Effects of household composition on infant feeding and mother–infant health in northern Kenya

Households with alloparents, individuals other than the mother who care for an infant, can shift members' roles and affect mother–infant health.


| INTRODUCTION
Infants require intensive care to thrive.Whereas a mother's role is centralized when considering their infant's needs, care is frequently provided by others as well.Alloparent care refers to childcare by an individual other than their mother (Hrdy, 2009;Waynforth, 2020;Wilson, 1975).Humans form 3% of alloparenting among the animal kingdom, where survival of young offspring and reproductive continuity is dependent on assistance from alloparent caregivers (Lukas & Clutton-Brock, 2012).With the aid of alloparents, the mother's responsibilities over herself and her child may change (Scelza & Hinde, 2019;Sharma & Kanani, 2006), influencing both maternal and infant health (Meehan et al., 2013;Waynforth, 2020).While the role of an alloparent can be taken on by anyone within the household, siblings, with their sustained presence in the household, are a likely source of alloparent care (Kramer et al., 2016).However, if the siblings are very young or require a lot of care themselves, they may serve an incompatible role, taking away resources from the other children in the house (Helfrecht & Meehan, 2016;Kramer et al., 2016).Additionally, food security levels can affect mother-infant health, since coping with food insecurity can entail compromised dietary quality (Hadley & Crooks, 2012).The health impacts of this are likely greater if limited foods are shared by more household members, reducing the allotted food for dyads.The combination of food insecurity and sibling/non-sibling alloparents in the household may affect dyads' diet and nutrition differently than when either is present alone.This study investigated whether infant feeding practices and maternal-infant health outcomes differ by household composition and food insecurity among Ariaal mother-infant dyads in agropastoral communities of northern Kenya.
There is evidence that alloparents contribute to less time spent breastfeeding as well as earlier weaning (Deang et al., 1988;Meehan & Roulette, 2013;Quinlan & Quinlan, 2008), although contrary evidence also exists where alloparents can contribute positively to the practice of breastfeeding and delay introduction to complementary foods (Negin et al., 2016;Scelza & Hinde, 2019).Some studies further link alloparenting to beneficial effects, including reduced maternal energy expenditure (Meehan et al., 2013) and lowered risk of infant infections (Waynforth, 2020).The mixed results in the effect of alloparents on infant feeding may be attributable to differences in subsistence economies and local ecologies between these populations (e.g., African foragers versus a UK cohort).They may also be due in part to who alloparents are.The effects of allocare by knowledgeable and experienced adults (e.g., grandmothers) may be different from those by less experienced individuals such as infant's siblings (Deang et al., 1988;Sear et al., 2000;V azquez-V azquez et al., 2021;V azquez-V azquez et al., 2022).While the existing literature is limited, studies have suggested that siblings can act as alloparents, aiding with infant care unless they are too young themselves to care for others (Helfrecht & Meehan, 2016;Kramer et al., 2016).Meanwhile, food insecure households have been associated with lower odds of exclusive breastfeeding (Orr et al., 2018) and lower likelihoods of meeting age-appropriate infant feeding practices (Macharia et al., 2018).
This study evaluated four sets of hypotheses summarized in Table 1.We hypothesized that households consisting of mother-infant dyads and other individuals (potential alloparents) would associate with lower breastfeeding frequency (Hypothesis 1a) and earlier introduction of complementary foods (Hypothesis 2a) compared to households without alloparents, similar to previous studies among subsistence-oriented populations in Africa (Meehan & Roulette, 2013;Vankayalapati & Fujita, 2022).Our reasoning was that under alloparental care, infants in such populations would have less time with the mother, as mothers may leave the infant's side to engage in household chores or other subsistence labor and, therefore, less access to consumption of mothers' milk (in the absence of stored milk).This would reduce the frequency and exclusivity of breastfeeding and increase reliance on alternative foods.We further hypothesized that the effects of siblings on infant feeding practices would be similar to those of non-sibling alloparents in direction but with lesser magnitudes (Hypothesis 1b and 2b), considering the generally higher dependability and quality of allocare provided by non-sibling (more often adults) than sibling (more often children) alloparents.
However, if such households are under heightened food insecurity, alternative foods may be unavailable, curtailing the shift away from exclusive breastfeeding.In this case, breastfeeding frequency might hold or even increase with the presence of alloparents (Hypothesis 1c) and complementary feeding would be less likely (Hypothesis 2c).
Additionally, we hypothesized that households with potential alloparents would associate with lower risk for infection among both infants (Hypothesis 3a) and mothers (Hypothesis 4a) with the reasoning that alloparent support would alleviate the mother's childcare workload and stress (Scelza & Hinde, 2019;Vankayalapati & Fujita, 2022), which would, in turn, promote dyads' health.We hypothesized that siblings would have less protective effects than non-sibling alloparents (Hypothesis 3b and 4b) because sibling alloparents may not be as vigilant with sanitation as adults.As for heightened food insecurity, we expected to find higher risk for infection in both infants (Hypothesis 3c) and mothers (Hypothesis 4c), due to potentially worse nutrition and overall health compared to households not facing food insecurity (Macharia et al., 2018;Ntshebe et al., 2019).

