Menu
Menu
Menu
Menu
Background: 50% of Malagasy children have moderate to severe stunting. In 2016, a new 10 year National Nutrition Action Plan (PNAN III) was initiated to help address stunting and developmental delay. We report factors associated with risk of developmental delay in 3 and 4 year olds in the rural district of Ifanadiana in southeastern Madagascar in 2016. Methods: The data are from a cross-sectional analysis of the 2016 wave of IHOPE panel data (a population-representative cohort study begun in 2014). We interviewed women ages 15-49 using the MICS Early Child Development Indicator (ECDI) module, which includes questions for physical, socio-emotional, learning and literacy/numeracy domains. We analyzed ECDI data using standardized z scores for relative relationships for 2 outcomes: at-risk-for-delay vs. an international standard, and lower-development-than-peers if ECDI z scores were > 1 standard deviation below study mean. Covariates included demographics, adult involvement, household environment, and selected child health factors. Variables significant at alpha of 0.1 were included a multivariable model; final models used backward stepwise regression, clustered at the sampling level. Results: Of 432 children ages 3 and 4 years, 173 (40%) were at risk for delay compared to international norms and 68 children (16.0%) had lower-development than peers. This was driven mostly by the literacy/numeracy domain, with only 7% of children considered developmentally on track in that domain. 50.5% of children had moderate to severe stunting. 76 (17.6%) had > = 4 stimulation activities in past 3 days. Greater paternal engagement (OR 1.5 (1.09, 2.07)) was associated with increased delay vs. international norms. Adolescent motherhood (OR. 4.09 (1.40, 11.87)) decreased children’s development vs. peers. Engagement from a non-parental adult reduced odds of delay for both outcomes (OR (95%CI = 0.76 (0.63, 0.91) & 0.27 (0.15, 0 48) respectively). Stunting was not associated with delay risk (1.36 (0.85, 2.15) or low development (0.92 (0.48, 1.78)) when controlling for other factors. Conclusions: In this setting of high child malnutrition, stunting is not independently associated with developmental risk. A low proportion of children receive developmentally supportive stimulation from adults, but non-parent adults provide more stimulation in general than either mother or father. Stimulation from non-parent adults is associated with lower odds of delay.
Our study population is based in Ifanadiana District, with a population of approximately 209,000 residents. Ifanadiana is a low-resource setting in a low-resource country. Approximately 85% of the population subsist from agriculture, 3% have access to improved latrines and only 20% have access to safe drinking water. Health indicators are low compared with Madagascar as a whole—in 2014, 34% of children had all appropriate vaccines by the time they were 23 months old (vs. 51% for Madagascar); under 5 mortality was estimated at 145/1000 live births (vs. 62/1000 for Madagascar); and 20% of mothers had trained assistance in delivering their last baby (vs. 59% for Madagascar). Children’s development in the district is similarly low compared to the country as a whole. Four percent of children ages 3 or 4 were attending preschool (vs. 7.7% nationally) and 27% of adults between ages 18–35 had no formal education (vs. 21% nationally) [14]. The IHOPE cohort was established as an extension of a representative baseline survey from 2014, which occurred at the outset of the health system strengthening intervention. The survey [14] used a 2-stage random sample in 80 clusters to select 1600 households. The IHOPE study revisits the same households every second year to collect data on health and economic indicators, and it is stratified to be representative of both within and outside the initial PIVOT catchment area [15]. The IHOPE study uses methods, techniques and questionnaires based primarily on Demographic and Health Surveys (DHS), with additional questions from the Multiple Indicator Cluster Surveys (MICS), and the Rwanda Questionnaire on Well-being [16]. Specifically, we used the early child development index (ECDI) modules from the MICS 4. The ECDI is designed to be an internationally comparable population screener for developmental risk [17, 18] (as opposed to a diagnostic tool for developmental delay), which includes questions about development of 3 and 4 year olds, including risk and protective factors. Specific questions include whether adults interact with the child in developmentally supportive ways (reading, counting or naming things, singing to the child, playing with the child, taking the child out of the home compound) (See Table 1 for specific questions in the ECDI). The ECDI focuses on 4 