The Impact of a School-Based Kitchen Garden Program Lisa Gibbs, Phd1; Petra K

Home , School garden

UNDER EMBARGO UNTIL MARCH 7, 2013, 12:01 AM ET Podcast available online at www.jneb.org Research Article Expanding Children’s Food Experiences: The Impact of a School-Based Kitchen Garden Program Lisa Gibbs, PhD1; Petra K. Staiger, PhD2; Britt Johnson, BHSc1; Karen Block, MPH1; Susie Macfarlane, BScPsych2; Lisa Gold, PhD3; Jenny Kulas, BA1; Mardie Townsend, PhD4; Caroline Long, PhD2; Obioha Ukoumunne, PhD5

ABSTRACT Objective: Evaluate achievement of the Stephanie Alexander Kitchen Garden Program in increasing child appreciation of diverse, healthy foods. Design: Comparative 2-year study. Setting: Six program and 6 comparison primary schools in rural and metropolitan Victoria, Australia, matched for socioeconomic status and size. Participants: A total of 764 children in grades 3 to 6 (8–12 years of age) and 562 parents recruited. Retention rates at follow-up included 85% children and 75% parents. Intervention: Each week of the school year, children spent 45 to 60 minutes in a garden class and 90 minutes in a kitchen class. Phenomenon of interest: Program impact on children’s willingness to try new foods, capacity to describe foods, and healthy eating. Analysis: Qualitative data analyzed using inductive thematic analysis. Quantitative data analyzed using random-effects linear regressions adjusted for school clustering. Results: Child and parent qualitative and quantitative measures (if never tried before, odds ratio 2.0; confidence interval, 1.06–3.58) showed increases in children’s reported willingness to try new foods. No differences in articulation of food descriptions (program vs comparison groups). Qualitative evidence showed that the program extended its influence to healthy eating, but this was not reflected in the quantitative evidence. Conclusions and Implications: Findings indicate program success in achieving its primary objective, meriting further program research. Key Words: program evaluation, primary schools, cooking, gardening, child (J Nutr Educ Behav. 2013;45:137-146.)

INTRODUCTION aiming to increase children’s willing- rary effect, which disappeared after 10 ness to try new foods are generally months, on willingness to try new Food choice in humans is often deter- education-based (ie, nutrition educa- foods. They concluded that extensive mined by taste.1 Bitter and sour tastes tion with some tasting component),2,3 recurring exposure to new foods was re- – are common in plant foods, consump- garden-based,4 6 or a combination quired to achieve any lasting change to tion of which is highly desirable in of both.7,8 Nutrition education food preferences rather than cognitive a balanced diet. However, encouraging generally is composed of descriptions education. children to try these foods as a means of foods and information regarding Kitchen garden programs are an- of promoting diversity in diet and nutrition content. Reverdy and other mechanism through which meeting nutritional recommendations colleagues3 conducted taste classes children can be introduced to differ- is often problematic. Interventions and found a strong positive but tempo- ent foods. The nature and extent of existing kitchen garden programs vary; the main differences are related 1Jack Brockhoff Child Health and Wellbeing Program, The McCaughey VicHealth Centre to whether there is a cooking compo- for Community Wellbeing, University of Melbourne, Australia nent in the program and the amount 2School of Psychology, Deakin University, Australia of student involvement in the garden- 3Deakin Health Economics, Deakin University, Australia ing and/or cooking activities. The 4School of Health and Social Development, Deakin University, Australia descriptive terminology varies accord- 5PenCLAHRC, Peninsula College of Medicine and Dentistry, University of Exeter, United ingly. For example, the ‘‘Edible Kingdom Schoolyard’’ describes a seed-to-table Address for correspondence: Lisa Gibbs, PhD, Level 5, 207 Bouverie Street, Carlton, Victoria, program for grade 6 children from 2 Australia 3053; Phone: þ61 3 8344 0920; Fax: þ61 3 9348 2832; E-mail: [email protected] schools in California,9 the ‘‘youth nu- Ó2013 SOCIETY FOR NUTRITION EDUCATION AND BEHAVIOR trition program’’ describes a nutrition http://dx.doi.org/10.1016/j.jneb.2012.09.004 education program conducted within Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013 137 138 Gibbs et al Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013