| Study population
A secondary analysis was conducted using data from a cross-sectional survey of breastfeeding mother-infant dyads in the Ariaal agropastoral communities of northern Kenya (Fujita et al., 2022).This population relies heavily on the rearing of mixed-species livestock and small-scale dryland farming for subsistence and cash cropping (Fratkin, 2019).They are formerly nomadic Ariaal camel and cattle herders who have settled on the eastern slopes of Mount Marsabit after becoming destitute by reduction in access to traditional grazing land and diminished herd size and health due to major droughts of 1970s and 1980s (Fratkin, 2019;Paredes Ruvalcaba et al., 2020;Shisanya, 1990).
The Ariaal society is organized by the patrilineal kinship system and follow patrilocal residence rule; motherinfant dyads generally live in the husbands' natal community (Fratkin, 2019).Women may briefly return to their natal community for childbirth, but this is limited to a relatively short duration (weeks).Many aspects of Ariaal peoples' lives are dictated by one's gender.They have distinct gender roles, with the women taking on almost all of the household responsibility, including childcare, food procurement and preparation, rearing of small livestock (goats and sheep), caring for the sick animals, milking cows and goats (Fratkin, 2019), collecting water and firewood, and selling livestock products in markets (the latter two often involving hours of walking) (Fratkin & Smith, 1995).Unmarried men herd cattle in remote camps referred a "fora" and married men engage in village and tribal decision making (Fratkin, 2019).Similarly, food norms differ for women and men: women's foods consist of animal milk and milk products and men's foods consist of animal meat and blood.Women (with her children) generally consume their meals separately from men (Fratkin, 2019).
While Ariaal cultural norm is that postpartum mothers stay home to take care of her infant and avoid heavy work outside the home in the initial months, in practice, mothers may return to strenuous work (e.g., collecting firewood) much earlier, especially if there is no help from others (Miller, 2011).Ariaal mothers generally breastfeed infants for two to 3 years.If the mother is at home, infants are breastfed "on demand" night and day during the initial months.However, if the mother needs to perform tasks outside the home, she may leave the infant asleep or under the care of relatives or friends and return home as quick as possible to feed (Miller, 2011).As postpartum time advances, most mothers begin complementary feeding at about 6 months, initially with cow milk and then potatoes and beans that have been cooked and mashed (Miller, 2011).Ecological and sociocultural determinants of the health and morbidity among infants, children and mothers of the study area have been described (Fratkin et al., 2004;Fujita et al., 2004;Miller, 2011;Nathan et al., 1996).

Outcome Hypothesis Prediction
Breastfeeding frequency 1a Non-sibling alloparents will associate with lower breastfeeding frequency.
1b Sibling alloparens will associate with lower breastfeeding frequency, but to a lesser extent than non-sibling alloparents. 1c In the presence of food insecurity, alloparents will associate with higher breastfeeding frequency.

Complementary feeding 2a
Non-sibling alloparents will associate with earlier introduction to complementary feeding. 2b Sibling alloparens will associate with earlier introduction to complementary feeding, but to a lesser extent than non-sibling alloparents.2c In the presence of food insecurity, alloparents will associate with later introduction to complementary feeding.

Infant infection 3a
Non-sibling alloparents will associate with lower risk of infant infection.
3b Sibling alloparents will associate with lower risk of infant infection, but to a lesser extent than nonsibling alloparents.
3c Elevated food insecurity will associate with higher risk of infant infection.