domains—learning, physical growth, social-emotional and literacy/numeracy. The IHOPE surveys are implemented and conducted by the Madagascar Institute of Statistics (INSTAT), the same organization that conducts the DHS and MICS in Madagascar. Early Child Development Index Questions, Development Domains, and Criteria for “on-track” development We defined development outcomes in 2 ways; one to compare to external cohorts in other countries, and one to consider risks to development in children within the cohort relative to one another. For external comparisons, we used the standard definitions from the MICS. Using these definitions, a child was considered to be developmentally “on track” if the child had a positive score in three of four domains in the ECDI, and “at risk for delay” if positive scores in fewer than 3 domains. For internal comparisons, a child was considered to have “low development relative to peers” if the ECDI score was at least 1 standard deviation (SD) below the IHOPE sample mean. We used the standard definitions in the MICS to create our stimulation indicators. A variable for “adult involvement” was created as a continuous score of number of reported developmentally stimulating activities conducted with a child by any adult (mother, father or other adult). We also created a binary variable for involvement—“adult disengagement” if no adult did at least 4 learning activities with the child in the last 3 days, and maternal, paternal and “other adult” disengagement to assess differences between caregivers. Home environment was assessed with a series of specific questions about number and source of playthings, number of books in the house, and whether and how long the child was left alone or with another child under the age of 10 years as a caregiver. Nutritional status was assessed using the World Health Organization measures [19]; we used both continuous weight-for-age (underweight), weight-for-height (wasting) and height-for-age (stunting) age-adjusted z scores compared to a normed international population, and binary variables to characterize moderate and severe stunting, wasting and underweight status. The number of other children under age 5 in the house was assessed as a continuous variable. A binary variable was created for maternal education—any years of formal education vs. no formal education. Household poverty was assessed using wealth indices as determined using DHS methods using principal components analysis [20]. Cutoff points for wealth quintiles were determined to be the values closest to but less than the 20, 40, 60 and 80th percentiles of the cumulative wealth index. We estimated, for all household members, injury or illness in the 4 weeks prior to the interview, whether household members had sought care for these illnesses or injuries, and whether they had missed work activities or school because of the illness/injury. Maternal age at birth was calculated from estimated maternal and child birthdates and was assessed as a continuous variable and a 4-category variable (age 15–19, 20–24, 25–34 and 35–49). Orphan status was presented as a categorical variable (single maternal; single paternal; both; neither; and unknown status if one or other parent’s survival was unknown); and fathers’ presence in the home was assessed as a binary variable. We conducted descriptive analysis using frequencies and percentages for binary and categorical data and means with standard deviations and medians with interquartile ranges for continuous data. We analyzed data using Stata 15 (Stata: College Station, Texas). We assessed factors associated with the outcomes (risk for delay or low development related to peers) using logistic regressions (for binary endpoints) and accounting for clustering at the sampling level using Stata’s “cluster” function. Factors in univariable analysis significant at an alpha of 0.1 were entered into a full multivariable model and reduced to a final model using backward stepwise regression with an alpha of 0.05 for retention. Wald tests were used to compare reduced to full models. We assessed factors in the final models for interaction. The protocol and tools for the IHOPE cohort were reviewed and approved by the Harvard Medical School Institutional Review Board and the Madagascar National Ethics Committee. Verbal consent was obtained from eligible adults (ages 15–59) and from parents or legal guardians for children’s participation. In common with most population health surveys, written consent was not required as the study was deemed to present no more than minimal risk of harm to subjects and involved no procedures for which written consent is normally required outside the research context. All data were de-identified data prior to analysis; investigators had only access to data identifiable at the cluster level.
N/A
