This manuscript is based on ENGINE birth cohort study data. The ENGINE birth cohort study is a prospective, community-based study within Empowering New Generations to Improve Nutrition and Economic opportunities (ENGINE) program. ENGINE was a five years nutrition intervention program funded by United States Agency for International Development (USAID) implemented from September 2011 to September 2016 in 100 selected districts in rural Ethiopia. Its main goal was to improve nutritional status of mothers’ and young children through a multi-sectoral approach targeting health, nutrition and agriculture. The program conducted operational researches as part of its rigorous evaluation to generate evidence on program performance and impact. The ENGINE birth cohort study was one of the operational researches conducted under this program. The study was led by Tufts University in partnership with Jimma and Hawassa Universities and Ethiopian Public Health Institute. The study aimed to investigate the benefits of an integrated nutrition program and its co-location with agricultural growth program on household agricultural production and productivity, food security, diet diversity, socio-economic status and livelihoods, as well as health and nutritional status of mothers and their children.
The study had an open cohort design, with recruitment and follow up of pregnant women happened for a period of two years. It was conducted from March 2014 to March 2016 in three Districts (Woliso, Tiro-Afeta and Gomma) in the South Western part of Ethiopia. Considering 30% attrition rate, a total of 4680 pregnant women were recruited between 12 and 32 weeks of gestation. The data was collected at the lowest administrative cluster (Kebele) level. A total of 117 clusters with a total sample size of 40 pregnant women per cluster were included in the study. All Kebeles within a district were sampled–with the exception of a few excluded due to inaccessibility–for a total of 1,560 pregnant women recruited in each of the three districts. In each study Kebele, study participants were recruited consecutively until the quota of 40 pregnant women was achieved. Mothers with serious medical conditions, early pregnancy termination, multiple pregnancies, still birth and newborns with congenital anomalies were excluded from the study/follow up.
Data was collected once during pregnancy for all women (twice for those in the first trimester), at birth, and then every three months until the child was 12 months old. Data collection was conducted by trained nurses electronically using Open Data Kit (ODK) software on handheld tablets and submitted to a secured server via an internet connection.
Measures
Infant Feeding Practices (IFPs): Infant feeding index was constructed using data collected at birth and then every three months until 12 months of age. Mothers were asked about timing of breastfeeding initiation, colostrum feeding, anything given to the infant before giving breast milk, whether the infant was still breastfed, number of times the infant was breastfed during the day and night yesterday, and what the infant ate yesterday. Based on these information, five separate IFP indices were prepared to assess age-specific infant feeding practices; namely, within three days of birth, at three months, six months, nine months and 12 months of child age.
The indices were computed following the methods suggested by Ruel and Menon [52]. Each item was scored depending on whether a practice was appropriate based on the WHO infant feeding recommendations [17,54]. A practice that was appropriate for a specific age group received a score of 1, and a practice that was inappropriate received a score of 0. Practices that are considered particularly relevant for a given time point received a score of 2 or 3. For example, breastfeeding received a score of 2 for an infant from birth to 12months of age. A score of 0 was given to non-breastfed infants. Use of bottle with a nipple was scored as 0 because the practice is considered inappropriate for all age groups; avoidance of infant bottles received a score of 1, indicating an appropriate practice. The dietary
diversity score was calculated by adding the number of food groups consumed in the last 24 hours and received a score of 0 if the child got below three food groups, 1 if the child got three food groups or scored 2 if the child got four or more than four food groups in the past 24 hours (Table 1).
The unstandardized total score could reach a maximum of 9-15 scoring points depending on the time point. The indices were standardized by converting each score into percentage of the maximum total score of the scales at each time point. A higher score in the feeding scales indicated a better infant feeding practice. The index was treated as continues variable. Table 1 below depicts the infant feeding practice variables and scoring system used in this paper.
Maternal depressive symptoms: Maternal depressive symptoms were assessed using the Patient Health Questionnaire (PHQ-9) [55]; once during pregnancy for all women (twice for those in the first trimester), within 72 hours after birth and three months postpartum. The PHQ-9 is a 9-item self-administered questionnaire designed to evaluate the presence of depressive symptoms during the prior two weeks. The nine items of the PHQ-9 are based directly on the nine diagnostic criteria for major depressive disorder in the DSM-IV (Diagnostic and Statistical Manual Fourth Edition) [56]. Each of the nine items can be scored from 0 (not at all) to 3 (nearly every day). Thus, the total score can range from 0 (absence of depressive symptoms) to 27 (most severe depressive symptoms).
