Participating in a Nutrition-Sensitive Agriculture Intervention Is Not Associated with Less Maternal Time for Care in a Rural Ghanaian District

Background: Nutrition-sensitive agriculture (NSA) interventions may increase farm-related work for mothers, with consequences for child nutrition. The Nutrition Links (NL) intervention provided mothers with poultry, gardening inputs, technical support, and education to improve livelihoods and child nutrition outcomes in rural Ghana. Objectives: Our objective was to compare time allocated to child care by a cross-section of mothers in the intervention group of the NL intervention with the control group (NCT01985243). Methods: A cross-section of NL mother-child pairs was included in a time allocation substudy [intervention (NL-I) n = 74 and control (NL-C) n = 69]. In-home observations of the mother-child pair were conducted for 1 min, every 5 min, for 6 h. Observations were categorized into 4 nonoverlapping binary variables as follows: 1) maternal direct care, 2) maternal supervisory care, 3) allocare, and 4) no direct supervision. Allocare was defined as care by another person in the presence or absence of the mother. Any care was defined as the observation of maternal direct care, maternal supervisory care, or allocare. Generalized linear mixed models with binomial data distribution were used to compare the child care categories by group, adjusting for known covariates. Results: Maternal direct care (OR = 1.07; 95% CI: 0.89, 1.28) and any care (OR = 1.56; 95% CI: 0.91, 2.67) did not differ by intervention group. However, there was a higher odds of allocare (OR = 1.36; 95% CI: 1.04, 1.79) in NL-I than in NL-C women. Conclusions: Maternal participation in an NSA intervention was not associated with a decrease in time spent directly on child care but was associated with an increase in care from other household and community members. The clinicaltrials.gov number provided is for the main NL intervention and not this current substudy. The study site, Upper Manya Krobo (UMK), is 1 of the 21 districts in the Eastern region of Ghana (12). This is a mostly rural agricultural district and has a total of 198 communities within 6 administrative subdistricts. Agricultural activities depend almost exclusively on 2 rainy seasons: early April to August, and September to October. Crop farming is the main livelihood in the majority of the communities. Farming is, however, at the subsistence level with limited use of mechanized agriculture technologies. The main food crops grown in UMK are cassava and maize. However, cowpea, mango, and other fruits and vegetables are also cultivated. The second most important livelihood in the district is trading. The district is a major commercial center for agricultural produce in the Eastern region, due to the presence of 3 large markets. Communities along the Volta Lake depend mainly on fishing as a source of livelihood. Basic infrastructure is generally inadequate. The road network is particularly poor, making transportation of people and market goods a major challenge. Access to potable water and electricity is limited to the more urban communities. The district is served by a hospital, maternal and child health clinics, and community-based health planning and services compounds. The Nutrition Links (NL) project commenced in 2013 as a partnership between McGill University, World Vision, the University of Ghana, and local nongovernmental (Heifer Ghana and Farm Radio International), governmental (Ghana Health Service, District Office of Agriculture, and National Commission for Civic Education), and private (Upper Manya Krobo Rural Bank) institutions. The design, setting, randomization, and primary outcomes of the NL intervention have been previously described (10). Briefly, it included a series of institutional and community-based activities including an integrated agriculture and nutrition education trial that was carried out sequentially in two 12-mo phases, ∼1 y apart, and involving mothers of infants and young children. This substudy involved mothers who were participating in the second phase of the trial. The first phase of infants and young children were aged 9.4 ± 3.9 mo, whereas the second were slightly older at 12.4 ± 6.3 mo. The intervention provided each woman in the second phase with 1) technical support and transfer of poultry husbandry (30 point-of-lay Swiss Brown chickens) and horticultural inputs (seeds; 5–10 kg sweet potato vines; tomato and green leafy vegetable seedlings) for home gardening; 2) weekly child nutrition and psychosocial stimulation education; and 3) community-wide health-related education (food demonstrations, mother-to-mother support groups on infant and young child feeding, and gender and diversity training). Mothers in the intervention also received continuous technical support with poultry farming and home gardens. The study participants were a cross-sectional sample of the NL trial phase 2 participants. The sampling procedure for the main NL intervention has been previously described (10). A census was first conducted in 3 subdistricts of the UMK district. A total of 89 communities that were organized into 16 clusters were assessed for eligibility for the NL intervention. To ensure the selection of a minimum of 14 households with infants or young children per cluster for participation in NL intervention activities, a total of 39 communities were selected in each of the 16 clusters. Eight clusters (19 communities) were allocated to the intervention group (NL-I), and 8 clusters (20 communities) were allocated to the control group (NL-C). Of the eligible households, 93 intervention and 91 control group mother-child pairs completed the baseline survey for the second phase and were eligible to participate in this present study. The flow of participants through the study is shown in Figure 1. Participant flow through the study. The 6-h direct observations of mother-child pairs were carried out between October and December 2016. The day of the week for the observations was randomly assigned to communities and excluded weekends. Field workers carried out an initial visit to inform mothers about the study and obtain informed consent. The mothers were then given information about the day of the visit for their respective communities and also informed to expect field assistants on that day of any of the coming weeks. Research assistants were standardized in their observations through a pretesting exercise. The in-home observations were carried out using focal person sampling (mother-child pair) involving 1-min sampling at 5-min intervals for a total of 6 h (13). Two research assistants were assigned to each mother-child pair; one observed the mother and the other the child with the aid of a hand-held watch. Each observation lasted 1 min; the remaining 4 min were used for recording observations. Every 50 min of observation