South West Detroit,10 and the ‘‘school environmentally sustainable garden- consultation with the SAKG Founda- garden program’’ describes a gardening. Children were actively involved tion. A social-ecological theoretical – based program in an elementary in all aspects, from garden design, pre- model31 34 and principles of effective school.11 For the purposes of this re- paring beds, planting seeds, trans- health promotion35 informed the port, the term ‘‘edible garden’’ refers planting seedlings, nurturing the evaluation. Each recognizes the inter- to programs that include active partic- growing plants (including weeding, dependence and impact of physical, ipation in developing and maintain- watering, fertilizing with homemade social, and policy environments on ing a garden with edible plants. The compost and ‘‘worm juice,’’ and apply- individual behaviors and outcomes. term ‘‘kitchen garden program’’ refers ing organic pest control) to harvesting Ozer’s34 social-ecological conceptual to programs that, in addition to a gar- the yield. The cooking component of model suggests that exposure to den with edible plants, include partic- the program included teaching the fresh produce constitutes a proximal ipation in preparing and cooking children to safely wield sharp chef kni- student-level effect of school gardens meals in the kitchen-classroom using ves in preparing and cooking 3- or 4- that in turn may be linked to distal ef- food harvested from the garden. course meals based on available fresh fects such as an increased intake of The increasing popularity of produce from the garden. Different fresh produce associated with reduced community-based kitchen garden pro- dishes prepared each week included risk of obesity and chronic disease. grams has extended into the school handmade pastry, bread, and pasta; A nonrandomized, pre- and post- setting, with widespread reports of salads; curries; and desserts. At the comparison study design was used to school community benefits but limited end of each kitchen class, the children, evaluate the impact of the SAKG Pro- evidence about impact and out- staff, and adult volunteers shared and gram over a 2.5-year period. A mixed – comes,6,12 16 particularly for programs enjoyed the multi-course meal that methods approach was adopted to en- with a dedicated cooking component.9 was prepared. sure the capacity of the evaluation to There are indications that cooking and A primary outcome of the program assess the feasibility, acceptability, gardening programs have positive and subsequently the program evalua- and effect of the SAKG Program.36 outcomes such as engagement in tion was children’s appreciation of The qualitative components of the – the program activities,17 20 increased a diverse range of foods, as indicated evaluation provided an understand- nutrition knowledge,19,21,22 increased by an increased willingness to try ing of the experience and impact of ecoliteracy,4,9 as well as beneficial new foods. A secondary outcome of the program on children, and their at- – impacts of experiential learning.22 24 the program and the evaluation was titudes to food. The quantitative mea- There are also some early indications an increase in children’s capacity to sures provided information about the that gardening and nutrition pro- describe foods, as a way of demon- extent of change occurring as a result grams can increase willingness to taste strating food knowledge and sharing of the SAKG Program. The rationale vegetables4,5,25 or increase preference experiences and appreciation of for this mixed-methods approach in- for vegetables.17,21,22,26 In a review food. This aspect of food appreciation, cluded the potential for triangulation, article by Robinson-O’Brien et al,6 3of assessed in this evaluation, has not or corroboration of findings from evi- 11 garden program evaluations in- been explored in the kitchen garden dence collected with different instru- cluded willingness to taste vegetables literature despite its high profile in ments and from different sources, as or other foods as a measurement of adult assessments of food and restau- well as complementarity, in which re- change. The results of these 3 evalua- rants. Changing child dietary intake sults from 1 method are enhanced tions differed: 1 study reported in- was not an outcome of the SAKG Pro- and clarified by combining with the creases in children’s willingness to gram. Nevertheless, any increase in other.37 Comparison also provides op- taste fruits and vegetables,25 asecond children’s willingness to try new portunities to reflect on and interpret study reported increases for only vege- foods would provide an opportunity findings from different methods and tables,5 and a third found little evidence to promote increased diversity and sources that are inconclusive or in- of an increase.21 More comprehensive healthy eating in their diets. For this consistent.38 The Human Research and rigorous evaluations are re- reason, the authors included assess- Ethics Committees at the relevant quired.6,14,15,27 The authors were ment of healthy eating as a secondary universities and the State Govern- unable to find other published studies outcome of the evaluation. Other ment Department of Education re- that examined the impact of cooking components of the evaluation, in- viewed and approved the project classes or kitchen garden programs on cluding increased knowledge and eco- design. willingness to try new foods. nomic, volunteer, and social impacts, This study provides a comprehen- will be reported elsewhere.29,30 sive evaluation of a school-based Participants and Recruitment kitchen garden program: the Stephanie Alexander Kitchen Garden (SAKG) Pro- METHODS The eligible sample of participants in- gram.28 The program model is embed- Study Design cluded all children in grades 3 to 6 ded in the school curriculum and from the 6 program and 6 comparison includes 45 min/wk in a garden class The SAKG Program was developed schools, a representative parent or with a garden specialist and 90 min/ without specific reference to a theoret- guardian for each participating child, wk in the kitchen with a cooking spe- ical model but the research team was as well as all principals, specialist cialist. The program is designed to able to identify multiple aligned theo- kitchen and garden staff, volunteer give children knowledge and skills in retical frameworks of relevance in adults who assist in the kitchen and Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013 Gibbs et al 139 garden classes, and classroom teachers in the comparison schools partici- also conducted with each of the 6 who are responsible for supervising pated in gardening (and some program school principals and all 10 the children during the program cooking) activities, although for sub- specialist kitchen and garden classes. stantially fewer hours and in a less staff from participating schools The sample size calculation for the structured way than in the program (Figure). quantitative measures was based on schools. This means that the evalua- As is consistent with accepted in- detecting a mean difference of 0.5 of tion needed to establish an impact of ductive methods of inquiry for quali- a standard deviation (effect size) be- the SAKG Program model beyond tative research,41 the semi-structured tween the program and comparison any of these alternative activities focus group discussions and interview groups on continuous measures with rather than compared with schools questions were designed to be open- 80% power at the 5% level of signifi- with no cooking or gardening pursuits. ended, to avoid leading participants cance. A trial in which individual par- Participation in the evaluation was toward particular answers. Partici- ticipants are allocated would require through a voluntary recruitment pro- pants were asked to describe their 63 children (and their parents) in cess in which all eligible staff and fam- level of involvement in the program, each trial group. As clusters (schools) ilies were sent information letters and their expectations, their experiences, were allocated here, this sample size consent forms via the school and re- and what they believed had been the was inflated by the design effect quested to return consent forms to impact of the program on themselves, [Design effect ¼ 1 þ (n – 1) Intra-cluster the school for the research team to on the children, and on the school correlation coefficient],39 where n, the collect. community. These questions elicited number of children sampled from data from all groups of participants each school, was assumed to be 40 relevant to the primary objective con- and the intra-cluster (intra-school) Qualitative Data Collection cerning the impact of the program on correlation coefficient for the out- Procedures and Analyses children’s willingness to try new come was assumed to be 0.05. The de- foods. It also provided additional in- sign effect was then 2.95 and each Qualitative data were collected formation regarding the social im- trial arm required at least 186 children through separate focus group discus- pacts of the program,29 impact on drawn from 5 schools under the in- sions for each of the stakeholder volunteers, and process informa- flated calculation. Adding an extra groups (teacher, parent, and volun- tion,30 reported elsewhere. Focus cluster to each arm to allow for ineffi- teer) at 4 of the program schools. All groups, interviews and observations ciency resulting from imbalance in teachers of grades participating in were not planned for comparison the number of children recruited the SAKG Program were invited to schools because at the beginning of from each school,40 the aim was to re- participate in the focus groups. Focus the evaluation they had no formal cruit 240 children drawn from 6 groups were held after school in place gardening or cooking program to schools in each trial arm. of a scheduled staff meeting to ensure discuss. The 21 SAKG Program schools maximum attendance. The 4 teacher Focus groups and interviews were operating at the time in Victoria, focus groups each had 5 to 10 partici- recorded and transcribed verbatim. Australia, were stratified and 6 were pants. Parents and volunteers were in- Transcripts were coded, categorized, selected to represent a range of charac- vited to participate in separate focus and analyzed using inductive the- teristics in terms of geographic loca- groups via notices in school newslet- matic analysis to identify emerging tion (metropolitan and rural), school ters and letters distributed by school themes and patterns, which were size (small, medium, and large), and administration staff. These focus then further analyzed according to socioeconomic status (represented by groups therefore were composed of their relationship to the existing evi- the percentage of school families re- self-selected parents and volunteers dence base and theoretical perspec- ceiving the government Education who chose to respond to the invita- tives. QSR NVivo qualitative data Maintenance Allowance). Six compar- tions. Four parent and 4 volunteer fo- analysis software (version 8, QSR In- ison schools were then individually cus groups were conducted with ternational Pty Ltd, Doncaster, Victo- matched according to the same crite- participation numbers ranging from ria, Australia, 2008) was used as ria. At the beginning of the study, 5 2 (owing to recruitment difficulties at a data management tool for handling of the 6 comparison schools had their 2 schools) to 13 participants, with oc- the large quantity of data. own edible gardens; by the end, the casional mixed membership because A neutral researcher observer also sixth school also had an edible gar- some parents were also volunteers. conducted class observation at 4 of den. In most of these cases, gardening Two children’s focus groups were the program schools at 3 points was offered on an ad hoc basis, de- held at each of the 6 program schools, over the course of the evaluation. De- pending on the interests of children which resulted in a total of 12 groups, tailed field notes consisted of obser- or teachers. One of the comparison each with 10 to 12 child participants. vations within program classes and schools, however, moved over the Teachers were asked to choose chil- description of children’s attitudes, course of the evaluation from using dren from grades 3 to 6 for these behaviors, interactions, and conver- their garden produce for occasional groups, representing a range of experi- sations relating to their cooking and cooking to developing their own ences with the program from among gardening tasks, and changes in all structured cooking program, in which those for whom consent to participate of these over time. Analysis included all children participated once or twice in the evaluation had already been the researcher’sreflections on the per term. Thus, many of the children obtained. Individual interviews were meanings of what had been 140 Gibbs et al Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013

