Complementary feeding practices and associated factors of dietary diversity among uncomplicated severe acute malnourished children aged 6–23 months in Burkina Faso

Nutritional treatment of children with uncomplicated severe acute malnutrition (SAM) is based on ready-to-use therapeutic foods (RUTF). With treatment provided at community level, children could have access to other foods, and a reduction in the dose of RUTF could further increase dietary diversity during treatment. We assessed the dietary diversity score (DDS), the minimum dietary diversity (MDD), the minimum meal frequency (MMF) and the minimum acceptable diet (MAD) of 459 infants and young children aged 6–23 months being treated for SAM with different doses of RUTF. We also investigated the factors associated with DDS. Dietary intake was estimated using a single 24-h multipass dietary recall, 1 month after starting treatment, from December 2016 to August 2018. The DDS was calculated on the basis of eight food groups. Differences between children receiving the reduced RUTF and the standard RUTF dose and factors associated with DDS were assessed by Poisson and logistic regression models. RUTF dose was not associated with DDS (4.07 ± 1.25 for reduced RUTF and 4.01 ± 1.26 for standard RUTF; P = 0.77). Food groups most consumed by children were grains, roots or tubers (96%) and legumes and nuts (72%). Eggs consumption was low (3%). DDS was positively associated with child’s age, mother’s education, household wealth index, urban residence and rainy season. The present findings show that children with SAM consumed a variety of foods during treatment in addition to the RUTF ration prescribed to them. Reducing the dose of RUTF during SAM treatment did not impact DDS. The MANGO study has been described in detail elsewhere (Kangas et al., 2019; Nikièma et al., 2021). In short, it was a randomized controlled non‐inferiority trial conducted in the health district of Fada N’Gourma, eastern region of Burkina Faso. The MANGO trial received the approval of the Ethical Committee (Comité d’éthique pour la Recherche en Santé) in Burkina Faso in December 2015 and the clinical trials board (Direction Générale de la Pharmacie, du Médicament et des Laboratoires) in September 2016. The study was conducted in 10 health centres. In 2018, in the eastern region, the prevalence of global acute malnutrition (GAM) defined by WHZ < −2 SD was estimated at 8.5% with 1.7% of SAM, defined as WHZ < −3 SD (MS, 2018). Coverage for SAM treatment was estimated at 48% (ACF, 2014). In the eastern region, 63% of children in general are breastfed exclusively up to 6 months (MS, 2018). Introduction of complementary feeding was not effective in all children from 6 to 8 months; only 70% of them had started to consume foods other than breastmilk, and the average of food groups consumed per day was two (MS, 2018). Less than a quarter (23%) of children had an adequate MDD; a little more than half of children (57%) met the requirement for MMF, and MAD was 18% (MS, 2018). Most of the households in the study area experienced a food surplus during the 2014/2015 crop year, and cereal production coverage was 97% (Nikièma et al., 2015) The main ethnic groups in the region are the ‘Gourmatché’ (66%) followed by the ‘Mossi’ (16%) and the ‘Fulani’ (13%) (Bahan, 2009). Before enrolment to study, all parents of eligible SAM children received information on the trial objectives and processes, and the possibility to withdraw at any time and signed informed consent of the parents was obtained before starting the trial. At admission, data were collected about the socio‐demographic characteristics, anthropometry and 2‐week retrospective morbidity of the child, and a clinical examination was performed. The household food insecurity access scale (HFIAS) was also assessed (Coates et al., 2007). As per national SAM treatment protocol, all mothers of the two arms were recommended weekly not to share the RUTF with other members of household to give complementary food to the child if they were still hungry after eating the RUTF and to breastfeed at all times on demand. The dietary intake study was a cross‐sectional study, nested in the MANGO trial, and started 2 months after launch of the MANGO study. Data collection took place 1 month after the admission of the children to treatment, on a total of 459 children aged 6–23 months (57% of MANGO main study sample). The sample for dietary intake was previously described (Nikièma et al., 2021). Complementary foods practice data were collected during Week 4 or 5 of treatment. Data on children's complementary food intake was collected using a 24‐h multipass dietary recall (24HDR) (Gibson & Ferguson, 2008) at Week 4 or 5 during treatment, from December 2016 to August 2018. The 24HDR was conducted as a face‐to‐face interview by experienced and trained investigators. First, the caregiver was asked to recall all the foods and drinks the child had eaten the previous day, including snacks. In the second pass of recall interview, the investigators repeated to the caregiver, in a chronological order, each of the foods and drinks consumed and probed for cooking methods and recipes. The caregivers were asked to describe the composition of the listed composite foods and beverages. During the third pass, the amount consumed by the child was estimated by the caregiver using calibrated local household utensils (glasses, cups, bowls and spoons). Before closing the interview, the investigators checked responses with caregivers to make sure nothing has been omitted. The average recipe method (Gibson & Ferguson, 2008) was used to convert amounts consumed from household units to grams. Food weights were measured using a TANITA KD electronic kitchen