Prelacteal feeding practice and its associated factors among mothers having children less than 2 years of age in East Africa: a multilevel analysis of the recent demographic and health surveys

Background Prelacteal feeding is a major public health problem that increases the risk of morbidity and mortality in children. It also result delayed breastfeeding initiation and interferes with exclusive breastfeeding. Although numerous studies have been done on prelacteal feeding in individual East African countries, most of them did not consider community-level factors that could affect the likelihood of prelacteal feeding. This study, thus, aimed to assess the pooled prevalence and associated factors of prelacteal feeding practice in East Africa. Methods We used pooled data from the 12 east Africa countries Demographic and Health Surveys (DHS). A total weighted sample of 33,423 women was included in the final analysis. We employed multilevel logistic regression analysis to assess factors associated with prelacteal feeding practice. Finally, the Adjusted odds ratio (AOR) with 95% Confidence (CI) interval was reported and variables with p value< 0.05, in the multivariable analysis, were declared to be significant predictors of prelacteal feeding practice. Result In this study, the pooled prevalence of prelacteal feeding practice was 11.85% (95%CI: 11.50, 12.20) with great variation between countries, ranging from 3.08% (95%CI: 2.35, 3.81) in Malawi to 39.21% (95%CI: 36.36, 42.06) in Comoros. Both individual and community-level factors were associated with prelacteal feeding practice. Of the individual-level factors, home delivery, multiple birth, cesarean delivery, non-exposure to media, delayed initiation of breastfeeding, and being a small-sized baby were associated with higher odds of prelacteal feeding practice. Among the community-level factors, rural residence and higher community-level of media exposure were associated with lower odds of prelacteal feeding practice. Conclusion In this study, the pooled prevalence of prelacteal feeding is high. Both individual and community level variables were associated with prelacteal feeding practice. Therefore, individual and community-level interventions that encourage mothers to deliver in the health facility and promoting timely initiation of breastfeeding are needed to reduce prelacteal feeding practices in east Africa. Moreover, media campaigns regarding this harmful traditional practice could be strengthened.


Conclusion:
In this study, the pooled prevalence of prelacteal feeding is high. Both individual and community level variables were associated with prelacteal feeding practice. Therefore, individual and community-level interventions that encourage mothers to deliver in the health facility and promoting timely initiation of breastfeeding are needed to reduce prelacteal feeding practices in east Africa. Moreover, media campaigns regarding this harmful traditional practice could be strengthened.

Keywords: Prelacteal feeding, East Africa, Multilevel analysis
Background Optimal breastfeeding is important for the immediate and long-lasting health of the child by preventing the most childhood killers: pneumonia and diarrhea [1]. Practicing optimal breastfeeding is very important for the prevention of undernutrition and the cognitive development of the newborn [2][3][4]. Moreover, optimal breastfeeding significantly reduces the risk of developing different infectious diseases and non-infectious inflammatory diseases such as allergy and asthma, as well as obesity and chronic non-communicable diseases such as diabetes mellitus [5,6].
Initiating breastfeeding within 1 h of birth, exclusively breastfed for the first 6 months of life, and continued breastfeeding up to the age of 2 years are recommended by the World Health Organization and United Nations Children's Fund [1]. However, in different countries including countries in East Africa, the majority of mothers offer suboptimal breastfeeding practices to their newborns [7][8][9][10][11][12][13][14][15][16].
Prelacteal feeding is giving foods or liquids (except recommended medications) to newborns before breastfeeding is established [1]. It is a major public health problem that increases the risk of acquiring respiratory tract infections, diarrhea, and malnutrition [5,17]. Furthermore, the practice of prelacteal feeding deprives newborns of taking colostrum that is rich in nutrients and immunoglobulins [18,19]. It has also shown that giving prelacteal foods delays breastfeeding initiation and interferes with exclusive breastfeeding [1,3,20,21].
Despite its great effects on the health of the newborn, prelacteal feeding is widely practiced in many countries in the world with the highest prevalence in the southeast and central Asia, and Latin America [7,[21][22][23]. In Africa, most of mothers provide prelacteal foods to their newborn, and in sub-Saharan Africa, about 32.2% newborns are exposed to prelacteal foods [15,24,25]. Works of the literature revealed that maternal education [22,26,27], antenatal care (ANC) utilization [15,27], home delivery [13,28], delivery by cesarean section [14,22,27], sex of the child [15,28], and late initiation of breastfeeding [28] are among the factors that are associated with prelacteal feeding practice.
Although numerous studies are done on prelacteal feeding practice in individual east African countries, most of them did not consider the community-level factors that could affect the likelihood of prelacteal feeding. Therefore, we aimed to assess the pooled prevalence and associated factors of prelacteal feeding practice in east Africa. Identifying various factors at both individual and community levels can have a key role in implementing policies and programs aimed at minimizing prelacteal feeding practices.

