Early feeding practices and associated factors in Sudan: a cross-sectional analysis from multiple Indicator cluster survey

The 2014 Sudan Multiple Indicator Cluster Survey (MICS) is a representative survey designed to provide national-level data as well as state-level estimates for maternal and child health in the 18 states in Sudan. Sudan Central Bureau of Statistics (CBS) in collaboration with the Ministry of Health, conducted this survey as part of the global MICS programme [26]. The website of the MICS programme makes the anonymised survey dataset freely available.

This study utilises cross-sectional secondary data from the 2014 Sudan MICS. This survey utilised an updated 2008 population census and randomly selected 40 enumeration areas in each State. The survey design employed two-stage cluster sampling after stratifying by urban and rural areas within each State. In the first stage, within each stratum, enumeration areas were selected with probability proportional to size. In the second stage, 25 households were selected in each enumeration area. Paper-based questionnaires were used for data collection using the Arabic language. Data were collected using three questionnaires, collecting information at the household level, on all eligible ever-married women aged 15–49 years and on all eligible children under-5 years of age. The survey sampled 18,000 households of which 16,801 were interviewed; 18,302 eligible women and 14,081 mothers/caretakers of children under-5 were interviewed. Detailed reports on the survey design, methods and findings of the survey can be found on the MICS website [27].

Data was collected from the three MIC questionnaires: breastfeeding information from the women questionnaire; child’s birth order from the children under-5 questionnaire and sociodemographic data from the household questionnaire, respectively. Analysis for this study was restricted to data from mothers with last-born children in the two years preceding the survey. The total sample size was 5622 mother-baby pairs.

Percentages and 95% confidence intervals (95% CI) of the three breastfeeding outcome indicators were reported; overall and by explanatory variables. Logistic regression analyses were used to model the odds of the outcome variables. Crude logistic regression analyses were performed as initial steps of qualifying factors to be included in the multivariable logistic regression analyses. Potential factors with p - values < 0.25 were considered to develop an initial reduced model [31]. Using the adjusted Wald test, variables that tested insignificant (with p - values > 0.05) were then eliminated from this model. Confounding was assessed; variables were included in the model if their coefficients changed by 20% or more. Identified a priori as a confounder, maternal age was included in all models. F-adjusted mean residual goodness-of-fit tests were used to assess any evidence to lack of fit of the final model [31]. Odds ratios (OR), adjusted odds ratios (AOR) and their 95% CI from logistic regression analyses were reported. Stata version 15.0 was used for all analyses. Svy commands using Taylor linearization for the variance estimation were used to account for the complex survey design using assigned weights, primary sampling units (clusters) and strata. The sample weights were the inverse of the probability that the observation was included because of the sampling design. Weighted sample sizes and percentage are reported unless otherwise indicated. Missing values in the outcome variables were handles similarly as in MICS methodology in order to produce similar indicators as in previously published reports [27]. Multiple imputation was used to account for missing data in the predictors.

A total of 5622 observations of mothers/last-born child pairs within two years preceding the survey were included in this study. Around 96% of mothers reported ever-breastfeeding their child. The majority of the sample resided in rural areas (73.5%), in Darfur (27.9%), and Central (29.1%) regions (Table 1). Half of the mothers were aged 25–34 years, three-quarters had primary, or no education and only 7.3% completed higher education. Only 9.1% of births were given birth by caesarean section, mostly born at home (71.3%) and attended by skilled health personnel (77.7%).

Early initiation of breastfeeding was reported by 68.7% (95% CI 66.7, 70.7) of women (Table 1). It was least prevalent in the Northern region 59.9% (95% CI 54.4, 65.2) and most prevalent in Eastern region 83.2% (95% CI 79.1, 86.7).

Univariate logistic regression analyses revealed that early initiation was significantly associated with region of residence, household wealth, father’s education, place of birth, birth assisted by a skilled attendant, birth by caesarean section, and birth order (Supplementary Table S1).

Compared to the capital city of Khartoum, the adjusted odds of early initiation was lower among mothers who were residents in all other regions; the Eastern region was an exception (Table 2). The adjusted odds of early initiation among mothers in this region was 1.79 higher compared to Khartoum (AOR 1.79; 95% CI 1.21, 2.64; p = 0.004). These odds were significantly higher with increasing father’s education. This result was similar for mother’s education up to secondary level where the adjusted odds of early initiation decreased for women with higher education compared to those with no education. The adjusted odds of early initiation was 34% higher among mothers whose births were assisted by a skilled attendant compared to unassisted mothers (AOR 1.34; 95% CI 1.06, 1.68; p = 0.014). Compared to mothers who gave birth by vaginal delivery, those who gave birth by caesarean section had 78% lower odds of commencing breastfeeding within the first hour after birth (AOR 0.22; 95% CI 0.17, 0.27; p <  0.001). Additionally, mothers of higher parity had higher odds of early initiation compared to mothers who had their first birth.

Nearly three-quarter of mothers did report not giving prelacteal feeds to their infants (71.7%; 95% CI 69.6, 73.8) (Table 1). Results from univariate logistic regression showed that region of residence, household wealth quintile, planned pregnancy, place of birth and birth by caesarean section were significant predictors of avoiding prelacteal feeding (Supplementary Table S1).

After adjusting for other covariates, region of residence, mother’s age and education, planned pregnancy, and birth by caesarean section were significant factors associated with avoiding prelacteal feeding practice in Sudan (Table 2).

