Children’s eating attitudes test (ChEAT): validation and reliability in Turkish children

Objectives This study aims to examine the validity and reliability of the children’s eating attitudes test (ChEAT) in Turkish children. Methods The participants were 331 children (137 boys and 194 girls, ages 8–15). Data was collected through face-to-face interviews using a questionnaire containing socio-demographic characteristics, the ChEAT, and the children's eating behaviour questionnaire (CEBQ). Exploratory factor analysis (EFA) was performed to determine the factor structure of the Turkish version of the ChEAT. Additionally, the reliability was examined in terms of internal consistency and test–retest reliability. All statistical analyses were performed using Mplus Trial Version and SPSS 11.5 (SPSS, Chicago, IL, USA). Results According to the goodness-of-fit statistic, a three-factor solution was appropriate and compatible with clinical considerations. The three factors explained 50.1% of the variance. Cronbach’s alpha coefficient was 0.75 for ChEAT-26, 0.67 for “Preoccupation with thinness and food”, 0.63 for “Social pressure to eat”, and 0.71 for “Dieting”. Furthermore, the test–retest reliability was 0.72, 0.62, 0.59, and 0.59 respectively. Statistically significant correlations between the ChEAT and CEBQ were found (p < 0.05). “Preoccupation with thinness and food” was significantly higher in obese children (p < 0.05), while “Social pressure to eat” was lower (p < 0.001). Sex, grade, BMI, parental education in addition to working status affected the ChEAT-26 scores. Conclusions The present study has provided preliminary evidence for the validity and reliability of a Turkish version of the ChEAT.


Introduction
Obesity and eating disorders (EDs) adversely affect physical and mental health. These two significant public health issues are also connected [1,2]. EDs are severe psychosomatic disorders common in adolescent girls. However, they can occur regardless of gender or age [3,4]. Anorexia nervosa (AN), bulimia nervosa (BN), and binge eating disorder (BED)) are the most commonly studied EDs [5]. AN incidence begins to increase at around age 10. However, it is also reported in children aged 7-8 [6]. In a Canadian study, it was determined that between the ages of 5 and 12, early-onset EDs were found in 2.6 (95% confidence interval [CI], 2.1-3.2) cases per 100 000 person-years. Also, 47.1% of girls and 54.5% of boys showed signs of growth delay, and 46% of children were below the 10th percentile for body mass index [7]. In a recent study, among children in the United States, the prevalence of subclinical AN, BN, and BED was 6.0%, 0.2%, and 0.5%, respectively [8]. EDs in childhood and adolescence are correlated with overweight/obesity, reduced body esteem, and poor mental health [9,10]. Considering that EDs occur early, it is crucial to screen for nutritional attitudes and behaviors which may lead to malnutrition [11].
Garner and Garfinkle [12] developed the 40-item Eating Attitudes Test (EAT) with the aim to evaluate adult eating attitudes and behaviors [12]. It was shown that the reliability and validity of the factor analysis remained unchanged when the number of items was reduced to 26. The EAT-26 is widely used worldwide to screen and evaluate symptoms along with the features of EDs [13]. Maloney et al. modified the EAT-26 to develop the Children's Eating Attitudes Test (ChEAT) in order to make it easier for children to grasp [14]. Therefore, some words were simplified, e.g., "terrified" was changed to "scared", and "preoccupied with" was reworded as "think a lot about" [14]. The Children's Eating Behavior Questionnaire (CEBQ) was developed by Wardle et al. [15], a parent-report measure designed to assess variations in children's eating behaviors. Yilmaz et al. [16] conducted a Turkish reliability and validity study for CEBQ. While the CEBQ is a scale that evaluates eating behaviors reported primarily by parents, ChEAT is important because it is a self-administered scale by children used to evaluate EDs.
The ChEAT scale has been adapted in various countries, and its validity and reliability have been examined [11,[17][18][19]. The original EAT had three subscales: "Dieting, "Bulimia and Food Preoccupation", and "Oral Control". [13], while the ChEAT had no subscales in [14]. However, different factor structures were found in the validity and reliability studies of ChEAT in other languages. Additionally, some items were deleted from the scale [11,17]. In Finnish children, the ChEAT was determined to have four factors with 24 items, and Cronbach's alpha value of the scale was found to be 0.84 [17]. Recently, in a Japanese study, the scale was determined as five factors with 25 items. Cronbach's alpha varies between 0.58 and 0.82 for subscales and 0.81 for the total scale [11]. A similar study in Spain revealed five factors with 26 items (Cronbach's alpha coefficient between 0.58 and 0.84 for subscales and 0.86 for the total scale) [19]. The different factor structures may be due to cultural differences. A similar structure may not be provided in every society and culture. Therefore, psychometric analyzes of ChEAT in Turkish culture are very important.
The prevalence of childhood and adolescent obesity is rising globally as well as in Turkiye [20]. The WHO European Regional Obesity Report 2022 has revealed that Turkiye has a higher obesity rate (32.1%) compared to that of other European nations (23.3%). On the other hand, no large sample-size studies evaluate the prevalence of eating disorders in children or adolescents in Turkiye [21]. Recent literature on the burden of eating disorders suggested that in individuals with eating disorders, quality of life is reduced, yearly healthcare expenditures are 48% greater than that in the general population, and mental health comorbidity is associated with 48% lower annual earnings [22]. Early diagnosis in addition to intervention of abnormal eating attitudes are crucial to minimizing adverse outcomes, which creates a need for a psychometrically reliable assessment tool. The ChEAT is the most widely used standardized self-report measure of symptoms and concerns characteristic of eating disorders worldwide [23]. This study aims to examine the validity and reliability of the ChEAT in Turkish children.

