**Development of RisObIn.Com, a Screening Tool for Risk of Childhood Obesity in the Community**

**Ana Catarina Moreira 1,2,\* , Patrícia Almeida Oliveira 1,3 , Rute Borrego <sup>1</sup> , Telma Nogueira 4,5 , Raquel Ferreira 1,6 and Daniel Virella <sup>7</sup>**


Received: 28 September 2020; Accepted: 23 October 2020; Published: 27 October 2020

**Abstract:** The prevalence of childhood overweight has increased considerably in the past three decades and there is evidence that childhood obesity can persist into adulthood. A simple tool to identify relevant risk factors may alert families and prevent overweight and obesity. This study aims to develop a pre-school screening tool to assess the risk of childhood obesity. Child anthropometric measurements and several risk factors for childhood obesity factors were obtained. The effect of the variables on the outcome of obesity (defined as increased anthropometry-estimated adiposity) was assessed by binary logistic regression analyses. The identified variables were submitted for expert panel validation and combined for the tool development. A total of 304 children were included. Eight items were included in the tool. A higher score of the tool indicates a greater risk for obesity in childhood with the cutoff point set at 0. The tool sensitivity for obesity was 95%, specificity was 74.4%, the positive predictive value was 37.3%, and negative predictive value was 98.9%. The Risk of childhood Obesity In the Community (RisObIn.Com) tool is proposed to be a comprehensive tool to identify children at high risk for late childhood obesity at admission to primary school. Further studies are needed to assess the performance of the tool.

**Keywords:** childhood overweight; children; risk; community; screening; tool

#### **1. Introduction**

The prevalence of childhood overweight and obesity has increased considerably in the past decades, mostly in high-income countries but recently also rising in low- and middle-income countries. Although in high-income countries, a recent decrease has been observed [1], prevalence remains very high [2,3]. Obesity is defined as an abnormal fat accumulation that impairs health [4] but it has been shown that obesity defined by Body Mass Index (BMI) alone is a remarkably heterogeneous condition with varying cardiovascular and metabolic manifestations across individuals, which may differ according to age and gender [5]. This is a chronic disease that increases heavily the burden on citizens, health care systems, productivity, cities, and society and should be considered a top priority and main target to combat the increasing non-communicable diseases epidemic [6]. There is evidence that childhood obesity can persist into adulthood [7], lead to physical obesity-related complications,

and affect psychological health and social and emotional well-being [8]. This emphasizes the importance of early intervention to prevent the onset of obesity in childhood. A comprehensive and proactive strategy to deal with the challenges imposed by the obesity epidemic is needed and requires the development and implementation of programs for prevention, early diagnosis, and treatment, especially in children [6]. However, interventions to reduce childhood obesity show limited effectiveness, particularly for weight-related outcomes [9].

Therefore, sustainable and effective interventions to prevent childhood obesity should target higher-risk children [10]. Obesity development involves a complex interplay between physiological environmental, psychological, social, and behavioral exposures [11,12]. There is evidence of epigenetic processes in utero that contribute to infant obesity, including DNA methylation, and gut microbiome alterations [13]. Breastfeeding is also associated with obesity protection [14,15]. Additional life course exposures include socio-economic status, food production and marketing, food insecurity, and obesogenic environments, that can promote unhealthy lifestyles. In this environment, some individuals are genetically more susceptible to develop obesity [16].

A simple tool to identify relevant obesity risk factors early in life may alert families and caretakers into positive changes, improving a child's weight trajectory and preventing overweight and obesity. Tools to identify children at risk for obesity have been published [17–19]. To our best knowledge, none of these include a large diversity of parameters known to affect weight gain trajectory; the broad variables related to obesity require a more comprehensive tool.

This study aims to develop a pre-school screening tool to assess the risk for childhood obesity based on a broad spectrum of risk factors considering peri-natal, anthropometric, sociodemographic, past eating habits, current eating habits, subjective anthropometry perception, subjective eating habits perception and physical activity, and sleeping habits, at a multivariable level.

#### **2. Materials and Methods**

Data from the community-based participatory research Sintra Grows Healthy (SGH) [20] was used for independent analysis, as a cross-sectional with nested case-control study. The study sample comprises schoolchildren aged 6–12 years attending the first to fourth grades of six public primary schools in Sintra municipality, Portugal. Anthropometric measurements of children were assessed and a wide set of data was obtained by the application of a questionnaire to the children's legal guardian, mainly their parents. For the purpose of the present study, an additional set of questions relevant for the study of obesity risk factors was collected by applying a second questionnaire to the children's legal guardian. Only children for whom both these questionnaires were filled, were selected for this study. Data were collected between 2017 and 2018. Written informed consent was obtained and the safety and confidentiality of all the collected and archived data were ensured. Approval was obtained from the National Commission of Data Protection and the Ethics Commission of Lisbon Academic Medical Center.

Anthropometric measurements were obtained directly by trained members of the SGH research team, using standardized anthropometric procedures [21]. Children were barefoot and wearing minimal clothing to assess height and weight. Height was assessed using a portable stadiometer to the nearest 0.1 cm (SECA 213®) in the vertical position, with feet together and the head in the Frankfort plane. Weight was assessed through a portable calibrated scale (SECA Robusta 813®, SECA Deutschland, Hamburg, Germany), expressed up to 0.1 kg. Body Mass Index (BMI) was calculated as weight (kg) divided by height squared (cm<sup>2</sup> ). BMI was classified according to age and gender z-scores of the World Health Organization for children aged 5 to 19 years old [22]. Waist circumference was directly measured on the skin to the nearest 0.1 cm according to the World Health Organization method with a non-extensible and flexible tape (SECA 201®, SECA Deutschland, Hamburg, Germany) [23]. Waist-to-Height Ratio (WtHR) was calculated as waist circumference (cm) divided by height (cm) and classified as an indicator for early health risk as ≤0.5 or >0.5 [24].

The set of questions specifically included for the present study were gathered and developed through a literature review regarding childhood obesity [11,12,16]. Children were asked to fill the questionnaire at home with their parents. To assure data confidentiality, each child was assigned a subject identification code. The questionnaires were collected by teachers and sent back to the SGH team. Data entry and revision was conducted through a standardized procedure. The information collected for the present analysis included (a) parental nationality; (b) parental level of education; (c) family type (nuclear/extended two-/one-parented); (d) number and age of siblings; (e) mean monthly income; (f) parental current employment status; (g) present parental weight and height (from which BMI was calculated and categorized into underweight, normal weight, overweight, or obese) and (h) father's and mother's body image perception. The caretakers were asked to recall peri-natal information: (a) maternal weight before and after pregnancy (from which BMI pre-pregnancy and adequacy of weight gain during pregnancy were calculated according to the Institute of Medicine recommendations [25]); (b) maternal tobacco consumption during pregnancy; (c) diagnosis of gestational diabetes and/or pre-eclampsia; (d) information given by the assistant physician during pregnancy on adequateness of the fetus for gestational age; and (e) the gestational age in complete weeks (to determine if the birth was pre-term or term). Information regarding anthropometric data throughout childhood was retrieved from the records in the child health bulletin: (a) birth weight and length [from which BMI was calculated through World Health Organization Anthro software for Windows, version 3.2.2. (World Health Organization—Department of Nutrition, Geneva, Switzerland), and classification into small for gestational age (SGA, < 10th percentile), appropriate for gestational age (10th–89th percentile), and large for gestational age (LGA, > 90th percentile) were obtained]; (b) weight and length at the ages of 12, 18, and 24 months and 3 and 5 years old [from which BMI z-score was calculated through the World Health Organization Anthro software for Windows, version 3.2.2. (World Health Organization—Department of Nutrition, Geneva, Switzerland)]. Overweight (including obesity) was defined according to age and sex z-scores (above 2) of the World Health Organization for children up to 5 years old [26]; since the medical visits from birth to 5-years-old may not have occurred at the exact dates selected to recall anthropometric data, the health record information closer to those ages and respective dates were collected to correctly assess z-scores. To classify anthropometric measurements at each visit, the exact age was calculated by the difference between birth date and the records visit date.

We collected data on child and family feeding patterns at two moments. We asked about breastfeeding (total and exclusive duration) and the introduction of solid foods (age, appetite, and type of meal first introduced). For present feeding pattern, we asked about: child's appetite, the Mediterranean diet pattern index of the child (KIDMED [27]) through adapted questions, and its family (PREDIMED [28]), and one question to both child and caretakers regarding child's intake when worried, irritated, or anxious, extracted from the from Child Eating Behaviour Questionnaire [29]. To assess the child's and caretaker's perception of the quantity of child usual intake, we used images of four meal plates with different portions of food (A to D, ascendingly). According to children's height-for-age, we determined the two images closer to their recommended portion (z-score ≤ 1 corresponded to images A and B; z-score > 1 corresponded to images B and C). Image D represented an excessive food portion for any children of our sample. We compared the adequacy of the caretaker's answers and classified them as adequate, excessive, or lower. Caretakers were asked to select an option regarding the child's nutritional status for age between "low weight," "adequate weight," or "excessive weight" and compared to the child's BMI. This perception was categorized as correct, underestimated, or overestimated.

