A mixed-methods study of factors influencing access to and use of micronutrient powders in Rwanda

The World Health Organization recommends point-of-use fortification with multiple micronutrients powder (MNP) for foods consumed by children aged 6–23 months in populations where anemia prevalence among children under 2 years or under 5 years of age is 20% or higher. In Rwanda, anemia affects 37% of children under 5 years. The MNP program was implemented to address anemia, but research on factors affecting the implementation of the MNP program is limited. We conducted a mixed-methods study to examine the factors influencing access to and use of MNP among mothers (N=379) in Rutsiro district, northwest Rwanda. Inductive content analysis was used for qualitative data. Logistic regression analysis was used to determine factors associated with the use of MNP. Qualitative results indicated that the unavailability of MNP supplies and distribution issues were major barriers to accessing MNP. Factors influencing the use of MNP included mothers’ perceptions of side effects and health benefits of MNP, as well as inappropriate complementary feeding practices. Mothers of older children (aged 12–23 months) were more likely to use MNP than those of younger children (aged 6–11 months) (adjusted odds ratio [aOR]=3.63, P<.001). Mothers whose children participated in the supplementary food program were nearly 3 times more likely to use MNP than those whose children had never participated in the program (aOR=2.84, P=.001). Increasing household hunger score was significantly as-sociated with lower odds of using MNP (aOR=0.80, P=.038). Mechanisms to monitor MNP supply and program implementation need to be strengthened to ensure mothers have access to the product. MNP program implementers should address gaps in complementary feeding practices and ensure mothers have access to adequate complementary foods.

