Study design

The study was conducted in Kenya-Nairobi City County at Kariobangi North-Health Centre. Ethical approval was obtained from The Kenyatta National Hospital/University of Nairobi Ethical and Research Committee (KNH/UoN-ERC Ref. P141/03/2014). All women gave written informed consent for themselves and their infants in Kiswahili prior to their inclusion in the study. Women with high score on EPDS or suicidal ideation or those presenting with mental health symptoms during the interview were referred for a follow up at a specialist mental health clinic in Mathare North. However the challenge in the system is often around follow-up as the social services system is neither too resourceful nor efficient.

Setting

Kariobangi is located approximately 15 km northeast of Nairobi city centre. It is a low-income residential estate that consists of both lower middle class and informal settlement. The informal settlement is characterized by high unemployment rate, poor housing, and poor public services such as sanitation, clean and water. Kariobangi North-Health Centre is the only public health center that provides services for the entire Kariobangi North area together with other few private dispensaries that offer services in this area. The facility offers curative outpatient services, a maternal and child health clinic with antenatal and postnatal care, family planning, growth monitoring and promotion, immunizations, Antiretroviral Therapy, and HIV counseling and testing. The maternal and child health department receives around 50–60 patients per day making 1500–1800 patiently monthly. Poor children living in urban areas remain disadvantaged, in 2012 an estimated under-five mortality rate in slums of Nairobi was 79.8 per 1000 compared to the 2009 estimate of 63.4 per 1000 [24, 26]. A large proportion of slum dwellers are adolescents and young adults from different tribal and cultural backgrounds who are between ages 15–45. The source of livelihood mainly comes from informal income-generating activities such a retail trade of consumable goods in small kiosks or paid casual work.

Study participants

Women were eligible for recruitment if they were 17–44 years of age with 6–16 weeks-old infants and attended the maternal and child health (MCH) clinic for infant immunization visits. Women with pre-partum cases of serious mental illness, as evidenced by documentation of diagnoses or psychotropic medication prescription were excluded as these could be a distinct group themselves contributing to more bias or greater heterogeneity. Kariobangi is a low-income residential estate in north eastern part of Nairobi, Kenya. It consists of both lower middle class and slum-type dwellings and split into two parts: Kariobangi North and Kariobangi South. A consecutive sampling method was used where every eligible mother-infant pair was recruited who attended the 6th and 16th week immunization clinics on active MCH clinic days between July 14 and August 15, 2014 and met the inclusion criteria. Our estimated sample size was 216 women and we were able to successfully recruit the number needed. All women who were eligible and invited to the study agreed to participate as the lead researcher well-integrated her work as part of the everyday MCH clinic functioning and offered extra hands and nutritional support however we had 7.4% N = 16 missing data of mother-infant pair, the missing data include information such as mothers age, income level, infant’s age etc., at times the clinic was bustling with women and there were oversights in keeping a tab on the mothers moving from one clinic to the next. Women who were not residing in the Kariobangi area were excluded. We did not screen for any preexisting illnesses or probe about HIV status and do recognize that these are limitations of our work. We tried as much as possible to make data collection efficacious, not difficult for mothers to respond to and one with least interruption to the flow of the clinic.

Data collection procedure and measures

The nurse in charge of the MCH clinic approached eligible mothers who came to the clinic and invited them to participate in the study. After registering their attendance with the nurse, mothers were approached for participation, the study purpose was explained to them and sought their consent and interest in participating, and after consent form signed the study tools were administered.

The lead author, who is a nutritionist and a public health researcher, conducted interviews with participating mothers in one section of the clinic. Mothers filled in a socio-demographic questionnaire which probed their age, marital, educational and occupational status including monthly income, dwelling type and the number of children. Postpartum depression was assessed using Edinburgh Postnatal Depression Screening [28]. The EPDS includes 10 items focusing on mood in the postpartum period in the past one week; it has a clinical postnatal depression score cut-off of 13 and above. Women with high score on EPDS or suicidal ideation or those presenting with mental health symptoms during the interview were referred to the mental health clinic held every Tuesday of the week at the health center or refereed to a nearby hospital which offer mental health services.