PROGRAM COMPARISON

Enrollment into Program Enrollment into Comparison

Schools with the SAKG Program Comparison schools selected selected using random sampling using random sampling and Enrollment methods to ensure representation matched comparison sampling of geographic location, methods to ensure similarity to socioeconomic status and school program sample. population size.

Allocated to Program Allocated to Comparison (n = 724) (n = 750)

All children in grades 3-6 All children in grades 3-6 Allocation received the SAKG Program. exposed to some sort of gardening and/or cooking experience, but not the SAKG method.

Recruitment and Baseline Recruitment and Baseline

Interviews with all Consent was obtained and data Consent was obtained and data Interviews with all principals (n = 6). collected from 475 children and collected from 289 children and principals (n = 6). 326 parents (Participation rate: Recruitment 236 parents (Participation rate: Teachers in grade child 65.9%, parent 49.6%). and child 38.5%, parent 31.5%). Teachers in grade 3-5 invited to Baseline 3-5 invited to participate in a Children not available for data Children not available for data participate in a survey (n = 44). collection (n = 12). Parent non- collection (n = 9). Parent non- survey (n = 31). response (n = 45). response (n = 43).

Follow-up Follow-up Interviews with all principals (n = 6) Follow-up data was collected Follow-up data was collected and kitchen and from 360 children and 264 from 252 children and 193 garden staff from parents (Retention rate: child parents (Retention rate: child all schools (n = 10 – 76%, parent 56%). 87%, parent 67%). in 2 schools a single person was doing Children lost to follow-up Children lost to follow-up Interviews with all both roles). (n = 115) (n = 37) principals (n = 6). - 65 no longer attend school Follow-up - 26 no longer attend school Teacher surveys - 9 declined participation - 2 declined participation Teacher surveys (n = 45) - 16 not available - 8 not available (n = 26) - 25 no reason provided - 1 no reason provided Focus Groups: Parent (n = 20) Parents lost to follow-up (n = 61) Parents lost to follow-up (n = 75) Child (n = 124) - 27 no response - 38 no response Volunteer (n = 17) - 34 child no longer attends - 26 child no longer attends Teacher (n = 26) school school

Program analyzed Comparison analyzed (352 children, 186 parents) (240 children, 130 parents)

Excluded from analysis due to Excluded from analysis due to Analysis incomplete data: incomplete data: Children (n = 8) Children (n = 12) Parents (n = 78) Parents (n = 63)

