An agriculture–nutrition intervention improved children’s diet and growth in a randomized trial in Ghana

listen audio

Study Justification:
– Stunting in Ghana is a significant issue, particularly in rural communities with high poverty rates and low education levels.
– Integrated agricultural interventions have the potential to address this problem by improving access to nutritious foods.
– This study aimed to test the effectiveness of a 12-month intervention that included inputs and training for poultry farming and home gardening, as well as nutrition and health education, on child diet and nutritional status.
Highlights:
– The study used a cluster randomized controlled trial design, with 16 clusters randomly assigned to either the intervention or control group.
– The intervention group showed significant improvements compared to the control group in terms of minimum diet diversity and length-for-age and weight-for-age z-scores.
– Sensitivity analyses confirmed the robustness of the findings.
Recommendations for a Lay Reader:
– Integrated interventions that increase access to high-quality foods and provide nutrition education can improve child nutrition.
– Poultry farming and home gardening can be effective strategies to address stunting in rural communities.
– Policy makers should consider implementing similar interventions to improve child nutrition and reduce stunting rates.
Recommendations for a Policy Maker:
– Allocate resources to implement integrated agricultural interventions that focus on improving access to nutritious foods and providing nutrition education.
– Support training programs for poultry farming and home gardening in rural communities.
– Collaborate with relevant stakeholders, such as government agencies, private sector service providers, and community organizations, to implement and sustain these interventions.
Key Role Players:
– Government agencies responsible for agriculture, health, education, and finance.
– Private sector service providers in the health, education, agriculture, governance, and finance sectors.
– Community organizations and leaders.
– Researchers and experts in nutrition and agriculture.
Cost Items to Include in Planning the Recommendations:
– Training programs for poultry farming and home gardening.
– Provision of agricultural inputs, such as seeds, tools, and equipment.
– Nutrition and health education materials.
– Monitoring and evaluation activities.
– Staff salaries and travel expenses.
– Community engagement and mobilization efforts.
– Infrastructure development, such as chicken coops and home garden plots.
Please note that the cost items provided are general suggestions and may vary depending on the specific context and implementation strategy.

The strength of evidence for this abstract is 8 out of 10.
The evidence in the abstract is strong, but there are some areas for improvement. The study design is a cluster randomized controlled trial, which is a robust design for evaluating interventions. The sample size is adequate, and the study includes intention-to-treat analyses and sensitivity analyses to assess the robustness of the findings. The study also provides specific effect sizes and confidence intervals for the outcomes of interest. However, the abstract could be improved by providing more details on the intervention components and the specific results of the sensitivity analyses. Additionally, it would be helpful to include information on potential limitations of the study and recommendations for future research.

Stunting in Ghana is associated with rural communities, poverty, and low education; integrated agricultural interventions can address the problem. This cluster randomized controlled trial tested the effect of a 12-month intervention (inputs and training for poultry farming and home gardening, and nutrition and health education) on child diet and nutritional status. Sixteen clusters were identified and randomly assigned to intervention or control; communities within clusters were randomly chosen, and all interested, eligible mother–child pairs were enrolled (intervention: 8 clusters, 19 communities, and 287 households; control: 8 clusters, 20 communities, and 213 households). Intention-to-treat analyses were used to estimate the effect of the intervention on endline minimum diet diversity (≥4 food groups), consumption of eggs, and length-for-age (LAZ)/height-for-age (HAZ), weight-for-age (WAZ), and weight-for-length (WLZ)/weight-for-height (WHZ) z-scores; standard errors were corrected for clustering. Children were 10.5 ± 5.2 months (range: 0–32) at baseline and 29.8 ± 5.4 months (range: 13–48) at endline. Compared with children in the control group, children in the intervention group met minimum diet diversity (adjusted odds ratio = 1.65, 95% CI [1.02, 2.69]) and a higher LAZ/HAZ (β = 0.22, 95% CI [0.09, 0.34]) and WAZ (β = 0.15, 95% CI [0.00, 0.30]). Sensitivity analyses with random-effects and mixed-effects models and as-treated analysis were consistent with the findings. There was no group difference in WLZ/WHZ. Integrated interventions that increase access to high-quality foods and nutrition education improve child nutrition.