This study was conducted in Rutsiro district, northwest Rwanda, approximately 140 km from the capital city, Kigali. The district has the highest prevalence of child stunting (54%) among children under 5 years.24 The majority (∼98%) of the district's population is rural, and agriculture on small plots of land is the main livelihood.25 The main subsistence crops are maize, beans, banana plantain, cassava, and sweet and Irish potatoes. The health system in the district consists of 1 hospital and 17 health centers.26 Each health center oversees community health workers (CHWs) who provide community-based nutrition and other health services to an average of 23,000 inhabitants living within the health center's catchment area.27 The services provided by CHWs include distribution of MNP, locally known as Ongera, to caregivers with children aged 6–23 months. In Rwanda, the Ministry of Health or UNICEF deliver MNP to district hospitals, which then distribute MNP supplies to health centers. MNP is then distributed by the health centers to CHWs, who in turn distribute MNP to caregivers during monthly child growth monitoring and promotion activities. Some nongovernmental organizations, mainly World Vision International (Rwanda) and Caritas Rwanda, support the MNP program implementation through training of CHWs and awareness activities related to child feeding. Every caregiver with a child aged 6−23 months is entitled to 30 sachets of MNP per month, which they receive free of charge. Using cooking demonstrations, CHWs also counsel caregivers on optimal complementary feeding practices, such as age-specific dietary diversity, consistency and quantity of complementary foods, and on MNP usage.8,13 Every caregiver with a child aged 6−23 months is entitled to 30 sachets of MNP per month, which they receive free of charge. This study used a cross-sectional convergent mixed-methods design,28 combining both quantitative and qualitative data. The data used in this study were collected as part of a survey conducted between September 2018 and January 2019 to investigate the factors associated with nutritional status of children aged 6–23 months. Details on the survey sample size estimation and participants recruitment are described elsewhere.29 Briefly, the district was first divided into 3 zones based on main roads connecting the district to its neighboring districts. In each zone, 3 health centers were purposely selected to maximize geographic distribution, for a total of 9 health centers. Within each of the selected health center's catchment area, 2 villages were randomly selected. In these villages, monthly growth monitoring lists were obtained from CHWs and used to compile a sampling frame from which participants were randomly selected. Mothers who refused to participate and those who were not found in their homes were replaced (11 mothers in total) by selecting the next name on the list. Eligibility criteria were (1) having a child aged 6–23 months; (2) child was apparently healthy (i.e., no overt signs of illness); and (3) being in the 2 lowest socioeconomic categories. Of the 400 survey participants, 21 (5%) of the children were excluded from the analysis due to premature birth (i.e., before 37 weeks of gestation) or low birthweight (i.e., less than 2.5 kg). The remaining 379 participants formed the basis of the present study. Quantitative and qualitative data were collected concurrently using a survey questionnaire. The questionnaire was developed in English, translated into Kinyarwanda, and programmed into a handheld tablet (Samsung Galaxy Tab 8.0 T295, Korea). It was pretested, and data were collected through face-to-face interviews. Qualitative data were audio-recorded. This study was approved by the Massey University Human Ethics Committee (reference: SOA 17/67) and the Institutional Review Board of the University of Rwanda's College of Medicine and Health Sciences (reference: 003/CMHS IRB/2017). Permission to collect data was also obtained from the Rutsiro District Public Health Office. Oral informed consent was obtained from all participants. “Ever using MNP” was the primary outcome variable. Mothers were asked if they added (yes/no) MNP to the target child's foods in the last 7 days prior to the survey. Mothers who had not used MNP were asked whether they had ever used MNP before (yes/no). A mother was categorized as “ever used MNP” if she had used MNP in the previous 7 days or before, and those who had not used MNP either within 7 days prior to the survey or before were categorized as “never used MNP.” Information related to participants' demographics, socioeconomic, household food security, and indicators of health system engagement were obtained through mothers' recall. Health cards were used for verification (e.g., child age and heath information). Demographic information reported by mothers included the child's age and sex and the maternal age at first birth. Mothers also reported presence of symptoms of child diarrhea (defined as ≥3 watery or loose stools per day) and upper respiratory infections (runny nose, coughing, or wheezing) in the previous 4 weeks. Socioeconomic variables included maternal education level (coded as none/incomplete primary education, complete primary education, secondary education) and household asset ownership (e.g., radio, land, domestic animals, housing characteristics). Fourteen household assets were used to create a household wealth index using principal component analysis.30 The first component was taken to represent the household wealth index and divided into terciles (lower, middle, and upper). A household hunger score—a proxy of a household's ability to access food—was measured using a validated cross-cultural household hunger scale (HHS).31 Adhering to HHS measurement guide, mothers were asked 3 questions intended to capture 3 situations (i.e., lack of food of any kind in the house; going to sleep hungry because there was not enough food; and going a whole day and night without eating) reflecting a household's experience of insufficiency of food supply and intake and physical consequences. Each question was followed by the frequency-of-occurrence question (i.e., how often the reported situation was experienced). The responses were coded and used to generate a household hunger score that ranged from 0 (indicating no hunger) to 6 (indicating severe hunger). Indicators of health system engagement are (1) attendance at growth monitoring site in the previous month (coded as yes/no); (2) the number of antenatal care visits when pregnant with the study child (coded as <4 visits or ≥4 visits; a minimum of 4 visits is recommended in Rwanda5); and (3) whether the child ever participated in the supplementary food program (coded as yes/no). A household hunger score—a proxy of a household's ability to access food—was measured using a validated cross-cultural household hunger scale. The questionnaire included an open-ended question that was used to collect in-depth information on the reasons for not using MNP. Mothers who had not used MNP in the previous 7 days (i.e., those who used MNP but not in the previous 7 days, and those who never used MNP) were asked to provide reasons for not using MNP. Probes (either open-ended or specific to the mothers' comments) were used to obtain additional information.32 Median (interquartile range [IQR]) values were determined for continuous data and percentages for categorical data. Bivariate and multiple logistic regression analyses were performed to examine factors associated with using MNP. The full model adjusted for the presence of diarrhea and respiratory infection in the past 4 weeks. We adjusted for these variables because our previous research in the same population showed that child illness has negative effects on how mothers feed their children, including withholding or restricting some foods from children's diets.33 Unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CI) were computed. Variables with a P value of <.05 were considered significant predictors. We did not perform a Bonferroni correction because, although the correction decreases the probability for type I error, such adjustment is vulnerable to type II error and can obscure important findings.34,35 All statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY). Mothers' responses were audio-recorded, transcribed verbatim in Kinyarwanda, and translated into English. Content analysis36 was used to analyze the data. An inductive content analysis approach, which is recommended when there is no prior research or little is known about the studied phenomenon, was used. The data analysis had 3 phases: preparation, organization, and reporting.37 The first phase consisted of careful reading of the data several times to become immersed in and familiar with the data. In the organization phase, each transcript was read carefully by the first author, highlighting the text (words or phrases) that appeared to describe the phenomenon under study (i.e., access to and/or use of MNP). The highlighted texts were openly and manually coded by giving each text a descriptive code. The second author read the data to confirm the descriptive codes. These codes were revised, and the codes that emerged from the revision were jointly reviewed before integrating them into the analysis. Final codes were examined, compared, and grouped into categories that represented similar meaning.38 The first, second, and last authors reviewed, discussed, and agreed on the final code categories. In the final phase of analysis, SPSS (version 25) was used to quantify the frequency of major categories and subcategories.39 To interpret and report the findings, examples of original textual responses representing specific code or category are presented.

• dusingizimana, theogene
• weber, janet l.
• ramilan, thiagarajah
• iversen, per ole
• brough, louise