Menu
Menu
Menu
Menu
Background: Inadequate food and water resources negatively affect child health and the efficiency of nutrition interventions. Methods: We used data from the SHINE trial to investigate the associations of food insecurity (FI) and water insecurity (WI) on mothers’ implementation and maintenance of minimum infant dietary diversity (MIDD). We conducted factor analysis to identify and score dimensions of FI (poor access, household shocks, low availability & quality), and WI (poor access, poor quality and low reliability). MIDD implementation (n = 636) was adequate if infants aged 12 months (M12) ate ≥ four food groups. MIDD maintenance (n = 624) was categorized into four mutually exclusive groups: A (unmet MIDD at both M12 and M18), B (unmet MIDD at M12 only), C (unmet MIDD at M18 only), and D (met MIDD at both M12 and M18). We used multivariable-adjusted binary logistic and multinomial regressions to determine likelihood of MIDD implementation, and of belonging to MIDD maintenance groups A-C (poor maintenance groups), compared to group D, respectively. Results: Low food availability & quality were negatively associated with implementation (OR = 0.81; 0.69, 0.97), and maintenance (ORB = 1.29; 1.07, 1.56). Poor water quality was positively associated with implementation (OR = 1.25; 1.08, 1.44), but inconsistently associated with maintenance, with higher odds of infants being in group C (OR = 1.39; 1.08, 1.79), and lower odds of being in group B (OR = 0.80; 0.66, 0.96). Conclusion: Food security should be prioritized for adequate implementation and maintenance of infant diets during complementary feeding. The inconsistent findings with water quality indicate the need for further research on WI and infant feeding.
The SHINE trial design and outcomes have been published previously [45]. Briefly, SHINE was a 4-arm cluster-randomized trial testing the independent and combined effects of infant and young child feeding (IYCF) and household water, sanitation, and hygiene (WASH) on child stunting and anemia. It was a community-based intervention implemented in rural areas of Shurugwi and Chirumanzu districts in Zimbabwe. The districts were divided into clusters, each defined as a catchment area serviced by 1–4 village health workers (VHW) of the Ministry of Health and Child Care. Between 22 November 2012 and 27 March 2015, pregnant women, 15–49 years old, who were permanent residents of those rural areas were enrolled. The infants born to these mothers were then followed over time to ascertain stunting prevalence. The analyses presented in this paper focus on the IYCF arm (n = 1148 live born infants). The IYCF intervention included six nutrition education modules delivered by VHW during 15 home visits to participating women. The intervention which promoted WHO recommended feeding practices, adapted to the local context, were delivered at monthly intervals starting at infant age 5 months (Table (Table1).1). From the 6-month home visit until the 18-month visit, a daily supply of 20 g of Nutributter® was provided to the caregiver to supplement the diet of the index infant. Additional detail on the IYCF protocol is available online. SHINE’s IYCF arm was chosen for our analyses for two reasons: 1) the children showed improvements in growth compared to the other SHINE arms [45], and 2) it allows the exploration of the effects of FI and WI on recommended feeding practices because nutrition education was targeted without direct intervention on food or water. Education modules in the Infant and Young Child Feeding (IYCF) Nutrition Intervention arm of the Sanitation Hygiene and Infant Nutrition Efficacy (SHINE) Trial Research nurses made home visits at multiple times to collect relevant information from mothers and infants: at baseline (during the pregnancy period) and at infant ages 1, 3, 6, 12 and 18 months. Since SHINE was household-based, the intermediate visits were conducted only when participants were available at the address where they consented. If the participants remained inaccessible after two attempts to reach out during the intermediate visits, the mother-infant dyad data were considered missing. At M18, participants were visited anywhere in Zimbabwe, even if they had moved from their initial residence. Our sample excluded infants who had died (n = 50), whose mothers rescinded consent (n = 2), and who were lost to follow-up (n = 33 at M12; and an additional n = 4 at M18). Twins were also excluded (n = 21 pairs) because diet was reported for only one of the infants, and it was not possible to distinguish which infant the data belonged to. Figure 1 illustrates the sample of included and excluded participants. Flowchart of eligible and included mother-infant dyads. aTwins were excluded since only one diet questionnaire was filled and it was not possible to determine which infant the information belonged to. bMissing covariates: unknown HIV−status (n=1), maternal age (n=20), maternal education (n=3), all other covariates (n=0). MHFI= Multidimensional Household Food Insecurity. MHWI= Multidimensional Household Water Insecurity Infant diet was assessed using the WHO infant diet assessment questionnaire, and the MIDD indicator was defined as infants who were fed at least four out of the following seven food groups: 1) grains, roots and tubers, 2) legumes and nuts, 3) dairy products, 4) flesh foods, 5) eggs, 6) vitamin-A rich fruits and vegetables, and 7) other fruits and vegetables [8]. During the post-partum period, when infants were aged M12 and M18, mothers were asked to recall what they fed their infants in the 24 h prior to the home visit interviews. The first main outcome, MIDD implementation, occurred if mothers reported feeding their infants a minimum of four food