EPDS is a well-established instrument for postpartum depression that has demonstrated acceptable clinical utility as a screening scale in SSA [18, 29–31]. Members of the University of Nairobi Maternal and Child mental health working group had recently carried out a formal Kiswahili translation of EPDS including discussion of face validity issues and cross-cultural emic-etic issues in translation [32] and it was this version of the tool that was administered orally to the women. The lead researcher first gave the socio-demographic tool followed by EPDS and then gave the breastfeeding questionnaire which was followed by the nutritional status. Infant feeding practices were determined via an interviewer-administered, adapted and modified USAID Tool Kit for monitoring and evaluating breastfeeding practices [33]. Based on WHO infant and young children feeding recommendations for infants 0–6 months of age, feeding practices were classified as “exclusive breastfeeding” or “non-exclusive breastfeeding” on the basis of whether or not the infant had consumed solid foods or liquids other than milk with the exception of oral rehydration solution, drops or syrups consisting of vitamins, minerals supplements or medicines [34] (Fig. 1).

The first author used anthropometric measurements to assess infant nutritional status. The weights for babies were taken using a weighing scale and interpreted through infant growth monitoring card as percentiles and z-scores. In the MCH clinic, a trained nurse or (sometimes this could be a nutritionist) was responsible in taking the weight following the WHO guidelines of taking height and weight. In this study, the focus was on weight for age “underweight only”. Weight-for-age is a composite index of height-for- age and weight-for-height. It takes into account both acute and chronic malnutrition. The birth weights, weight at six weeks were used along with weight at the point of assessments to rule out if the infants were underweight or growing well. All children in this study with weight for age of-1SD, -2 SD and -3SD were considered underweight. For the premature or underweight unborn babies the direction of the growth curve was more important than the position of the curve on the chart from birth weight to the age of the assessment. A flat growth curve or a downward growth curve indicated danger which could indicate malnutrition. We do recognize that in focusing only on underweight categories we were not able to differentiate between stunting or wasting or a combination of both. It would have been interesting to know whether underweight was due to stunting, wasting or both.

Statistical methods

Sample size was determined based on World Health Organization estimate prevalence of depressive symptoms amongst women in developing countries at anywhere from 15% onwards [14]. Minimum required sample size was calculated using the prevalence formula by [35]. Pearson’s Chi square tests were used to evaluate associations between predictor and outcome variables. Using regression, we created a model with both feeding practices and PPD as potential predictors of infant nutritional status to explore the possibility that feeding practices explained or mediated part of the association between PPD and the infant nutritional status outcome. In this study analyses p < 0.05 was considered statistically significant. SPSS version 20 was used for analysis.

Social demographic characteristics of mothers

Prevalence of postpartum depression

Feeding practices

When queried about their feeding practices, 27% of mothers had supplemented their infants particularly with water (31%), formula (8%), cow milk (6%), and soft porridge (17%). On further probing, many women reported that their children were not getting satisfied with the breast milk alone. Others described low volumes of breast milk due to lack of food while others reported that their children suffered digestive complications like constipation prompting them to give water. Socio-demographic factors such as marital age, marital status, parity and income were identified a priori to be adjustment variables in the association between non-exclusive breastfeeding and PPD. Women with post-partum depression were significantly more likely than women without PPD to have practiced non-exclusive breastfeeding (UOR 6.99, 95% CI 2.88–16.96; P < 0.001).

Infant’s nutritional status

Using the weight-for-age indicator of malnutrition, 34% (95% CI 27.9–41.1%) of infants were determined to be underweight. As mentioned before socio-demographic factors were identified a priori to be adjustment variables in the association between infant weight and post-partum (these included marital age, marital status, parity and income). Infants of women with PPD were significantly more likely than infants of women without PPD to be underweight [OR 5.43; 95% CI (2.22–13.27); (P = 0.001)].

