A free lunch or a walk back home? The school food environment and dietary behaviours among children and adolescents in Ghana

Food environments can play an important roles in shaping nutrition and health outcomes. One such environment that has potential to affect youth is the school food environment. In contrast to higher-income countries, however, there is a critical evidence gap on the role of school food environments on children and adolescents in low- and middle-income countries. This mixed-methods study contributes to filling this gap by investigating the role of school food environments on dietary behaviours of children and adolescents in Ghana. It draws on data from household and school questionnaires as well as focus group discussions collected as part of the baseline for an impact evaluation of the Ghana School Feeding Programme (GSFP). Multi-level regression models were fitted with random intercepts at the individual, household and community levels. Excerpts from the focus group discussions provided a deeper understanding of quantitative findings. Children and adolescents who received free school meals provided by the GSFP or who lived further away from school were less likely to go home for lunch. More than half of sampled schools reported offering foods for sale by independent vendors, the most common being meals followed by confectionery, fruit and sugar-sweetened beverages. Predictors of bringing money to school to buy food included non-receipt of free school meals, adolescence, greater commuting distance from home, household asset score, and urban location. Policy efforts focusing on the school food environment may contribute to healthy dietary behaviours for children and adolescents with positive impacts over the lifecourse.

countries, however, there is a critical evidence gap on the role of school food environments on children and 23 adolescents in low-and middle-income countries. This mixed-methods study contributes to filling this gap by

27
Multi-level regression models were fitted with random intercepts at the individual, household and 28 community levels. Excerpts from the focus group discussions provided a deeper understanding of quantitative 29 findings. Children and adolescents who received free school meals provided by the GSFP or who lived further away 30 from school were less likely to go home for lunch. More than half of sampled schools reported offering foods for 31 sale by independent vendors, the most common being meals followed by confectionery, fruit and sugar-sweetened 32 beverages. Predictors of bringing money to school to buy food included non-receipt of free school meals, 33 adolescence, greater commuting distance from home, household asset score, and urban location. Policy efforts 34 focusing on the school food environment may contribute to healthy dietary behaviours for children and adolescents

44
The present study investigates the role of the school food environment on dietary behaviours among 45 children and adolescents in the context of Ghana, a country undergoing the nutrition transition. In doing so it seeks 46 to follow the definition of food environment presented by Herforth and Ahmed (2015). Quantitative data from national survey data primarily shed light on how availability and affordability might shape behaviours. The 48 qualitative data from focus group discussions provide insights into these dimensions as well as convenience and 49 desirability.

50
The national school feeding programme, which provides free meals to selected schools, is a defining

56
Research from high-income countries indicates that schools are an important food environment for children 57 and adolescents (Kubik et al. 2003;Wechsler et al. 2000). In particular, school feeding programmes, which provide 58 food to children in schools on a regular basis, can contribute to improved diets and health (USDA, 2012; Story et al.

59
2002). Food advertising and placement, nutrition and health education, sales of meals and snacks by independent 60 vendors and peer influences may also play a role (Story et al. 2002).

61
Less is known about school food environments in low-and middle-income countries, especially in sub-

72
The bulk of the literature from low-and middle-income countries regarding the school food environment 73 focuses on how the provision of school feeding may impact intra-household food reallocation (Greenhalgh et al. 74 2007). The provisions of food may represent a significant transfer to households and the evidence suggests that 75 households do not respond by providing the child with less food at home, a finding known as the 'flypaper effect' 76 (Greenhalgh et al. 2007). Fewer studies have investigated other dietary behaviours such as bringing money to school 77 to buy food and the types of foods purchased. This behaviour may be associated with the consumption of less 78 healthy foods and worsen overall dietary quality, especially with the widespread shift from under-to over-nutrition 79 and related non-communicable diseases known as the nutrition transition (Popkin 1998). In one study from Jamaica, 80 children who received school meals were as likely to bring money to school as children who did not receive school 81 meals, however purchase patterns were not analysed (Powell et al. 1998). More broadly, several reviews of the 82 literature find that school feeding can contribute to nutrition, health and educational outcomes for children and

84
More research in low-and middle-income countries is merited as the school food environment may offer 85 promising policy levers to counteract the nutrition transition (Popkin 1998

91
Our multi-methods study was guided by the framework set by Herforth and Ahmed (2015), which defines measure and analyse these dimensions using ecological models and social cognitive theory (Glanz 2005

