Drivers of stunting reduction in Senegal: A country case study

Background: Senegal has been an exemplar country in the West African region, reducing child stunting prevalence by 17.9% from 1992 to 2017. Objectives: In this study, we aimed to conduct a systematic in-depth assessment of factors at the national, community, household, and individual levels to determine the key enablers of Senegal’s success in reducing stunting in children <5 y old between 1992/93 and 2017. Methods: A mixed methods approach was implemented, comprising quantitative data analysis, a systematic literature review, creation of a timeline of nutrition-related programs, and qualitative interviews with national and regional stakeholders and mothers in communities. Demographic and Health Surveys and Multiple Indicator Cluster Surveys were used to explore stunting inequalities and factors related to the change in height-for-age z-score (HAZ) using difference-indifference linear regression and the Oaxaca-Blinder decomposition method. Results: Population-wide gains in average child HAZ and stunting prevalence have occurred from 1992/93 to 2017. Stunting prevalence reduction varied by geographical region and prevalence gaps were reduced slightly between wealth quintiles, maternal education groups, and urban compared with rural residence. Statistical determinants of change included improvements in maternal and newborn health (27.8%), economic improvement (19.5%), increases in parental education (14.9%), and better piped water access (8.1%). Key effective nutrition programs used a community-based approach, including the Community Nutrition Program and the Nutrition Enhancement Program. Stakeholders felt sustained political will and multisectoral collaboration along with improvements in poverty, women’s education, hygiene practices, and accessibility to health services at the community level reduced the burden of stunting. Conclusions: Senegal’s success in the stunting decline is largely attributed to the country’s political stability, the government’s prioritization of nutrition and execution of nutrition efforts using a multisectoral approach, improvements in the availability of health services and maternal education, access to piped water and sanitation facilities, and poverty reduction. Further efforts in the health, water and sanitation, and agriculture sectors will support continued success. Am J Clin Nutr 2020;112(Suppl):860S–874S. In this mixed methods study we applied several complementary approaches to inform study objectives, including a systematic literature review, a retrospective quantitative analysis, a program and policy analysis from 1990 to 2017, and qualitative data collection and analyses. An adapted conceptual framework (Figure 2) was designed based on the UNICEF nutrition framework and Lancet nutrition framework to guide all analyses (49, 50). Refer to the methods paper in this series [Akseer et al. (56)] for information on the development of a framework. Our analyses focused on Senegal's periods of stunting decline, which was gradual from the early 1990s to 2000, faster between 2000 and 2005, and then followed by a more marginal decline. Conceptual framework showing distal, intermediate, and proximal determinants of stunting. Note: Maternal height, maternal BMI, maternal anemia, and vitamin A supplementation have been removed from analyses as these data were not available in all survey rounds, particularly in the DHS 1992/93 and 2017. Complementary feeding and inadequate dietary intake variables (including: minimum dietary diversity, grains, legumes, dairy, flesh foods, eggs, vitamin A–rich fruits and vegetables, other fruit and vegetables) were considered for inclusion, however, were only measured for the 6–23-mo old population; however, this subgroup was not analyzed. The total mean HAZ change for the 6–23-mo age group was too small for our model's results to be meaningful, and analysis resulted in a model with unstable results. Exclusive breastfeeding was also considered for inclusion, however it was only measured for the < 6-mo-old population, which was omitted from analysis due to small sample size. DHS, demographic and health survey; DPT3, diphtheria, pertussis, and tetanus; HAZ, height-for-age z score. Ethics approval for the study, inclusive of primary data collection, was obtained through the Ministry of Health Ethics Board of the Cheikh Anta Diop University. Ethics approval for the broader stunting case study was also obtained through the Research Ethics Board at the Hospital for Sick Children (SickKids), in Toronto, Canada. We conducted a systematic scoping literature review between November 2017 and December 2018 of published peer-reviewed and grey literature to synthesize information on contextual factors, national and subnational interventions, policies, strategies, programs, and initiatives that may have contributed to reductions in child stunting in Senegal from 1990–2017. Using the search categories of 1) stunting (e.g., stunting, linear growth, linear growth stunting, HAZ, height, height-for-age, LAZ, length, length-for-age, undernutrition, malnutrition, nutr*), 2) child (e.g., child*, infan*), and 3) Senegal (Senegal*), we searched >15 online databases and grey literature sources. After deduplication, 1728 records were found, of which 44 articles remained after full text review. Full methods and results of the systematic review are included in Supplementary