Participants
The study was carried out in elementary, middle, and high schools in Erzurum, Turkiye. Inclusion criteria include children aged 8-15 with no chronic or mental disorders. In order to conduct factor analysis, it is specified that the sampling must be taken at least five and preferably ten times the scale items number [24]. Our goal was to conduct 260 participants as there are 26 items in ChEAT. In total, 331 children participated in the study in case extreme and missing values would emerge. This study comprised 194 (58.6%) girls and 137 (41.4%) boys. The mean age of the children was 11.73 ± 2.19 years. In order to assess test-retest reliability, a subset of 93 children was invited to recomplete the ChEAT after 15 days.

Procedure
Maloney et al., who developed the scale, were contacted via email for permission to adapt the Children's Eating Attitude Test into Turkish and exemine its reliability and validity in Turkish children [14]. Simple random selection was used to recruit participants from randomly selected schools in Erzurum (one of the metropolitan cities of Turkiye). The data was collected utilizing surveys and inperson interviews. The children along with their parents received two surveys (one child and one parent). Uncompleted surveys were not included.
Following the standard procedure recommended by Brislin [25] and Prieto [26], the scale was translated from English to Turkish by researchers proficient in both English and Turkish. Five academics proficient in English as well as Turkish in the field of nutrition and dietetics in addition to eating attitudes and behavior in children translated the original scale into Turkish. They contributed their opinions regarding the intelligibility of the scale and its relevance to Turkish culture. Following receipt of expert comments, all of the scale's items were revised. Then, the Turkish form of the scale was translated to English by five other Turkish academics specializing in nutrition and dietetics along with eating attitudes and behavior in children who had never seen the English form of the scale before and knew both languages and cultures. A professional translator was then consulted for approval on the Turkish and English translations of the ChEAT. Following consultation, the Turkish version was a close translation of the original version. The preapplication of the scale was undertaken with 30 children to assess the intelligibility of the questionnaire. In line with the opinions of the researchers, the ChEAT's items were easily comprehensible to children (see Appendix for Turkish translation).
Ethical permission was obtained from the Erzurum Technical University Ethics Committee (Meeting Number: 8; Decision Number: 4; and 29.08.2022) in addition to the Erzurum Provincial Directorate of National Education (19.09.2022). The study was carried out following the principles outlined in the Helsinki Declaration. Written informed consent was obtained from parents and verbal consent of the children was obtained.

Measures
The survey for children contained general information, the Children's Eating Attitudes Test (ChEAT) [14], and anthropometric measurements. The children themselves responded to this survey. The survey of the parents contained general information and the Children's Eating Behavior Questionnaire (CEBQ) [16]. One of the parents (mother/father/guardian) responded to this questionnaire with questions regarding their children.

The children's eating attitudes test
Maloney et al. [14] modified the EAT-26 to develop the Children's Eating Attitudes Test-26 (ChEAT-26) with the aim to make it easier for children (8)(9)(10)(11)(12)(13)(14)(15) years old) to grasp [14]. The ChEAT-26 self-administered questionnaire assesses dieting behavior and eating attitudes of children [27]. It consists of 26 items which are scored on a six-point Likert scale with the categories "never", "rarely", "sometimes", "often", "usually", and "always". The score is calculated by recording "never", "rarely", and "sometimes" categories as zero, "often" as one, "usually" as two, and "always" as three. Items 19 and 25 were reverse-coded, as suggested by other researchers [11,17,27]. The total original score ranges from 0 to 78. Higher scores indicate the severity of the eating disorder. Cronbach's alpha coefficient was 0.76 in [14]. This study utilized the Turkish translation of the original "The Children's Eating Attitudes Test".