Data source, data collection, and study population
We used pooled data from the 12 east Africa countries Demographic and Health Surveys (DHS) that were conducted from 2008 to 2019. All these surveys used a stratified two-stage cluster sampling technique. The key demographic and health indicators were collected in each DHS [29]. Five questionnaires: the Household Questionnaire, the Woman's Questionnaire, Man's Questionnaire, the Biomarker Questionnaire, and the Health Facility questionnaire were used in each survey to collect the demographic and health indicators. A pretest was performed before collecting the data and a debriefing session was held with the pre-test field staff [30]. Further information regarding the data collection procedure is found in each countries survey report.
For our study, we used a kid's data set with a total weighted sample of 33,423 women ( Fig. 1).

Variables of the study Dependent variable
The outcome variable was prelacteal feeding practice, which is defined as giving anything other than breast milk for the newborn in the first 3 days after delivery [31].

Independent variables
Both individual and community level explanatory variables were incorporated in this study (Table 1).
Individual-level variables Women age, educational level, maternal occupation, wealth quantile, ANC visit, place of delivery, delivery by cesarean section, the timing of initiation of breastfeeding, perception of distance from the health facility, media exposure, parity, sex of the child, birth order, and size of the child at birth.
Community-level variables In this study, place of residence was a non-aggregate community-level variable while community level of women education, community level of media exposure, community level of ANC utilization, and community poverty level were constructed through the aggregation of individual-level factors to conceptualize their neighborhood effect on prelacteal feeding practice.

Data management and statistical analysis
The data were appended, recoded, and analyzed using Stata 14 software. The sample was weighted using the primary sampling unit variable, stratification variable, and the weight variable, to restore its representativeness and to get a better estimate throughout the analysis [32]. Both the weighted and unweighted results were presented and compared. The proportion of prelacteal feeding practice per each independent variable and the absolute risk difference was calculated. The pooled data have a hierarchical structure with individuals nested within clusters. Therefore, we employed a multilevel logistic regression analysis. To conduct the multilevel logistic regression analysis, four models were fitted. These are: the null model (a model containing only the outcome variable), model 1 (a model with the outcome variable and individual-level variables), model 2 (a model with the outcome variable and the community variables only), and model 3 (a model with the outcome variable and both the individual and community level variables).
The random effect analysis that is a measure of variation of prelacteal-feeding practice across communities or clusters, were assessed using intra-class correlation coefficient (ICC), median odds ratio (MOR), and a proportional change in variance (PCV) [33][34][35]. Since the models fitted are nested models, deviance was used for model comparison and model fitness. Both bivariable and multivariable multilevel analyses were done and variables with p-value < 0.20 in the bivariable analysis were eligible for multivariable analysis. Finally, the Adjusted Odds Ratio (AOR) with 95% Confidence Interval (CI) was reported and variables with p value< 0.05, in the multivariable analysis, were declared to be significant predictors of prelacteal feeding practice. Variance inflation factor (VIF) was used to test Multicollinearity and there was no Multicollinearity between independent variables.

