The influence of maternal psychosocial circumstances and physical environment on the risk of severe wasting in rural Gambian infants: a mixed methods approach

Based on the preliminary analysis of the quantitative data, maternal illiteracy, history of a child death and maternal depression were associated with severe wasting in infancy. We therefore used these findings to develop a sampling framework for the in-depth interviews (IDIs) (Fig. 1). We used stratified purposive sampling where mothers of infants who were in both groups i.e. cases and controls, were selected, in order to capture a diversity of views and experiences about psychosocial, cultural, economic and environmental factors that influenced their infant feeding and rearing practices and their perceptions on how these impacted on the nutritional status of their children. It also allowed us to look for any major differences across the 2 sets of carers i.e. those with or without severely malnourished children.

Purposive sampling is commonly used in qualitative research to identify a particular group of people who possess certain characteristics or who reside in circumstances pertinent to the phenomenon being studied and are therefore “information-rich” [39]. Stratifying the sampling framework also allowed us to identify any major variations in practices or perceptions within each strata [39]. This iterative process led us to undertake IDIs with fathers of some of the cases and controls, and with the research staff. We stopped recruiting participants for IDIs when we got to the point of “data saturation” i.e. the point in data collection and analysis when “new information does not generate any new themes or variability within the themes in the dataset” [40]. In total, we undertook 19 IDIs with carers (8 with mothers and 2 with fathers of cases, 8 with mothers and 1 with a father of controls) and 4 with the research staff (3 male, 1 female).

Baseline maternal demographic data including maternal age, parity, residence, marital status, socio-economic status (assessed using household income, maternal education, household assets, housing quality and access to water and sanitation) and details of her spouse were collected at recruitment into the ENID trial [38]. In addition, data on the infant anthropometry, feeding and illness episodes were collected prospectively during the trial [30].

The development of the qualitative data collection tools (i.e. observation tool and IDI guide) was an iterative process that involved the revision of the tools in order to facilitate further exploration of factors that were found to be significantly associated with severe wasting in the infants in this study (i.e. increased frequency of complementary feeds) in the quantitative data analysis. We also added probing questions to explore the reasons for why certain factors that we had hypothesised as being important contributors to severe wasting in this population were not found to be so in the quantitative analysis.

Data were analysed using Stata (version 12.0, StataCorp, College Station, TX). We incorporated available prospectively collected data for the following variables: infant feeding data on timing of the introduction of complementary feeds, parity, infant morbidity and wealth indices. Descriptive statistics were computed to report number and percentage for categorical data, and the mean or median and standard deviation (SD) or interquartile range (IQR) for continuous data. The two sample Student T-test was used to assess the difference WHZ at the point when the cases growth faltered and at 12 months of age between the cases and controls. A principal component analysis (PCA) was used to determine household socioeconomic status using an asset based index in which we included the following six socioeconomic indicators: ownership of television, car, electricity, motorcycle, bicycle, or animal cart (yes/no), that were prospectively collected in the ENID trial. The adequacy of these indicators was assessed for inclusion in the PCA using the Kaiser-Meyer-Olkin measure [47], and they explained over 22% of the variability in the combined socioeconomic factor score (Further details in Additional file 1). The households that the children came from were classified into quintiles based on Filmer and Pritchett’s method [48].

The association between the demographic, psychosocial, infant feeding and WASH characteristics and severe wasting in infancy was assessed using conditional logistic regression, where cases were only compared to controls in the same set for crude odds ratios (ORs) and then, adjusted for confounders (including maternal age that was correlated to both severe wasting and maternal education/parity/number of child deaths; and paternal age that was correlated to both severe wasting and socioeconomic status) and collinearity in the multivariable model [49]. Using a backward stepwise model selection criteria [49], where explanatory variables with a sparse amounts of data (e.g. sibling deaths) and those with unreliable estimates where the 95% confidence intervals were ≥10, were dropped from the model. In addition, we incorporated the prospectively collected data variables where the retrospectively collected data was either not available or was not robust e.g. age of introduction of complementary feeds and household assets in order to improve the accuracy of our findings.