This study was a cluster randomized controlled community trial carried out within the context of a 5‐year capacity‐building and research programme (Nutrition Links [NL]) in the Upper Manya Krobo District (UMKD) of the Eastern Region of Ghana. The NL programme provided training on nutrition, gender and diversity, data management and analysis, and evidence‐based decision making to government and private sector service providers in the health, education, agriculture, governance, and finance sectors of the district. The NL team stratified the six UMKD subdistricts by population size and randomly selected three subdistricts to serve as the study site for this trial. In the three selected subdistricts, we completed a census of communities (n = 86) with GPS location of all households. Three additional communities were included in the study site (total n = 89) because they received services from the Ghana Health Service subdistrict personnel even though they were slightly outside the subdistricts’ political boundaries. Based on census data generated, the 89 communities were then organized geographically into 16 clusters. Our aim was to have at least 14 households with infants/young children in each cluster, that is, the minimum target for group activities for the intervention. The clusters consisted of either one distinct community or multiple adjacent small communities (range of 2–10). Within each cluster, we randomly chose communities until we reached a minimum of our target number of eligible households per cluster. A total of 39 communities were selected (range: 1–6 communities/cluster) as the study area. The 16 clusters were randomly assigned to treatment group (sequential, using random numbers). The eight intervention clusters had 19 communities (range: 1–6), and the eight control clusters had 20 communities (range: 1–4; Figure 1). Given the nature of the intervention, it was not possible to mask the treatment assignment; therefore, the project maintained separate field staff for the implementation of the intervention and survey data collection. The clusters were geographically distant enough from each other to avoid direct contamination—that is, no control community participants received inputs or took part in educational activities planned for intervention participants. Flow of participants through the agriculture–nutrition cluster randomized controlled trial in Upper Manya Krobo District, Ghana Given limited human and financial resources, enrolment and intervention implementation were carried out in two phases. In Phase 1 (2014–2015), all women with infants (0–12 months) who lived in the selected communities and who planned to remain in the community for the duration of the project were invited to enrol in the trial. In Phase 2 (2016–2017), the age range was expanded to target young children <18 months to include the planned sample size. For both phases, additional eligibility criteria for the intervention participants included the timely preparation of (a) a chicken coop that met project specifications and (b) a fenced home garden plot. Although the trial was directed to women, the project staff encouraged the woman's household and community to support the activities. All eligible households (n = 277) in the selected communities of the eight intervention clusters were invited to enrol in Phase 1 in 2014 (Figure 1). After the end of the first phase and the completion of the 12 months of trial activities, we identified newly eligible households (n = 95) from the same communities and invited them to enrol in Phase 2 in 2016. Two of the intervention clusters had no newly eligible households, so only six intervention clusters were active in Phase 2. A total of 34 eligible households were not enrolled, and 51 were enrolled, but baseline data were lost due to a malfunction of the electronic tablets. We considered it untenable to enrol participants a second time from control cluster communities that had received no benefit. Thus, the order of including the control clusters was randomly assigned. To mimic the intervention enrolment, five control clusters were used in Phase 1 (135 eligible households) and three control clusters (114 eligible households) in Phase 2. Among the control clusters, 36 households were not enrolled. Ethics approval for the trial was obtained from the institutional review boards of McGill University (# 822‐0514) and the Noguchi Memorial Institute for Medical Research at the University of Ghana (#060/13‐14). All participants provided written informed consent for themselves and their children. The trial was registered at Clinicaltrials.gov ({"type":"clinical-trial","attrs":{"text":"NCT01985243","term_id":"NCT01985243"}}NCT01985243). The 12‐month intervention was an integrated package of agricultural inputs and training as well as education in nutrition, health care, and child stimulation for participants. Beekeeping was introduced for interested households only in Phase 1 for honey harvesting after the end of the trial. The relevant intervention components are described below in more detail. Household and maternal sociodemographic data (e.g., maternal ethnicity and education) had been collected through the NL district‐wide baseline survey (November 2013–June 2014) and were incorporated into the data set for this analysis. The intervention‐specific data were collected using electronic tablets through baseline and endline surveys completed during the months before and after each phase of the trial. Household data included characteristics such as family composition and demographics, household assets, water and sanitation