was followed by 10 min of rest. The research assistants recorded observations on structured paper templates (Supplemental Appendices A and B). The research assistant observing the mother recorded all activities during the 1-min window, including whether the mother could see or hear her child, the location of the activity, and the persons present during the activity. Research assistants followed the mother everywhere she went (e.g., farm, market, riverside, clinic) to ensure that all activities were captured. Similar information was collected on the child, with critical attention paid to who was providing care to the child. This resulted in 2 sets of data: 1) maternal observations, and 2) child observations. The observations in both datasets were identical unless the mother and child were separated. Observations of the mother’s hygiene practice were carried out for 4 key activities (meal preparation, mother eating, feeding the child, and cleaning the child after defecation) when they happened during the 1-min observation window. Whenever one of these activities was recorded, field assistants noted whether the mother washed her hands with just water, soap and water, or not at all. The hygiene score was calculated by counting the number of times the mother washed her hands with soap and water before meal preparation, eating food, and feeding the child. The number of times the mother washed her hands with soap and water after she attended to the child after defecation was also counted. The total number of times handwashing with soap was observed with the 4 activities was then divided by the total number of times these activities were observed in the 6-h observation period. The Home Observation for Measurement of the Environment (HOME)—a tool associated with mental development—was used to assess the psychosocial stimulation of the child during the 6-h observation period (14, 15). This tool, adapted for use in low- and middle-income countries, has 45 simple binary response questions about the amount and quality of interactions in the home and the presence of learning and play materials available to the child. As recommended, the HOME was not assessed if the mother and child were not together for ≥45 consecutive minutes. The HOME score was the sum of positive responses out of the total of the 45-item questionnaire. In addition to the direct observation data, a wide range of household-, maternal-, and child-specific information was available through the NL baseline survey. Relevant baseline data for this substudy included household (ownership of assets and household size), maternal (depressive symptoms and anthropometry), and child (diet intake and anthropometry) information. The household asset index was calculated based on binary questions about the ownership of 13 household assets: floor material, wall material, cooking fuel, electricity, and ownership of a telephone, radio, television, video player, DVD/CD player, refrigerator, sewing machine, motorcycle, and car. The first component of the principal component analysis was then used as the wealth index (10). The 20-item Self Reporting Questionnaire (SRQ-20) was used to measure depressive symptoms in mothers (16). It included questions with binary responses on feelings of worthlessness, fatigue, difficulty concentrating, depressive moods, and other mental depressive symptoms that fall under “common mental disorders” (17). A depressive symptoms score was calculated as the sum of positive responses to the SRQ-20 questions (range 0–20). Validation studies of the SRQ-20 in low- and middle-income countries have demonstrated the scale’s internal consistency (Cronbach α = 0.84) and have suggested a cut-off point of 5–6 out of 20 to provide the best balance between specificity and sensitivity (18, 19). The child’s dietary diversity score was assessed with a binary scale list-based FFQ, which was adapted for the local context. The answers to the food frequency questions were recategorized into 7 food groups (grains, roots, and tubers; legumes and nuts; dairy products; flesh food; eggs; vitamin A–rich foods and vegetables; and other fruits and vegetables). The percentage of children who met the WHO’s recommended cut-off for minimum dietary diversity of 4 out of the 7 food groups was then calculated (20). Weight and height measurements for both the mother and child were taken in duplicate to the nearest 0.1 kg and 0.1 cm, respectively, using digital scales (Tanita Corp) and stadiometers (Shorr Productions), using recommended WHO standards (21). Child care was coded using the data from both the maternal and child observations. The coding reflected whether the child received care at any observation time point, and who provided the care, irrespective of any other activity happening at the same time. This allowed for a maximum of 61 one-minute care observations for each child. Five variables were created to describe child care: 1) maternal direct care (child care by the mother only), 2) maternal supervisory care (mother is not providing direct care but can see or hear the child), 3) allocare (child care by another person in the presence or absence of the mother), 4) any child care (maternal direct care, maternal supervisory care, or allocare), and 5) no supervision (child is left unattended to). Each child care variable was coded into a binomial variable (present/not present at each observation event), and the prevalence of each category was estimated. The covariates used in our analysis were child age, maternal age, maternal education, working status, maternal BMI, depressive symptoms score, household wealth index, and household size. These were selected based on literature and factors that could potentially influence the amount of care a child received in a household. The binomial child care outcomes [1) maternal direct care, 2) maternal supervisory care, 3) allocare, 4) any child care, and 5) no supervision] were used in separate generalized linear mixed models to compare the differences in child care by intervention group. Specifically, SAS PROC GLIMMIX (SAS Institute Inc) with the logit function was used while accounting for the random effect of cluster and predictor variables (child age, maternal age, maternal education, working status, maternal BMI, depressive symptoms, household wealth index, and household size). For predictor categorical variables with 3 levels (child age, maternal age, and education), the Dunnett test was used to adjust the P values for multiple comparisons (22). The institutional review boards of McGill University and the Noguchi Memorial Institute for Medical Research at the University of Ghana provided ethics approval for the trial. Informed consent was obtained from all mothers in the study. The trial was registered at clinicaltrials.gov ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT01985243″,”term_id”:”NCT01985243″}}NCT01985243).