Maternal infection 4a
Non-sibling alloparents will associate with lower risk of maternal infection.
4b Sibling alloparents will associate with lower risk of maternal infection, but to a lesser extent than non-sibling alloparents.
4c Elevated food insecurity will associate with higher risk of maternal infection.
At the time of the survey, the 2006 Horn-of-Africa drought was underway, exhausting livestock grazing and upkeep of crops (in limited higher-elevation areas with arable land) and skyrocketing food prices.This had adverse effects on households' food security although these effects were more severe for poorer families whose livelihoods relied heavily on climate sensitive livestock and farm products than other families who were more integrated in sectors less directly affected by droughts (e.g., husbands earning wages in cities; Fujita, 2008).

| Data collection
The present analysis focused on the convenience subsample dyads having complete data for household composition, food security, and infant feeding practices (n 208 of original 241), and a convenience subset (n 83) having additional morbidity recall data.The original survey (Fujita, 2008) was approved by the institutional review boards of the University of Washington and Kenya Medical Research Institute.The present study used deidentified data requiring no further approvals.
Household composition was coded using two dummy variables (0/1-absent/present): one for the presence of a non-sibling individual, and the other for the presence of siblings.We preferred dichotomous variables over continuous variables because the distribution of the number of non-sibling individuals in each household was highly skewed, with very few households having more than one.
The food insecurity score was based on the accessibility of 10 different nutrient-dense foods (higher scores representing heightened insecurity).The score was based on an interview with the mothers.They were asked to name the barriers, if any, to consuming milk, eggs, meats, organ meats, tallow, Kimbo (vitamin A-fortified cooking fat), sukuma (a dark-green leafy vegetable), yams, carrots, oranges, mangoes, and papaya.The mother scored one point on the scale for every food item they reported with financial or affordability barriers.In the context of the ongoing severe drought and decimated agropastoral productivity, access to these food items was largely dependent on purchase from the market.
For infant feeding data, mothers were asked to report the frequency with which they nursed their infant during the preceding 24 h.They further indicated whether the reported frequency was typical for their usual routine.The typical breastfeeding frequency (nursing bouts) was used in this study (natural log transformed values were used for regression modeling; see below).Dyads were classified as complementary feeding (0/1 = no/yes) if the mother answered positively to the question whether they feed the baby any foods or liquids other than mothers' milk, and the mothers listed food items in the subsequent interview regarding complementary feeding.
Infection (0/1 = no/yes) was based on a 10-day morbidity recall interview, administered by a local nurse.Maternal and infant infection was indicated if there was a report of the following symptoms.For mothers: cough, headache, chest pain, runny or stuffy nose, abdominal pain, fever, and diarrhea were assessed.For infants: cough, fever, diarrhea, rapid respiration, chest/breathing problems, conjunctivitis, ear symptoms, stomatitis, and scabies were assessed.

| Data analysis
Stata version 15 was used for statistical computation.The alpha-level was set at 0.05.To evaluate our hypotheses, we utilized robust regression models for breastfeeding frequency (Hypothesis 1a-c) and logistic regression models for complementary feeding (Hypotheses 2a-c) and infection (Hypotheses 3a-c and 4a-c), with household composition variables as the primary predictors and food insecurity as the secondary predictor.Interaction between food insecurity score and each household composition variable on the infant feeding outcomes was evaluated by adding an interaction term at a time.When the interaction was apparent, we conducted a joint F-test to evaluate the joint significance of the main-effect and the interaction.We did not evaluate these interactions on dyads' risk for infection because our sample size for these outcomes was too limited and because existing literature is inadequate to meaningfully hypothesize the nature of such interactions.To visualize the characteristics of regression models, we produced forest plots for coefficients and margins plots to assess the nature of interaction.All final models were adjusted for infant age, infant sex, and maternal age.There was no evidence of variance inflation.

