Rural-urban differences on the rates and factors associated with early initiation of breastfeeding in Nigeria: further analysis of the Nigeria demographic and health survey, 2013

The outcome variable for this study was EIBF, defined as introducing breastfeeding to newborns within the first hour of delivery in line with the WHO recommendation [1]. Responses to the question on ‘when child was put to breast’ following delivery, in the 2013 NDHS [9], was re-categorized as ‘within one hour’ (early initiation, coded as ‘1’) and ‘beyond one hour’ (late initiation, coded as ‘0’) for use in logistic regression analyses. We selected independent variables based on the objective of this study and the review of previous studies [11‐13] with consideration of the information available in the 2013 NDHS. The variables were broadly categorized into three groups; socioeconomic, biodemographic and health/support factors, in line with practice in previous studies [17‐19].

Socioeconomic variables assessed in this study included mother’s and father’s education level (secondary/higher, primary and none*), mother’s occupation [re-categorized as not working (not working, households and domestic works), agriculture (paid and unpaid agricultural works) and paid works* (office, business, clerical, both skilled and unskilled manual) [11, 20], and wealth index. Wealth index, a proxy for household’s socioeconomic status, was derived in the survey through the principal component analysis of available households’ assets. Based on the similarities between the five categories captured in the 2013 NDSH, the variable was re-categorized into three in the present study as follows: rich = richer + richest, middle = middle, and poor* = poorer + poorest.” [18].

Biodemographic variables assessed were birth type (single* and multiple), mother’s age (re-categorized as: < 20*, 20–34, ≥ 35 years), sex of child (male and female*), mother’s religion (re-categorized as: Christianity, Islam* and Traditional/other) and birth size (re-categorized as follows large = very large + larger than average, average = average and small* = very small + smaller than average). ‘Birth size’ describes mothers’ perception of the size of their babies at birth. The variable was used as a substitute for ‘birthweight’ given that information on birthweight was substantially missing in the 2013 NDHS, a common feature of most population-based surveys in developing countries like Nigeria [9, 21]. There is evidence, however, that birth size estimates are usually closely related to the values of birthweight [22], thus, justifying the substitutionary use of the variable.

Other biodemographic factors included the region of residence (categorized according to the geo-political zones in Nigeria: North-Central, North-East, North-West, South-East, South-South and South-West*), mother’s marital status (never married, formerly married and currently married*) and birth order (1*, 2–3, and ≥4). Health seeking/support factors, including antenatal attendance (none*, 1–3 and ≥4), mode of delivery (cesarean section* and non-cesarean section vaginal) as well as the place of delivery (home* and health facility) were equally assessed in this study. These variables and their categorization compares well with those of previous studies in Nigeria and internationally [11, 12, 17‐19]. (Note: * is the reference category used in analyses).

To ensure that no important explanatory factors were missed, variables with p ≤ 0.20 in the Chi-Square test were selected for inclusion in the initial multivariable regression model in line with practice in previous studies [18, 23]. This cut-off point was chosen following a critical appraisal of evidence in the literature [24‐26]. A backward elimination regression analysis was then performed to obtain the final parsimonious model, which only retained explanatory variables significantly associated with EIBF at 5% level (p - value <0.05). Adjusted odds ratios in the final parsimonious models together with their 95% CI and p - values were reported. To reduce possible statistical errors, analyses were double-checked, and all variables that satisfied inclusion criterion were included in our models. Also, our backward elimination modelling was tested by including potential confounders. All these analysis procedures were carried out separately for data disaggregated by rural-urban residence. All data management and analyses were performed using the Statistical Package for Social Sciences (SPSS), version 21. The sample weight and the multistage cluster design of the 2013 NDHS were accounted for using the Complex Sample Analysis method [27]. Where appropriate, the term ‘residence’ and ‘areas’ were used interchangeably in this study.

The demographic details of the sample, together with the percentage of mothers who fed breast milk to their newborns within the first hour of delivery are shown in Table 1. The rates of EIBF were 30.8% (95% CI 29.0, 32.6) and 41.9% (95% CI 39.6, 44.3) in rural and urban residence (p < 0.001), respectively (Table 1). Regionally, the highest rates of EIBF occurred in the North-Central region both in urban (63.5%, p < 0.001) and rural (43.5%, p < 0.001) residences. On the other hand, the South-West (28.3%, p < 0.001) and the North-West (21.6%, p < 0.001) regions, had the lowest rates of EIBF in urban and rural residences, respectively (Fig. 1). Compared to their counterparts with no education (26.1%), the rate of EIBF in rural residence was higher among mothers who had at least a secondary school education (41.9%, p < 0.001). Conversely, in urban residence, mothers with no education had a higher rate of EIBF (51.2%) than their counterparts who had a secondary/higher education (39.1%, p < 0.001).