Children and caretakers identified the child's body figure through body image scales [30]. We compared that perception with the corresponding child's BMI and categorized it as incorrect, relatively correct, or correct to child's BMI, as previously done [31]. The perceptions were additionally categorized as correct, underestimated, or overestimated. Caretakers also identified their own body figures through body image scales [32]. We compared that perception with the corresponding parental

BMI and categorized it as incorrect, relatively correct, or correct to child's BMI, as previously done [33]. The perceptions were additionally categorized as correct, underestimated, or overestimated.

We questioned about the physical activity and sedentary behaviors of the child, the frequency of consumption of meals in front of a screen, and the number of hours of sleep.

Variables were grouped into eight dimensions: peri-natal, anthropometric, sociodemographic, previous eating habits, current eating habits, subjective anthropometry (perception), subjective eating habits (perception), physical activity, and hours of sleep.

#### *2.1. Development of the Risk Index Tool and Scoring*

As the BMI, as a single measurement of obesity, does not reflect the whole complexity of the condition [6], the European Association for the Study of Obesity proposed to improve the diagnostic criteria for obesity with the inclusion of other dimensions, including the degree of adiposity [34]. Therefore, to increase the accuracy of the outcome measure, to reflect an adiposity-based condition, a composite variable was created using BMI and WtHr. Thus, the primary outcome measure in this study is increased adiposity, defined as overweight (including obesity) with WtHr > 0.5, while in primary school. To test the effect of the factors under investigation on the primary outcome, binary logistic regression analyses were performed. Exposures were tested within the aforementioned dimensions (dependent variable: overweight (including obesity) with WtHr > 0.5; factors: all risk factors by dimension).

The regression analysis was used to identify factors associated with the primary outcome on each of the eight dimensions. The studied risk factors that showed an association with our primary outcome were presented to an expert pediatric panel (nutrition, education, nursing, pediatrician, and exercise physiology experts) for external construct validation with the purpose of developing the proposed tool: RisObIn.Com (Risk of childhood Obesity In the Community).

The most agreed risk factor variables were then combined to develop the RisObIn.Com tool. At least one item from every considered dimension was included in the score if any of the items revealed significant statistical relevance. The Beta (β) values to a decimal case were used to generate the scoring system as an indicator of the association between each variable, and 0 and 1 scores were assigned to the response option regarding their association with the outcome variable (overweight including obesity with WtHr > 0.5). As an example, on the physical activity item, a score of 0 was assigned to the response option "plays regular and programmed physical activity," and a score of 1 to the response option "doesn't play regular and programmed physical activity." This score was then multiplied by 1.6 to obtain the item's score, as β was 1.642. The final score of the RisObIn.Com tool was obtained by the sum of all item scores and a correction factor was added to obtain zero as the cut-off value.

#### *2.2. Statistical Analysis*

All data were checked for entry errors. Statistical analysis was done using IBM SPSS Statistics for Windows, version 26 (IBM Corp, Armonk, NY, USA). OpenEpi Version 3.01 was used to calculate confidence intervals (CI) [35]. Continuous data were checked for normal distribution using the Kolmogorov-Smirnov test and graphically by evaluating histograms and expressed as mean and standard deviation. Non-normally distributed data were expressed as median (Min-Max). Comparisons between the participants studied and those not included were made by using *t*-tests for normally distributed continuous variables, Mann-Whitney rank-sum tests for non-normally distributed continuous variables, and χ2 tests for categorical variables.

Cutoff point analysis was performed to identify the optimal value that differentiates the risk of obesity from non-risk of obesity in children. The threshold was defined by the largest distance from the diagonal line of the receiver operating characteristic (ROC) curve (sensitivity × (1 − specificity)). Using the cutoff point obtained, both sensitivity and specificity and positive and negative predictive

values were calculated, with their 95% CI. All *P* values reported were based on two-sided hypotheses and compared to a significance level of 5%.

#### **3. Results**

#### *3.1. Study Sample Characteristics*

Data was collected from 593 subjects. From those, 289 had incomplete data on crucial information to proceed with the analysis (for example, sex information) and therefore were excluded. The remain 304 gathered anthropometric measurements and data regarding our set of questions and therefore were included, despite some were not complete. In the portion of not included subjects, the children's median age was 8.0 (5.8–10.8) years old (missing 162), 7.8 (5.9–10.8) for girls, and 8.1 (5.8–10.4) for boys. The mother's median age was 39.0 (25.0–54.0) years old (missing 62), and father's was 41.0 (25.0–66.0) years old (missing 62). There were no significant differences between the child's age (U = 2371.0, *p* = 0.166), mother's age (U = 41,256.5, *p* = 0.153), child's BMI z-score (U = 23,156.0, *p* = 0.204), mother's (U = 38,080.0, *p* = 0.240) or father's nationality (U = 33,481.5, *p* = 0.207), parental current employment status (U = 30,097.5, *p* = 0.308), mean monthly income (U = 33,703.5, *p* = 0.510), and father's level of education (U = 36,689.0, *p* = 0.939) between the children included in the sample and those not included. The father's age (U = 33,362.5, *p* = 0.021) and the mother's level of education (U = 39,273.0, *p* = 0.035) was significantly different between the children included in the sample and those not included. The sociodemographic characteristics of the sample are presented in Table 1.


**Table 1.** Sociodemographic characteristics of the study sample.


**Table 1.** *Cont*.

The overall prevalence estimates of underweight, normal weight, overweight, and obesity, and central adiposity are shown in Table 2. Overall, the prevalence rate was 20.7% for overweight and 9.5% for obesity. Most children (80.8%, *n* = 244) had WtHr ≤ 0.5. Combining BMI and WtHr, 16.8% [95% CI 13.0–21.4] (*n* = 51) children were overweight (including obesity) with WtHr > 0.5. Only seven

children had a WtHr > 0.5 with a normal weight and four children had a WtHr ≤ 0.5 with obesity.

#### *3.2. Risk Factors for Overweight (Including Obesity) with WtHr* > *0.5*

Risk estimation models for overweight (including obesity) with WtHr > 0.5 were explored within each dimension of variables. The significant risk factors on each of the eight dimensions are presented in Table 3 and Supplementary Table S1.

#### 3.2.1. Peri-Natal Dimension

The only variable retained in the final peri-natal dimension estimation model was maternal pre-pregnancy BMI; being classified as overweight increases, in mean, 2.6-fold the risk of overweight (including obesity) with WtHr > 0.5, and being classified as obese increases, in mean, 4.1-fold the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.2. Anthropometric Dimension

For the anthropometric dimension, the retained variable was the BMI at 5 years old; being classified as overweight increases, in mean, 4.2-fold the risk of overweight (including obesity) with WtHr > 0.5.




#### **Table 3.**Variables retained in the statistical models.

#### 3.2.3. Sociodemographic Dimension

The sociodemographic variable retained in the final estimation model was paternal BMI; paternal BMI reflecting overweight decreases, in mean, 33% the risk of overweight (including obesity) with WtHr > 0.5, and paternal BMI reflecting obesity increases, in mean, 4-fold the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.4. Past Eating Habits Dimension

For the past eating habits dimension, the only variable included in the final model was the type of meal used for solid foods introduction; if soup (rather than cereals) was the first solid food introduced, it decreases, in mean, 60% (the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.5. Current Eating Habits Dimension

For current eating habits, the variables child's appetite, the caretaker's perception of the child's intake through image, and the PREDIMED questions regarding vegetable daily intake and butter, margarine, and cream daily intake were included in the final model. The child's appetite decreases, in mean, 83% the risk of overweight (including obesity) with WtHr > 0.5. The caretaker's perception of the child's intake through image increases, in mean, 1.5-fold the risk of overweight (including obesity) with WtHr > 0.5. The PREDIMED question regarding vegetable daily intake decreases, in mean, 38% (the risk of overweight (including obesity) with WtHr > 0.5, and the PREDIMED question regarding butter, margarine, and cream daily intake increases, in mean, 1.5-fold the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.6. Subjective Anthropometry Perception Dimension

For subjective anthropometry perception, the two variables retained in the final model were (1) the adequacy of father's own body image perception in comparison to his real BMI; and (2) the adequacy of the caretaker's opinion regarding the child's nutritional status in comparison to the child's real BMI. The relatively correct adequacy of father's own body image perception compared to real BMI increases, in mean, 4.9-fold the risk of overweight (including obesity) with WtHr > 0.5, and the correct adequacy of father's own body image perception compared to real BMI increases, in mean, 2.6-fold the risk of overweight (including obesity) with WtHr > 0.5. Regarding the adequacy of the caretaker's opinion regarding the child's nutritional status compared to the child's real BMI, correct adequacy decreases, in mean, 31.6-fold the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.7. Subjective Eating Habits Perception Dimension

In the eating habits subjective data, the variables retained in the final estimation model were the caretaker's perception of child's food intake when worried, irritated, or anxious and the caretaker's perception of the adequacy of the child's food intake for age. The caretaker's perception that the child's food intake when worried, irritated, or anxious is not affected decreases, in mean, 74% the risk of overweight (including obesity) with WtHr > 0.5. The caretaker's perception that the child's food intake is inferior or adequate for age decreases, in mean, 92% the risk of overweight (including obesity) with WtHr > 0.5.