Figure. Flow diagram of recruitment and data collection procedures for the Stephanie Alexander Kitchen Garden (SAKG) Program evaluation. Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013 Gibbs et al 141 observed, and their explanations. ity of the descriptions (eg, use of eval- comparison schools, respectively, These data contributed to the evalua- uative and descriptive words relating whereas the parent retention rates tion findings related to the primary to temperature, color, taste, texture, were 55.6% (n ¼ 264) for program objectives of willingness to try new smell, and emotional impact) and schools and 66.7% (n ¼ 193) for com- foods, and the capacity to describe coded on a 4-point scale (1 ¼ unclear parison schools. The number of food and food experiences, as well [preference only, eg, ‘‘yummy’’]; 2 ¼ teachers who participated in the as to process evaluation findings re- limited [1 dimension, eg, ‘‘cold’’]; teacher survey was relatively consis- ported elsewhere. 3 ¼ basic [more than 1 dimension de- tent at baseline and follow-up for pro- scribed, eg, ‘‘smooth and salty’’]; 4 ¼ gram (n ¼ 44 vs 45) and comparison Quantitative Data Collection sophisticated [more than 1 dimen- schools (n ¼ 31 vs 26). However, Procedures and Analyses sion, clear and sophisticated descrip- many teachers changed roles, and so tion, eg, contrast between crunchy were not able to be involved in the Parent questionnaires included a ques- nuts and a sweet, spicy sauce]). Re- program and evaluation at both base- tion about the child’s willingness to sponses were dichotomized for analy- line and follow-up. The Figure pro- try new foods. The child question- sis by combining levels 1 and 2 into vides a flow diagram of the sample naire extended this to 3 items, which a ‘‘simple’’ category and levels 3 and for each stage of the evaluation. asked about children’s willingness to 4 into a ‘‘more complex’’ category. At baseline, demographic charac- try new foods if they had (1) never Outcomes were compared between teristics were similar between the pro- tried it before, (2) cooked it them- the program and comparison groups. gram and comparison groups. Overall, selves, and (3) grown it themselves. Demographic and baseline character- 54% of participating children and This was modified from a validated istics were summarized using num- 89% of participating parents were fe- scale by Cunningham-Sabo and col- bers or percentages for categorical male, and children had attended their leagues.42 Responses were provided variables, and means and standard respective schools for a mean of 4 on a 4-point Likert scale, from ‘‘never’’ deviations for quantitative variables. years. The majority of parent respon- to ‘‘always.’’ Responses were dichoto- Binary outcomes were compared be- dents were born in Australia or New mized to ‘‘always’’ vs the other 3 cate- tween study groups using the method Zealand and spoke English at home gories combined. Dichotomization of marginal logistic regression models (Table 1). was necessary here and where noted using Generalized Estimating Equa- in the analyses described below, be- tions with information sandwich cause there were insufficient data in standard error (specifying an ex- Qualitative Findings some of the categories to support valid changeable correlation structure), statistical analyses. Parent question- which allowed for within-school clus- The qualitative findings provided an naires also included demographic tering. All analyses were carried out opportunity to capture the perspec- questions and questions about dietary using Stata 10.1 software (release 7.0, tives of children, teachers, parents, intake. Stata Statistical Software, College Sta- volunteers, school principals, and To describe children’s food choices tion, TX, 2000). Fruit and vegetable kitchen and garden specialist staff and food descriptions, children were consumption data were compared about the nature of changes they also asked in the child questionnaire with the 2 servings of fruit and 5 serv- experienced and observed through be- to name and describe their favorite ings of vegetables per day recommen- ing part of the SAKG Program. Four ‘‘savory’’ and ‘‘fruit and vegetable’’ ded in the Australian Guide to Healthy primary themes emerged from the foods. The sophistication of children’s Eating.43 data analysis: children eating and ap- food choices was assessed on a 4-point preciating new foods, the impact of scale according to whether they were the program on student engagement takeaway or processed foods, the RESULTS and learning, the social impacts of number of foods listed, and the com- Participants the program on the broader school plexity of ingredients or flavor (1 ¼ community, and the transfer of pro- takeaway [eg, french fries]; 2 ¼ limited A total of 26 classroom teachers, all gram impacts to the home environ- [common food, few ingredients, and school principals and specialist ment. Qualitative data also provided simple flavors, eg, pasta]; 3 ¼ basic teachers, 17 volunteers, 20 parents, information on processes and imple- [more complex flavors and more and 124 children participated in inter- mentation of the program as well as ingredients, eg, curry chicken]; 4 ¼ views and focus group discussions participants’ willingness to pay for sophisticated [complex combinations (Figure). At baseline, child participa- the program. These latter data arose or less common foods or flavors, eg, tion rates were 65.9% (n ¼ 475) for in response to specific questions de- quinoa, chicken and coriander salad]). program schools and 38.5% (n ¼ signed to inform the economic com- Responses were dichotomized for 289) for comparison schools, whereas ponent of the evaluation. This article analyses by combining levels 1 and 2 parent participation rates were 49.6% reports findings pertaining to the pri- into a ‘‘simple’’ category and levels 3 (n ¼ 326) for the program group and mary objective of the program: its im- and 4 into a ‘‘more complex’’ cate- 31.5% (n ¼ 236) for the comparison pact on children’s appreciation of and gory. The sophistication of the lan- group. Retention rates at follow-up willingness to try new foods. guage children used to describe their for children were 75.8% (n ¼ 360) Children, teachers, parents, volun- favorite foods was assessed according and 87.2% (n ¼ 252) (of baseline par- teers, school principals, and kitchen to the clarity, number, and complex- ticipants) for program schools and and garden specialist staff reported 142 Gibbs et al Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013

