The effect of school feeding programme on class absenteeism and academic performance of schoolchildren in Southern Ethiopia: A prospective cohort study

Abstract Objective: Ethiopia recently scaled up the implementation of a school feeding programme (SFP). Yet, evidence on the impact of such programmes on academic outcomes remains inconclusive. We evaluated the effect of the SFP on class absenteeism and academic performance of primary school students (grade 5-8) in Sidama zone, Southern Ethiopia. Design: This prospective cohort study enrolled SFP-beneficiary (n 240) and non-beneficiary (n 240) children 10-14 years of age from sixteen public schools and followed them for an academic year. School absenteeism was measured as the number of days children were absent from school in the year. Academic performance was defined based on the average academic score of the students for ten subjects they attended in the year. Data were analysed using multivariable mixed effects negative binomial and linear regression models. Setting: Food insecure districts in Sidama zone, Southern Ethiopia. Participants: SFP-beneficiary and non-beneficiary children 10-14 years of age. Results: The mean (sd) number of days children were absent from school was 4·0 (sd 1·5) and 9·3 (sd 6·0), among SFP beneficiaries and non-beneficiaries, respectively. Students not covered by the SFP were two times more likely to miss classes (adjusted rate ratio = 2·30; 95 % CI 2·03, 2·61). Pertaining to academic performance, a significant but small 2·40 (95 % CI 0·69, 4·12) percentage point mean difference was observed in favour of SFP beneficiaries. Likewise, the risk of school dropout was six times higher among non-beneficiaries (adjusted rate ratio = 6·04; 95 % CI 1·61, 22·68). Conclusions: SFP promotes multiple academic outcomes among socio-economically disadvantaged children.