facilities, agricultural practices including raising of poultry, use of district services, and food insecurity. Household food insecurity was measured with the 15‐item Latin American and Caribbean Food Security Scale (Food and Agriculture Organization, 2012). Maternal‐ and child‐specific information included diet, anthropometric measurements, haemoglobin concentration, health behaviours, and symptoms of physical and mental (mother only) health. Weight was measured to the nearest 100 g with a digital scale (Tanita Corporation of America, Inc., Arlington Heights, IL, USA) and length/height to the nearest 0.1 cm with a stadiometer (Shorr Productions, Olney, MD, USA). All measurements were done using standard procedures, and weight and length/height measurements were taken in duplicate. A third measurement was taken if the discrepancy was above the World Health Organization (WHO) cut‐off for acceptable difference in repeated measurements (WHO Multicentre Growth Reference Study Group, 2006). The sample size was calculated with an α = 0.05, power = 0.80, effect size d = 0.35, and variance inflation factor = 1.79, resulting in 227 households/group. Assuming a loss‐to‐follow‐up of 10%, the sample size estimate was 250 per treatment group or a total of 500 mother–child pairs. The data were analysed with STATA version 13 (StataCorp, 2013). The primary outcome measures of interest were endline diet quality (minimum dietary diversity [≥4 out of 7 food groups] and consumption of eggs during the previous day) and endline nutritional status (WAZ, length‐for‐age [LAZ]/height‐for‐age [HAZ], and weight‐for‐length [WLZ]/weight‐for‐height [WHZ]). A nonquantitative list of foods consumed yesterday was used to identify children's intake of seven food groups: grains, roots, and tubers; legumes and nuts; dairy products (milk, yogurt, and cheese); flesh foods (meat, fish, poultry, and organ meats); eggs; vitamin A‐rich fruits and vegetables; and other fruits and vegetables (WHO, 2008). The minimum diet diversity score of children was coded as a dichotomous variable (<4 food groups [not minimally diverse] or ≥4 food groups [minimally diverse]). Weight and length/height data were transformed into standardized deviation scores using the WHO age and sex growth references (WHO Multicentre Growth Reference Study Group, 2006). The wealth variable was derived from a principal components analysis of 13 household asset variables (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). Wealth scores were extracted from the first component and categorized by tertiles (low, medium, and high). A food security score was constructed with the 15 questionnaire items (Food and Agriculture Organization, 2012). Households were categorized by the number of affirmative answers: food secure (0), mildly food insecure (1–5), moderately food insecure (6–10), and severely food insecure (11–15). Unadjusted bivariate analyses were performed to test the relationship between outcomes and possible covariates using independent Student's t test for continuous variables and Pearson's goodness‐of‐fit chi‐square for categorical variables. Factors were included initially in the multivariable models if baseline group comparisons had a P value < 0.20 or if factors were considered to be important to child diet or growth based on previous research. We completed an intention‐to‐treat (ITT) analysis first. We estimated the size of the effect of the intervention on continuous outcomes (WAZ, LAZ/HAZ, and WLZ/WHZ) and dichotomous outcomes (minimum diet diversity and consumed eggs) using linear regression models with cluster‐robust standard errors based on the Eicker–Huber–White robust approach as implemented in the “cluster()” option to the “regress” and “logit” commands in STATA (Cameron & Miller, 2011). For all outcomes, we conducted an initial model without covariates and then a second model that included phase of enrolment and covariates for the child (baseline age, sex, baseline value of the outcome, and time elapsed between measurements), mother (education, marital status, and ethnicity), and household (food security, wealth, and raised poultry previously). Endline diet diversity was also included initially in the models for anthropometric outcomes. Backward elimination stepwise covariate selection procedure was used to select covariates with a P value of <0.10 (testing across categories) to adjust for the final models. No interaction terms with intervention were significant and therefore were not included in the final models. Adjusted odds ratios (aOR) or beta coefficients from the models are reported. Statistical significance was set at P < 0.05 unless otherwise indicated. To assess the robustness of the findings, we used different statistical models and indicators of the outcomes of interest in sensitivity analyses (Thabane et al., 2013). We have included two additional statistical approaches run for each of the final ITT models: (a) a random‐effects model (using “xtreg” and “xtlogit”) and (b) a mixed‐effects model (using “mixed” and “melogit”). In addition, we ran an as‐treated analysis that replaced “intervention” with “received inputs” as an indicator of level of participation in the trial. Among the 287 participants who were enrolled in intervention clusters, 233 received the poultry and garden inputs (144 in Phase 1, 89 in Phase 2). The primary reason for not receiving the inputs was because participants had not prepared the coop and garden. The three models (linear regression with cluster‐robust standard errors, random effects, and mixed effects) were used for the as‐treated analyses.