#### 3.2.8. Physical Activity and Sleeping Habits Dimension

For physical activity and hours of sleep, the variables included in the final model were the child's participation in programmed sports activity, the number of sedentary hours in a weekday, the number of sedentary hours on a weekend day, and the total number of sedentary hours in a week. The child's participation in programmed sports activity decreases, in mean, 81% the risk of overweight (including obesity) with WtHr > 0.5.

#### *3.3. Scoring and Risk Index*

The set of 13 variables identified for the items of the screening tool were submitted to the pediatric expert panel.

The variable father's BMI from the sociodemographic dimension was excluded as two members of the panel did not agree with its inclusion in the tool, for lack of evidence of the variable impact in childhood obesity.

The current eating habits dimension variables for PREDIMED vegetable servings daily intake and butter, margarine, and cream servings daily intake, and the caretaker's perception of the child's intake through image were excluded. One member of the panel did not agree with the inclusion of the butter, margarine, and cream daily intake variable in the tool, for lack of evidence of the variable impact in childhood obesity.

The adequacy of the father's body image perception vs. BMI from the subjective anthropometry perception was excluded as two members of the panel did not agree with its inclusion in the tool, for lack of evidence of the variable impact in childhood obesity.

For the physical activity and sleep hours dimension, only the programmed physical activity practice variable was included. The remaining were excluded due to the observed lack of measurable effect.

Eight items were included in the tool and all dimensions except the sociodemographic were included. The items and respective scores are presented in Table 4.


**Table 4.** Items included in the Risk of childhood Obesity in the Community (RisObIn.Com) tool, categorization, and scoring.

Applying the tool items to every child who gathered responses to all the tool items (*n* = 145), the sum of the items ranged from −4.20 to 4.60. Higher values of the tool indicated a greater risk of obesity in childhood. The area under the ROC curve was 0.897 [95%CI 0.825-0.968; *p* < 0.001] for girls and 0.779 [95%CI 0.612–0.947; *p* = 0.016]—Figure 1. The uncorrected optimal cutoff point of the RisObIn.Com tool was -1, thus, a correction factor (+1) was applied to obtain a cutoff point value of zero. The tool sensitivity, based on the optimal cutoff point, was 95.0%; that is, 95.0% of children who had overweight (including obesity) with WtHr > 0.5 while on primary school, got a score greater than 0 in the transition from pre-school to primary school had overweight (including obesity) with WtHr > 0.5 while on primary school. The specificity was 74.4%, meaning that 74.4% of children who did not have overweight (including obesity) with WtHr > 0.5 while on primary school, got a score equal or

less than 0 in the transition from pre-school to primary school did not have overweight (including obesity) with WtHr > 0.5 while on primary school. The positive predictive value was 37.3%, meaning that among those who had a score greater than 0, the probability of having the condition was 37.3%. The negative predictive value was 98.9%, meaning that among those who had a score equal or less than 0, the probability of not having the condition was 98.9%.

−

**Figure 1.** Performance of the proposed screening tool (RisObIn.Com) to identify girls (**a**) and boys (**b**) in the transition from pre-school to primary school that had overweight (including obesity) with WtHr > 0.5 while in primary school. The receiver operating characteristic (ROC) curves were calculated for the cutoff point value of zero. The area under the ROC curve was 0.897 (95%CI 0.825–0.968; *p* < 0.001) for girls; and the area under the ROC curve was 0.779 (95%CI 0.612–0.947; *p* = 0.016) for boys. CI: Confidence Interval; WtHr: Waist-to-Height Ratio.

#### **4. Discussion**

Preschool and primary school ages are among the most critical periods for determining obesity later in life [36]. A screening tool applied at this occasion will signal cases that will benefit from general and customized intervention, improving resources to prevent child obesity.

This new proposed tool combines elements that reflect the multifactor nature of obesity: maternal BMI before pregnancy; the child's own BMI at 5 years old; the first solid food introduced for diversification; the caretakers perception of the current appetite of the child; the parental opinion regarding the child's intake adequacy; the child's food intake when worried, irritated, or anxious; the parental perception of the child's nutritional status; and the regular practice of programmed physical activities. Through this weighted combination, RisObIn.Com provides a score that estimates the risk of obesity through school age.

To the best of our knowledge, this is the first screening tool to assess the risk of obesity in children at the entrance of primary school that includes a set of parameters from different dimensions and specific periods, acknowledged to affect weight gain trajectory. The development of this tool contrasts with other existing tools, which selected items through the opinion of experts [17] or literature review [37], focusing only on a specific period of childhood, such as the peri-natal period [18] or the present moment of assessment [19]; others attempted to associate a large number of variables to childhood obesity risk through data mining [38]. The inclusion of all these dimensions can justify the large areas under the ROC curve obtained by RisObIn.Com, larger than other tools [18,19,38]. We observed a difference between boys and girls regarding the ROC findings considering the sex variable, but we consider that it is not possible to infer this difference is maintained when the tool is applied in samples

differ in dimension or population characteristics. RisObIn.Com achieves 95% sensitivity and 74.4% specificity despite the tool having been developed using a smaller sample than other studies.

#### *4.1. Peri-Natal Dimension*

Maternal pre-pregnancy BMI was the only variable retained in the final estimation model from the peri-natal dimension set. This variable has been shown to be positively associated with infant adiposity, as well as childhood obesity and overweight [12]. A recent meta-analysis identified significantly higher odds for childhood obesity with higher pre-pregnancy maternal BMI, 89% for the offspring of overweight women before pregnancy and 264% for those who were obese before pregnancy [39].

#### *4.2. Anthropometric Dimension*

On the anthropometric dimension, the retained variable was the BMI at 5 years of age. The growth patterns of BMI during childhood, particularly during critical periods, are closely related to adult obesity risk [40]. The second physiological rise in BMI occurs, in general, between 3 and 7 years [41]. Pei et al., in 2013, on the German birth cohorts GINIplus—The German Infant Study on the Influence of Nutrition Intervention plus Air pollution and Genetics on Allergy Development, and LISAplus—Influence of Life-style factors on Development of the Immune System and Allergies in East and West Germany plus Air Pollution and Genetics on Allergy Development, also found that BMI at 60–64 months of age was significantly associated with overweight at the age of 10 years [42]. Children with higher BMI at 5 years of age probably have already experienced an adiposity rebound, and early age at adiposity rebound is known to be a risk factor for later obesity [43].

#### *4.3. Previous Eating Habits Dimension*

Evidence shows that breastfeeding is a protective factor for obesity [14,15]. However, in this study, regarding previous eating habits, we found that only the type of first complementary solid food introduced was associated with overweight (including obesity) with WtHr > 0.5 while in primary school. Recent data does not support the hypothesis that the quality of complementary foods has a direct effect on the risk for later obesity [44]. Soup can be prepared using many different vegetables as ingredients; the quality and quantity of vegetables and fat used have a strong influence on its nutritional value. By opposition, infant cereals do not allow nutrient composition modifications, thus having a more constant energy value. In countries [45] such as Portugal, the traditional recommendation of the first complementary food to be introduced is infant cereals [46]. One can only speculate that, in infants that present a higher BMI or rapid weight gain [47], health professionals will recommend that the first food to introduce should be soup, since it allows to manipulate the amount of vegetables and fat content, in the attempt to reduce energy intake.

#### *4.4. Current Eating Habits Dimension*

In the current eating habits dimension, the parameters retained in the final model were the child's appetite and the caretaker's perception of high, low, or adequate appetite of the child. These measures relate to appetite and reflect self-regulation of energy intake. Biological regulation of appetite is very complex, engaging a number of tissues, organs, hormones, and neural circuits with several feedback pathways between the brain and peripheral tissues [48]. These mechanisms can be influenced and modulated by several factors, beyond the aim of the development of a screening tool. Adequate nutritional status children eat smaller portions at lunch/dinner and may eat more snacks throughout the day; the energy of those snacks is usually greater than the energy of lunch or dinner meals [49].

When parents perceive their child as having increased appetite, they may try to implement restrictive feeding practices. These practices can increase the preoccupation of the child with food and affect eating behaviors, eventually leading to paradoxical weight gain [50]. As both extremes of feeding practices could shape children's relation to food, we cannot exclude that, in the past, parents may have forced their children to eat after they are satisfied, promoting a dysregulation on this equilibrium, once the tendency to encourage children to clear their plate is reported to be associated with obesity [51].