In qualitative research data analysis is an inductive process that seeks to learn about the perceptions that participants hold about a problem or issue by identifying patterns or themes [50]. HMN used the information from the informal observations, interview transcripts and summary information from the debriefing sessions, to develop summary sheets of caregivers in both groups. She then used headings and sub-headings based on the study questions and emerging themes to categorize information from the summary sheets. These ensured that we could comprehensively describe each group of mothers. This process formed the preliminary stages of data analysis as it involved reading and re-reading of the observation tools, interview transcripts and field notes, and thorough familiarization with the data from each caregiver.

We undertook data analysis alongside data collection in order to allow questions to be refined and new avenues of inquiry to develop [51]. We used NVivo 10 software (QSR International Pty Ltd. 2012) in subsequent data management and analysis using codes. After extensive familiarization with the data, HMN developed a coding scheme [52]. Two of the authors (HMN, MKM) agreed on the coding scheme who together reviewed the first 10% of respondents’ transcripts independently and through multiple iterations came to an eventual consensus. We revised the coding scheme further during the analysis. These multiple iterations are a key inductive process in the development of data coding schemes [53]. HMN undertook the NVivo coding of all the 23 transcripts. Using a thematic analysis approach, we grouped the data into themes and sub-themes and evaluated key emerging themes and how the themes/subthemes were interconnected [54‐56]. To guide our analysis, we used a conceptual framework (Fig. 3) that was based on the UNICEF Conceptual framework for undernutrition in children (Additional file 1) [11].

The mean WLZ for the cases and controls were −3.5 (SD 0.5) and −0.6 (SD 1.1), p < 0.005 respectively at the age they were matched. The mean WLZ of the cases and controls at 12 months remained significantly different (−2.1 [SD 1.3] and −0.8 [SD 1.1], p < 0.005). There two groups were similar with regard to age, gender, distance from the MRC clinic and village size. At the start of this follow up study, all the children were over 6 months of age with a mean age of 2.9 years (SD 0.8) for cases and controls. The primary carers during the first 12 months for all the children in the study were their mothers (Table 1).

The overall prevalence of maternal depressive symptoms was 13% and this was similar between the mothers of the cases and of the controls (13% vs 12%). There maternal level of education was also similar between cases and controls and the majority had attended Arabic school (Islamic studies) (Table 1).

Overall traditional pit latrines were the commonest form of sanitation within compounds (95% of participants) and the source for clean drinking water was less than 30 min away for most mothers (>96%). Only a third of mothers undertook water purification and this was often filtration using a cloth and over 90% reported that they practised hand washing with soap during their daily activities (Table 1).

Infant feeding choices were often made by the mothers. Infants were often breastfed up to 24 months, with the introduction of complementary feeds at an average of 5.2 months. The commonest complementary feed was maize meal (“coos”) porridge. Infants were often fed 3 complementary meals a day and mothers in both groups most frequently reported practising scheduled complementary feeding (70% cases vs 72% controls). The mothers of the cases reported a higher frequency of complementary feeding (mean 4 vs 3) (Range 1–8) (Table 1).

Overall children in both groups had a median of 2 (IQR 1, 4) diarrhoea episodes in the first 12 months of life and this was similar for cases and controls (10 [IQR 6, 14] and 9 [IQR 6,13] (Table 1).

Overall, a third of mothers had experienced the death of a child other than the index child, with the most recent loss being closer to the neonatal period in the controls (median 2 [IQR 1, 5] vs 6 [2, 7] months (Table 1) (Further details in Additional files 1, 2, 3, 4 and 5).

In the univariable analysis, the only factor that was associated significantly with severe wasting in the infants was increasing frequency of complementary feeds (OR 1.51 [1.13–2.01], p = 0.005) (Table 2). In the multivariable model, this association remained significant (OR 2.06 (1.17–3.62), p = 0.01). The treatment of water showed a trend towards a protective effect but this was not statistically significant (OR 0.31 (0.09–1.04), p = 0.06). Maternal depressive symptoms were not significantly associated with severe wasting in infants in this model (OR 1.37 (0.32–6.00), p = 0.67) (Table 3).