| RESULTS
The sample characteristics are summarized in Table 2.The mean (±SD) age of mothers (n 208) was 28 ± 7 years, ranging from 18 to 48.The mean age of infants was 8 ± 4 months, ranging from 1 to 20.The percentage of households with non-sibling alloparents was 85.1%, and the percentage of households with sibling alloparents was 76.0%.The mean food insecurity score of the households was 3 ± 1.8, ranging from 0 to 10.The mean breastfeeding frequency was 9 ± 4 bouts per 24 h, and 62.5% of the dyads were complementary feeding.For the subsample dyads (n 83) who participated in the morbidity recall, 27.7% of mothers and 63.9% of infants had infection symptoms.The subsample dyads characteristics did not differ from the overall dyads in maternal age, parity, and the proportion of dyads living with non-sibling alloparents or siblings.However, the subsample dyads had a higher proportion of male infants (61.5 vs. 50%, p = .037)and older average infant age (9.6 vs. 8.0 months, p = .002)with fewer breastfeeding frequency (7.4 vs. 9.1 bouts, p < .001)and was more likely to be complementary feeding (74.7 vs. 62.5%, p = .022),compared to the overall dyads.They also had a slightly lower food insecurity score (2.7 vs. 3.0, p = .004).
The regression models are summarized in Table 3. Household type alone was unassociated with breastfeeding frequency (Table 3, Model 1A, Figure 1A).However, there was interaction between the presence of non-sibling alloparents and food insecurity (Coeff.0.14; 95%CI 0.05-0.23,β .68,Joint F test p = .003,Table 3, Model 1B): among dyads of non-sibling alloparent households, breastfeeding frequency increased as food insecurity intensified, while it did not among their counterparts without alloparents (Figure 1B).The presence of siblings was unassociated with breastfeeding frequency.Households with non-sibling alloparents were associated with reduced odds of complementary feeding (Coeff.À1.98; 95% CI À3.38 to À0.58, Table 3, Model 2A, Figure 1C).There was no interaction between non-sibling alloparents and food insecurity on complementary feeding (Figure 1D).The presence of siblings was unassociated with complementary feeding.

| DISCUSSION
The test of hypotheses revealed an intriguing mix of associations and lack thereof, supporting some of our predictions but not others.Nonetheless, on balance the data support that infant feeding practices and motherinfant dyads' infection rates differ by household composition and food insecurity in this population, with the presence of potential alloparents and who the alloparents are-siblings or non-siblinghaving different and complex effects on these outcomes.For infections, households with siblings were associated with a lower risk for mothers and infants.For complementary feeding, it was households with non-sibling alloparents that had an association with lower odds.Likewise, it was among these households that we found breastfeeding frequency increased as a function of household food insecurity.In this section, we discuss insights from these complex results drawing upon the existing literature.

| Breastfeeding frequency
Our results did not support our expectation that breastfeeding frequency would be lower in alloparent-available households (Hypothesis 1a and 1b, Table 1).We found no overall association between non-sibling alloparents and breastfeeding, contrary to other studies in which breastfeeding was influenced by alloparents, specifically maternal grandmothers (Negin et al., 2016;Scelza & Hinde, 2019;Sharma & Kanani, 2006).It is possible that the lack of association in our study is due in part to the patrilocal residence rule Ariaal families follow; motherinfant dyads typically live with or near the husband's blood relatives, rarely including maternal grandmother (Fratkin, 2019).In this context, alloparents' impact on breastfeeding practices may be different from those found in populations where maternal grandmothers are in closer proximity to dyads.However, our findings supported our expectation that the alloparent effect would depend on household food insecurity level (Hypothesis 1c, Table 1).Namely, dyads in non-sibling alloparent-available households had breastfeeding frequency that increased as food insecurity intensified.While the existing literature addresses the impact of alloparenting and food insecurity on breastfeeding individually (Macharia et al., 2018;Negin et al., 2016;Scelza & Hinde, 2019;Sharma & Kanani, 2006), these are rarely investigated simultaneously.By studying the interactive effects of food insecurity and alloparents, we found that food insecurity could positively moderate the effect of alloparents on breastfeeding frequency.It is unclear what underlies this moderation, but the co-occurrence of non-sibling alloparents and heightened food insecurity might somehow alter the household's social dynamics to increase mother-infant face time (Sharma & Kanani, 2006) or facilitate maternal opportunity to learn how to breastfeed better from experienced alloparents such as grandmothers (Scelza & Hinde, 2019).This moderation could also arise if intensified food insecurity culminates in diminished infant condition, to which mothers respond by increasing breastfeeding effort (Tracer, 2009).For T A B L E 3 Regression models for infant feeding practices by non-sibling and sibling alloparent availability alone (A) and their interaction with food insecurity score (B/C) adjusted for maternal age and infant age and sex.instance, households facing food insecurity may experience less dietary diversity and diminished dietary quality (Macharia et al., 2018).This might lead to diminished infant nutrition and health, which might be exacerbated by the presence of alloparents, who not only consume a portion of the little food available in the household, leaving less for the infant and mother, but also might introduce an additional risk of infection (Waynforth, 2020).Under such a scenario, mothers might increase breastfeeding frequency to mitigate these issues.