There was no statistically significant difference in the rates of EIBF between home (42.4%) and health facility (41.7%, p = 0.761) delivery in urban residence. However, in rural areas, delivery in a health facility was significantly associated with higher rates of EIBF (42.4%) than home delivery (27.3%, p < 0.001). Similarly, rich households in rural areas had significantly higher rate of EIBF (44.2%) than their counterparts in poor households (26.4%, p < 0.001). In urban areas, there was no statistically significant difference in the rate of EIBF between the rich (42.2%) and poor households (37.5%, p = 0.408).

Furthermore, rural mothers with at least four antenatal attendance had higher EIBF rates (36.5%) than those who had no antenatal attendance (25.0%, p < 0.001). The reverse was the case in the urban residence where the rates of EIBF was significantly higher among mothers who had attended no or 1–3 times antenatal care compared to their counterparts with ≥4 attendances (p = 0.027). Mothers who were delivered by cesarean section had lower rates of EIBF than their counterparts with non-cesarean delivery, regardless of living areas (Table 1). In rural areas, mothers professing Christianity had higher EIBF rates (40.0%) than their Muslim counterparts (26.7%, p < 0.001). On the other hand, Muslim mothers in urban areas had higher EIBF rates (48.3%) than Christian mothers (35.8%, p < 0.001).

Table 2 presents the results of the multivariable regression analyses. For rural residence, three socioeconomic factors, wealth index, mother’s occupation and husband educational level, were significantly associated with EIBF. Similarly, three biodemographic (birth order, birth size and region of residence), as well as two health/support factors (place and mode of delivery), were found to be significantly associated with EIBF in the residence. Specifically, the odds of EIBF was 35% higher among rural mothers in rich households compared to their poor counterparts (AOR 1.35, 95% CI 1.06, 1.71), and 35% greater among mothers who were not working or those engaged in households/domestic works relative to their counterparts in paid jobs (AOR 1.35, 95% CI 1.15, 1.60). Mothers whose husband obtained at least a secondary school education had 28% increased odds of early initiation of breastfeeding (AOR 1.28, 95% CI 1.07, 1.57). Similarly, rural mothers of birth order ≥4 were 29% more likely to breastfeed within their first hour of delivery compared to those of birth order of one (AOR 1.29, 95% CI 1.10, 1.60). Further, the odds of EIBF was increased significantly among rural mothers living in the North-Central (AOR 1.84, 95% CI 1.34, 2.52), North-East (AOR 1.48, 95% CI 1.05, 2.08), and South-South (AOR 1.51, 95% CI 1.11, 2.10) regions. Also, the odds of EIBF were 33% and 46% significantly higher among mothers who had delivered a large sized baby and in a health facility, respectively (Table 2). Lastly, in the rural residence, the odds of EIBF were 3.5-fold higher among mothers with a non-cesarean delivery compared to those with a cesarean section (AOR 3.50, 95% CI 1.84, 6.62).

In urban residence, no socioeconomic factor was significantly associated with EIBF. However, one biodemographic factor (region of residence) and one health/support factor (mode of delivery) attained statistical significance (Table 2). Compared to the South-West region, urban mothers living in the North-Central (AOR 4.40, 95% CI 3.15, 6.15), North-East (AOR 3.50, 95% CI 2.55, 4.83), North-West (AOR 2.08, 95% CI 1.54, 2.80), and South-South (AOR 2.84, 95% CI 2.03, 3.97) regions had increased odds of EIBF. As the case in rural residence, urban mothers who had a non-cesarean delivery had 2.5-fold increased odds of EIBF compared to those who had undergone a cesarean section (AOR 2.48, 95% CI 1.60, 3.80).

Our study presents a snapshot of the rates and factors associated with early breastfeeding initiation in rural and urban Nigeria. National representativeness, high response rate, application of complex sample statistics in all analyses (to adjust for sample weights and cluster design of the survey) and low missing data are some of the strengths of this study. Others include large sample size and the use of rural-urban data disaggregation method. Nevertheless, this study is limited in that the cross-sectional design of the survey does not allow the causal relationship to be estimated. Also, data were self-reported, collected retrospectively and so prone to social desirability and recall biases. However, restricting our study to data in the two years preceding the 2013 NDHS reduces the chances of recall bias. To the best of our knowledge, this is the first study to investigate rural-urban differences in EIBF rates and associated factors using nationally representative data in Nigeria.