEASURE DHS Program. The survey adopts a two-stage sampling design. The first stage involves the selection of clusters consisting of enumeration areas delineated for the Population and Housing Census preceding the survey [12]. The second stage involved the selection of households from each cluster. A total of 9396 women (97.3% response rate) were interviewed for the survey. The survey provides a complete birth history of women and their children. Further details of the methodology can be found in the final report [12]. Actual sample of mothers with children having complete cases on the variables used in this study were 4219 children. Permission to use the dataset was obtained from MEASURE DHS. Dataset is available to the public at www.​measuredhs.​org. We relied on the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) statement in writing the manuscript.

The outcome variable was derived from the question “How long after birth did you first put (NAME) to the breast?” Responses were recorded in number of hours or days [14]. Our outcome variable “early initiation breastfeeding” was defined as initiation of breastfeeding within 1 hour of birth and was expressed as a dichotomous variable with category 1 for initiation of breastfeeding within 1 hour (early) and category 0 for initiation of breastfeeding after 1 hour (late). The independent variables were chosen based on previous studies and availability in the dataset [1, 13]. Some of the variables were recoded while others were adopted as reported in the 2014 GDHS. These included sex of child, size at birth, birth order, twin status, type of delivery assistance, place child was delivered (child factors). Others were mother’s age, marital status, religion, ethnicity, educational level, working status, frequency of listening to radio, frequency of listening to radio, frequency of watching television, number of ANC visits (maternal factors) and region , place of residence and wealth index (community/household factors) (see Table 1).

Analysis was done using Stata version 14. The first step of the analysis involved the distribution of early initiation of breastfeeding across the child, maternal and community/household level characteristics of women using frequency and percentages (see Table 1). This was followed by a bivariate analysis using logistic regression to assess the association between early initiation of breastfeeding and the child, maternal and community/household factors. This was presented in three models, with the child factors in model I, maternal factors in Model II and community/household factors in Model III. The results were presented as unadjusted odds ratios (UORs) at p < 0.05 (see Table 2). Variables that were significant at the bivariate level (5% margin), were used to build three models in a multivariable logistic regression analysis. Model I was made up of the child factors and early initiation of breastfeeding only. In Model II, the maternal factors only were added to the variables in Model I. Finally, Model III was a complete model that had the child, maternal and community/household factors and early initiation of breastfeeding. The results were presented as adjusted odds ratios (AORs) at p < 0.05 (see Table 3). Sample weight (v005/1000,000) was applied to correct for over-and under-sampling while the SVY command was used to account for the complex survey design and generalizability of the findings.

The DHS Program reports that ethics approval was sought from the National Public Health Reference Laboratory of the GHS and Noguchi Memorial Institute for Medical Research all in Ghana. It was further reported that informed consent was provided by all study participants.

Table 1 presents results of the distribution of early initiation of breastfeeding across the child, maternal and community/household characteristics of women. We found that 55.7% of women initiated breastfeeding within the first hour of delivery. In terms of the child characteristics, most of the children who went through early initiation of breastfeeding were males (51.5%), larger than average size at birth (51.6%), 2–4 birth order (54.3%), were born single (97.3%), were delivered with the help of health professionals (77.8%), and at the health facility (77.0%). With the maternal factors, most of the women who practiced early initiation of breastfeeding were aged 25–34 (49.0%), married (62.7%), Christians (76.0%), Akans (48.8%), had secondary education (49.2%), never read newspaper (87.7%), listened to radio at least once a week (49.3%), watched television at least once a week (44.2), and had 4+ ANC visits. With community/household factors, women from the Ashanti region (15.4%), those in rural areas (54.4%) and those of poor wealth index had the highest prevalence of early initiation of breastfeeding.

Table 2 shows the bivariate results of early initiation of breastfeeding in Ghana. At p < 0.05, birth order (UOR = 0.74^*** CI = [0.63–0.86]), place of delivery (UOR = 0.82^** CI = [0.72–0.94]) and type of assistance during delivery (UOR = 0.79^*** CI = [0.69–0.91]) were the child factors that showed statistically significant relationship with early initiation of breastfeeding in Ghana. With maternal factors, religion (UOR = 0.66^* CI = [0.48–0.91]), ethnicity (UOR = 0.69^* CI = [0.51–0.93]), frequency of reading newspaper (UOR = 1.39^* CI = [1.02–1.90]) and frequency of watching television (UOR = 1.26^** CI = [1.08–1.49]) showed statistically significant associations with early initiation of breastfeeding at p < 0.05. Region was the only community/household variable that was significant at the bivariate level.

As shown in Table 3, children of first birth order [AOR = 0.71, CI = 0.61–0.84], those who were delivered by non-professionals [AOR = 0.51, CI = 0.30–0.88] and those whose mothers were Traditionalists [AOR = 0.65, CI = 0.46–0.92] and Mole-Dagbanis [AOR = 0.69, CI = 0.54–0.89] were less likely to go through early initiation of breastfeeding compared to those of 2–4 birth order, those who were delivered by health professionals, those whose mothers were Christians and Akans, respectively. Conversely, children born to mothers who read newspaper/magazine at least once a week were more likely to go through early initiation of breastfeeding compared to those who never read newspaper/magazine [AOR = 1.40, CI = 1.01–1.95]. Children born to mothers who watched television less than once a week were more likely to go through early initiation of breastfeeding compared to those who watched television at least once a week [AOR = 1.40, CI = 1.01–1.95]. Finally, women from the Northern [AOR = 2.40, CI = 1.77–3.26] and Upper East regions [AOR = 2.57, CI = 1.86–3.56] practiced early initiation of breastfeeding compared to those from the Ashanti region (see Table 3).

This study is a cross sectional survey thereby limiting causal inference. We admit the possibility of recall bias especially among multiparous women. Yet, its national representativeness and large sample enhance generalisability prospects to other low-and middle-income countries especially in Africa. In addition, we could not include other variables such as maternal nutrition status, cultural taboos, and health system related factors that were not captured in the DHS.