Socio-demographic characteristics of the study population
Among 38,270 mothers, with under 2 year living children, who ever breastfeed or are breastfeed their child, 33,423 mothers (weighted) were included for the final analysis (Fig. 1). The majority of the study participants were from Mozambique, Kenya, Tanzania, Zambia, and Ethiopia. Regarding place of residence, more than three fourth (77.23%) of respondents were rural dwellers. The median age of mothers was 27 (IQR ± 10) years. About half (50.36%) of the respondents had a primary level of education and 45.83% of respondents were from Educational level It was the minimum level of education the mother achieve and it was coded as no formal education, primary education, and secondary and above (combining secondary and higher education categories).
Mother's occupation It was the current working status of a woman, which was grouped into non-working and working.
Wealth quantile In the DHS, it was based on wealth quintiles, with the first (lowest) quintile being the poorest, second (poorer), middle, fourth (rich), and the highest quintile being the richest. For our study, it was re-categorized as poor (includes the lowest and the second wealth quantiles), middle, and rich (includes the fourth and the highest wealth quantiles).

ANC visit
It was defined as the number of ANC visits while a mother was pregnant and it was coded as no ANC visit, 1-2 ANC visits, 3 ANC visits, and 4 and above ANC visits.
Place of delivery Grouped into delivery at home and health facility.
Delivery by cesarean section Categorized as yes (if the mother gave birth by CS) and no (otherwise).

Timing of initiation of breastfeeding
It was the time at which the mother gave breast milk to the newborn baby and categorized as early (at/within 1 h) and delayed (after 1 h).

Perception of distance from the health facility
It was based on whether the mother perceives distance from the health facility as a big problem or not a big problem media exposure Created by combining whether a respondent reads a newspaper, listens to the radio, and watches television and coded as yes (if a woman had been exposed to at least one of these media) and no (otherwise).

Multiple birth
It was coded as no (if women gave a single birth) and yes (otherwise) Sex of the child The sex of the child was categorized into female and male.
Size of the child at birth It was based on the mother's perception and categorized as very small, smaller than average, average, larger than average, and very large in the DHS. This variable was re-categorized into small (very small, smaller than average), average (average), and large (larger than average and very large) for our study.
Community-level variables

Residence
The place of residence for the mother and coded as rural and urban

Community-level women education
It was the proportion of women with a minimum of primary level of education derived from data on mothers or respondent's level of education. Then, it was categorized using national median value to values: low (if the mother was from communities in which ≤50% of women had at least primary education) and high (if the mother was from communities in which > 50% of women had at least primary education) community educational level.
Community poverty level It was the proportion of women in the poorest and poorer quintiles derived from data on the household wealth index. Then, it was categorized, based on national median value, into: low (if the mother was from communities in which ≤50% of women had poor socioeconomic status) and high (if the mother was from communities in which > 50% of women had poor socioeconomic status) community poverty level.

Community-level of ANC utilization
It was the proportion of women with at least one ANC visit and categorized using national level quartiles to: low (if the mother was from communities in which ≤25% of women utilizing ANC), middle (if the mother was from communities in which 25-75% of women utilizing ANC), and high (if the mother was from communities in which ≥75% of women utilizing ANC) community-level ANC utilization.
Community-level media exposure The proportions of mothers who were exposed to media within a specific cluster. It was categorized in the same fashion as the community level of women's education into low and high community-level of media exposure.
Note: ANC = Antenatal Care, CS=Cesarean Section, DHS=Demographic and Health Surveys households with poor socioeconomic status. Regarding the timing of breastfeeding initiation, the majority (80.67%) of respondents initiate breast milk within 1 h. More than half (51.7%) and three-fourth (78.86%) of respondents had four and above ANC visits and gave their last birth at the health facility, respectively ( Table 2).
Proportion of prelacteal feeding practice by sociodemographic characteristics and the absolute risk difference Table 3 revealed the weighted and unweighted proportion of prelacteal feeding by each independent variable and their absolute risk difference. The weighted percentage of prelacteal feeding among mothers who gave a multiple birth was 20.23% while in those who gave single birth was 11.73% with an absolute risk difference of 8.56%. The proportion of prelacteal feeding among mothers who initiated breast milk within an hour and after an hour was 7.99 and 27.93%, respectively, with an absolute risk difference of 19.94%. The absolute risk difference of prelacteal feeding among those mothers who gave birth at home and at the health facility was 8.46%. Regarding country, the highest absolute risk difference (36.13%) was found between Comoros and Malawi ( Table 3).