98
Food supplied by school feeding programmes and independent vendors reflect the availability dimension, 99 which may depend on community characteristics such as urbanicity. Affordability is reflected in the cost of these 100 foods in relation to the household income of children and adolescents. Convenience is also likely to influence 101 decision-making, especially for children who live further away from school. Lastly, desirability may be enhanced by 102 factors shaping preferences such as cultural norms that may be established by food consumption patterns in the community and by peers, as well as nutrition education and advertising. Herforth

110
This mixed methods study draws on quantitative and qualitative data from surveys and focus groups. The

116
Ghana was a lower-middle income country at the time the study was conducted. The population includes 117 more than nine major tribes or ethnic groups with over 100 sub-groups. Members of the various ethnic groups share 118 a common cultural heritage, history, language, and origin. The 10 regions of Ghana correlate to some degree with 119 ethnic groups, and can generally be grouped into the North (Upper East, Upper West and Northern regions), which 120 is generally less economically developed than the South (anywhere from the Brong Ahafo Region southwards).

121
Ghana launched a school feeding programme in 2005 that sought to provide a daily, hot meal to children attending 122 selected schools. In 2012, it was estimated that one in three children attending public, primary schools benefited 123 from the programme. The model is known as home-grown school feeding (HGSF) as foods are procured from the 124 community with the objective of promoting the incomes of local smallholder farmers, as well as the nutrition of 125 children. The meals, which typically include a locally produced staple such as rice, cassava, or yam with a local side 126 dish such as groundnut soup or okro stew 1 , are intended to be well-balanced and appeal to local tastes and preferences (Parish and . In 2012, the Ghana School Feeding Programme (GSFP) was retargeted to 128 districts with the highest levels of food insecurity and poverty (Gelli A. et al. 2016

141
(UNICEF 2012). Another investigation using the data from the present study found that average BMI-for-age z-142 score was -0.592 while the average height-for-age z-score was -0.925 .

177
The respondent to the household questionnaire was the head of household or caregiver. As the measures of 178 dietary behaviours were not reported by the children and adolescents themselves, the responses were considered as 179 perceptions. In the school questionnaire, head teachers or caterers reported whether or not the school was in the 180 GSFP, and the availability of certain foods for sale at the school by independent food vendors. The types of food 181 reported for sale were meals, confectionery, fruit and sugar-sweetened beverages.

182
The household questionnaire included a number of modules. In the education module, the respondent 183 reported the dietary behaviours of the child or adolescent during the past week. If the child or adolescent attended 184 school, the respondent was asked if he/she received free meals at school. If the response was positive, she/he was 185 asked how many days a meal was received by the child or adolescent in the past week and whether they consumed 186 less food at home on the days on which they received a free meal at school. The household head was also asked to 187 report the number of days in the past week that the four dietary behaviors were exhibited, as well as the amount of

191
Caregivers also reported if the child or adolescent received free meals at school.

192
The set of variables also included other individual, household and community characteristics that may be  The time and distance to travel to and from school was also reported in the household questionnaire.

208
Children and adolescents who live further away from school may benefit more from having free school meals or 209 foods offered for sale than children who live closer and who can go home more easily for lunch, reflecting the  inclusion of fixed effects in the models was tested with the log-likelihood ratio test. Several sensitivity analyses 231 were undertaken for the multivariate findings. Binary constructions of the outcome variables were also tested which 232 noted if the diet behavior occurred at least once in the past week. In addition, alternative age groupings for 233 childhood and adolescence were tested to investigate the robustness of findings.

234
3 School food environment measures were effectively at the community level as there was one school per community in the data. who reported still being in school and attended kindergarten up to sixth grade in a primary or public school at least 236 one day in the previous week. The sample was then restricted to children and adolescents who reported whether or 237 not they received a free school meal at least once in the previous week as well as other dietary behaviours.

238
Respondents with missing information were more likely to be from urban areas, but otherwise did not significantly 239 differ from non-respondents. Observations with missing values for other variables from the household and school 240 survey were also excluded.

241
In total, the analytic sample included 4,258 children ages five to 17 years from 1,951 households located in        questionnaire. Independent vendors offered foods for sale in more than half (53 percent) of schools. Availability of 301 foods for sale through vendors appeared to be more common in schools without the GSFP (57 versus 39 percent), 302 however, these differences were not statistically significant, perhaps due to a limited sample size.