We observed that within households, mothers were preoccupied with domestic chores (including food preparation) and ‘gardening’ (farming) in order to provide food for their families. In addition, they also bore the responsibility of caring for their aging parents-in-law. Apart from mothers who were still breastfeeding, they were often unable to supervise feeding or play sessions with infants or young children and relied on a family member, e.g. paternal grandmother, or an older sibling for support with this. When these support networks were not available, these sessions were unsupervised. This lack of supervision sometimes posed a risk to the infants, e.g. being too close to a fire, or sharing a feeding bowl with a domestic animal.

“Yes, because if the mother leaves the child at home with the younger ones who cannot take care of that child, for example...I met the child eating with a dog. You see, and whilst the elder sister was there who could not even take care of that child. When I asked, she said the mother had gone to the garden...” (Field worker, 150211_003)

Under these circumstances, mothers were therefore unable to quantify the amount of food their infants had consumed and often assumed they were full when they stopped crying or feeding.

In this challenging context for mothers, the support of their husbands appeared to be a crucial mitigating factor. Both the male and female respondents perceived the role of the father as being a provider even when he was not permanently based in the home, i.e. bringing in resources to support the mothers’ infant feeding and rearing strategies such as purchasing ingredients for enriching complementary feeds (sugar and butter for enriching “pap” (maize meal porridge)) or equipment for maintaining hygiene such as soap, as well as transport to access health care facilities for the infant and other family members.

“Fathers actually have a role to play, because fathers should stand in the care of both the mother and the child. Both the mother and father own the child. … father’s role is to buy soap for the mother, in order to launder and if he (infant) starts to eat, you can buy food stuff which will benefit him (infant) and give to the mother to give him. But for today if you want to leave him in the care of the mother only, “baa fele a,amang semboo soto” (the mother is not financially strong)”…. (Mother, control_MAL130J, Bajana)

In addition, mothers also thrived on a supportive marital relationship with their husbands and this seemed to enhance their ability to make decisions in line with the infant feeding and rearing strategies. Mothers were the primary decision makers on matters pertaining to infant feeding and rearing, including accessing preventative or minor illness health care services that were often accessible on foot. This is because, their husbands, apart from those who worked for the MRC or were government health care workers (HCWs), had limited knowledge of the infant feeding and rearing guidance and therefore relied heavily on the mothers to make these choices. However, when their infants were very sick and required admission to the MRC Keneba clinic, mothers sought the financial support of their husbands for transportation.

“Yes, if they involve themselves in, it will give better care to the children. When a mother doesn’t have something and you the father have it then you put it there. …it will also add the child care. But if the mother doesn’t have it and you (husband) too did not do it, caretaking will not take place. He (infant) will be left out to get sick.” (Mother, case_MAL019P, Kuli Kunda)

Although, the regular pastime for men in the West Kiang is to sit in the “bantaba” (a traditional meeting place for the men of the village), when not engaged in agriculture or male-led work, some husbands recognised the domestic pressures that the mothers were under, reflected on this and considered/admitted to offering practical assistance to their wives (mothers) so that the care of the infant was optimised, despite this being culturally alien. For them the aims were to support the mother but also their infants.

“Yes I (husband/father) helped the mother in that because the reason why I helped her, I like myself that is why. I know that she alone cannot do all that… If she is at one place, I can also be at the other place because to leave it with her alone, she is not a slave. We are all marriage partners. … I do not leave my wives. I always help them. And in terms of hygiene, I helped them there too. Because I clear my environment, I tell them “make this place and make the other place”. If you want the children’s health, you have to take care of your place. If you do not take care of them, there will be no health.” (Father, case_MAL008F, Jali)

Mothers also felt that the practical involvement of their husbands in the care of their infants enhanced their nutrition. Mothers mentioned support such as feeding, bathing and taking the infant to the clinic to enabled the mother to complete her tasks and therefore had time to give appropriate attention to the feeding of their infants.

“Yes they also have a role to play. That is because a child is closer to the wife that is why caretaking starts with the wife. But the husband also has a role. … The child’s needs should also be the role of the husband. The needs such as make him clean, and should feed him. Like in the case of feeds, whatever your earnings can allow you to afford, you should give to the child because that will improve his health.” (Mother, case_MAL015R, Jali)

However, when this support from their husbands was not forthcoming e.g. due to abandonment of mothers and their children because of marital tension or the cultural union of widows with a close male relative; the mothers relied more heavily on the support of the maternal grandmother/parents or peers. This support was less reliable as often these parties themselves relied on financial support from elsewhere. Under these circumstances, mothers found that this constrained their ability to adhere to the infant feeding recommendations, by limiting their options for enriching the popular complementary food pap. In desperation, some mothers opted to increase the breastmilk content in the diets of their over 6-month-old infants and gave less or no complementary food, which was presumably detrimental to the nutritional status of their infants.

