Determining factors associated with breastfeeding and complementary feeding practices in rural Southern Benin

Abstract This study aimed at characterizing breastfeeding and complementary feeding practices in a food‐insecure area of Benin and identifying factors associated with these practices. A cross‐sectional study was conducted in the districts of Bopa and Houéyogbé among n = 360 mother–child pairs. Children aged 0–17 months were considered. Socioeconomic characteristics among children and mothers, Breastfeeding on demand, Breastfeeding frequency during children illness, and Positioning and Attachment of children while breastfeeding were assessed using semi‐structured interviews and observations. Qualitative 24‐hr recalls were administered to mothers to compute WHO recommended complementary feeding practices indicators namely minimum dietary diversity (MDD), minimum meal frequency (MMF), and minimum acceptable diet (MAD) among 6–17 months old children (n = 232). Associations between each feeding practice and mothers' socioeconomic characteristics were tested using multivariate generalized linear models. Breastfeeding on demand and good positioning and attachment for breastfeeding rates were 59% and 66%, respectively. Only 26% of mothers used to increase breastfeeding frequency when their children were ill. The proportions of children who met MDD, MMF, and MAD were 51%, 75%, and 44%, respectively. Children living in Houéyogbé were less likely to be breastfed on demand compared with those living in Bopa; however, they had better breastfeeding frequency during illness and meal frequency. Socioeconomic factors with significant association with breastfeeding practices were children age and sex and mothers’ education, ethnicity, and employment status. Complementary feeding practices were positively associated with children's age but not with other socioeconomic characteristics. Breastfeeding and complementary feeding practices were almost suboptimal or medium and still need to be improved through well designed nutrition intervention program including nutrition education.


| INTRODUC TI ON
A diagnostic study conducted in two districts of southern Benin (namely Bopa and Houéyogbé) located in a high food-insecure area reported that infants and young children feeding practices were suboptimal (Mitchodigni, et al., 2017a;Mitchodigni, et al., 2017b). Authors suggested that multisectoral interventions should be implemented in this area to improve feeding practices for young children. Following this recommendation, a nutrition education program was planned in this region aiming at optimizing mothers' knowledge on nutrition and subsequently, breastfeeding and complementary feeding practices. Nutrition education is meant to improve knowledge, skills, motivation, and behavior of individuals or communities, leading potentially to subsequent positive effects on nutritional status and health (FAO, 2005). In sub-Saharan Africa, nutrition education interventions showed good contribution to improving breastfeeding practices (Aidam et al., 2005;Tylleskär et al., 2011), complementary feeding practices (Waswa et al., 2015) and also nutritional status of <2 years children (Lassi et al., 2013).
In order to refine the messages' content and take into account the contextual factors as part of the intervention, an in-depth understanding of the situation toward breastfeeding, complementary feeding practices, and associated factors in the area was necessary.
The present study, which was conducted before the implementation of the nutrition education program, aimed at (a) characterizing breastfeeding and complementary feeding practices in the intervention area using indicators which addressed directly the limitations among children feeding practices identified during the diagnostic survey and (b) identifying factors that were associated with these practices specifically in this intervention area. This could help to identify factors other than those relative to the intervention which could also influence the effect of the intervention by affecting the ability of the beneficiaries to adopt or reject recommended practices.

| Setting
The study was conducted in the districts of Bopa and Houéyogbé which present the highest rates of household food insecurity in the Mono department, respectively 40% and 34% (INSAE & PAM, 2014).

| Sampling
The data presented in this paper were related to the baseline survey of the impact evaluation of the nutrition education intervention implemented in the districts of Bopa and Houéyogbé in Southern Benin (Bodjrenou et al., 2020). A total of eight villages were randomly selected after stratification by district. In each village, 45 mother-child pairs were randomly selected from an exhaustive list of 0 to 17-months-old children living in the villages using a random number technique. We targeted this age-group firstly to focus on breastfeeding and complementary feeding practices. Secondly, the intervention and its impact evaluation were meant to last for 6 months; hence, children would be 6-23 months old at endline. In total, 360 mother-child pairs were surveyed.