Results of multivariate analyses

Prevalence of postpartum depression

Our study found a 13% prevalence of post-partum depressive symptoms among women attending MCH clinic based in an urban settlement of Nairobi. This estimate lies within the wide range of prior PPD prevalence estimates (6.1–30.6%), reported amongst African mothers [1, 3, 31, 36]. A few studies conducted in South Africa found higher rates of PPD, for example Tomlinson and colleagues estimated prevalence of postpartum depression as 34.7% in a peri-urban settlement in Cape Town [37]. Explaining the differences among these prevalence estimates could be methodological differences in the ways the studies were conducted, the settings where the studies were carried out, timing of post-partum assessment, screening instruments used and cutoff values used to classify mothers as depressed [5, 18, 31]. As is apparent the EPDS is the most widely used PPD measure, but a range of cut-off points have been set, ranging from 10 to 13. In this study the prevalence estimate would have been 21%, (95% CI 15.3–26.7%) had the clinical cutoff had been lowered to a score of 10.

Associations between PPD and socio-demographics characteristics

In this study postpartum depression was significantly associated with low socio- economic status where women with low levels of income were more likely to be depressed than mothers who had higher income. Marital status was also strongly associated with depression. Single mothers living in low-income urban settlements lack economic and child-rearing support from husbands and their extended families, thereby struggling more with life stress and consequently depression. Our findings are consistent with several studies that have reported how among mothers from lower socio-economic status, especially in the context of poor social support, are at increased risk of parenting difficulties and adverse child outcomes [15, 16, 22, 23, 36, 38].

Infant feeding practices and postpartum depression

Our study found that women with PPD were more likely to introduce supplementary foods earlier than women who were not depressed. These results compare well to the results reported by [38] in Brazil who evaluated the association between postpartum depression and interruption of exclusive breastfeeding in the first 2 months of life and found that children of mothers with postpartum depressive symptoms were at higher risk of early interruption of exclusive breastfeeding. In Australia [27] reported early cessation of breastfeeding to be significantly associated with postnatal depression. In Nigeria, [23] reported that depressed mothers were more likely to stop breastfeeding earlier, and their infants were more likely to have episodes of diarrhea and other infectious illnesses. In our study two social determinants associated with non-exclusive breastfeeding included income level and marital status.

Infant’s nutritional status and postpartum depression

This study provides further evidence of an association between postpartum depression and poor infant nutritional status. Low income Kenyan mothers living in urban settlements struggle to obtain enough food. When they have insufficient breast milk to nourish their infants, they supplement their infants’ diets with water and other foods, that then exposes infants to infectious diseases such as diarrhea, and, hence, their becoming become underweight [15, 39]. Our study findings suggest that post-partum depression may contribute to this vicious cycle. WHO recommends that mothers in LMIC settings exclusively breastfeed infants for the first six months to achieve optimal growth and development [34]. In under-resourced settings, where sanitation and safe water are often lacking, breastfeeding can be life-saving. Breastfeeding protects against infectious diseases, especially gastrointestinal infections, which largely contribute to child morbidity and mortality in low income countries [39].

Our results linking PPD and poor infant nutritional status are consistent with findings of previous studies. In Nigeria [23] showed that infants of depressed mothers had significantly poorer growth than infants of non-depressed mothers at the 3rd month. In Zambia, [16] found adverse infant health outcomes were proportionately greater in infants of depressed mothers and the associations with adjusted mean difference in infant weight and length too were statistically significant. Other independent associations with episodes of diarrhea, maternal education, infant age, supplementary feeding) and incomplete vaccination were identified as risk factors. Similarly in Malawi, [22] reported that mean length-for-age z-score for infants of mothers with PPD was significantly lower than for infants of mothers without PPD. The multivariate results show that maternal income, marital status, and employment status do not explain away the strong associations between maternal depression and infant nutritional outcomes.