try new foods and showed more Table 1. Baseline Characteristics for Program and Comparison Groups awareness of issues of health and nutrition. Children were reported to be Program Comparison willing to try new dishes and were Schools Schools making healthier choices and con- School characteristics suming more vegetables and fruits. School sample (n) 6 6 In some cases, the changes were re- Geographic location of schools ported to be dramatic. One parent Metropolitan 3 3 reported that her child, who had pre- Rural/regional 3 3 viously been reluctant to eat vegeta- Socioeconomic status bles, would now happily help Medium-high 3 3 prepare vegetable soup and discuss Low 3 3 all the vegetables in it while enjoying Size of school (number of students) it. One parent volunteer who spent 2 Medium-high (> 200) 4 4 sessions a week in the kitchen elabo- Low (# 200) 2 2 rated on why the program had made Duration between baseline and follow-up, 17.1 (4.6); 12–25.5 9.7 (1.6); 6–11 such a difference: mo (mean [SD]; range) The speed [of change] has been im- Child characteristics mense. The way they share, dis- Participating students (n) 463 280 cuss, and try food is the standout Female (%) 51.6 55.7 point, the willingness—they are Grade never forced— the fact that they Grade 3 (%) 35.1 23.6 know they can try without having Grade 4 (%) 34.3 35.4 to finish everything on their plate Grade 5 (%) 30.6 23.6 is really important .the peer group Grade 6 (%) 0 17.5 discussion around that promotes Years at school (mean [SD]; range) 3.9 (1.6); 0–6.5 4.1 (1.7), 0–7.5 the willingness to ‘‘have a go.’’ They had silver beet today, and Parent characteristics the other day, a bright pink beet- Participating parents (n) 281 193 root raita without any comment. Female (%) 89.3 88.1 Parent education level Others described how much they Completed # 10 y (%) 13.6 17.6 enjoyed watching children react and Completed 11–12 y or technical 41.9 33.0 change: qualification (%) I had 1 child and he wouldn’t eat Completed university degree (%) 36.8 42.0 the salad, just wouldn’t touch it. Other (%) 7.7 7.5 ‘‘I’m not eating that, it’s leaves!’’ Parents’ country of birth: Australia or 84.2 84.4 and then he tried them and thought New Zealand (%) they were actually quite nice. The Main language spoken at home: 96.8 94.8 week after, he came back and he English (%) said, ‘‘I made that leaf thingy that we made last week and I made it for my mum and she liked it too.’’ How good is that! in the focus group discussions and in- food that children had been bringing terviews that 1 of the most striking to school for snacks and lunches since With few exceptions, children in things evident since the introduction the program had been introduced. focus groups reported they were en- of the SAKG Program was children’s Teachers described feedback from joying trying new foods, were more increased appreciation of and willing- parents about how much more adven- confident in trying new foods, and ness to try new foods. One principal turous the children had become with were now eating a wider range of reported how a young boy in a kitchen both food and cooking. One teacher de- food than previously. They talked class had exclaimed in amazement: scribed a mother’s reaction as follows: about eating more vegetables in par- ‘‘This tastes better than Maccas ticular. They thought the food they She was really surprised. Just last [McDonalds]!’’ were eating now was healthier, and week she was in the kitchen help- According to teachers, the program many also said they were eating less ing and she is in raptures that the introduced children to new ingredi- ‘‘junk food.’’ They reported their par- kids just sit down and eat all of ents and tastes, and within a short ents were happy about these changes. these green, leafy vegetables and time almost all children were prepared Children also discussed how they en- enjoy their salads . to at least try a new dish. Teachers at joyed trying foods from different cul- several schools also reported that Comments from parents and vol- tures, and mentioned Mediterranean, they had seen a noticeable improve- unteers in focus groups agreed that Asian, and Moroccan as examples. ment in the nutritional quality of the children had become more willing to Children often indicated their Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013 Gibbs et al 143 appreciation of the fact the food they At follow-up, adjusting for baseline ment of the class increased children’s grew was organic. The children also and grade, children’s reports showed willingness to try new foods. This in- commented that they could taste increased odds of willingness to try cluded sitting down together to share ‘‘the freshness’’ and the fruits and veg- new foods if they had never tried it and enjoy the meal that they had pre- etables tasted better than those from (odds ratio [OR] 1.95, 95% confidence pared, with encouragement to taste the supermarket. interval [CI]: 1.06–3.58, P ¼ .03), had but no pressure to eat. This supports Many of the children stressed that cooked it (OR 2.37, 95% CI: 1.45– Birch and colleagues’ suggestion that changes to their eating habits had oc- 3.90, P ¼ .001), and had grown it social environments may also have curred since they had begun partici- (OR 2.25, 95% CI: 1.47–3.47, P < a role in shaping this behavior. It is pating in the program. Children .001) (Table 2). Parent questionnaire also consistent with the principles of reported they were not only eating data did not show evidence of a differ- ecological theory and health promo- new foods, they were also enjoying ence between the program and com- tion in terms of the interdependence the experience: ‘‘It was fun eating all parison groups regarding whether between individuals’ health behaviors of the new foods;’’ ‘‘I just like to children were ‘‘always’’ willing to try and their physical and social environ- know that you’re eating your own new foods (OR 1.69, 95% CI: 0.93– ment.31,35 stuff that you’ve been growing and 3.09, P ¼ .09) (Table 2). The quantitative evaluation mea- to be able to try new things that There was little evidence that the sures also supported this finding, you’ve never tried before .’’ SAKG Program was associated with which showed evidence of differences The class observation sessions also an increase in the capacity of children between children from program demonstrated that most children to describe foods, or in the proportion schools and comparison schools in were willing to try all of the new of children reported by parents to be child-reported willingness to try new foods. Only 1 child was observed to meeting consumption guidelines for foods, especially if they had grown eat nothing and only a handful would fruit, vegetable, and soft drink con- or cooked it themselves. This proxi- bypass even 1 item of food. In the ini- sumption (Table 2). Questionnaire re- mal effect of the school garden repli- tial round of class observations, sults showed that whereas over 70% cates Ozer’s34 model of the proximal teachers and volunteers were over- of children from both program and and distal effects of school gardens, heard urging the children to taste comparison groups were eating 2 serv- and is consistent with previous the food. By the third round of class ings of fruit at both baseline and research24 indicating a positive corre- observations (roughly a year later), follow-up, fewer than 10% were eat- lation among children between plant- no such urging was heard; there was ing 5 servings of vegetables per day ing and harvesting vegetables and an no need, because children who did at either baseline or follow-up43 increase in their willingness to taste not try the food were clearly the ex- (Table 2). them. The difference between pro- ception. It was interesting to observe gram and comparison schools’ base- during this last round of observations line child reports for willingness to that at 1 school, the children at the ta- DISCUSSION try new food may reflect the fact that ble asked the adult sitting with them children in some program schools why 1 child did not taste the food. Children’s increased willingness to try had just started participating in In contrast, the program appeared new foods emerged as a dominant program activities at the time of base- to have little impact on children’s ca- theme in all of the interviews, focus line data collection. These baseline pacity to describe foods, despite obser- group discussions, and class observa- differences were adjusted for in the vation of class exercises and specific tions from the program schools. Statis- analyses. education about language to describe tical evidence in the child-reported In contrast to the child question- tastes, textures, and smells. The chil- data of this program effect supported naires, parent questionnaire findings dren tended to use simple words these qualitative findings and showed on whether children were ‘‘always’’ such as ‘‘beautiful,’’ ‘‘nice,’’ and that the odds of reporting they were willing to try new foods showed little ‘‘yummy.’’ Chocolate tasted ‘‘choco- always willing to try new foods was evidence of a true difference between laty,’’ ice cream was ‘‘cold,’’ and spa- around twice as great for children in the program and comparison groups. ghetti was ‘‘stringy.’’ the program group compared with Both qualitative4,25 and quantitative5 the comparison group. Cooking and evidence of this effect has been docu- then sharing the meals was a core mented in previous studies of edible Quantitative Findings and therefore recurring feature of the gardens. The measure in the parent kitchen classes. However, the variety questionnaire was simplified com- At baseline, reports of child fruit and of the weekly menus meant that the pared with the child questionnaire. vegetable intake or descriptions of children were being exposed to It did not account for when children food were similar between the pro- a wide diversity of foods, rather than had grown or prepared the food. gram and comparison groups, but repeat tastings of the same food, which Therefore, it may not have had the there were differences in baseline child Birch and colleagues44 recommended sensitivity to reflect changes arising reports for willingness to try new food as an essential component of interven- from the program. if they had cooked it (38.5% vs 31.7%, tions to increase willingness to try new Teachers, parents, volunteers, and respectively; P ¼ .06) or grown it in the foods. Teacher and volunteer qualita- children reported in focus groups and garden (33.4% vs 25.1%, respectively; tive data and class observations interviews that they noticed a change P ¼ .02). instead suggest that the social environ- in children’s attitude toward and 144 Gibbs et al Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013