scale (model KD‐400SV, Tanita Corporation, precision 1 g, maximum load: 5 kg). Rainy season was defined as June to October and dry season as November to May. Morbidity was defined as the presence of fever and/or diarrhoea during the last week. Safe source of drinking water was defined as water from taps, boreholes and protected wells. Urban living environment was defined as living less than 30 min from the town of Fada on foot. Caregivers were considered to have formal education when they had at least attended formal primary school. Ethnicity was categorized as Gourma, Mossi, Fulani and others. Household wealth index was defined using principal component analysis and included variables describing income and possessions. Three categories of household wealth index were defined: low, medium or high. Complementary foods were ‘any food or liquids, whether manufactured or locally prepared, consumed as a complement to breast milk or as a breast‐milk substitute, fed to infants during the complementary feeding period’ (WHO & UNICEF, 2008). DDS was defined as the number of various food groups consumed by the child out of the eight recommended groups during the preceding 24HDR period. According to the United Nations Children's Fund (UNICEF) 2020 guidelines (UNICEF, 2020), the eight food groups include (i) breastmilk; (ii) grains, roots and tubers; (iii) legumes, nuts and seeds; (iv) dairy (milk, yoghurt, cheese); (v) flesh foods (meat, fish, poultry and liver or organ meats); (vi) eggs; (vii) vitamin A‐rich fruits and vegetables (carrots, mangoes, dark green leafy vegetables, pumpkins, orange sweet potato); and (viii) other fruits and vegetables. Vitamin A‐rich foods were defined as containing at least 60 retinol activity equivalents (RAE)/100 g (FAO, 2015; WHO, 2010). The definition of other revised standard indicators used in this study for complementary feeding (UNICEF, 2020) were as follows: The statistical software package Stata Version 15.0 (Stata Corp.) was used to perform the statistical analyses. Anthropometric z‐scores were calculated using the 2006 WHO growth standards (WHO, 2006). Three household food security categories were created from the HFIAS scores: food secure, mildly food insecure and moderately or severely food insecure (Coates et al., 2007). To compute food intakes, the dietary data were converted to food groups using a food composition database (FCDB) previously described (Nikièma et al., 2021). FCDB was inserted into Lucille Software (Ghent University, Belgium). The proportion of each of eight food groups consumed and their weight were also calculated. The data analysed concerned 459 children, including 219 in the reduced RUTF dose arm and 240 in the standard RUTF dose arm. Descriptive analysis was summarized as percentage and means (±SDs) in order to present the general profile of children at baseline and at time of the recall concerning their socio‐demographic, maternal, and household characteristics. Differences between intervention and control arms were assessed using standard Poisson regression model for DDS and number of meals per day, linear mixed models for each food group intake weight and mixed logistic regression models for all other binomial variables. All models used study site (health centre) as random effects and were adjusted for child's sex, age at recall, breastfeeding status, morbidity prior to recall and caregiver's education. Collinear variables (variance inflation factor > 10) were omitted from analyses. Possible predictors of DDS at individual, household, community and environmental levels have been selected to determine factors significantly associated with DDS (Abizari et al., 2017; Arsenault et al., 2014; Dafursa & Gebremedhin, 2019; Edris et al., 2018; Iqbal, 2017; Kuche et al., 2019). Given this, we selected at individual level some potential predictor variables such as child’s age (6–11 months or 12–23 months), sex (boy or girl), morbidity at last week before recall (yes or no), stunting at admission (yes or no), caregiver’s age (<25 years or > = 25 years), caregiver’s education (no or yes) and caregiver’s ethnic group (Gourma, Mossi, Fulani, others). For household factors level, we selected food security status (food secure, mild food insecurity or moderate or severe food insecurity), household’s wealth index (low, medium or high), household’s water source (safe or unsafe) and household size (<5 and > = 5). Household’s residence (rural or urban) constituted a community‐level variable, and season of interview (dry or rainy) was considered an environmental‐level variable. To determine the factors significantly associated with DDS, multivariate standard Poisson regression analysis using a stepwise backward approach to model construction was computed (Dangura & Gebremedhin, 2017; Issaka, Agho, Page, et al., 2015b; Joshi et al., 2012; Mitchodigni et al., 2017) to estimate the coefficient and 95% confidence interval (CI). The association between each independent variable and DDS was initially assessed in unadjusted regression model; then variables with P‐value < 0.2 were entered for adjusted model. Assumptions of the adjusted regression model (linearity, absence of multicollinearity and homoscedasticity of error term) were checked (Casson & Farmer, 2014; Marill, 2004). Model fitness was assessed using Pearson goodness‐of‐fit P‐value (adjusted R‐squared value) and was satisfying. The outputs of the analyses are presented via crude and adjusted unstandardized Poisson regression coefficients (β). Findings were considered significant at P < 0.05. This study has been approved by the Burkina Faso Ethics Committee of Health Research. This trial was registered on 13 May 2016 at the IRSCTN registry (http://www.isrctn.com/ISRCTN50039021).