The PHQ-9 scale had been validated for Afaan Oromo Language in a similar population prior to the commencement of the ENGINE birth cohort study and possessed good psychometric properties. A PHQ-9 score of 8 or above was taken as a cut off to define depressive symptoms [57]. For this study maternal depressive symptoms were classified as prenatal, postnatal and persistent. Only few mothers have time point two data and hence only time point one depressive symptoms data were used to define prenatal depressive symptoms. Depressive symptoms measured within three days of birth were used to define postnatal depressive symptoms. Whereas, persistent depressive symptoms were defined as mothers screened positive for depressive symptoms during all the three assessment periods; during pregnancy, at birth and three months postpartum. Only 1.2 percent of the participating women had persistent depressive symptoms and this category of depressive symptoms was not considered in the final model.
Household Food Insecurity: The household food insecurity was measured using the Household Food Insecurity Access Scale [58] at baseline (recruitment), at infants 6 months and 12 months of age. For this article we used the baseline measurement. The index women were asked nine questions (yes/no) to determine if anyone in their household had experienced problems of food access over four weeks preceding the interview. An affirmative response to any of the nine questions was followed by a question to determine how often the condition happened: rarely (1-2 times), sometimes (3-10 times), and often (> 10 times). Responses were coded as 0 = never (i.e., no experience), 1= rarely, 2 = sometimes, or 3 = often. Household food insecurity was categorized into four severity levels: food secure, mildly food insecure, moderately food insecure, and severely food insecure as per the algorithm described by Coates et al [58].
Intimate Partner Violence (IPV): A screening tool called HITS (Hurt, Insult, Threaten and Scream) was applied to assess intimate partner violence [59]. This data was collected from mothers within three days of birth. The scale has four items and each item was scored on a scale of 1 (never) to 5 (frequently) with total score of 20 possible. Then, sum score was computed and treated as a continuous variable in the model.
Maternal Social Support: Maternal Social support was measured using the Maternity Social Support Scale (MSSS) developed by Webster and colleagues [60] within three days of birth. The scale contains six items. Each item has measured on a five-point Likert scale of 1 (never) to 5 (frequently) and a total score of 30 was possible. Similarly, the score was treated as continuous variable in this study where a high score corresponds with a high level of perceived social support.
Socio-demographic characteristics: Educational status of the mother was categorized into four as illiterate, primary, junior and secondary and above for analysis purpose. Marital status was dichotomized into married (married monogamous and married polygamous) and unmarried (single, widowed, divorced, and separated). Religion was categorized into three as Muslim, Protestant and Catholic and Orthodox. Similarly, mothers’ age was categorized as <25 years, 25-35 years and above 35 years. Gestational age at birth was dichotomized as term (37 weeks and above) and preterm (<37 weeks). Birth weight dichotomized as normal (2500gm and above) and low birth weight (<2500gm); however, birth weight was treated as a continuous data in the model. A wealth index was created following the methods described by the Demographic and Health surveys for Ethiopia [21] using polychoric principal component analysis to represent a composite measure of a household’s cumulative living conditions and then separated into quintiles.
Statistical Analysis
We examined whether missing data on feeding practices and maternal depressive symptoms differed from those who were not missing these data. We compared these two groups on infants’ birth weight, household food security, and other key baseline sociodemographic variables. For the continuous variables, we used a t test for equality of means, and, for the categorical variables, we used Pearson’s chi-square tests.
Participants’ characteristics, IFPs and maternal depressive symptoms were summarized using descriptive statistics. To assess longitudinal relationship of infant feeding practice (IFP) and maternal depressive symptoms, we assumed that the repeated measurements of IFPs taken from each infant, overtime, are correlated and it is expected that study participants changed feeding practices over time as infants gets older. To examine differences in IFP within individual subjects over the follow up period, a linear multilevel mixed effects (fixed effects and random effects) model with a random intercept and a random slope was fitted with maximum likelihood estimation method. The fixed effects describe a population intercept and population slopes for a set of covariates, which include exposures and potential confounders. Random effects describe individual variability in IFP and changes over time. By considering individual random slopes and intercepts, this model allows to examine the influence of covariates on the change in IFP over time. Subjects with IFP data from at least two assessment intervals were included in the analysis.