Factors associated with prelacteal feeding practice in East Africa
Fixed effect analysis Table 4 revealed a multilevel analysis for the final model, both for the weighted and unweighted data. We considered the weighted data to assess factors associated with prelacteal feeding practice in East Africa since it gives an appropriate parameter estimate to draw a valid conclusion. Therefore, the interpretations here are based on the weighted data. In the multivariable multilevel analysis; multiple birth, the timing of breastfeeding initiation, media exposure, place of delivery, delivery by cesarean section, size of the child at birth, residence, and community level of media exposure were significantly associated with prelacteal feeding practice (p < 0.05  (Table 4).
Random effect analysis and model comparison Table 5 revealed the random effect analysis for the model with the weighted data. The ICC value in the null model indicates 9.3% of the total variations of prelacteal feeding practice were due to the difference between clusters. Besides, the high MOR value in the null model which was 1.74 revealed that when we randomly select mothers from two clusters, mothers from a high-risk cluster had 1.74 times more likely to practice prelacteal feeding as compared to mothers from a low-risk cluster. Moreover, the PCV in the final model revealed that about 13.4% of the variability in prelacteal feeding practice was explained both by individual and communitylevel factors. Regarding model fitness, model 3 was the best-fit model since it had the lowest deviance (Table 5).

Discussion
This study aimed to assess the pooled prevalence and associated factors of prelacteal feeding practice in east Africa. The pooled prevalence of prelacteal feeding was 12%. The prevalence in this study is in line with a study done in Ethiopia [36], however, it is lower than reports from other studies [7,8,[37][38][39] (Table 6). This discrepancy might be due to the difference in the study population, the variation in living conditions, and the difference in access to media and information across countries. This suggests strategies concerning suboptimal feeding patterns are decreased over time due to the expanded utilization of maternal health services. This study also found the high heterogeneity, from 3% in Malawi to 39% in Comoros, of prelacteal feeding practice across east African countries. This may be due to the difference in the study period. For example, the data for Comoros was collected in 2012, while the data for Malawi was collected in 2015. Besides, the high   heterogeneity in prelacteal feeding practice may be due to the sociocultural and socioeconomic differences among mothers in east African countries. This study identified different factors that were associated with prelacteal feeding practice. In the unweighted data analysis, factors such as maternal education, multiple birth, media exposure, parity, ANC visit, place of delivery, delivery by cesarean section, birth size, residence, community-level of media exposure, communitylevel of ANC utilization, and community poverty level were associated with prelacteal feeding practice.
However, weighed data analysis identified multiple birth, the timing of breastfeeding initiation, media exposure, place of delivery, delivery by cesarean section,  size of the child at birth, residence, and community level of media exposure as predictors of prelacteal feeding practice. This finding is consistent with different studies done elsewhere (Table 6). We prefer to discuss the results we get from the weighted data, which is necessary when we analyze DHS data [32]. Weighting preserves the representativeness of data and it helps to get standard and appropriate statistical estimate (robust standard error) [32]. Therefore, we give stress to the findings from weighed analysis and the interpretations and discussions, in this paper, are based on the weighted data. Institutional delivery was associated with lower odds of prelacteal feeding practice. This is consistent with studies done in Ethiopia, Nigeria, and Pakistan [26,[40][41][42]. This might be justified as many health centers and hospitals ensure breastfeeding counseling during   pregnancy, delivery, and postpartum periods to deter prelacteal feeding practices [45][46][47][48]. Another possible reason for introducing prelacteal feeding might be since mothers who delivered at home have no the opportunity to access health information about safe breastfeeding practices. Besides, a woman who delivered by cesarean section was more likely to provide prelacteal feeding. This is in agreement with studies conducted elsewhere [7,12,36,39,40]. This might be because those mothers may be difficult to give breast milk since they are still recovering from pain, immobilization, and tiredness. This indicates that physicians may not be equipped with the appropriate skills to support mothers under such circumstances.