The variables from the PREDIMED index, collected in the SGH study, while explored in the analysis, were excluded, because this Mediterranean diet assessment tool is not yet validated for Portugal; therefore, it may not be adapted to Portuguese food habits. The caretaker's perception of the child's intake through image was also excluded. We presented caretakers a set of four images of a lunch/dinner plate. By presenting an even number of options instead of an odd number, we avoid the tendency for the selection of the central option, improving the reliability of the answer. However, since the need to use images will be a challenge for the application of this tool and the removal of the item had a small effect in the final model, it was removed.

#### *4.5. Subjective Perception of Anthropometry Dimension*

Regarding the subjective perception of anthropometry, this study found a high proportion of parents that misclassified the nutritional status of their child. It has been speculated that the high prevalence of overweight and obesity in children in the last decades can influence the misinterpretation of the child's normal weight [52], leading to the perception of "normal" weight despite the BMI indication of overweight. If parents cannot recognize their child as overweight, they will not act to change behaviors and the situation can exacerbate, leading to obesity.

#### *4.6. Subjective Perception of Eating Habits Dimension*

Adequate dietary habits are important for health throughout life, but particularly during childhood, considering that the dietary habits during this specific period are potentially perpetuated into adulthood [53]. In the dimension related to the subjective perception of eating habits, the variables retained in the final estimation model were the caretaker's perception of the child's food intake while worried, irritated, or anxious and the caretaker's perception of the adequacy for age of the child's usual food intake. Previous studies have shown psychopathology to be associated with overweight in children [54,55]; validity data indicate that children as young as 4 years old can report on their own anxiety symptoms [56]. In reaction to anxiety, emotional eating acts as a biological response that provides temporary feelings of gratification/satisfaction [57]. The intervention approach for children with anxiety symptoms and emotional eating should be adapted to address the negative emotions underlying eating behaviors as well as teaching healthy coping strategies for these emotions [58]. Parents and their own perception of the child's dietary habits is one of the most important factors for the dietary habits of children [59]. Childhood obesity experts recommend that childhood overweight prevention should focus on parents, according to the growing evidence of the role of parental practices and family environment in promoting effective changes [11,12].

#### *4.7. Physical Activity and Sleeping Habits Dimension*

For the dimension concerning physical activity and sleeping habits, the variables included in the final model were the child's participation in programmed sports activity. The levels of physical inactivity are rising in many countries with major implications for general health and the prevalence of non-communicable diseases, such as obesity. The association had already been identified in a similar population living in Portugal's capital (Greater Lisbon) [60]. A recent multinational cross-sectional study demonstrated that low levels of moderate-to-vigorous physical activity or high sedentary levels during weekdays and weekends were associated with higher odds of obesity in 9–11-year-old children in 12 countries [61].

The RisObIn.Com tool seems to be a comprehensive tool to identify, at school entrance, 5-to-6-year-old children at higher risk for late childhood obesity. It was conceived to be applied either by the parents or by teachers or school health professionals, such as a school nurse, school nutritionist, or school social worker, with parental feedback and the child's health bulletin for easier recall of mother pre-pregnancy BMI and anthropometric measures at the age of 5 years. The tool carries a

small and simple set of instructions for its effective use. A web-based tool allows a quick, simple, and automated form of application. RisObIn.Com is hosted in Health & Technology Research Center webpage available at https://htrc.estesl.ipl.pt/risobin-com/.

The synergy between the authors of this study and the SGH research team benefited from the logistics associated with data collection and allowed the enhancement and efficiency of resources. The methodology included direct anthropometric measurements, assessment of a set of variables identified as influent in obesity development (from peri-natal to the present moment, including sociodemographic, anthropometric, past and current eating habits, subjective anthropometry, and eating habits perception, and physical activity and hours of sleep) and had the endorsement of an expert panel group composed by skilled professionals from different areas. As a positive asset, the selected outcome measure reflects adiposity by the cumulative outcome of overweight (including obesity) with WtHr > 0.5, which allows both a better characterization of the nutritional status and better accuracy. On the other hand, the mixed design of case-control study nested on a cross-sectional study is a limitation, due to recollection bias, and causal inferences to be made, just epidemiologic and statistic associations. Another source of bias was related to maternal weight before and after pregnancy that was reported and not measured. The local nature of the sample and the differences observed between the children included and not included in the study do not allow immediate generalization of the findings to other populations; therefore, external validation of the screening tool and the study of its performance in different samples is required. Eventual ethnic differences were not explored, due to national ethical and legal restrictions related to data protection. The effective sample size was smaller than expected and it probably affected the ability to identify more exposure variables significantly associated with the study outcome; a larger sample would probably have improved the observed results. Longitudinal data analysis would allow evaluating the tool's ability to predict BMI change over time.

The RisObIn.Com screening tool is proposed to be routinely used by teachers and other school personnel with the participation of parents or caretakers for early identification of children who might benefit from preventive actions, but its use could also be extended to health care professionals such as nurses, family physicians, or pediatricians. The RisObIn.Com screening tool is a simple and inexpensive tool that can provide an evaluation of the risk factors for pediatric obesity and may identify those in need for healthy lifestyle changes.

#### **5. Conclusions**

The RisObIn.Com tool is proposed to be a comprehensive tool to identify children at high risk for late childhood obesity at admission to primary school, by the age of 5 to 6 years old. Further studies are needed to assess the external validity and the generalization of the findings, as well as to confirm both the performance of this tool to identify children with obesity risk at admission into primary school and the effect of the subsequent intervention to prevent obesity in children.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/12/11/3288/s1, Table S1: Variables in the statistical model.

**Author Contributions:** Conceptualization, A.C.M.; methodology, A.C.M., R.F., T.N., P.A.O., and D.V.; software, D.V., and P.A.O.; formal analysis, D.V.; investigation, A.C.M. and R.B.; resources, R.F. and T.N.; data curation, T.N. and P.A.O.; writing—original draft preparation, P.A.O., A.C.M., and D.V.; writing—review and editing, P.A.O., A.C.M., and D.V.; supervision, A.C.M. All authors agreed on the submitted version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We would like to acknowledge the members of the experts panel whose contribution was valuable for the development of the present tool (Bandola, L.; Barrigas, C.; Cunha, S.; Neves, C.; Sousa, J.). The present study is part of the MSc thesis in Clinical Nutrition of one of the authors (P.A.O.) (supervisor A.C.M.), from Faculty of Medicine of Lisbon and Lisbon School of Health Technology, Lisbon, Portugal. The authors are grateful to the SGH team and to its coordinator Joana Sousa, to whom we express our recognition. H&TRC authors gratefully acknowledge the FCT/MCTES national support through the UIDB/05608/2020 and UIDP/05608/2020.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Play with Your Food and Cook It! Tactile Play with Fish as a Way of Promoting Acceptance of Fish in 11 to 13-Year-Old Children in a School Setting—A Qualitative Study**

### **Rikke Højer 1,2,\* , Karen Wistoft <sup>3</sup> and Michael Bom Frøst <sup>2</sup>**


Received: 23 August 2020; Accepted: 15 October 2020; Published: 17 October 2020

**Abstract:** Despite a tradition of consuming fish in Denmark and despite the health benefits of eating fish, Danish children consume only one-third of the officially recommended amount of fish. The objective of this study was to explore an experiential and sensory-based exercise in a school setting with focus on tactile play and cooking as a way of promoting 11- to 13-year-old children's acceptance of fish. The design was a qualitative exploratory multiple-case design using participant observation in a school setting. Six classes were recruited from the Eastern part of Denmark (*n* = 132). Based on an exercise with cooking fish and gyotaku (fish print), four meta-themes were identified by applying applied thematic analysis: rejection, acceptance, craftsmanship, and interaction. Rejection and acceptance appeared along a rejection–acceptance continuum related to how the fish was categorised (animal, non-animal, food) in different phases of the experiment. Rejection was promoted by mucus, smell, animalness, and texture, whereas helping each other, tactile play, and craftsmanship promoted acceptance. In conclusion, this study found that tactile play combined with cooking could be a way of promoting acceptance of fish. The findings also support a school setting as a potential gateway in promoting healthy food behaviour.

**Keywords:** food acceptance; tactile play; cooking; children; fish; health promotion

#### **1. Introduction**

#### *1.1. Background*

Children aged 11 to 13 years are in the early adolescent life phase [1], a phase defined by a developmental plasticity [2], where lifelong habits can be established [3]. The adolescent life phase is critical when it comes to behavioural changes in, for example, dietary habits [4]. The changes in dietary habits are due to, for example, an increase in autonomy and a decrease in family influence [5,6].