consumption of vegetables after the introduction of the SAKG Program, ICC 0.014 0.023 0.034 but the parent questionnaire results * * did not reﬂect these changes. Previous *

P studies indicate a perceived increase in .001 daily vegetable consumption, based < on child participant survey results,45,46 24-hour recall records,23 and lunchroom observations.47 How- ever, some researchers have noted that there is limited evidence of nutritional behavioral changes as a result of edible and kitchen garden programs.24 The strong qualitative ﬁndings in the SAKG Program evaluation may have occurred because the changes reported qualitatively reﬂected the more dra- Unadjusted Adjusted matic changes affecting the subgroup of children who initially showed greater reluctance to try new foods, rather than changes at the population level. This may mean that the program effect lies more in reducing inequities than in overall population shift. It Follow-up may also be because the observed changes were still occurring mostly within the school environment in the context of the SAKG Program, and the home environment was offer- ing only limited opportunities for children to demonstrate their willingness to try new foods. Further research is required to explore whether the impact

Baseline of the program extended to the home environment. The qualitative evaluation findings % n % n % n % n OR OR 95% CI Program Comparison Program Comparison may also reflect the beginning of a change and may become more evident over time. One study suggests that far fewer hours of nutrition education are required to change participants’ health knowledge (about 15 hours) relative to the time requisite to alter health behavior (about 50 hours).48 Therefore, other changes such as increased healthy eating are more likely to become evident over a longer period of exposure to the pro- refers to the comparison between the program group and the comparison group at follow up in analyses that are unadjusted and then fi

P gram. The qualitative nding that children appreciated the freshness of fruit and vegetables, as well as that they were grown organically, is consistent with that of Somerset and Markwell,45 who found evidence of an increase in the number of students (grades 4 and 6) who reported that edible garden produce tasted better than