Appendix 2. Senegal has had multiple sequential DHS performed over the past 3 decades, including serial DHS since 2012. We selected DHS that were considered good quality and that aligned with our study period (and stunting prevalence reduction trajectories) as the primary quantitative datasets used in this study. Available anthropometry data for children <5 y old by survey round are presented in Table 1. Sample size by survey year based on valid child anthropometric data1 1Based on index (youngest) child data. DHS, demographic and health survey; MICS, multiple indicator cluster survey. We studied child HAZ and stunting prevalence (HAZ < −2SD) estimated using WHO child growth standards as the main study outcomes (51). Potential determinants or “covariables” were selected in line with our conceptual framework (Figure 2), as those that were distal, intermediate, and proximal factors to the child stunting outcome. Covariables were identified as individual- or household-level factors from DHS and Multiple Indicator Cluster Surveys (MICS). Although we searched for ecological variables (at the region level), we were unable to track meaningful and available indicators. We estimated child HAZ kernel density plots and HAZ compared with age plots (“Victora curves”) (52) using smoothed local polynomial regressions to examine population shifts in growth faltering and growth trajectories by period of life. We estimated piecewise linear splines to quantify the slopes and inflection points of growth trajectories (53). We used previously published standardized methods for understanding stunting prevalence by wealth quintile (Q1–Q5), maternal education, area of residence (urban or rural) and child gender (54, 55). Household asset data were used to estimate wealth scores (later organized into quintiles) using principal components analysis. To account for the cumulative distribution of the wealth index, we also calculated slope index of inequality (SII) and concentration index (CIX) to measure absolute and relative socioeconomic inequalities (54, 55), respectively. We examined relative declines in child stunting prevalence for each region in Senegal using compound annual growth rates (CAGRs). To understand potential determinants of stunting reduction, we conducted two complementary sets of multivariable analyses to study relationships between child HAZ (outcome) and covariables. Both analyses used a series of step-wise linear regression models and hierarchical modelling of distal, intermediate, and proximal level variables as suggested by Victora 1997 (49). We harmonized variables in the DHS 1992/93 and DHS 2017 rounds and applied difference-in-difference methods with time*covariable interaction terms to examine if change in a proposed predictor of HAZ leads to a change in HAZ. Oaxaca-Blinder decomposition methods were applied to statistically decompose changes in mean child HAZ between 1992/93 and 2017 into potential determinants (covariables). Please see Supplementary Appendix 3 and the Methods paper in this series [Akseer et al. (56)] for full methods detail. All analyses were conducted with Stata 14.0 and accounted for survey design and weighting. Key nutrition-specific and -sensitive policies and programs that took place between 1990 and 2017 were assembled into a timeline using an iterative approach. The Senegal study principal investigator and research team members proposed a timeline based on a systematic desk review of literature. This timeline was reviewed by expert stakeholders and iterated on until agreement between the Senegal research team and country experts was reached. Qualitative interviews were conducted to gain perspectives of key national stakeholders, regional stakeholders (childcare and health workers), and mothers in the community. Participants were identified and selected using purposive sampling strategies (57), including snowball sampling in the regions of Louga, Diourbel, and Kaolack due to their substantial progress in reducing under-5 stunting (57, 58). First, 21 interviews were conducted with experts in Dakar who were officials from the Ministry of Health and other specialized agencies, international nongovernmental organizations (NGOs), and professors. Second, 20 interviews were conducted with resource persons, including nursery school teachers, health staff (i.e., doctors, nurses, midwives, and community health workers), NGO workers, imams, and village chiefs. In each of the 3 regions, 2 interviews with resource persons were conducted in rural areas and 2 in urban areas. Last, in each region, 2 focus groups were held with mothers who gave birth between 1992 and 1997 and 2 focus groups with younger women who gave birth between 2012 and 2017. National and regional interviews were conducted in French while focus groups were conducted in the main local language of Wolof, and were audiorecorded, transcribed for analysis, and translated into English. Data generated during focus group discussions and semistructured interviews were analyzed using our study framework. Thematic analysis was conducted to explore key themes that emerged based on stunting determinants, including socioeconomic status, migration, hygiene and sanitation, and nutrition and eating behaviors. Responses from national stakeholders, regional stakeholders, and mothers at the community level were analyzed separately. Full qualitative analysis methodology can be found in Supplementary Appendix 4.