This prospective cohort study compared class absenteeism and academic performance of students enrolled in SFP beneficiary and non-beneficiary second-cycle primary schools in four rural districts of Sidama zone, Southern Ethiopia. The study was conducted in the 2017/2018 academic year. Baseline information was gathered at enrolment, and class attendance and academic performance were measured at the end of the first and second semesters of the year. The study was conducted in sixteen schools from four SFP-targeted rural districts (Borecha, Dara, Bona and Loko Abaya) of Sidama zone. Sidama is located approximately 300 km south of the national capital Addis Ababa. The zone covers nearly 10 000 km2 area and is administratively divided into ten districts. In 2017, the zone had approximately four million inhabitants of whom 95 % were rural dwellers(17). Sidama is characterised by three agroecological zones: lowlands (20 %), midlands (48 %) and highlands (32 %). The economy in the area is reliant on rain-fed subsistence agriculture(18). According to a recent study, 18 % of the households in the area were food insecure and the prevalence of severe food insecurity was as high as 48 %(18). In the Ethiopian education system, primary education is divided into first (grade 1–4) and second (grade 5–8) cycles. At the time of the survey, 111 s-cycle public schools (second-cycle primary schools) were functional in the four study districts, of which 27 were targeted by the SFP. Inclusion of schools into the programme is determined based on extent of food insecurity in the school catchment area as judged by administrative bodies and donors. Students enrolled in SFP-targeted schools daily receive a free cooked school meal prepared from cereals, legumes and vegetables. The nature of the SFP implemented in all the schools was more or less the same in terms of food served and frequency of feeding. Students enrolled in sixteen rural second-cycle primary schools (eight schools with SFP and eight schools without SFP) in the aforementioned four districts were eligible for the study. Students registered in SFP-targeted schools were considered as SFP beneficiaries, whereas those enrolled in non-targeted schools were assumed otherwise. The sample size was calculated using G*Power 3.1 programme(19), assuming that the two primary outcomes (class absenteeism and academic performance) would be compared between the two groups using one-tailed mean difference test. The sample size was computed with 95 % confidence level, 80 % power, one-to-one allocation ratio between the two groups, medium effect size (d = 0·4) and design effect of 2. Further, 20 % compensation for possible dropout was added. Ultimately, the sample size of 480 (240 SFP beneficiaries and 240 non-beneficiaries) was determined. From each of the four districts, two SFP-targeted and two non-targeted schools, in total sixteen schools, were included in the study. In each district, two schools with SFP were selected at random among the SFP-targeted schools and matching schools without SFP were identified using predefined matching criteria – being within the same district and having comparable agroecological features. In each school, thirty students were selected from the available sections using a proportional stratified sampling technique. Ultimately, based on class rosters and a table of random numbers, simple random sampling (SRS) was performed to select the students. With the intention of maximising the sample size of the study, at enrolment study, subjects who were not willing to take part in the study were replaced by randomly chosen eligible children from the same section (Fig. ​(Fig.11). Flow chart of the study The exposure of interest was SFP status (beneficiary v. non-beneficiary), while the two primary outcomes were total number of school-days missed in the year and average academic score of the students in the academic year. School dropout rate (yes/no) was also considered as a secondary outcome. Factors considered as potential confounders included: age and sex of the child, maternal and paternal educational status, household wealth index, monthly income and food insecurity, head of the household (male v. female) and enrolment of the household in the Productive Safety Net Program. Socio-demographic and economic characteristics of the study participants and their caregivers were assessed at enrolment using interviewer-administered questionnaires prepared in the local Sidamu Afoo language. The data were primarily collected from the parents of the index children through home interviews by trained and experienced enumerators and supervisors. Household food insecurity was assessed using the Household Food Insecurity Access Scale and categorised as food secure or mild, moderate or severe food insecurity(20). Class absenteeism rate and academic performance were measured based on school administrative records. Class attendance was measured as the total number of days children were absent from school in the year, while class performance was quantified based on the average score of the students (minimum 0 and maximum 100) for all ten courses they attended in that year. The ten courses were English language, Amharic language, Sidamu Afo language, Mathematics, Social Science, Sport Education, Civics, Integrated Basic Sciences (grades 5–6), Art (grades 5–6), Music (grades 5–6), Physics (grades 7–8), Chemistry (grades 7–8) and Biology (grades 7–8). Each course was rated with a scale ranging 0–100. We used STATA version 14 for data analysis. Principal Component Analysis was performed for computing household wealth index – a composite index of living standard – based on multiple variables including materials used for house construction, access to improved drinking water source and sanitation facilities, and ownership of livestock, private house and agricultural land. Principal Component Analysis was done using varimax rotation. The Kaiser–Meyer–Olkin measure of sampling adequacy was acceptable (0·64), and Bartlett’s test of sphericity was significant. Only variables that had communality scores above 50 % were retained in the analysis. Ultimately, the factor with the highest eigenvalue was taken and divided into three equal tertiles: poor, middle and rich. Pearson’s χ 2 or Fisher’s Exact tests were used to check the presence of systematic differences between students retained in the study (n 463) and lost to follow-up (n 17) depending on whether the assumptions of χ 2 test were fulfilled or not. χ 2 test was also performed for comparing SFP beneficiaries and non-beneficiaries based on socio-economic characteristics. Factors that were significantly different (P value < 0·05) between the two groups were statistically adjusted using multivariable regression models. Mixed effects negative binomial regression for count outcome was fitted to measure the effect of the SFP on school absenteeism. Poisson regression for count outcome was not used because the assumption of equidispersion, as evaluated by comparing the mean and variance, was violated. Mixed effects linear regression was performed to assess the effect of the SFP on average academic performance, and the assumptions of the model (normality and homoscedasticity of error terms and linearity of relationship) were assessed using partial plots and found to be satisfied. School dropout was modelled using mixed effects Poisson regression for binary outcome. In all multivariable models, absence of multicollinearity was evaluated using variance inflation factor and found to be within the acceptable range (variance inflation factor < 10). In all the mixed effects models, random intercept was set for schools. In these mixed effects models, we did not consider school-level variables as predictors because all schools had more or less similar profile – all were public schools, second-cycle primary schools and the teacher to student ratio was very comparable.

N/A

• desalegn, tsion a.
• gebremedhin, samson
• alemayehu, fikadu reta
• stoecker, barbara j.