Results of Outcomes at Baseline and Follow-up and Adjusted Differences between Program and Comparison at Follow-up store-bought alternatives. Such reports may indicate an enhanced en- .05 is considered significant. gagement, awareness, and apprecia- < Grown it 33.4 458 25.1 275 38.7 351 22.9 244 2.10 2.25 1.47-3.47 Never tried it before 29.8 463 25.3 277 35.4 353 24.0 245 2.05 1.95 1.06-3.58 .03 Cooked it 38.5tion 460 31.7 of 278 the 51.0 354 fruit 33.6 245 and vegetables 1.92 2.37 1.45-3.90 that .001 Parent report that child alwaysOutcome willing 1.2: to Food try choices new andChild foods food provides descriptions ‘‘more complex’’ descriptionChild of provides favorite ‘‘more savory complex’’ food descriptionChild of reports favorite more fruit/vegetable complex favoriteChild savory reports food ‘‘more 6.0 complex’’ 9.4 favoriteOutcome fruit/vegetable 404 1.3: 451 Food and beverageChild intakes 10.8 eats at 6.3 least 23.9 2Child servings eats 280 of at 278 fruit/d least 279 5Child servings drinks 23.8 of no 56.9 vegetables/d sweet 13.6 drinks/d 347 344CI 193 indicates 53.1 confidence interval; 10.4 ICC,* P 34.2 intra-cluster correlation 19.6 260 coefficient; 240 OR,adjusted odds 226 440 for ratio. baseline 25.8 16.3 outcome scores 23.6 and 442 grade and school 158 clustering. 0.99 18.8 274 1.08 53.0 278 351 1.11 1.50 22.8 1.31 0.46-2.65 0.46-3.78 56.2 355 .78 19.7 1.69 244 .62 7.7 0.93-3.09 273 0.085 244 84.2 0.457 268 .09 5.9 0.80 74.6 0 1.13 189 189 0.98 0.44-2.15 7.3 79.8 74.1 1.40 256 255 275 0.75-2.61 .95 72.5 76.2 9.5 .29 0.095 155 188 153 0.037 75.6 251 1.50 68.1 0.75 158 1.68 0.87 0.90-3.14 0.54-1.42 .11 1.76 .59 0 0 1.33 0.70-2.5 .38 0.073 Outcome 1.1: Willingness to tryChild foods will always try a new food if has: Table 2. participants grow, cook, and Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013 Gibbs et al 145 consume. Ultimately, this may shift There could also be a participant re- ucation on willingness to taste novel food their preferences to homegrown or lo- cruitment bias in the study, especially in children. Appetite. 2008;51:156-165. cally grown organic fruit and vegeta- the qualitative aspects, because those 4. Lautenschlager L, Smith C. Beliefs, bles if they are accessible in the with more positive comments may knowledge, and values held by inner- home environment. be more likely to participate. Potential city youth about gardening, nutrition, The lack of evidence of a difference sources of bias of parent-reported and cooking. Agriculture and Human between program and comparison child food records include those re- Values. 2007;24:245-258. groups in children’s capacity to de- lated to social desirability, serving 5. Morris JL, Neustadter A, Zidenberg- scribe food suggests that the capacity size estimation, and parental recall of Cherr S. First-grade gardeners more increases with age, in line with in- the child’s diet.49 However, it is antic- likely to taste vegetables. California Ag- creasing cognitive and language abil- ipated that the level of any of these riculture. 2001;55:43-46. ity, and is not affected by the SAKG potential biases would be similar be- 6. Robinson-O’Brien R, Story M, Heim S. Program. These findings demonstrate tween the program and comparison Impact of garden-based youth nutrition the importance of comparative evalu- groups. As such, although the intervention programs: a review. JAm ation to ensure that changes that oc- amounts reported may not be com- Diet Assoc. 2009;109:273-280. cur naturally are not attributed to pletely accurate, the analysis of the 7. Jaenke RL, Collins CE, Morgan PJ, the intervention. differences between the groups Lubans DR, Saunders KL, Warren JM. Several limitations exist within this should be valid. The impact of a school garden and cook- study. This study was not a random- ing program on boys’ and girls’ fruit and ized, controlled trial (RCT). Based on vegetable preferences, taste rating and the findings of this comparative IMPLICATIONS FOR intake. Health Educ Behav. 2012;39: study, an RCT is recommended to RESEARCH AND 131-141. confirm or refute the results of this PRACTICE 8. Morgan PJ, Warren JM, Lubans DR, study. Saunders K, Quick G, Collins CE. fi Enthusiasm in the program schools The strength of the ndings in terms The impact of nutrition education ’ for the SAKG Program associated with of children s increased willingness to with and without a school garden on the celebrity factor of Stephanie Alex- try new foods provides preliminary knowledge, vegetable intake and pref- ander’s involvement may have in- evidence of the success of the SAKG erences and quality of school life among creased recruitment and resulted in Program in achieving its primary ob- primary-school students. Public Health a potential bias in responses. In con- jective, and is consistent with relevant Nutr. 2010;13:1931-1940. trast, some teachers and parents re- theoretical frameworks and the exist- 9. Murphy JM. Education for Sustainability: ported that they were initially ing limited evidence base. Neverthe- Findings from the Evaluation Study of concerned that the program would less, the link between increased The Edible Schoolyard. Berkeley, CA: take time away from academic sub- willingness to try new foods and in- Centre for Ecoliteracy; 2003. jects. However, they found that the creased fruit and vegetable intake is 10. Pothukuchi K. Hortaliza: youth ‘‘nutri- program actually supported children’s inconclusive. Although changed die- tion garden’’ in Southwest Detroit. Chil- personal and academic development. tary intake is not an aim of the SAKG dren, Youth and Environments. 2004;14: Only schools that had established Program, the seriousness of the cur- 124-155. their own edible gardens, or were in rent prevalence of child overweight 11. Robinson CW, Zajicek JM. Growing the process of doing so, agreed to par- and obesity and the success of the pro- minds: the effects of a one-year school ’ ticipate as comparison schools. In ad- gram in increasing children s willing- garden program on six constructs of dition, 1 comparison school moved ness to try new foods merit further life skills of elementary school chil- from using its garden produce for investigation of the program using dren. HortTechnology. 2005;15: occasional cooking to developing its an RCT as the program is extended 453-457. own more structured cooking pro- to schools throughout Australia and 12. Lautenschlager L, Smith C. Under- gram over the course of the evalua- to the United Kingdom, to explore standing gardening and dietary habits tion. This means that the evaluation its impact on changed dietary intake among youth garden program partici- needed to establish the impact of the over time. pants using the Theory of Planned Be- SAKG Program was greater than any havior. Appetite. 2007;49:122-130. alternative program being conducted 13. Murphy JM, Schweers E. Evaluation of in comparison schools. The delay in REFERENCES a Food Systems-based Approach to Fostering recruiting comparison schools re- Ecological Literacy: Final Report.Cam- duced the duration between baseline 1. Birch LL, Fisher JO. Development of bridge: Massachusetts General Hospital, and follow-up data collection, and re- eating behaviors among children and ad- Harvard Medical School; 2003. sulted in program schools having an olescents. Pediatrics. 1998;101:539-549. 14. Ozer EJ. The effects of school gardens average of 7 months more before 2. Martins Y, Pelchat ML, Pliner P. ‘‘Try on students and schools: conceptualiza- follow-up assessment. Because chil- it; it’s good and it’s good for you’’: ef- tion and considerations for maximizing dren’s developmental advances could fects of taste and nutrition information healthy development. Health Educ Be- also have been responsible for the im- on willingness to try novel foods. Appe- hav. 2007;34:846-863. provement seen in program schools, tite. 1997;28:89-102. 15. Phibbs EJ, Relf D. Improving research this discrepancy was adjusted for in 3. Reverdy C, Chesnel F, Schlich P, on youth gardening. HortTechnology. the analysis. Koster€ EP, Lange C. Effect of sensory ed- 2005;15:425-428. 146 Gibbs et al Journal of Nutrition Education and Behavior Volume 45, Number 2, 2013