groups at M12 as described above. MIDD implementation was determined at M12 because the last nutrition education module was delivered at 9 months, with a reminder of all previous modules provided at M12 (Table (Table2).2). The second main outcome, MIDD maintenance, was categorized into four mutually exclusive groups based on the combined MIDD practices at M12 and M18 as follows: Group A = unmet MIDD at both M12 and M18, Group B = unmet MIDD at M12 only, Group C = unmet MIDD at M18 only, and Group D = met MIDD at both M12 and M18. Description of multidimensional household food insecurity and water insecurity variables Type of drinking source, non-drinking source (improved (piped, protected ground), unimproved ground (unprotected boreholes/ wells), surface) Drinking water satisfaction (3-level) MHFI Multidimensional Household Food Insecurity MHWI Multidimensional Household Water Insecurity The Multidimensional Household Food Insecurity (MHFI) and the Multidimensional Household Water Insecurity (MHWI) measures were used as primary FI and WI exposures, respectively [46]. These measures were developed specifically for the rural Zimbabwean households and their validities were tested to ensure robustness and usefulness. In brief, factor analyses were run on groups of food- and water-related variables. The process identified multiple dimensions representing the different aspects of each FI and WI. The resulting MHFI measure was characterized by 1) poor food access, 2) household shocks, and 3) low food availability & quality, whereas MHWI was characterized by 1) poor water access, 2) poor water quality, and 3) low water reliability. Each dimension of the MHFI and MHWI was made up of aggregated groups of variables as summarized in Table Table2.2. Standardized scores were obtained from post-estimation commands using the PCAmix package in R4.0.2 (R Foundation for Computational Statistics, Austria, Vienna). At baseline, a structured questionnaire was used to collect information on socio-demographic characteristics such as maternal age, maternal education, marital status, parity, religion, maternal employment outside the home and household size. Maternal depression, based on Edinburgh’s Postnatal Depression Scale and Mothering Self-Efficacy were collected using validated scales as described previously [47, 48]. The HIV status of women was determined using the rapid test algorithm; those who tested positive were directed to local clinics for follow-up and treatment. Socio-economic status (SES) was based on a wealth index created specifically for this population and reported in a prior publication [49]. Season at enrolment was characterized as calendar quarter during the baseline interview. Infant characteristics such as date of birth, sex, birthweight, and premature birth (gestational age < 37 weeks) were abstracted from health facility records by trained nurses. Our analyses excluded the mother-infant dyads who had incomplete information on MHFI (n = 95), MHWI (n = 148), MIDD implementation at M12 (n = 17), MIDD maintenance from M12 to M18 (n = 8) and the above-mentioned covariates (n = 24). We used descriptive statistics to summarize the characteristics of participants included in the analysis. We report medians and interquartile ranges (IQR) for the distributions of FI and WI dimensions; means and standard deviations (SD) for normally distributed variables; and frequencies and percentages for categorical variables. We fit binary logistic regression models to investigate the association of MIDD implementation at M12 (yes vs. no) with FI and WI. Since MIDD maintenance from M12 to M18 was a non-ordered categorical outcome (groups A-D), we used multinomial logistic regressions for the assessment of the relationship between MIDD maintenance and household-level FI and WI. All regression models utilized cluster-robust estimations to account for clustering of participants within study districts. All analyses were performed with all FI, and WI dimensions included simultaneously. To identify relevant covariates for inclusion in the models, three groups of variables were defined. Group 1 included only variables that were considered theoretically critical given the exposures and population: season at baseline interview (calendar quarter), wealth index (tercile), and household location (Chirumanzu vs. Shurugwi). Group 2 included variables that are commonly controlled for in this population and in nutrition behavior change interventions: maternal HIV status (positive vs. negative), maternal age (years), maternal education (some primary, some secondary, completed secondary), maternal religion (Apostolic, other Christian, other religion), and infant sex (male vs. female). Group 3 included mothering self-efficacy (scores: 1–5), maternal depression (scores: 0–30), household size, and parity (parous, nulliparous, missing). To select model covariates, we implemented backward elimination logistic regression by forcing retention of all Group 1 and Group 2 variables and setting retention of Group 3 variables at p < 0.2. MIDD outcomes were then modelled using all variables retained. However, since none of the identified Group 3 variables changed the measures of associations by ± 10% or more, the final models were the most parsimonious models with Group 1 and Group 2 variables only, determined by AIC and BIC. The main analyses described in this section were conducted in Stata IC v.16 (StataCorp LP, College Station, TX). The Medical Research Council of Zimbabwe (MRCZ) and the Institutional Review Board (IRB) of the Johns Hopkins Bloomberg School of Public Health reviewed and approved the SHINE trial protocol. Written informed consent was obtained from women in the local languages (Ndebele, Shona and English).