Consumption of fish provides valuable nutrients. Especially fatty fish have a high content of vitamin D, which is important for e.g., calcium (Ca) absorption, bone health, and childhood growth stages [7,8]. Regular consumption of fish, especially those high in n-3 poly unsaturated fatty acids (PUFA), also reduce incidences of, for example, diabetes mellitus, systemic arterial hypertension, central obesity and hyper-lipidemia [9,10], and seem to positively influence intestinal microbiota [11]. Furthermore, the macro nutrient content of fish with regard to protein is 15–20% and fish contains all the essential amino acids [12], which is beneficial for the diet as the sulphur-containing amino acids, cysteine and methionine, are absent in plant protein. Furthermore, proteins from fish have a high degree of digestibility i.e., 85–95% [12,13]. Studies have shown positive health effects as a result of fish protein intake e.g., by decreasing the risk of metabolic syndromes and increasing insulin sensitivity [14–17].

#### 1.1.1. Acceptance and Rejection of Food

This study focuses on fish as part of a healthy diet. According to a national study, Danish children aged 10 to 17 years eat only 105 g of fish per week [18], one-third of the Nordic recommendations of 350 g per week [19]. The intake of fish among Danish early adolescent phase children corresponds with international observations [20–22]. Furthermore, to the authors' knowledge, little research has been conducted in the area of early adolescent phase children's acceptance of fish.

Rozin and Fallon [23,24] have developed a framework in which they have identified three principal motivations within the taxonomy of food acceptance and rejection, which drive food acceptance and rejection: sensory-affective factors (e.g., liking/disliking taste or smell), anticipated consequences (e.g., negative/positive physiological or social), and ideational factors (e.g., knowledge of the nature or origin of a food). These motivations and attributes can lead to either rejection or acceptance: the psychological rejection categories are distaste (the concept distaste includes all sensory characteristics, real or imagined [25,26]), danger, inappropriateness, and disgust, and the acceptance categories are good taste, beneficial, appropriate, and transvalued [25,27]. Furthermore, Rozin and Vollmecke [27] point out that the influence of culture and context are predominant factors influencing acceptance and rejection, and that acquired likes can be promoted by social encounters with people outside the family, especially peers. The framework of rejection and acceptance developed by Rozin and colleagues [23,24,27] has been applied repeatedly in studies investigating food behaviour (e.g., [25,26,28,29]).

Based on the limited research conducted within and around the target group of this study, Prell, Berg, and Jonsson [30] identified a negative attitude towards the smell, the fear of finding bones, the accompaniments, and friends' behaviour as primary barriers to eating fish. In a study focusing on foods in general, Frerichs et al. [31] found that appearance and texture were primary drivers for accepting or rejecting food. Furthermore, Mitterer-Daltoé, Latorres, Treptow, Pastous-Madureiraa, and Queiroz [32] and Latorres, Mitterer-Daltoé, and Queiroz [33] found that young children had a higher acceptance of fish than older children. This might be due to the older children's cognitive maturation, leading to food-related cognitions increasing and becoming more complex [34]. The animal origin of fish could also play a role in rejection, since foods of animal origin tend to promote an attitude of disgust more than those of vegetable origin [24,25,35,36]. Increasing acceptance of food through tactility (the sense of touch by using the hands) or tactile play is a research area that has yet to be explored in greater depth. Five recent studies have been conducted in this research area [37–42], but these studies all fall outside the age-related sample of this study. Nevertheless, the results are interesting and relevant to this study as they point to a positive impact of tactile play on food neophobia and/or food acceptance.

Another way of influencing food behaviour and promoting acceptance of healthy foods has been sought through a hands-on approach and cooking programmes. A review of the effect of cooking programmes by Utter, Fay, and Denny [43] concluded that cooking programmes may have a positive impact on food-related beliefs, knowledge, skills, and behaviours. Of the 20 studies included in the review, only three were on children in the age range of the sample group in the present study. However, none of the studies included in the review focused on foods of animal origin. Furthermore, observations of children's food behaviour and learning processes have been included in studies by, for example, Block et al. [44], Fisher and Birch [45], and Gibbs et al. [46]. The relevance of applying observation as a research method relates to the objective of revealing actual behaviour.

Nelson, Corbin, and Nickols-Richardsson [47] argue that culinary skills education offers a unique opportunity for experiential learning, which they illustrated through the use of the Kolb Cycle of Experiential Learning [48] combined with culinary skills education (Figure 1). ‐

‐ **Figure 1.** Model of culinary skills education as a process for Kolb's cycle of experiential learning developed by Nelson, Corbin, and Nickols-Richardsson [47]. Figure by first author R. Højer.

According to Nelson et al. [47], culinary skills education promotes knowledge through experience, as illustrated in Figure 1. As students move from observational to experiential learning stages and engage in culinary concepts, a foundation for promoting critical thinking and learning skills and technical proficiencies is laid out, all aimed at promoting healthy food behaviour. Furthermore, Nelson et al. [47] conclude that nutrition knowledge alone, aimed at promoting healthy food behaviour, seems incomplete without the dimension of experiential learning via interactions with food and cooking equipment.

#### 1.1.2. The Subject Food Knowledge

In 2014, the subject Food Knowledge replaced the subject Home Economics as part of a reform of the Danish compulsory primary and lower secondary schools. The subject is mandatory for one year and can be taken in 4th, 5th, or 6th grade. In the subject Food Knowledge, students focus on four areas of competencies: Food and Health, Knowledge of Food, Cooking and Dining, and Food Cultures. The purpose of the reform was to ensure that Food Knowledge provides students with an opportunity to work with senses and experiences. Experimentation, creation, and communication in relation to food and meals are also key elements, as is the development of, for example, new skills and knowledge through motor skills, cognition, and perception [49].

#### 1.1.3. Gyotaku Explained

Gyotaku is a traditional Japanese art form (see Figure 2); gyo is the Japanese for fish and taku for rubbing or printing: fish rubbing or fish printing [50].

Gyotaku was used by Japanese fishermen more than a hundred years ago [51]. To avoid misunderstanding, the fishermen used it to replicate the correct size of the fish, whereby it became a documentation method. During the twentieth century, the practice of gyotaku has been turned into an art form.

As an example of an experiential exercise, gyotaku was adapted to firstly include a tactile art exercise, which was the traditional part of the exercise to be explored in this study. Secondly, after the art part of the exercise in which the fish served as an art medium, the fish would then be included in a cooking exercise. The gyotaku exercise was chosen for its novelty in a Danish context and for its tactile hands on approach to the fish.

**Figure 2.** Gyotaku of flounder *(Platichthys flesus)*, artist: R. Højer, photo: Marilyn Koitnurm.

#### *1.2. Study Aim*

The aim of this study was to promote children's acceptance of fish. Based on the hypothesis that through hands on experience with fish it is possible to promote acceptance of fish, the objective of this study was, through an intervention, to explore the potential of a sensory-based experiential exercise in a school setting with focus on cooking and tactile play as a way of promoting 11- to 13-year-old children's acceptance of fish. The two main research questions to be answered were: (1) how do children respond to handling, preparing and cooking fresh fish? and (2) how does the process of the sensory-based tactile experiment gyotaku affect children's acceptance of fish?

#### **2. Materials and Methods**

#### *2.1. Study Design*

‐ ‐ ‐ ‐ This study was an intervention with a multiple-case study design [52]. Six cases in six different classes from six different schools were included in the intervention. All participating classes underwent the experiential gyotaku exercise one class at a time. The qualitative method used to collect data consisted of participant observation [53].

‐

 ‐ The gyotaku exercise was integrated into the (in Denmark) compulsory subject Food Knowledge (*Danish: Madkundskab*) [49] in the fifth and sixth grades and it meets the official learning goals (for 2017–2018 and 2019) for the subject Food Knowledge set by the Ministry of Children and Education [54].

#### Ethics Approval

Ethics approval for this study was given by the joint Research Ethics Committee of the Faculty of Science and the Faculty of Health and Medical Sciences, University of Copenhagen, Denmark (reference 504-0005/17-5000).

#### *2.2. Participants*

‐ ‐ We recruited six classes from fifth and sixth grades (11 to 13 years of age) from six different Danish public schools (*n* = 132). Four classes were from the capital region and two from the region of Zealand (see Table 1 for participant characteristics). Recruitment was geographically limited to the eastern part of Denmark due to convenience. The recruitment was done by sending out information letters via e-mail to schools in the eastern part of Denmark addressed to the school's Food Knowledge teachers. For all participating children, written informed consent was given by the legally appointed caregiver parent or either parent if the parents were married or had joint custody. Children's refusal to touch, handle, and/or taste the fish was respected by the researchers.


**Table 1.** Participant characteristics.

\* Number of teachers present during the gyotaku exercise. \*\* Schools SA, SB, and SC are schools from workshops during Science Week 2016. \*\*\* Schools MB, MC, and MD are schools from the main study 2017.

#### *2.3. Setting and Gyotaku Exercise*

School SA, SB, and SC took part in gyotaku workshops in a teaching kitchen at the Department of Food Science at the University of Copenhagen, Frederiksberg, Denmark, in a field trip setting. School MB, MC, and MD were in their natural educational setting, since the gyotaku exercise took place on three different occasions at schools in the ordinary school teaching kitchen. This differentiated setup was due to practical organization as the classes SA, SB and SC participated as part of Science Week 2016, a yearly returning science festival in Denmark, whereas the classes MB, MC, and MD did not participate in Science Week 2016 and data were collected during early spring 2017. All classes carried out the gyotaku exercises based on the same exercise guide.