| Data collection
Mothers or primary caregivers of children were interviewed during home visits by trained enumerators. Mothers' socioeconomic characteristics including age, ethnic group, employment status, marital status, level of education, and participation in nutrition education programs were collected. Breastfeeding on demand and attitudes toward breastfeeding frequency during child illness were recorded, while correct positioning and attachment of the child during breastfeeding was observed. We considered six parameters during the observations: breasts held by mother's hand, mother's hand doing a "C-shape," child's chin attached to the breast, child's lip in eversion, child in straight position, and child's whole body facing the mother's chest (Unicef, 2012;Vinther & Helsing, 1997;WHO & UNICEF, 1993). Enumerators observed mothers breastfeeding their children; each correct breastfeeding positioning and attachment parameter adopted by mothers was marked "1," and "0" otherwise.
The enumerators stayed in the study villages for about 2 weeks. This gave them several opportunities to observe mothers' practices. If an enumerator did not have the opportunity to observe a breastfeeding episode when interviewing a mother, she/he did not ask the mother to breastfeed her child, as this might disturb the habits. Enumerators had taken advantage of their stay in the village and would come back to make the observation as soon as he would have the opportunity to see the child at the breast. There is no specific duration of a breastfeeding episode. We have observed that youngest children breastfed longer and more frequently than older ones. Enumerators recorded different parameters of breastfeeding positioning and attachment as soon as mother positioned child and started breastfeeding.
The children aged 6-17 months feeding practices were also assessed using a qualitative 24 hr recall (WHO et al., 2008). All foods and drinks that had been used to feed the child the previous day and their constitutive ingredients were listed by the mother or the primary caregiver.

| Data management and statistical analysis
The following binary variables were computed and analysed: (a) Breastfeeding on demand (yes/no) (b) Breastfeeding frequency when the children were ill (increased versus unchanged or decreased). For each mother, a breastfeeding positioning and attachment score was computed by summing up the mark attributed to each of the six positioning parameters that were considered during the observations. The score, which theoretically ranged from 0 to 6 points, was recoded as a binary variable using the median score as a threshold: (a) mothers with a score less than the median and (b) mothers with a score equal or higher than the median (hence considered as "good positioning").
From the 24 hr recall data, foods were categorized into seven food groups as recommended namely: (a) grains, roots, and tubers; (b) legumes and nuts; (c) dairy products; (d) flesh foods; (e) eggs; (f) vitamin-A rich products; (g) fruits and vegetables different from those rich in vitamin-A (Kennedy et al., 2010;WHO et al., 2008).
The minimum dietary diversity (MDD), minimum meal frequency (MMF), and minimum acceptable diet (MAD) were computed following WHO and UNICEF guidelines (Kennedy et al., 2010;WHO et al., 2008). Children aged 6-23 months were considered having met MDD if they had consumed foods from at least four different food groups out of the seven recommended over the day prior to the survey. Children who received solid, semi-solid, or soft foods (including milk feeds for nonbreastfed children) the minimum number of times or more the day prior to the survey were considered having met MMF. A child who met both MDD and MMF was considered having met MAD (WHO et al., 2008). We, then, determined the proportions of children having met MDD, MMF, and MAD.
Characteristics of mothers and feeding practices indicators were described and presented using descriptive statistics: percentages for categorical variables and mean ± SD for continuous variables.
Comparisons among the two districts were also performed using chi-square tests for categorical variables and Mann-Whitney test or Student's t tests for continuous variables. Associations between socioeconomic characteristics of mothers and breastfeeding and feeding practices outcomes were analysed through a two-step approach.
Bivariate analyses (chi-square, Mann-Whitney, and Student's t tests) were first performed to identify factors with significant association with each practice. Variables that were significant at 20% (Mickey & Greenland, 1989;Trekpa et al., 2005) were kept for the next step.
Secondly, we used generalized linear model (GLM) with binomial distribution probability and logit link function to build models presenting the contribution of the variables (those that were significant in bivariate analysis) to each outcome. All independent variables were entered together into each model in one step. Data presented included odd-ratios, confidence interval associated, z-value, and p-value.
The relevance of regression models was assessed by displaying the goodness of fit parameters (X-squared and p-value) associated with Hosmer-Lemeshow test (p-value should be higher than 5%) and the error rate (quality of model's predictions) derived from the confusion matrix (error rate should be <0.5). Where Hosmer-Lemeshow test was not significant, we performed Pearson's residuals test (pvalue should be <5%).
Descriptive analyses and bivariate analyses were carried out using SPSS 23.0 (IBM-SPSS, 2014) while GLM and relative tests were conducted using R (Zuur et al., 2013). All analyses took into account the study design.