Based on the information provided, the following innovations were used in the study to improve access to maternal health:

1. Integrated agricultural interventions: The study implemented a 12-month intervention that included inputs and training for poultry farming and home gardening. This approach aimed to address the problem of stunting in rural communities by increasing access to high-quality foods.

2. Nutrition and health education: Along with the agricultural interventions, the intervention also included education on nutrition and health. This component aimed to improve knowledge and awareness among mothers about proper nutrition and healthcare practices for themselves and their children.

3. Cluster randomized controlled trial: The study used a cluster randomized controlled trial design to evaluate the effectiveness of the intervention. This approach allowed for the comparison of outcomes between intervention and control groups, providing evidence on the impact of the innovations.

4. Capacity-building and research program: The study was conducted within the context of a 5-year capacity-building and research program called Nutrition Links. This program provided training on nutrition, gender and diversity, data management and analysis, and evidence-based decision making to government and private sector service providers in various sectors relevant to maternal health.

Overall, these innovations aimed to improve access to maternal health by addressing the underlying factors contributing to poor maternal and child nutrition, such as poverty, low education, and limited access to high-quality foods.
AI Innovations Description
The recommendation based on the study is to implement an integrated intervention that combines agricultural inputs and training with nutrition and health education. This intervention can improve access to high-quality foods and nutrition education, leading to improved child nutrition. The study found that the intervention resulted in children in the intervention group meeting minimum diet diversity and having higher length-for-age (LAZ)/height-for-age (HAZ) and weight-for-age (WAZ) z-scores compared to children in the control group. The intervention included inputs and training for poultry farming and home gardening, as well as nutrition and health education. It also involved community engagement and support to ensure the success of the intervention. By implementing similar integrated interventions, access to maternal health can be improved, leading to better maternal and child health outcomes.
AI Innovations Methodology
Based on the provided description, the study implemented an agriculture-nutrition intervention in Ghana to improve children’s diet and growth. The intervention included inputs and training for poultry farming and home gardening, as well as nutrition and health education. The impact of the intervention was assessed through a cluster randomized controlled trial, where 16 clusters were randomly assigned to either the intervention or control group.

To simulate the impact of recommendations on improving access to maternal health, a similar methodology could be applied. Here is a brief description of the methodology:

1. Study Design: Conduct a cluster randomized controlled trial, similar to the Ghana study, where clusters are randomly assigned to either the intervention or control group. Clusters could be defined as geographical areas or communities.

2. Intervention: Develop an intervention that focuses on improving access to maternal health. This could include various components such as improving healthcare facilities, training healthcare providers, increasing awareness and education about maternal health, and providing resources for maternal health services.

3. Randomization: Randomly assign clusters to the intervention or control group. This ensures that the groups are comparable and any differences observed can be attributed to the intervention.

4. Data Collection: Collect baseline data on maternal health indicators such as maternal mortality rates, access to prenatal care, skilled birth attendance, and postnatal care. This data will serve as a baseline for comparison.

5. Implementation: Implement the intervention in the intervention group, while the control group receives standard or existing maternal health services.

6. Data Analysis: After a specified period, collect post-intervention data on the same maternal health indicators. Analyze the data using appropriate statistical methods to compare the outcomes between the intervention and control groups.

7. Impact Assessment: Assess the impact of the intervention on improving access to maternal health by comparing the post-intervention data with the baseline data. Calculate measures such as relative risk, odds ratios, or mean differences to quantify the impact.

8. Sensitivity Analysis: Conduct sensitivity analyses using different statistical models and indicators to assess the robustness of the findings, as done in the Ghana study.

By following this methodology, it would be possible to simulate the impact of recommendations on improving access to maternal health and evaluate the effectiveness of the intervention.

Yabelana ngalokhu:
Facebook
Twitter
LinkedIn
WhatsApp
Email