| Complementary feeding
Exclusive breastfeeding was more prevalent in households with potential non-sibling alloparents than those without, echoing existing research associating the presence of supportive grandmothers with a lower likelihood of complementary feeding (Meehan & Roulette, 2013;Negin et al., 2016;Sharma & Kanani, 2006).While this contradicts our expectation that alloparenting promotes complementary feeding (Hypothesis 2a and 2b, Table 1), it partially supports our expectation that alloparenting under food insecurity would delay introduction to complementary feeding (Hypothesis 2c, Table 1).

| Recent infections
Households with non-sibling alloparents were unassociated with maternal and infant infections, refuting our prediction (Hypothesis 3a and 4a, Table 1).One prior study found that alloparents, specifically grandmothers, may reduce mothers' workload (Meehan et al., 2013), and another found that infants living with fewer adults experienced less infection symptoms, like diarrhea (Ntshebe et al., 2019).It is possible that, in our study, we did not detect any effects on infection symptoms because the benefits of alloparents, such as increased assistance and stress alleviation, were offset by increased pathogen exposure (Waynforth, 2020) or reduced food resources due to more household members, who must be fed.
It is curious that the presence of siblings predicted lower likelihoods of both infant and maternal infection rates, while non-sibling alloparents did not.This protective effect of siblings nonetheless is partially consistent with our hypothesis that siblings' association with reduced dyadic infection risk (Hypothesis 3b and 4b, Table 1) and other studies (Helfrecht & Meehan, 2016;Kramer et al., 2016) reporting that while young siblings (e.g., toddlers) might serve as resource competitors that negatively impact other siblings' nutrition status, older siblings either cooperate to positively impact other siblings' nutrition status (Kramer et al., 2016) or have an insignificant (but not adverse) impact (Helfrecht & Meehan, 2016).Although we could not examine sibling age in our models because the data were incomplete (many mothers were unclear of siblings' numeric age), we have some evidence that the siblings in this study might have been generally old enough and capable of making a positive contribution to the dyads' health and wellness rather than contradicting them.For the subset of dyads with complete sibling age data (n 177), the mean sibling age was 7 years, ranging from 1 to 25.With older siblings present in the household, it is possible they take on alloparent behaviors, like helping to feed or care for the infant, positively contributing to the dyads' nutrition and health.Additionally, similar to non-sibling alloparents, older siblings may take on household chores, freeing the mothers from heavy chore burden and possibly increasing the time available for the mother to directly care for her infant, as observed elsewhere (Helfrecht & Meehan, 2016;Meehan & Roulette, 2013).
However, it is unclear why this protective effect was unique to sibling alloparents.It may be the case that siblings introduce less additional pathogen exposure compared to non-sibling alloparents (who may have broader ranges of social interaction and therefore more pathogen exposure).Similarly, due to many sibling alloparents being children, they may consume less food, taking away less of the household resources than non-sibling alloparents.In this way, sibling alloparents may contribute the benefits of stress alleviation as discussed above without the associated risks.It is also possible that the observed effect is due to the likelihood of healthier mothers having healthier and a greater number of offspring (Hill, 1993).This phenotypic correlation (Hill, 1993) could lead to healthier infants having more siblings/sibling alloparents.In this case, these lower infection rates could be attributed to better overall health (e.g., through shared genetic predisposition) than to the direct involvement of siblings in the household.
Additionally, male infants were associated with reduced rates of maternal infection, alluding to how the presence of strong gender roles within this population may affect the health of the mother (Fratkin, 2019).The Ariaal society's patrilineal descent and inheritance systems may be in part attributable to the observed relationship between male infant sex and maternal health.Among the Ariaal and many Kenyan pastoralists, a male heir (usually the eldest son) inherits the family's livestock.Women without a male heir lose the right to livestock and can become impoverished and isolated (Fratkin & Smith, 1995).Having a male infant and the prospect for enduring access to livestock may therefore increase women's prestige and social capital (Straight et al., 2022).This in turn may increase their autonomy in health-related decision making and therefore better maternal health outcomes (Brunson et al., 2009;Caldwell & Caldwell, 1993).Further research on household composition and maternal infection should pay attention to the role of infant sex in a given cultural context.