The sensory-based experiential exercise was a four-phase exercise consisting of a) gyotaku (fish printing), which also gave its name to the complete experiment, b) filleting a fish, c) cooking the fish fillets by a commonly used Danish method, and d) tasting.

Materials for the gyotaku experiment (per group of four children): one fresh whole flatfish with head (either dab (*Limanda limanda*) or flounder (*Platichthys flesus*)), one lemon, squid ink diluted with tap water in a cup, a small sponge, five A4 pieces of paper cut into eight equal parts, paper towels, printing paper, a cutting board, a sharp filleting knife, rye flour, salt, pepper, butter, rye bread, a frying pan, a stove and written experimental instructions.

General organisation: all of the children worked in groups of four. Each group received one fresh fish to be shared during printing, filleting, and cooking (1 fish = 4 fillets).

Phase a: Gyotaku (printing): The printing procedure was the actual gyotaku exercise. The children chose and picked up their group's fish from a box containing fresh fish on ice. The fish was then cleaned by washing it under cold running water while rubbing it with a slice of fresh lemon (this dissolves the fish's natural mucus cover). The fish was then dried with paper towels and placed on a cutting board. Paper squares were placed around the edge of the fish to avoid getting squid ink on the cutting board. Diluted squid ink was applied with a sponge to the surface of the fish until it was covered with ink. The paper squares around the fish were removed, and printing paper was placed on top of the fish. The print was transferred to the paper by stroking the fish on top of the paper. The paper was gently pulled off the fish, and a mirrored print of the fish had been transferred to the paper (see Figure 2).

Phase b: Filleting: If they wanted to, each child in the group filleted their own fish fillet by following the handout picture instructions. After the child had felt the fillet with his/her fingers to ensure that no fish bones were present, the fillet was ready to be cooked.

Phase c: Cooking: The fish fillets were turned in rye flour containing salt and pepper and were then fried in butter on a hot pan. This is the traditional way of cooking fish fillets in Danish cuisine.

Phase d: Tasting: The fried fish fillets were served on a slice of rye bread with butter and a slice of lemon. Tasting/eating was voluntary. This is a common way of serving fish fillets in Danish cuisine.

After the experiment, the children could take the gyotaku home, or the school could use it, for example in an art exhibition.

#### *2.4. Data Collection—Participant Observation*

The participant observation was primarily concept-driven [55] and based on the framework of Rozin and Fallon's [23] and Rozin and Vollmecke's [27] taxonomy of food rejection and acceptance. Therefore, a loosely structured observation guide, with room for exploratory inquiry, was constructed based on the main framework of acceptance and rejection with the following themes: (1) the social/group interaction element, (2) the children's interaction with the fish, (3) the process of the exercise, and (4) development/changes in attitude throughout the experiment. Documentation methods used during the participant observation were in the form of written field notes and situational photos to document the setting, various situations, and child–fish interactions. The field note strategy was inscription and transcription [55], in which descriptions of behaviours (inscriptions) and informants' own words and dialogues (transcription) were recorded in an observational journal based on the loosely constructed and pre-thematised observational guide.

The same researcher participated in all gyotaku exercises by observing and interacting with the children through informal conversations based on the observation guide. In all cases except two (school MB and MD), observation assistants were present throughout the gyotaku exercise. At schools SA, SB, SC, three observation assistants were present, and at school MC one observation assistant was present. In all cases, the observation assistants had a semi-participatory role while also documenting the gyotaku exercise through photos. During the participant observations, researchers and assistants interacted with the children through informal conversations based on the situation while the children were working with the fish. Questions were based on "free narrative" [56] to promote situational comfort and to get and keep the conversation flowing. The questions were directed towards the children's perspectives of the situational experiences; for example (to the whole group): "How is it going here?" and "How do you feel about filleting a fish?". Probing [53] was used to follow up on short answers, for example "Can you tell me some more about that?". The focus was on informality and conversations steered by the children and their point of view. If a child asked what had been written down during a conversation, he/she was given the opportunity to read it. After each observation session, observational journals and photos were compared and evaluated. Post-intervention notes were documented by the research group. Furthermore, the field notes in the observational journal were immediately after the observation separated into direct observations of behaviour, dialogues based on children's peer-to-peer dialogues and researcher–child dialogues, and researcher reflections. A pre-coding was conducted based on concept-driven coding [55]; for example based on the framework of acceptance and rejection [25,27], fish handling, sensory aspects, and group work.

#### *2.5. Data Analytical Method*

Data analysis was conducted by using Applied Thematic Analysis (ATA) developed by Guest, MacQueen, and Namey [57]. ATA was applied to identify themes and to analyse patterns of meaning in relation to the research questions under study and was chosen for its flexibility with regard to type of texts, for example field notes [57], and its ability to highlight similarities and differences across cases [58].

Through a concept-driven [55] processing of data based on the research questions, four meta-themes were identified by organizing the pre-coded text into a matrix based on the frequency of re-occurrence of documented observed behaviours and dialogues. The identified meta-themes were rejection, acceptance, craftsmanship, and interaction. A thematic map was constructed to create a visual outline

of possible sub-themes [57,58]. Finally, themes were re-considered to ensure accurate representation by re-reading the data set [57,58]. (See Figure 3 for presentation of the ATA data processing. This resulted in the appearance of sub-theme clusters as situational events, behaviours, etc. (see Figure 4). ‐ ‐ ‐ ‐ ‐

‐ ‐

‐ ‐

‐

‐ ‐ **Figure 4.** ATA frame for presentation of data: main category, meta-themes, sub-themes, and related clusters.

Data not relevant for the research questions were excluded from the data set and analysis after being re-read to ensure lack of relevance. Furthermore, the ATA frame (analysis, results, and discussion hereof) was read by and discussed with researchers within the research group, but for those who had not been present at the interventions the frame was read by and discussed with an experienced researcher outside of the research group.

 The essence of meta-themes and sub-themes are presented in Table 2. Data were not only sorted by meta-theme and sub-theme but also by exercise phase (see Table A1: Data set).


#### **3. Results**

In Figure 4 two main categories, four meta-themes, nine related sub-themes, and sixteen clusters *(italic)* are presented.

Data are presented according to the ATA frame (Figure 4) by including relevant examples from the data set to support the ATA. Abbreviations applied in the analysis: Obs: observation, ic: informal conversation. Phases of the exercise: #1 = Before printing; #2 = During printing; #3 = Between printing and filleting; #4 = During filleting; #5 = Frying; #6 = Tasting.

#### *3.1. Meta-Theme 1: Rejection*

#### 3.1.1. Sub-Theme: Distaste

Rejection based on distaste, which includes all sensory characteristics, both real or imagined [25,26], was based on two main sensory characteristics: smell and texture. Rejection based on smell was primarily present in two phases of the experiment. Firstly, at the beginning of the printing phase when the children were presented with the fresh fish:

#1 When the lid is removed from the fish on ice, several children say: *"Ugh, it smells fishy"* [in a bad way] (School all, obs.)

Secondly, smell was a source of rejection based on distaste in the final experiment phase (tasting):

#6 Some children do not want to taste the fish. Int.: *"Why?"* Response: *"It smells of fish. We know we do not like fish because it feels weird in the mouth"*. A girl says: *"That is also why my dad does not like fish"* (School MC, ic).

Furthermore, the texture of the fish in the mouth was a factor in rejecting the fish based on distaste:

#6 A girl nibbles on the fried fish: *"I don't like the fish. It is kind of* . . . *mushy"*. (School MD, obs).

#6 Everyone in the class tastes the fried fish, but three boys spit it out and agree that they do not like to chew it as it is too mushy and soft in the mouth. (School MB, obs).

#### 3.1.2. Sub-Theme: Disgust

*‐* ‐

Apart from behaviours and verbal expressions promoting rejection based on distaste, rejection was also observed for the affective response of disgust.

Fear of contamination was observed primarily in two situations. Firstly, at the beginning of the experiment (phase a) when children picked up the fresh fish using only the tips of their thumb and index finger as shown in Figure 5. Most often the task of picking up the fish would be done by two children going to the fish box. One would pick up the fish (as illustrated in Figure 5) while the child not picking up the fish would often stand in the background in order to not get too close to the fish, although still leaning forward to have a look. ‐ 

**Figure 5.** A display of disgust: picking up the fish, photo: R. Højer.

 Secondly, in relation to filleting (phase b):

#4 Several children put on latex gloves before starting filleting (School MB, obs.).

Rejections driven by disgust also appeared as a reaction to the idea of "animalness". These reactions were also predominant at the beginning of the experiment (phase a) and during the filleting phase (phase b):

#1 Girl, when fish has been collected: "*Yuck! Look, it has eyes"* [pinches her nose] (School SC, obs).