Mother with a multiple birth was associated with a higher likelihood of prelacteal feeding practice. This is in agreement with a study done in sub-Saharan Africa [15]. This might be because the mother with multiple births perceives their breast milk as insufficient and more likely to practice prelacteal feeding. Regarding the timing of breastfeeding initiation, mothers who had delayed initiation of breastfeeding had higher odds of prelacteal feeding practice compared with their counterparts. This is supported by studies done in Ethiopia, Uganda, and Pakistan [20,28,[41][42][43]. This may be justified, as the time between birth and breastfeeding initiation increase, there would be more room for malpractices such as prelacteal feeding.
The study at hand revealed that mothers with a smallsized baby had a higher likelihood of prelacteal feeding practice compared to mothers with large-sized babies. This is in concordance with different studies done elsewhere [10,12,15,27]. This may be due to the misconception that small-sized babies will benefit from other foods and liquids.
Mothers who had exposure to different media and mothers from communities with a higher level of media exposure had lower odds of prelacteal feeding practice as compared to their counterparts. This finding is in line with a study done in Nigeria [44]. This may be because disseminating information about the impacts of prelacteal feeding through different media could prevent prelacteal feeding practice. This suggests that printing and electronic mass media play a significant role in fostering optimal breastfeeding practices.
Moreover, women from rural areas had lower odds of prelacteal feeding practice compared to those from urban areas. This is in line with a study done in Egypt [10]. This might be due to the recent expansion of health extension programs among rural people, which Gao, 2020 [44] Trends in prelacteal feeding practices in rural Bangladesh from 2004 to 2019 88.0% Being having maternal education, improvements in socioeconomic status, and exposure to media decreases the odds of prelacteal feeding.
Note: ANC Antenatal Care, AOR Adjusted Odds Ratio, CI Confidence Interval increases women's level of understanding about the impact of prelacteal feeding on child health.

Strength and limitations of the study
This study was based on the pooled analysis of the East Africa countries DHS. It was based on a multilevel analytical approach that can able to identify both individual and community-level factors that were associated with prelacteal feeding practice. Also, appropriate estimation adjustments such as weighting were applied. Therefore, the findings of this study will provide important insights to policymakers and governmental and nongovernmental organizations to design the most appropriate interventions at both individual and community levels. However, this study was not without limitations, in which while interpreting the study findings should be with caution. First, the outcome variable, prelacteal feeding practice was assessed based on the maternal selfreport and therefore there might be a recall bias. Second, DHS did not collect some information such as maternal beliefs, misconceptions, and knowledge towards breastfeeding that were evidenced to influence prelacteal feeding practice. Third, the influence of medicines, including those used for cesarean sections are not assessed. Finally, since it was a cross-sectional study we are unable to assure the temporal relationship between prelacteal feeding practice and important independent variables such as the timing of initiation of breastfeeding.

Conclusion
In this study, the pooled prevalence of prelacteal feeding is high and still needs strengthening of interventions on appropriate breastfeeding practices. Both individual and community level variables were associated with prelacteal feeding practice. Of individual-level factors, home delivery, multiple birth, cesarean delivery, non-exposure to media, delayed initiation of breastfeeding, and being small-sized baby were associated with higher odds of prelacteal feeding practice. Among community-level factors, rural residence, and higher community-level of media exposure were associated with lower odds of prelacteal feeding practice. Therefore, individual and community-level interventions that encourage mothers to deliver in the health facility and promote timely initiation of breastfeeding are needed to reduce prelacteal feeding practices in east Africa. Moreover, media campaigns regarding this harmful traditional practice are recommended.