Menu
Menu
Menu
Menu
Introduction Stunting or linear growth retardation in childhood is associated with delayed cognitive development due to related causes (malnutrition, illness, poor stimulation), which leads to poor school outcomes at later ages, although evidence of the association between the timing and persistence of stunting and school outcomes within the sub-Saharan African context is limited. Methods Anthropometric data around birth (0–4 months), early (11–16 months) and late childhood (ages 4–8 years) along with school outcomes up until the age of 11 were analysed for a cohort of 1,044 respondents, born between 2002–2004 in Karonga district, Northern Malawi. The schooling outcomes were age at school enrolment, grade repetition in Standard 1 and age-for-grade by age 11. Height-for-Age Z-scores (HAZ) and growth trajectories were examined as predictors, based on stunting (<-2SD HAZ) and on trajectories between early and late childhood (never stunted, improvers, decliners or persistently stunted). Multinomial and logistic regression were used to estimate the association between stunting/trajectories and schooling, adjusted for socioeconomic confounders. Results The effects of stunting on schooling were evident in early childhood but were more pronounced in late childhood. Children who were stunted in early childhood (9.3%) were less likely to be underage at enrolment, more likely to repeat Standard 1 and were 2–3 times more likely to be overage for their grade by the age of 11, compared to their non-stunted peers. Those persistently stunted between early and late childhood (7.3%) faced the worst consequences on schooling, being three times as likely to enrol late and 3–5 times more likely to be overage for their grade by the age of 11, compared to those never stunted. Compared to improvers, those persistently stunted were three times as likely to be overage by two or more years by the age of 11, with no effect on enrolment or repetition. Conclusion Our findings confirm the importance of early childhood stunting on schooling outcomes and suggest some mitigation by improvements in growth by the age of starting school. The nutritional and learning needs of those persistently stunted may need to be prioritised in future interventions.
Continuous birth registration was set up as part of the baseline census for a demographic surveillance carried out between 2002 and 2004 in the southern part of Karonga district, in northern Malawi. Trained staff collected anthropometric data during the first visit after birth, which was usually within 2–6 weeks. Repeat anthropometry measures were collected during a follow-up visit after one year. Anthropometric data were also collected in later survey rounds on all children under the age of 10 between 2008–2011, so data were available for the 2002–4 birth cohort at ages 4–8. For those measured more than once between 2008–11 the earliest record was used. Socio-economic and schooling histories were collected in the original census and updated annually from 2007 to 2015. Routine training was provided to staff prior to collecting anthropometric data using methods recommended by the USAID’s Food and Nutrition Technical Assistance(FANTA) project[27]. For children below age 2, recumbent length was measured using a SECA210 polyurethane plastic measuring mat (with 0.5mm increments) while weight was measured using a spring scale (100g increments). Height of children older than two years was measured using the Leicester height measure. Maternal malnutrition, measured by the mother’s mid-upper arm circumference (MUAC), is a determinant of foetal growth restriction and early growth faltering [7,28]. In this study, MUAC was measured using a steel tape (1mm increments) and a cut-off of <21cm was used to define maternal malnutrition, as used previously in the same setting [29]. Early and later linear growth failure or stunting was defined as the Height-for-Age Z-score (HAZ) < -2 SD (termed as moderate/severe stunting) based on the WHO growth references for children below and above age 5[30,31]. The Z-score represents the difference in a child’s height from the median height of children within the reference population (at a given age and sex), divided by the standard deviation of the reference population. Growth trajectories between early and late childhood were defined as being never stunted, improvers (stunted in early childhood but not stunted in late childhood), decliners (not stunted in early childhood but stunted in late childhood), or persistently stunted (stunted in early and late childhood). In Malawi, primary education is free and is for eight grades, with the official age of entry being 6 years. With the introduction of free primary education in Malawi in 1994, enrolment is nearly universal, though school quality is poor, with frequent grade repetitions and students progressing slowly through school[32]. Under or over age enrolment is possible. Household poverty, long distances to school and perceptions of school readiness may prompt parents to delay enrolment[33–35]. However, parents may also enrol children at an earlier age, to allow younger children to accompany their older siblings to school; to provide a head-start in school; or to optimise free child-care provision in school while parents work[33]. In our analyses, those who enrolled in school prior to or after the official age of entry of 6 were categorised as being underage or overage at enrolment. Age-for-grade is the number of years a child is ahead/behind in class based on the official age-for-grade (Age-for-Grade = Current Age-Current Grade-5) and provides a cumulative measure of school performance irrespective of the highest grade achieved. Given the follow-up time available for this cohort, the analyses focuses on age-for-grade at age 11, which is the age up until when most respondents were seen. The effects of stunting on grade repetition in Standard 1 is also examined. Principal Component Analysis (PCA) was used to estimate relative household wealth at birth using data on dwelling characteristics (quality of walls, roof), ownership of consumer durables (clock, mosquito nets and bank account), and access to utilities (water, electricity). Categorical variables were made into dummy binary variables, while continuous variables (number of mosquito nets owned by a household) were re-scaled to have mean as zero and standard deviation 1 [36,37]. The first component explained 36% of the variation between households. The household wealth score was divided into thirds (most to least poor). Data on household assets collected between 2007–2011 were also used to construct asset indices for the follow-up period (early and late childhood) using PCA. Variables selected for inclusion in the asset index (bicycle, radio, oxcart, clock, mattress, bed and chair) were based on what was consistently available across all household survey rounds. Data on parental educational qualifications were collected during the socio-economic surveys under the assumption that education levels remained unchanged since the child’s birth. A few other variables, including season at birth, mother’s age at birth, mother’s MUAC, birth order, were initially explored but omitted from the final analysis, as they did not appear to confound the relationships. Maternal height was not included because it can have a direct effect on foetal growth [9] and we wanted the growth measure to include any pre-natal growth deficit. Father’s height was explored as a possible confounder. Logistic regression was used to conduct the analysis for the grade repetition outcome. Multinomial logistic regression was used for the analyses on age at enrolment and age-for-grade at age 11. Significance of the relationship between stunting and grade repetition was assessed by performing a Wald test of each estimated OR being equal to 1. For the analyses of outcomes age at enrolment and age-for-grade at age 11 that were modelled using multinomial logistic regressions, Wald tests for the ORs for each level of outcome being equal to 1 were performed. In each case, we refer to these as tests for heterogeneity as these capture whether stunting at each age-interval has different odds of each outcome level. Ethics approval for the study and the consent procedures were obtained from the National Health Sciences Research Committee in Malawi and the Research Ethics Committee of the London School of Hygiene and Tropical Medicine. Permissions to conduct the study were also granted by the traditional authorities, village headmen and traditional advisers in the study catchment area. For the demographic surveillance (which included schooling level recording and anthropometry at birth) the research purpose of surveillance was explained to each household. Verbal consent was sought (as was the norm for demographic surveillance studies at that time) and recorded by staff in the field register, along with any refusals to participate. Only those participants who consented to participate were included in the study. For the anthropometry study written informed consent was sought from the parent/guardian of each child.