| Socioeconomic characteristics of the sample
The average ages were 8.4 ± 4.9 months for children and 27.5 ± 5.9 years for mothers. Boys represented 53.1% of children in the sample. Most of mothers were from the Sahouè ethnic group, had no schooling, and lived with their husbands (Table 1). Children's characteristics (gender and age) were similar across districts.
Mothers' age, marital status, and ethnic group were also similar across districts. However, the proportion of mothers with no schooling was significantly lower in the district of Houéyogbé than Bopa, with 45.0% and 77.2% respectively (p-value <.001).
Only 13.9% of mothers had no income generating activity.
Less than 6% of mothers had benefitted from a nutrition education program in the past with a difference between districts (Bopa: 2.2%; Houéyogbé: 9.4%; p-value =.003) ( Table 1).

| Feeding practices
All children in our sample were breastfed. Percentage of mothers who used to breastfeed on demand their children were 58.7% ( Figure 1); this proportion was slightly higher in Bopa than Houéyogbé (63.9% and 53.4%, respectively; p-value = 0.043). The proportion of mothers who had met good positioning and attachment for breastfeeding was 66.1% with no significant difference across districts ( Figure 1). Only one-quarter of mothers declared increasing breastfeeding frequency when their children were ill.
The percentages of children meeting the requirements of MDD, MMF, and MAD were respectively 50.9%, 75.4%, and 43.5% (Figure 2). In Houéyogbé, 82.5% of children met the MMF while they were only 68.6% in Bopa (p-value =.015). The other indicators did not differ across districts.

| Factors associated with breastfeeding practices
Mothers who lived in Houéyogbé were less likely to breastfeed on demand (OR = 0.59, CI 95% = [0.36; 0.98], p = .04) than mothers who lived in the district of Bopa (Table 2). Mothers who had higher school level (at least secondary school) were 3.08 more likely to breastfeed on demand than mothers with no schooling (OR = 3.08, CI 95% = [1.40; 6.79], p = .005). Likewise, mothers who practiced trading and those who were not from the Sahouè ethnic group had higher odds to breastfeed on demand than others (Table 2).

| Factors associated with complementary feeding practices
Very few factors were associated with the three indicators on complementary feeding practices (

| D ISCUSS I ON
Results from our study showed that children feeding practices were suboptimal or medium and this situation requires actions for im- We also found that mothers living in Bopa, who were mostly involved in home-based activities, were more likely to breastfeed on demand than those living in Houéyogbé. The fact that Houéyogbé is more urbanized than Bopa may explain these results (Mitchodigni, et al., 2017b).