Compared to mothers who lived in Khartoum, the adjusted odds of avoiding prelacteal feeding in the first three days after birth was generally higher in almost all regions with significant odds ratio in Kordofan (AOR 1.68; 95% CI 1.19, 2.36; p = 0.003), Eastern (AOR 2.06; 95% CI 1.46, 2.90; p <  0.001) and Darfur (AOR 2.49; 95% CI 1.80, 3.45; p <  0.001) regions. Maternal age and education up to secondary level were positively associated with avoiding prelacteal feeding. The adjusted odds of avoiding prelacteal feeding was lower among mothers who gave birth by caesarean section compared to those who gave birth by vaginal delivery (AOR 0.57; 95% CI 0.45, 0.73; p <  0.001). Likewise, the adjusted odds of avoiding prelacteal feeding was 1.51 times higher for mothers who planned pregnancy of their last child compared to those who did not (AOR 1.51; 95% CI 1.18, 1.93; p = 0.001).

In this study, about half the children had optimal early feeding, i.e. had both early initiation within one hour of birth and did not have any prelacteal feeding in the first three days after birth (51%; 95% CI 48.9, 53.1) (Table 1). Eighteen percent of infants had early initiation within one hour of birth but received prelacteal feeding in the first three days after birth, and 20% of infants did not receive any prelacteal feeding but were not put to the breast within one hour of birth. Optimal early feeding varied by region (Fig. 1), ranging from 33% in the Northern region to 65% in the Eastern region. Univariate logistic regression indicated that optimal feeding was significantly associated with region of residence, wealth, maternal age, planned pregnancy, place of birth, birth by caesarean section, sex of the child, and birth order (Supplementary Table S1).

Results from multivariable logistic analysis (Table 2) revealed that region of residence, mother’s age and education, planned pregnancy, place of birth, birth assisted by a skilled attendant, birth by caesarean section, sex of the child, and birth order were significant predictors of optimal early feeding practice.

The adjusted odds of optimal early feeding practice was two times higher in the Eastern region (AOR 2.02; 95% CI 1.44, 2.85; p <  0.001) and 43% lower in Northern region (AOR 0.57; 95% CI 0.39, 0.83; p = 0.003) when compared to Khartoum. Maternal age was significantly associated with optimal feeding with older women having about 25% higher odds of practicing optimal feeding related to the youngest women aged 15–24 years (e.g. for women aged 34+ years AOR 1.25; 95% CI 1.03, 1.51; p = 0.024). Mothers who had secondary or higher education had a 62% (AOR 1.62; 95% CI 1.30, 2.02; p <  0.001) and 43% (AOR 1.43; 95% CI 0.99, 2.06; p = 0.057) increased odds of optimal early feeding respectively compared to those with no education. Mothers with a planned pregnancy had 1.33 times higher odds to optimally feed their infants (AOR 1.33; 95% CI 1.09, 1.62; p = 0.005). In comparison to home births, children who were given birth in a health facility were less likely to be fed optimally (AOR 0.75; 95% CI 0.60, 0.94; p = 0.013). Compared to vaginal delivery, caesarean section was significantly associated with a much lower odds of optimal feeding (AOR 0.34; 95% CI 0.25, 0.47; p <  0.001). On the contrary, skilled attendance at birth increased the odds of optimal early feeding (AOR 1.48; 95% CI 1.19, 1.83; p <  0.001). Sex of child was also significantly related to optimal feeding practice in this study with 16% higher odds of optimal feeding for mothers who had female vs male children (AOR 1.16; 95% CI 1.02, 1.33; p = 0.028). Moreover, compared to mothers with 1st.

birth, the odds of optimal early feeding among mothers with 2nd and 3rd or higher birth order were (AOR 1.28; 95% CI 1.08, 1.52; p = 0.005) and (AOR 1.36; 95% CI 1.04, 1.77; p = 0.025), respectively.

About two of three Sudanese mothers initiate breastfeeding early after childbirth. A similar number of Sudanese mothers avoid prelacteal feeding in the first three days after childbirth. Nonetheless, only half of these mothers follow both practices at the same time. The National Strategy for Infant and Young Child Feeding (IYCF) 2015–2024 established by the national nutrition program of Sudan Federal Ministry of Health address the challenges of the significant suboptimal IYCF practices in Sudan. The strategy calls for improvement of the implementation of BFHI and improvement of knowledge and skills of health service providers. This study provides evidence-based information to benefit the implementation of the strategy.

This research is the first in Sudan to analyse a nationally representative sample to explore early feeding indicators. The data set was nationally representative with low levels of missing data. Another strength in this study is the use of a new optimal feeding indicator which captures two important early feeding practices (early initiation and avoidance of prelacteal feeding). The study had some limitations that should be considered while interpreting the results. Firstly, although the MICS data provided the opportunity to control for several confounding factors, it did not allow to control directly for other important dimensions like social and cultural factors like the importance of advice from grandmothers and ethnic beliefs on infant early feeding [12, 40]. Nonetheless, apart from the diverse capital city, considering the cultural diversity in Sudan, this aspect may have been implicitly controlled for when assessing regional differences. Secondly, as with cross-sectional studies, this study does not allow inferring causality. Thirdly, recall bias may be potential when mothers recall information on early feeding practices of their children that happened up to two years preceding the survey. Lastly, the way in which missing values in the outcomes were handled may produce slight underestimation in the early breastfeeding indicator and overestimation in the avoidance of prelacteal feeding indicator.