16. Skelly SM, Bradley JC. The impor- 27. Blair D. The child in the garden: an Research. London, United Kingdom: tance of school gardens as perceived evaluative review of the benefits of Arnold; 2000. by Florida elementary school teachers. school gardening. J Environ Educ. 40. Eldridge SM, Ashby D, Kerry S. Sam- HortTechnology. 2000;10:229-231. 2009;40:15-36. ple size for cluster randomized trials: ef- 17. Heim S, Stang J, Ireland M. A gar- 28. Stephanie Alexander Kitchen Garden fect of coefficient of variation of cluster den pilot project enhances fruit and Foundation. About the Program. http:// size and analysis method. Int J Epide- vegetable consumption among chil- www.kitchengardenfoundation.org.au/ miol. 2006;35:1292-1300. dren. J Am Diet Assoc. 2009;109: about-the-program. Accessed November 41. Patton MQ. Qualitative research and 1220-1226. 12, 2012. evaluation methods. Thousand Oaks, 18. Morris JL, Koumjian KL, Briggs M, 29. Block K, Gibbs L, Staiger P, et al. CA: Sage Publications; 2002. Zidenberg-Cherr S. Nutrition to Growing community: the impact of 42. Cunningham-Sabo L, Marr-Lyon L, grow on: a garden-enhanced nutrition the Stephanie Alexander Kitchen Gar- Hood N, Walters L. Classroom cooking education curriculum for upper- den Program on the social and learning program’s impact on children’s cooking elementary schoolchildren. J Nutr Educ environment in primary schools. Health attitudes, cooking self-efficacy, and food Behav. 2002;34:175-176. Educ Behav. 2012;39:419-432. neophobia: development and testing of 19. Newell S, Huddy AD, Adams JK, 30. Gibbs L, Staiger PK, Townsend M, et al. a quantitative evaluation instrument. Miller M, Holden L, Dietrich UC. Methodology for the evaluation of the JNutrEducBehav. 2004;36(suppl 1). S69. The tooty fruity vegie project: chang- Stephanie Alexander Kitchen Garden 43. Smith A, Kellett E, Schmerlaib Y. Aus- ing knowledge and attitudes about Program. Health Promot J Austr. In press. tralian Guide to Healthy Eating.Canberra: fruits and vegetables. Aust N Z J Public http://dx.doi.org/10.1071/HE12905. Commonwealth of Australia; 1998. Health. 2004;28:288-295. 31. Bronfenbrenner U. The Ecology of Hu- 44. Birch LL, McPhee L, Shoba BC, 20. Alexander J, North M-W, man Development: Experiments by Nature Pirok E, Steinberg L. What kind of ex- Hendren DK. Master gardener class- and Design. Cambridge, MA: Harvard posure reduces children’s food neopho- room garden project: an evaluation of University Press; 1979. bia? Looking vs. tasting. Appetite. 1987; the benefits to children. Children’s Envi- 32. Hawe P, Riley T. Ecological theory in 9:171-178. ronments. 1995;12:123-133. practice: illustrations from a commu- 45. Somerset S, Markwell K. Impact of 21. Morris JL, Zidenberg-Cherr S. Garden- nity-based intervention to promote the a school based food garden on attitudes enhanced nutrition curriculum im- health of recent mothers. Prev Sci. 2005; and identification skills regarding vege- proves fourth-grade school children’s 6:227-236. tables and fruit: a 12-month interven- knowledge of nutrition and preferences 33. McLaren L, Hawe P. Ecological per- tion trial. Public Health Nutr. 2008;12: for some vegetables. J Am Diet Assoc. spectives in health research. J Epidemiol 214-221. 2002;102:91-93. Commun Health. 2005;59:6-14. 46. Hermann JR, Parker SP, Brown BJ, 22. Liquori T, Koch PD, Contento IR, 34. Ozer EJ. The effects of school gardens Siewe YJ, Denney BA, Walker SJ. Af- Castle J. The Cookshop Program: out- on students and schools: conceptualiza- ter-school gardening improves chil- come evaluation of a nutrition education and considerations for maximizing dren’s reported vegetable intake and tion program linking lunchroom food healthy development. Health Educ Be- physical activity. J Nutr Educ Behav. experiences with classroom cooking hav. 2007;34:846-863. 2006;38:201-202. experiences. J Nutr Educ. 1998;30: 35. World Health Organization. Ottawa 47. Parmer SM, Salisbury-Glennon J, 302-313. Charter for Health Promotion.Ottawa, Shannon D, Struempler B. School gar- 23. McAleese JD, Rankin LL. Garden- Canada: Department of Health and Wel- dens: an experiential learning approach based nutrition education affects fruit fare, World Health Organization; 1986. for a nutrition education program to in- and vegetable consumption in sixth- 36. Baum F. Researching public health: be- crease fruit and vegetable knowledge, grade adolescents. J Am Diet Assoc. hind the qualitative-quantitative meth- preference, and consumption among 2007;107:662-665. odological debate. Soc Sci Med. 1995;40: second-grade students. J Nutr Educ Be- 24. Morris JL, Briggs M, Zidenberg- 459-468. hav. 2009;41:212-217. Cherr S. School-based gardens can 37. Bryman A. Integrating quantitative 48. Connell DB, Turner RR, Mason EF. teach kids healthier eating habits. Cali- and qualitative research: how is it Summary of findings of the school fornia Agriculture. 2000;54:40-46. done? Qualitative Research. 2006;6:97. health education evaluation: health 25. Cason KL. Children are ‘‘growing 38. Moffatt S, White M, Mackintosh J, promotion effectiveness, implementa- healthy’’ in South Carolina. J Nutr Howel D. Using quantitative and qual- tion, and costs. J Sch Health. 1985;55: Educ. 1999;31:235-236. itative data in health services research— 316-321. 26. Lineberger SE, Zajicek JM. School gar- What happens when mixed method 49. Livingstone MBE, Robson PJ, dens: can a hands-on teaching tool af- findings conflict. BMC Health Serv Wallace JMW. Issues in dietary in- fect students’ attitudes and behaviors Res. 2006;6:1472-6963. take assessment of children and ado- regarding fruit and vegetables? Hort- 39. Donner A, Klar N. Design and Analysis lescents. Br J Nutr. 2004;92(suppl 2): Technology. 2000;10:593-597. of Cluster Randomization Trials in Health S213-S222.