TA B L E 3
Results from the GLM on complementary feeding practices to the recommended positions for breastfeeding. Mothers explained during informal discussions that as children were growing up, they became stronger and more agitated; consequently, it was more difficult to control them during breastfeeding.
Regarding complementary feeding, as expected, we observed that children's age was positively associated with the MDD and MAD.
Generally, mothers used to replace breast milk with porridge and then family foods progressively as children are growing up. Traditional porridges that are first given to young children consist of a cereal mixed with water and sometimes sugar, which provide low dietary diversity.
The family meals that are eaten by older children offer higher dietary diversity. Thus, as children are growing up and start eating family meals, their dietary diversity increases (Amoussa Hounkpatin et al, 2014).
Living in Houéyogbé also seemed to be more favorable to higher meal frequency among children than living in Bopa; probably because of the difference of urbanization (Mitchodigni, et al., 2017b).
No other factors were found to be associated with the complementary feeding indicators we investigated.
We observed that some intercepts were significant (Increasing

Breastfeeding Frequency during Illness, Good Positioning, and
Attachment for Breastfeeding, minimum acceptable diet). Thus, there were possibly other factors influencing feeding practices, but these factors were not collected in the present study. Mitchodigni, et al. (2017b) showed also that agriculture, especially the diversity of food groups grown by households, increased the likelihood of meeting the MDD among 6-23 months old children.
Moreover, we expected that participating in a nutrition education program in the past would have been associated with present feeding practices. Indeed, many studies had shown that nutrition education interventions targeting breastfeeding or dietary habits had good contribution to improving breastfeeding and complementary feeding practices in low-and middle-income countries especially in Africa (Aidam et al., 2005;Lassi et al., 2013;Tylleskär et al., 2011;Waswa et al., 2015). In other cases, nutrition education helped to increase knowledge and attitudes but not practices (Mojisola et al., 2019;Ruzita et al., 2007). However, in the present study few mothers (<6%) participated this program and this could not allow good analysis or conclusions. Moreover, participating a nutrition program does not necessary lead to behavior change and achieving the adequate knowledge does not guarantee the adoption of good practices. Some other factors such as limited access to nutritious foods (Dang et al., 2005), financial constraints (Aborigo et al., 2012;Otoo et al., 2009), and pressures or support from families including food taboos (Kakute et al., 2005;Amoussa Hounkpatin et al, 2014;Issaka et al., 2015) play an important role. In the other hand, nutrition education programs have to be well planned and delivered (Chapman-Novakofski, 2014) and intensive (Bukusuba, 2005;Roy et al., 2005) to ensure their efficiency .

| Limitation
We could have used qualitative method in the present study to have in-depth understanding of the situation. Qualitative research approaches allow a detailed description of participants' feelings, opinions, experiences, and interpretations of their actions (Rahman, 2017). However, they have some limitations. Results cannot be generalized to the entire research population with the same degree of certainty as quantitative measures. Sample size used in qualitative research is often smaller than that used in quantitative research methods (Dworkin, 2012). Qualitative research results are not tested to determine whether they are statistically significant or due to chance (Atieno, 2009). Therefore, we decided to use quantitative method to ensure measurability and comparability (statistical) of indicators since we aim to collect data related to the same indicators at the end of the intervention.

| CON CLUS ION
Breastfeeding and complementary feeding practices were not optimal in the districts of Bopa and Houéyogbé, department of Mono, Southern Benin. These results supported the importance of a nutrition education program at the community level in order to improve mothers' and community members' knowledge and attitudes toward children feeding practices. Moreover, socioeconomic, cultural, and demographic factors such as age and sex of children, district of residence, ethnic group, education level, and employment status identified in the present paper as influencing infants and children feeding practices will be taken into account when designing the community-based nutrition education program. According to these factors, different categories of people within our target population will be defined. Thus, key messages will be adapted to address each of these categories considering their specific backgrounds in terms of knowledge, occupation, food habits, culture, and beliefs.

ACK N OWLED G M ENTS
Research activities were funded by the Ministry of Foreign Affairs of Finland and CGIAR; The corresponding authors received also a PhD allowance from IRD.

CO N FLI C T O F I NTE R E S T
"No conflicts of interest."

E TH I C A L A PPROVA L
The study was conducted in accordance with the Declaration of