Effect of school feeding program on the anthropometric and haemoglobin status of school children in Sidama region, Southern Ethiopia: a prospective study

Ethiopia recently scaled up the implementation of a school feeding program (SFP) as a targeted intervention for protecting disadvantaged school children from hunger and food insecurity. However, the contribution of the program to advancing the nutritional status of children has not been adequately explored. We assessed the effect of SFP on the anthropometric and haemoglobin status of school children in Sidama Region, Southern Ethiopia. Our prospective cohort study compared the height-for-age z-score (HAZ), BMI-for-age z-score (BAZ) and haemoglobin concentration of SFP beneficiary (n 240) and non-beneficiary (n 240) children, 10-14 years of age. The children were recruited from 8 SFP implementing and 8 control schools using a multistage sampling procedure and were followed for an academic year. The SFP intervention and control schools were matched one-to-one based on agro ecological features and geographical proximity. Exposure, outcome and pertinent extraneous variables were collected through baseline and end-line surveys. Multilevel difference-in-differences (DID) analysis was used to measure the net effect on the outcomes of interest. In the multivariable DID model adjusted for potential confounders including maternal and paternal literacy, household monthly income, wealth index and household food insecurity, the SFP did not show significant effects on the haemoglobin concentration (β = 0·251, 95 % confidence interval (CI): -0·238, 0·739), BAZ (β = 0·121, 95 % CI: -0·163, 0·405) and HAZ (β = -0·291, 95 % CI: -0·640, 0·588) of children. A prospective cohort study was used to compare the anthropometric and haemoglobin status of SFP beneficiary and non-beneficiary school children 10–14 years of age. SFP beneficiaries and non-beneficiaries of the program were enrolled from 16 schools and followed for one academic year. Data on exposure, outcome and other pertinent control variables were gathered over two rounds of baseline and end-line surveys. The net effect of the program on the outcomes of interest was estimated using Difference-in-Differences (DID) analysis. The study was conducted in 16 second-cycle primary schools (SCPS) found in four SFP-targeted rural districts (Borecha, Dara, Bona and Loka Abaya) of Sidama region, southern Ethiopia. The region is located approximately 300 km south of the national capital, Addis Ababa, and has about four million inhabitants, of whom 95 % were rural dwellers(14). The zone covers nearly 10 000 km2 area and is characterised by diverse agroecological features. The region is divided into 23 administrative districts. In Ethiopia, the primary education system is divided into first- (grade 1–4) and second- (grade 5–8) cycles. During the study, 111 SCPS including 27 SFP-targeted schools were operational in the four districts. Schools are typically enrolled in the program in consideration of the severity of the food insecurity situation in their catchment areas, as evaluated by the regional education bureau, donors and other partners. Students registered in SFP-targeted schools receive a daily free cooked school meal prepared from cereals and legumes. The sample size of 480 school children (240 SFP beneficiaries and 240 non-beneficiaries) was determined using G*Power software(15) assuming mean BAZ, HAZ and haemoglobin would be compared between the two groups using a one-tailed mean difference t-test at 95 % confidence level and 80 % power. Furthermore, one-to-one allocation ratio, medium effect size (d 0⋅4), design effect of 2 and 20 % allowance for possible dropout were assumed. The sample size calculation formulas in consideration of each objective of the study were presented later (Table 1). Sample size calculation for each objective of the study Students were identified from eight SFP-targeted and eight control schools from the above-mentioned four districts. Students registered in SFP-targeted schools were considered as SFP beneficiaries, whereas, those from non-targeted schools were taken as non-beneficiaries. From each of the four districts, two beneficiary and two non-beneficiary schools, totally sixteen schools, were included in the study. In every district, two schools with SFP were selected at random from schools enrolled in the program. Then each of the selected schools with the feeding program was one-to-one matched with a control school based on predefined matching criteria: similar agro-ecology features (similar in agricultural production, climatic zone and landscape) and geographical proximity (being within the same district). Ultimately 480 children, 30 eligible students per school, were selected across all the grades using a simple random sampling technique. Purposely to increase the sample size, those who were not willing to take part in the study were replaced by randomly selected eligible children from the same section. The exposure variable of interest was SFP enrolment status (beneficiary v. non-beneficiary) and the outcome variables were BAZ, HAZ and haemoglobin concentrations of the children. Control variables were socio-economic status indicators including maternal and paternal literacy, household monthly income, household wealth index, household food insecurity, head of the household (male v. female), type of water source at home (improved v. non-improved), age and sex of the child. Data were collected in two rounds of surveys implemented at the beginning (September 2017) and end (June 2018) of the 2017/2018 academic year. A similar set of variables were collected in the surveys. Haemoglobin status was measured in the field from capillary blood using HemoCue analyzer (HemoCue Hb 301, Ängelholm, Sweden). Haemoglobin measurements were adjusted for altitude according to the recommendation of the World Health Organization (WHO)(16). Height and weight measurements were taken following standard procedures without heavy clothing and shoes using a portable stadiometer and calibrated digital UNICEF scale. Haemoglobin and anthropometric measurements of children were taken in private settings in each school. Parents of the index children were interviewed at home about socio-demographic, economic, access to social services and household food security situations using a structured and pretested questionnaire prepared in the local Sidamu Afoo language. The parts of the questionnaires on socio-demographic and economic information were adopted from the standard Demographic Health Survey (DHS) questionnaire(17). Household food insecurity was measured in accordance with the Food and Nutrition Technical Assistance (FANTA) indicator guide and classified as food secure and mild, moderate or severe food insecurity(18). In accordance with the criteria of WHO-UNICEF, type of drinking water source was categorised as improved or unimproved(19). SPSS version 23 software was used to analyse the data. Anthropometric indices were generated using WHO AnthroPlus software based on WHO-2007 population growth reference data. Adolescent children with z-score below −2 for BMI-for-age and height-for-age indices were considered as thin and stunted, respectively. As commonly done in national Demographic and Health Surveys(20), household wealth index – a proxy indicator of household cumulative living standard – was developed using Principal Component Analysis (PCA) based on ownership of durable household assets and materials used for household construction. DID analysis was used to estimate the net effect of SFP on the three outcomes of interest. DID is an analytic technique for estimating the effect of an intervention based on observational data by taking into consideration the baseline difference between the intervention and control groups and changes overtime in the groups(21). In the present analysis, DID was estimated using simple and multiple mixed-effects linear regression model with a random intercept defined for each school. Initially, the balance of the control variables between the intervention and control groups was assessed using χ2 and independent sample t-tests and in the multivariable models, significantly unbalanced variables were adjusted. The study was implemented in accordance with the Declaration of Helsinki for research involving human subjects. The study protocol was approved by Hawassa University Ethics Committee (Reference No; IRB/003/10) and data were collected after taking written informed consent from the parents of the children. Assent was also secured from the children themselves.