| Limitations
This study suffers some limitations inherent in a secondary analysis of archival data.Most notably, we did not have observational data on alloparenting behaviors.Rather, we compared households with and without potential alloparents to address the alloparenting effect.
We do not have information on who these individuals were nor their role within the home.Nonetheless, it is reasonable to assume that non-dyad individuals in the household will engage in alloparenting, given strong ethos of cooperation within Ariaal households (Fratkin, 2019).The proportion of households without non-sibling alloparents was small (15%).The food insecurity score capturing the accessibility to specific foods utilized here, may have underestimated the true extent of food insecurity during the major drought (food unavailability).Similarly, water insecurity, which we could not quantify, might have been an important determinant of infant feeding and dyads' health.The morbidity recall data were from a convenience subsample with limited sample size and therefore statistical power, and over-representing male and older infants compared to the overall sample, potentially reflecting self-selected mothers who were more concerned with or attuned to the health of their infants.These issues may hinder the generalizability of our findings.

| CONCLUSION
Studies have been inconclusive in determining whether and how the presence of alloparents affects infant feeding practices or mother-infant health.This study incorporated food insecurity, a variable largely neglected in this literature, in nuanced hypotheses to fill these knowledge gaps.Among the Ariaal agropastoralists, mothers in households with non-sibling alloparents breastfed more frequently as food insecurity intensified.Additionally, households with non-sibling alloparents delayed infants' introduction to complementary foods.Households with siblings were protective against both infant and maternal infection.Through investigating the interaction between household composition and food insecurity, this study characterized a complex relationship indicative of the context-dependent nature of the effects of alloparenting on infant feeding.Further research should enroll mother-infant dyads in diverse cultural settings, monitoring their health prospectively to better understand the role of alloparenting and its possible interaction with household food security.

F
I G U R E 1 Household composition and food insecurity as predictors for infant feeding practices.(A,B) Breastfeeding frequency.(A) Regression coefficients for variables in Model 1B for ln(breastfeeding frequency).(B) Predictive margins of non-sibling alloparents on ln(breastfeeding frequency) as a function of food insecurity score per Model 1B.There was interaction between non-sibling alloparents and food insecurity score.Breastfeeding frequency increased as food insecurity score increased only among household with non-sibling alloparents.Translating this in the scale of nursing bouts, dyads living with non-sibling alloparents having food insecurity scores <3 and >5, respectively have the mean predicted bouts 7.4 (SD 0.8, n 63) and 12.8 (SD 1.7, n 14).These values for their counterparts living without nonsibling alloparents are 5.9 (SD 0.1, n 3) and 7.9 (SD 0.95, n 12).(C,D) Complementary feeding.(C) Regression coefficient for variables in Model 2A for the probability of complementary feeding.(D) Predictive margins of non-sibling alloparents on the probability of complementary feeding as a function of food insecurity score per Model 2A.Households with non-sibling alloparents were associated with reduced probability of complementary feeding.Unlike breastfeeding frequency, there was no interaction between household type and food insecurity.In panels A and C, the circles are point estimates and bars are 95% CIs.The vertical dashed lines represent the coefficient zero therefore no association.(B,D) The circles are marginal effects and bars are 95% CIs.

F
I G U R E 2 Household composition and food insecurity as predictors for infant and maternal infection.(A) Infant infection: The presence of siblings was associated with lower probability of infant infection, and food insecurity was associated with elevated probability of infant infection.(B) Maternal infection: The presence of siblings was associated with lower probability of maternal infection while raising male infant was associated with lower probability of maternal infection.The vertical dashed lines represent the coefficient zero therefore no association.
Sample characteristics.Note: p values are for the differences between the subsample and the overall sample per the one-sample t test or the one-sample proportion test.Bold values signify p < .05.
T A B L E 2 p Mean (SD) or n (%) Mean (SD) or n (%) Regression models for dyad health outcomes by nonsibling and sibling alloparent availability.