A behavior change communication intervention, but not livelihood interventions, improves diet diversity and animal-source food consumption among Ghanaian women

Background: Women of reproductive age (WRA), especially in sub-Saharan Africa, are vulnerable to micro-nutrient deficiencies driven largely by poor quality diets. Intervening into food value chains, on which many households in low-and middle-income countries depend for their livelihood, may be a promising approach to improving diets in these contexts. Objective: In this pilot-scale randomized trial, we evaluated whether a multisectoral, food value chain intervention improved the diet diversity and the consumption of animal-source foods (ASFs) among WRA in Ghana. Design: Twelve fish-smoking communities in two regions of Ghana with 296 eligible women were randomly as-signed to one of three 9-month treatment arms: 1) behavior change communication (BCC) to promote improved diet quality through twice-weekly audio messages and bi-weekly peer-to-peer learning sessions; 2) BCC with microcredit to increase women’s incomes; or 3) BCC with provision of new smoke-oven technology. We assessed baseline-endline and between-treatment arm differences using a 10-food group diet diversity score (DDS), the Minimum Dietary Diversity for Women (MDD-W) indicator, and 7-day frequency of ASF consumption. Results: Among 118 participants (39 in both treatment arm 1 and treatment arm 3, and 40 in treatment arm 2, with no participant refusals), DDS increased from a mean (SD) of 4.0 (1.3) at baseline to 5.1 (0.9) at endline (P-value < 0.0001). The proportion of women achieving the MDD-W indicator nearly doubled from baseline (35.6%) to endline (69.5%) (P-value < 0.0001). Frequency of ASF consumption similarly increased for meat and poultry (2.7 (4.1) to 4.7 (5.3); P-value < 0.0001) and eggs (1.5 (3.1) to 2.3 (4.9); P-value = 0.02). Few differences in these outcomes were observed among treatment arms. Conclusions: A BCC intervention improved diet diversity and consumption of ASFs among participants. However, neither a group-based microcredit nor improved smoke oven intervention, both of which increased women’s income, led to additional dietary improvements.