“…I just noticed that when I asked him to do… he did not do it. When I ask from him once, twice and thrice, and he does not do it, I never bother myself to tell him again. When I go to my people if on a particular day they too do not have it, then I leave it. I do not cook for him on that day and keep it for him, I then breastfeed him.” (Mother, case_MAL019P, Kuli Kunda)

“You see a man; he may sit without digging toilet at home. He will not buy soap and will not buy anything. You will be responsible of putting everything in your food. You will buy soap and there is no toilet in the compound. That can bring difficulty.” (Mother, case_MAL041S, Nyorro Jattaba)

Within the homesteads paternal grandmothers were often elderly and mothers perceived their contribution to the care of infants as being mainly to carry them while the mothers undertook their domestic chores. Their role in infant feeding and rearing therefore appeared to be marginal with mothers opting to use the recommended infant feeding information from health care workers (HCWs) in their decision-making processes.

Mothers also highlighted the importance of peer support in encouraging them to adhere to the recommended infant feeding and rearing strategies. Within villages the mothers met regularly and shared their knowledge and experiences of infant feeding with peer counsellors who had been trained by the Gambia National Nutrition Agency (NaNA) to promote exclusive breastfeeding under 6 months.

“Yes, I spread the information among the women at our “bantabato” (meeting ground) to tell the women that we should all do it that way. ... I myself get up and ask them to give me my child to breastfeed.” (Mother, control_MAL256C, Jali)

Peer support also proved critical when a mother was bereaved and therefore not able to care for her infant and young children. In these situations, her peers would take on the care of the children until she was able to herself but duration of this support was variable and appeared to depend on a mother’s ability to maintain good social interactions with her peers.

Having a sickly infant who required several admissions to health care facilities was stressful for mothers. The admissions particularly when the infant was unwell were not conducive for health promotion as HCWs focused on treating and saving the infant’s life during which time the mothers were distressed and preoccupied with their infant’s illness. By contrast mothers whose infants were healthy could access and assimilate health promotion messages regularly.

The death of a child was a common experience for mothers. They found it more difficult to recover from the death of an older child who they had formed a relationship with, than from the death particularly of a newborn infant. Their experience of miscarriages was secretive and often not shared with their husbands. However, although distressing, mothers showed resilience by rationalising these losses and therefore did not feel that these bereavements affected their parenting capacity and therefore the growth and nutrition of their infants.

“When a human being dies, and now moves from you, it makes you sad. But it is God that created death. And also in terms of Islam, if you want to do it in another way, it diminishes your Islam. I was very used to him (deceased child) truly speaking when he was with me. When he was taken from me, I felt that someone was taken from me but I then held onto this that it is God who gave me and when He was giving to me I was not aware of it. Therefore, now if He is in need of him and I can take it to be God’s property.” (Mother, case_ MAL015R_Jali)

In this community, widows had limited autonomy in decision making for themselves or their children. In addition to grieving for the loss of their husband, the cultural practice that required them to remarry a male relative of the deceased (levirate marriage) sometimes exposed them and their children to adverse psychosocial circumstances as they were often remarried as the third or fourth wife to husbands who did not have the resources to support them adequately. This in turn limited a mother’s ability to care for her infant optimally.

Although the norm in this community was for mothers to breast feed children until 2 years of age, short birth spacing reduced the length of time a mother could feed her index infant as the belief in the population is that pregnant women should not breastfeed. Mothers found this stressful as it constrained their ability to adhere to the societal norms of breastfeeding whilst increasing their work load i.e. increasing number of young children to care for.

“… because you may have a child less than 2 years and you later have another one who is less than 1 year you then conceive another one. Now if you want to take care of them, one is just removed from breastfeeding, the other is lactating and pregnant with another. That caring becomes difficult.” (Mother, control_MAL138F, Jiffarong)