#4 Int.: *How is it going with filleting the fish?* The girl cutting responds: *"I think that sound when you kind of hit the bone with the knife and that sound it makes* . . . *ugh"* [shrugs] (School MD, ic).

#4 Girl, during filleting: "*Yuck, it has fish guts inside* [viscera]" [she pinches her nose and turns away, holding her hands in front of her mouth] (School MD, obs).

#### *3.2. Meta-Theme 2: Acceptance*

#### 3.2.1. Sub-Theme: Tactility

Acceptance through tactility was observed in two forms: "sensing a transformation" and "reduction of animalness" through the sense of touch and a re-categorisation of the fish from animal to non-animal. The former displayed itself at the beginning of the experiment (phase a) after the fish's natural mucus layer had been washed and removed:

#1 Boy group after washing the fish: they stroke it and agree that it is weird because it was so slimy before but now it is soft to the touch (School MD, obs).

When the children started the printing process (phase a), it seemed like the fish had been re-categorised from animal to an art medium. Touching the fish was no longer an issue:

#2 During the printing process, great attention is given to getting the right amount of ink on the eyes, fins, and the mouth to get them onto the paper. This is done by unfolding the fins with their fingers and dabbing the sponge lightly on the eyes, fins, and the mouth (School all, obs.).

#2 Between prints, the fish is gently patted and stroked by several children; it is "tickled" between the eyes and around the mouth (School all, obs).

‐

‐

#### 3.2.2. Sub-Theme: Exploration

*‐* ‐

‐

Exploration was predominant in two main scenarios: exploring the fish before and after filleting (phase b). There were clear signs of curiosity, as shown in the following example: ‐

#3 A girl is exploring the fish. She opens the fish's mouth and looks into it: *"I just had to look inside. You can see its teeth* . . . *I just had to touch"*. Another girl in the group: *"Ohh yes, its mouth can get really big"*. The first girl replies: *"Yes, it can eat big fish"* (School MC, ic). 

This exploratory scenario is also seen in Figure 6 with children putting their fingers in the fish's mouth to feel its teeth.

**Figure 6.** Children exploring the fish, photo: R. Højer.

After filleting, children explored the fish:

#4 Roe in fish: at first the children do not want to touch or even look, but after a while they start to pick at it with the knife tip, and then cut it, mash it, and study the small eggs (School SA, SB, SC, obs.).

Both exploratory scenarios led to a greater child interaction with the fish.

#### 3.2.3. Sub-Theme: Liking

Acceptance due to liking was primarily driven by the sensory characteristic "taste" (the fish tasted good). It also seemed like taste familiarity was a factor in liking it.

#6 A girl is eating her fish fillet: *"Mmm, I love fish fillet"* Int.: *"Why?"* Girl: *"It is kind of a little bit sweet but also just good. We also get it at home"* (School MD, ic).

#6 A girl tastes a little bit of roasted fish roe and says: *"Mmm, it actually tastes like cod roe* . . . *but it is a little bit grainy and dry in the mouth"* (School MD, obs).

#### *3.3. Meta-Theme 3: Craftsmanship*

#### 3.3.1. Sub-Theme: Autonomy

Throughout the experiment, autonomy was a sub-theme, since all of the assignments were carried out through group negotiation and decision-making; there was freedom to organise the work themselves (no teacher involvement), for example, who should pick up the fish, who should fry the fish etc. Pride in their work was especially evident during printing (phase a) and filleting (phase b):

#3 After the printing, children show their self-made print to teachers and other groups (School all, obs).

#4 They want to try to fillet the fish themselves. The experimenter (first author) is not allowed to help too much, only to correct them if they have made a wrong cut (School all, obs).

#### 3.3.2. Sub-Theme: Skills

Skills were developed, particularly in the filleting process (phase b). It was observed that the children initially had difficulties in holding the knife correctly and actually filleting the fish. During the filleting process, they became more confident in using the knife and in how to fillet the fish (School all, obs.). During cooking (phase c), skills were developed when they were trained how to cook a fish for the correct amount of time:

#5 While frying the fish, the children are very preoccupied with cooking it for the right amount of time so it is not raw, but they are also focused on not cooking it for too long. They comment on the colour and use it as a way of telling if it is done (School all, obs).

A clear indication of the acquired skills can be seen in the following extract:

#5 After frying the fish, a girl says: *"Ah, now I know how to make fish fillet. I would like to try it at home if mom will buy a fish"* (School MC, ic).

#### *3.4. Meta-Theme 4: Interaction*

#### 3.4.1. Sub-Theme: Helping Each Other

The sub-theme "helping each other" appeared primarily as "us against them/the fish" and giving advice. A concept of "we are in this together" and "us against the fish" appeared, particularly at the beginning of the experiment (phase a), where the children had to pick up the fish and prepare it for printing:

#1 Two girls are washing and drying a dab before printing. They help each other by holding the fish at each end and carrying it together to the printing table (School MC, obs.).

Children also helped each other when washing the fish to remove the fish skin mucus prior to the printing (phase a). For example, one child supported the fish's tail, while another rubbed it with a lemon slice to remove mucus from the fish. Furthermore, helping each other was observed when, for example, applying the ink and giving advice on how to apply ink to the fish during printing, and giving advice on how to make a correct cut with the knife during the filleting phase (phase b):

#4 The girls give advice on how and where to cut: *"You have to start with the moon-shaped cut there"*. The boys correct each other more often (School MB, obs).

#### 3.4.2. Sub-Theme: Peer Influence

Peer influence was observed throughout the exercise and resulted in the other children in the group reacting either positively or negatively to the fish. The following two extracts illustrate peer influence leading to a positive reaction to the fish (the first extract) and a negative reaction to the fish (the second extract):

#3 After printing, two girls in a group of four are touching the fish, while the other two do not want to touch it. After observing the girls touching the fish for a little while, the other two girls change their mind and come over to the fish and try to touch it (School MB, obs.).

#6 Everyone in the class tastes the fried fish, but three boys from the same group spit it out and agree that they do not like to chew it as it is too mushy and soft in the mouth (first one boy spits it out, then the rest of the group) (School MB, obs).

The ATA is summarised visually in Figure 7, which shows meta-themes, sub-themes, and predominant clusters within identified sub-themes related to the different phases in the experiment.


**Figure 7.** Applied thematic analysis (ATA) summary visualised.

#### **4. Discussion**

 Based on thematic analysis, we propose the following diagram to explain a rejection–acceptance continuum (Figure 8).

**Figure 8.** A rejection–acceptance continuum based on fish categorisation with examples of elements driving acceptance forward or backwards. Developed by first author Højer, inspired by and Rozin and Fallon [24].

Figure 8 illustrates elements which drive either rejection or acceptance along the continuum. Our observations on the categorisation and re-categorisation of the fish as animal, non-animal, and food were in line with what Rozin and Fallon [63] and Martins and Pliner [25] refer to as "animalness". This animalness can be reduced in a fish by, for example, removing the head and bones, cutting it up, cooking it, and serving it without it resembling what it is: a fish and an animal [63].

At the beginning of the exercise, the fish was categorised as an animal due to its smell, slimy texture, and visual appearance (whole animal with head, fins, blood, etc.) and thereby promoted rejection. The whole fish represented a high degree of animalness, since it did not resemble what the children would typically eat (a breaded fish fillet). According to the results of a Danish citizen science project on Danes' fish eating habits, the hot fish dish most often eaten by children was breaded fish fillet (39%), and in the form of cold cuts (eaten on rye bread) the favourites were mackerel in tomato sauce (33%) and fish cakes (21%) [64]. According to Fischler [65], this could be categorized as gastro-anomie, because the consumer has problems in identifying food and food origin as a result of processing [65]. Rejection based on animalness was also found in a Norwegian study on adolescents' (16 to 17 years of age) attitudes towards meat from farm animals. Females, in particular, rejected meat due to its association with, for example, blood and animal parts [66]. The study also found that participants in regular contact with farm animals displayed no disgust reaction and had a more relaxed attitude towards meat production [66]. In an empirical study on what motivates food disgust, Martins and Pliner [25] found that animalness was not the complete explanation for a food disgust reaction, as non-animal food products were also capable of promoting disgust. According to Martins and Pliner [25], an explanation could be found in the experienced texture of, for example, slime, as it could be related to decay. Through multidimensional scaling analysis, they were able to identify independent (i.e., unique) dimensions, suggesting that both aversive textural properties and the reminders of animalness are primary variables accounting for perceptions of food disgust [25,67]. Egolf, Siegrist, and Hartmann [28] also found in their study on how people's food disgust sensitivity shapes eating and food behaviour that surface texture of food was capable of promoting disgust.