This study was carried out among 12 fish-smoking communities in the Central and Volta regions of Ghana, representing marine and freshwater fisheries, respectively. We used available government census data on the population and number of households within each region as well as information from key informants on the presence of small-scale fish-smoking enterprises to identify eligible communities. Eligible communities were defined as those located in districts not prone to conflict, in which fish smoking was carried out during 8 or more months of the year, and which had at least 15 households participating in small-scale fish smoking. We selected six communities from each region, ensuring within-region variation in community size and market access. Households in these communities were commonly engaged in wild capture fishing and in activities along the fish value chain as a main economic activity (e.g. smoking fish and selling fresh or processed fish to local and other domestic markets). Following community selection, we conducted a census in each community, and using the collected census data, we randomly selected 10 households per community that had a woman aged 18–49 years who met the following inclusion criteria: 1) self-reported as not pregnant at the time of enrollment; 2) engaged in small-scale fish smoking as her principal livelihood or had earned income from fish smoking in the past 12 months (‘small-scale’ was defined as an operation with four or fewer functional smoke ovens and no more than three hired laborers); 3) did not plan to move from her community of enrollment for more than 3 weeks during the subsequent 11 months; 4) was willing to accept a no-interest loan for her fishing smoking business and perceived that such a loan would benefit her business; and 5) did not already own an improved ‘Ahotor’-style smoke oven at the time of enrollment. If more than one woman in the same randomly selected household was eligible for the study, we randomly selected just one woman from the household for participation in the study. In total, 296 women across the 12 communities met these inclusion criteria and were considered eligible for participation in the study. The 12 study communities were subsequently randomly assigned to one of three treatment arms (TAs), such that two communities in both regions participated in each of the three TAs. The interventions were implemented for 9 months in each of the TAs. The first treatment arm (TA1) was an anemia behavior change intervention. Participants received twice-weekly audio messages on provided mobile phones promoting several anemia-mitigating behaviors (e.g. dietary improvements; infection prevention practices; and water, sanitation, and hygiene (WASH) best practices). Participants also joined in bi-weekly, facilitated peer-to-peer learning sessions to reinforce the behavior change practices communicated through the audio messages and to ensure participants listened to all received audio messages. The specific dietary change messages promoted through TA1 included increasing the overall diversity of diets of women to include a broad range of legumes, vegetables, and ASFs as well as special emphasis on increased consumption of ASF (including fish, meat, organ meat, and eggs) to ameliorate deficiencies of iron and key vitamins that underlie nutritional anemia (7, 29). The messages were adapted from educational and health promotion materials used by the Disease Control and Nutrition Departments of the Ghana Health Service as well as the National Malaria Control Programme. Importantly, all study participants, regardless of TA assignment, received these interventions included in TA1. One-third of the participants (n = 40 women) received only the TA1 interventions, while other participants in TA2 and TA3 received the TA1 interventions plus the additional interventions described below that were a part of TA2 and TA3, respectively. Treatment Arm 2 (TA2) participants were additionally provided with assistance aimed at increasing the profitability of their fish-smoking businesses through a group-based microcredit scheme. TA2 participants received two, separate, interest-free loans for their fish-smoking businesses (the second loan included a partial rebate depending on the timing and extent of repayment of the first loan, such that this second loan effectively functioned as a cash transfer). The first loan amount was 1,000 GH¢ (~$US 250 at the time of the study), and the second loan amount was 1,500 GH¢. Participants in TA2 also received entrepreneurship training focused on marketing, book-keeping, and saving strategies for their business as well as twice-monthly text messages conveying market price information for smoked fish. Participants in TA3 were given an improved ‘Ahotor’-style fish-smoking oven (valued at $US550) to replace the oven that was currently being used, in addition to the TA1 anemia behavior change intervention. The Ahotor oven is designed to reduce emissions from biomass fuel combustion, decrease polycyclic aromatic hydrocarbon levels of smoked fish, and increase fuel efficiency vis-à-vis the traditional Chorkor oven commonly used throughout Ghana. TA3 participants also received training on the use of the Ahotor oven with the aim of reducing harmful respiratory exposures associated with the wood burning and fish-smoking processes. During the pre-intervention ‘baseline’ period (May–June 2018), a pre-tested, quantitative, household survey instrument was administered by trained enumerators to collect data on household sociodemographic characteristics, women’s fish-smoking practices, empowerment status, and other information. For the index participant, we also collected data from a 24-h dietary recall, with a repeat recall on a nonconsecutive day during the same 7-day period. All these data were again collected from each participating woman at the ‘endline’ after interventions had been completed (May–June 2019). The 24-h diet recalls were collected using standard protocols involving the multiple-pass method and context-specific serving dishes and utensils (30). Data from the second 24-h dietary recalls obtained at both baseline and endline were used to calculate a continuous diet diversity score (DDS) for the baseline and endline periods, respectively. The first dietary recall at baseline was not used due to errors in quantifying specific serving sizes that were corrected in subsequent recalls. The DDS is based on recent consumption (i.e. past 24 h) of foods from 10 different food groups: 1) grains, white roots and tubers, and plantains; 2) pulses (beans, peas, and lentils); 3) nuts and seeds; 4) dairy; 5) meat, poultry, and fish; 6) eggs; 7) dark green leafy vegetables; 8) other vitamin A-rich fruits and vegetables; 9) other vegetables; and 10) other fruits (7). The consumption of each food group was coded as 1 if any amount of that food was consumed in the past 24 h, and as 0 otherwise. Based on this, the Minimum Dietary Diversity for Women (MDD-W) indicator was created by summing the values for each of the 10 food groups. The MDD-W indicator, ranging from 0 to 10, served as a proxy for the micronutrient adequacy of diets of WRA (7). Women who ate five or more of the 10 food groups were considered to have met the MDD-W indicator (7). ASF consumption was measured from each participant’s self-reported frequency of consumption of any ASF in the past 7 days and used to create variables representing different groups of ASF consumed. Meat and poultry consumption, for example, was based on the number of days and average number of times per day items from each of nine categories of meat and poultry were consumed. This same approach was used to assess fish consumption (based on eight categories of fish and shellfish), egg consumption (based on two categories), and dairy consumption (based on five categories). To complement these ASF consumption frequency data, we also assessed household expenditures (GH¢) ASF items in the past 7 days for the same ASF categories based on respondent recall of expenditures on each food item. Recent household income was assessed based on a self-reported questionnaire documenting all income-generating activities by all household members in the past 30 days. Additional earnings from land sales or other sources (e.g. remittances and social welfare benefits) were added to calculate the total earnings per household. Characteristics of the index woman in each household included self-reported age and education level. At baseline and endline, we further assessed participants’ knowledge of the dietary causes of anemia and strategies to prevent anemia through dietary change. These knowledge assessments were aimed at understanding participants’ uptake of the mobile phone audio messages focused on dietary change that were shared as part of the TA1 intervention. For each knowledge question pertaining to diet and anemia, we coded respondents’ answers as ‘correct’ or ‘incorrect’ based on a pre-specified answer key. For all questions, there were multiple correct responses. If the respondent provided at least one correct response for the question, it was coded as ‘correct’. Differences between baseline and endline values were evaluated by treatment group for DDS, recent consumption of specific foods group, and frequency of consumption of ASFs in the past 7 days. For continuous variables (i.e. DDS, ASF consumption meat and poultry, ASF consumption fish, ASF consumption eggs, and ASF consumption milk and dairy), the Wilcoxon signed-rank test for paired data was used for non-normally distributed data, and ANOVA was used for normally distributed data. To assess differences in categorical variables (i.e. MDD-W achieved and each food category that comprises the DDS), Pearson’s chi-squared test, McNemar’s test for correlated data, and Fisher’s exact test were used. Differences in DDS, MDD-W, and frequency of ASF consumption across treatment groups at baseline and endline, respectively, were assessed using ANOVA for continuous variables and Pearson’s chi-squared test for categorical variables. We further analyzed the association of treatment arm with DDS and ASF consumption at study endline using ordinary least squares linear regression models, and of treatment arm with MDD-W using logistic regression models. For each set of analyses, three models were constructed, adjusting for: 1) no covariates, 2) the baseline value of the dependent variable, and 3) the baseline value of the dependent variable as well as baseline household income and maternal education status. Statistical analysis was carried out in SAS version 9.4 (Cary, NC). We report statistical significance at the P < 0.05 and P < 0.01 levels as well as at the P < 0.1 level to indicate potentially meaningful trends in this pilot-scale study. The study protocol was approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board and the Ethics Committee for Basic and Applied Sciences at the University of Ghana, Legon. Comprehensive informed consent was obtained from all study participants. The trial is registered at clinicaltrials.gov ({"type":"clinical-trial","attrs":{"text":"NCT03498755","term_id":"NCT03498755"}}NCT03498755) (31).

• ludwig-borycz, elizabeth f.
• wilson, mark l.
• colecraft, esi k.
• jones, andrew d.