303
The most common foods sold were meals (48 percent), followed by confectionery (26 percent), fruit (21 304 percent), and sugar-sweetened beverages (11 percent). An estimated 19 percent of schools reported vendors selling 305 one option, the most common being meals, while 18 percent sold two items, the most common being meals and 306 snacks or meals and fruit, and 14 percent sold three items, the most common being meals, confectionery and fruit. beveragesone of which was located in Greater Accra.

370
More than 80 percent of recipients of free school meals from the survey sample, however, continued to eat 371 the same amount of food at home. A range of reasons were provided in the focus group discussions to explain this 372 result. Some caregivers cited parental responsibility to feed their children, while others expressed concern that 373 school meals might be not sufficient to "carry" them until the evening. Some also wanted to ensure that their 374 children were provided with a balanced diet, especially if children found the food given them at school to be

444
The availability of foods for sale, in particular meals (p<0.001) and sugar-sweetened beverages (p<0.05),

445
was also positively associated with bringing money to school in the regression analysis (see

461
Child gender and having a younger sibling less than five years of age did not emerge as significant in any 462 of the regression models, nor did interactions between child gender and age. In the focus group discussions,

495
Household income is also important, reflecting the affordability dimension of the school food environment.

496
The provision of free meals at school may be especially important for households from lower socioeconomic strata

521
As the household questionnaire was conducted at the homes of respondents in the community, some of the 522 children and adolescents in these households may not have been pupils at the schools sampled in the school

523
questionnaire. An analysis of the 2011-2012 EMIS data noted an average of 1.4 schools per community suggesting responses from the household and school surveys, a high correlation (r=0.77) was noted between school 526 participation in GSFP as reported in the school survey and household reporting of free school receipt in the 527 household survey. Differences in reporting may also be due to the regularity in the provision of meals through the 528 GSFP. Schools may indeed be part of the GSFP but not providing meals due to delays in payments to caterers, 529 which was a significant concern at the time of the data collection. 5

530
While the survey data was not nationally representative, the sample drew from all 10 regions of Ghana.

531
Due to the targeting of the GSFP to more food insecure areas, in particular the north and rural areas of Ghana, it is 532 likely that children and adolescents who attended schools that offered GSFP were from households of lower SES

545
The developing evidence base on nutrition value chains, which relate to both supply and demand pathways, can be 546 useful for this purpose . In addition, a food environments framework could be leveraged to other 547 critical issues in school nutrition in low-and middle-income countries such as food safety. More investigation is also  Feeding strategies used to keep children healthy Various food groupings and their health benefits for children; Food hygiene; Provision of meals at home to ensure children eat well, as opposed to buying street food, parental responsibility.

Household feeding
Factors influencing food allocation such as age, gender, monetary contribution, and health status; Advantages and disadvantages of the school feeding programme from the perspective of the children and caregivers; How school feeding may have influenced mode of feeding at home or not. Vitamins and food fortification Awareness of food sources of vitamins; Access and barriers to these foods, and their roles in the diet; Familiarity with commercially packaged vitamin supplements, the concept of food fortification and openness to the concept of the use of MNPs in school meals. Note: Main themes were defined before the data collection while sub-themes emerged after review of the transcripts. 1,683 N/A = not applicable. Obs (N) = number of observations reporting in the affirmative. a Reference period is past week. b Average number of days ranges from 0 to 5 days; c Question posed only to children and adolescents who reported receiving school meals at least once in the past week. The frequency of its occurrence was not reported.  Statistically significant coefficients noted as follows: * p<0.05, ** p<0.01, ***, p<0.001. Coeff = coefficient; SE= standard error; LL= Log-likelihood; ICC=Intra-class correlation; GHS = Ghana cedis. a Dependent variables range from 0 to 5 days and reference period is previous week b Reported in household survey. Ranges from 0 to 5 days and reference period is previous week. c School food environment variables reported in the school survey. d Reference group is children ages 5 to 10 years of age. Young adolescents are 11 to 14 years old and older adolescents are 15 to 17 years old. e A sibling 5 years of age or less. f Commuting distance to and from school in minutes. g Household respondent reported level of education achieved of each household member including household head. Reference group is none or some formal schooling. h Estimated household assets. Calculated using principal components analysis based on reported ownership of durables and access to service.