According to our observations during the printing phase, the fish was re-categorised from animal to non-animal, because it was perceived as an art/play medium and rejection cues were not evident. In this phase, tactility through touching the fish, as part of the assignment of printing, appeared to promote acceptance of the fish. This observation correlates with the findings of Coulthard and Sealy [38], who found that pre-school children tried more fruits and vegetables after participating in a sensory play activity with real fruits and vegetables than children in a non-food sensory play task (*p* < 0.001) and in a visual exposure task (*p* < 0.001). Similar results were also found by Nederkoorn, Theiβen, Tummers, and Roefs [41]: tactility increased the acceptance of food with the same texture.

The observed acceptance could also be promoted by the reduction in mucus on the fish after washing, which would reduce the texture-induced disgust as proposed by Martins and Pliner [25]. Nevertheless, this does not account for the following tactile exploration of the fish, where the children, driven by curiosity about something unfamiliar, put their fingers in the fish's mouth, touched the gills, eyes, tongue, etc., with all parts of the fish still covered by or containing mucus (see Figure 6).

At the beginning of the filleting phase of the exercise, the fish was again categorised as an animal, and rejection was promoted. A behavioural example was the observation of the children putting on latex gloves in this phase, although this behaviour could also be a result of peer influence.

Rejection was primarily due to the cutting through of the skin of the fish and cutting close to the bones. Both sound and visual cues reminded them that they were cutting into an animal, thereby increasing the perceived animalness. Later in the filleting phase, the fish were re-categorised from animal to food, because the fish was now fish fillets. The bones, skin, viscera, head, etc. were disposed of. What remained was a form of the fish that was familiar to the children: fish fillets. Applying Lévi-Strauss' [68] concept of nature-culture, we see that, through the filleting process, the fish had gone from a natural form to a more cultivated form, and through the frying of the fish the final step in the cultivating process had been reached.

Furthermore, during the filleting and cooking phase, the children started to learn technical skills, and they clearly took pride in their work, which could be an expression of what Sennett [59] calls the emotional reward for attaining a skill and doing it well, like a craftsman. This finding of promoted self-efficacy, as defined by Bandura [69], is supported by the findings of Cunningham-Sabo and Lohse [70] in an interventional study with fourth-graders. Not only did they find an increase in cooking and food self-efficacy but also an increase in fruit and vegetable preference. Most notable is the finding that non-cookers particularly benefitted from the intervention [70]. An increase in cooking efficacy was also confirmed in a similar study including an experiential approach by Jarpe-Ratner, Folkens, Sharma, Daro, and Edens [71], although no definition of the concept experiential was given.

At the end of the experiment, the fish were fried, and the re-categorisation from animal to food was complete. Observations showed that the majority of children chose to taste and eat the fish fillet on rye bread; the reason given was the good taste, a reason corresponding to the findings of Sick, Højer, and Olsen [29] in a study on children's self-evaluated reasons for accepting and rejecting foods.

Rejection was promoted by, for example, the texture of the cooked fish in the mouth. Rejection of fish based on texture was found by Donadini, Fumi, and Porretta [72], where fish was rejected due to softness, a jelly-like texture, fast melting, and tendency to fall apart easily textures. Texture was also found to be a key rejection characteristic by Sick, Højer, and Olsen [29]. For the children that showed reluctance throughout the experiment and ended up tasting the fish fillet, an "I filleted and cooked it myself" effect could be a possible explanation. A similar effect of "I cooked it myself" was found by Dohle, Rall, and Siegrist [73] and Allirot, da Quinta, Chokupermal, and Urdaneta [74]. Allirot et al. [74] and van der Horst, Ferrage, and Rytz [75] also point to the context or atmosphere in which the food exposure took place and the "cooking together" factor as relevant factors impacting food likes and dislikes and thereby promoting acceptance or rejection. Since the gyotaku experiment took place in a school(-like) setting, the "cooking together" and "helping each other" factors promoted acceptance of fish. According to Lukas and Cunningham-Sabo [76], there is a difference between cooking with friends and classmates. In a qualitative study, they found that classmates were typically associated with rules, structure, and restrictions, while friends are defined by fun and freedom [76]. Yet, when Lukas and Cunningham-Sabo [76] compared data across focus groups, they found that the cooking and tasting group did not make a clear distinction between classmates and friends, and the children in this group seemed to consider their classmates as friends in this "cooking together" context. This was not the case in the two other groups. However, other studies [77,78] have not found a correlation between an experience-based approach and positive change in acceptance, preference or liking of foods.

#### **5. Strengths and Limitations**

This section considers the credibility, transferability, dependability, and confirmability [53,79] of the study findings.

Credibility was sought in this study by comparing the findings with those of previous studies that have focused on similar research. Furthermore, to reduce observer bias, observer assistants were present in all but two cases, and after the experiment had ended, dialogues took place between the experimenter and assistant regarding what had been observed. Dependability was sought via a thorough description of the study design and the gyotaku experiment itself in order to ensure that other researchers are able to execute a study in a similar way. Even though true objectivity of the researcher rarely exists, confirmability was sought through a sampling process, whereby the participating classes entered the study according to the rule of "first responders to the information letter" sent out via email (regional) and shared in a Food-Knowledge-specific Facebook group for teachers (national).

Geographically, the data were only collected in the eastern part of Denmark. Therefore, in terms of transferability, it could be said that this case study is not representative of the general population of children in Denmark. Nevertheless, the findings can be seen as indications transferable to similar contexts, since the observations seemed stable and comparable across the cases (schools and classes). Furthermore, even though the experiment varied in terms of setting, the observations across the two

settings seemed stable and comparable. Additionally, a research participation effect [80] cannot be completely eliminated as the presence of the research and assistant group in the gyotaku experiment situation is an addition to the typical setting situation. Furthermore, analysis and ATA frame was validated by a researcher not part of the study, but with extensive research experience to reduce bias.

Even though more research in this area is needed, this finding opens up the possibility of transferring the gyotaku exercise and the expected outcome to settings outside the conventional school setting.

We recognise that the present study holds certain limitations investigation-wise that need to be addressed in future research. One such limitation is the aspect of how children categorise and re-categorise fish and how this is connected to the experimental context, the school arena. Furthermore, a more focused investigation of how tactile play might influence children's acceptance of food, especially outside the area of fruit and vegetables, is warranted.

#### **6. Conclusions**

With regard to how children responded to handling, preparing, and cooking the fish and how the process of the gyotaku experiment affected the acceptance, we identified that response of rejection and acceptance moved back and forth on a continuum. Rejection was driven by slimy touch, whole animal, smell, cutting through skin, texture of fish meat in the mouth, and taste, and acceptance was promoted by togetherness, helping each other, tactile play, re-categorisation of the fish, exploration, pride, skills, and was self-made. Furthermore, the movement back and forth was determined by how the fish was categorised (as animal, non-animal, or food). The study revealed that autonomy, skills, pride, and helping each other in the groups were important factors in promoting acceptance, whereas the texture of the fish, for example, led to rejection. Furthermore, we found that using the fish as a creative medium for tactile play became an important motivator in promoting acceptance. The findings in this study highlight that cooking combined with tactile play could be a way of promoting acceptance of fish, and as such serve as a potential strategy in promoting healthy food behaviour. The same exercise could be used with other food groups as well, for example with vegetables, fruit, chicken (e.g., print of feet or wings before preparing) etc. where the squid ink is substituted with berry juice or beet root juice. At the same time, our findings support the importance of the school setting and the subject Food Knowledge as a potential experiential learning gateway to promoting healthy food behaviour through focusing on children's food and culinary knowledge and skills, which has also been recommended by Nelson et al. [47].

**Author Contributions:** Conceptualization and methodology, R.H., K.W. and M.B.F.; data collection and data analyses, R.H., interpretation of data, R.H., K.W. and M.B.F., writing of the manuscript, R.H., review and editing, K.W. and M.B.F., supervision and funding acquisition, M.B.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is part of the research project Smag for Livet (Taste for Life) and was partly funded by the Nordea Foundation, Denmark, and by University College Absalon, Center for Nutrition and Rehabilitation, Nutrition and Health, Slagelsevej 70-74, 4180 Sorø, Denmark. The funding parties had no involvement in the work.

**Acknowledgments:** A special thank you to Margit Dall Aaslyng, University College Absalon, Center for Nutrition and Rehabilitation, Nutrition and Health, Sorø, Denmark, for invaluable comments provided in relation to the applied thematic analysis process.

**Conflicts of Interest:** The authors declare no conflict of interest with regard to authorship, research, funding and/or publication of this article.

#### **Appendix A**


**Table A1.** Data set: Meta-themes, sub-themes, and data extracts from field note journal (obs: observation \*, ic: informal conversation \*\*).

**Table A1.** *Cont.*


**Table A1.** *Cont.*


**Table A1.** *Cont.*


Phase in the experiment: #1 = Before printing; #2 = During printing; #3 = Between printing and filleting; #4 = During filleting; #5 = Frying; #6 = Tasting; #all = All phases of the experiment. obs: observation \*, ic: informal conversation \*\*.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*