The children's eating behavior questionnaire
The Turkish Children's Eating Behavior Questionnaire (CEBQ) consists of 35 Likert-type items assessed on a 5-point Likert scale by parents (1 = never, 5 = always).
In the original study [15], in which the scale was developed, an eight-subscale factor structure formed during the scale's development. The eight subscale Cronbach alpha coefficients ranged from 0.74 to 0.91. The subscales are as follows: Food responsiveness, Enjoyment of food, Emotional overeating, Emotional undereating, Desire to drink, Slowness in eating, Satiety responsiveness, and Fussiness. Yilmaz et al. [16] adapted the Turkish version of this scale. Cronbach alpha coefficients for the Turkish version varied between 0.61 and 0.84 [16]. Each subscale is assessed independently. This scale was utilized to determine the external construct validity of the Turkish version of the ChEAT, as it evaluates eating behavior in children.

Anthropometric measurements
The body weight and height of the parents were taken based on the self-reports. The researchers carried out measurements in children. The height and weight of children were measured using the methods given by Lohman et al. [28]. According to age, weight, height, BMI, and Z-scores were calculated according to age, using the World Health Organization's growth standards [29] with use of the WHO AnthroPlus software (version 1.0.4, February 2011). The BMI of children was categorized according to the Z-score junctions [29].

Statistical analysis
Exploratory factor analysis (EFA) with varimax rotation was performed to determine the factor structure of the Turkish version of the ChEAT-26. The number of factors was established based on the Scree plot and clinical considerations of factor structures. Exploratory factor analysis (EFA) for categorical data was applied using the mean and variance-adjusted weighted least squares (WLSMV) estimator which is an alternative method for ordinal data, in particular, which is not distributed normally, highly skewed or kurtic, or both [30]. Items with factor loadings above 0.30 were examined as salient. The associations between the factors of the Turkish version of ChEAT-26 and the factors of the Turkish Children's Eating Behavior Questionnaire were assessed with Spearman Correlation Coefficient regarding external validity. Reliability was evaluated in terms of internal consistency and test-retest reliability. Cronbach's alpha coefficient tested internal consistency [31]. Test-retest reliability was evaluated using the Intraclass Correlation Coefficient (ICC). A Cronbach alpha coefficient of < 0.39 indicates that it is not reliable, a value of 0.40-0.59 indicates low reliability, a value of 0.60-0.79 indicates that it is very reliable, and a value between 0.80 and 1.00 indicates that it is highly reliable [32].
Following determination of the factor structure, a group comparison was conducted using the subtotal scores for each factor based on EFA. A Mann-Whitney U Test and a Kruskal-Wallis variance analysis were used to compare the subscale scores regarding gender, class, parental education and working status as well as BMI-Zscore classification.
The post-hoc test for Kruskal-Wallis variance analysis was used to perform pairwise comparisons. Median (min.-max.) was used as descriptive statistics. All statistical analyses were performed using Mplus Trial Version and SPSS 11.5 (SPSS, Chicago, IL, USA).

ChEAT-26 factor analysis
We initially performed EFA with varimax rotation assuming three, four, and five-factor solutions. According to the goodness-of-fit statistic, a three-factor solution was appropriate and compatible with clinical consideration. The Kaiser-Meyer Olkin (KMO) coefficient was 0.825; the Bartlett X 2 was 2085.576, and p < 0.001. Items and factor loadings are given in Table 1. All 26 items loaded 0.30 or higher. The three factors were labeled as "Preoccupation with thinness and food", "Social pressure to eat", and "Dieting" (explained 50.1% of the total variance).
In assessing the associations between subscales of the Turkish version of ChEAT-26 and subscales of the Turkish Children's Eating Behavior Questionnaire, statistically significant weak correlations were found between "Preoccupation with thinness and food", and "Food responsiveness", "Emotional overeating", "Enjoyment of food", "Desire to drink", "Fussiness" (p < 0.05). There were statistically significant weak correlations between "Social pressure to eat" and "Enjoyment of food", "Satiety responsiveness", and "Slowness in eating". In addition, there were statistically significant weak correlations between the 'ChEAT-26 score' and Satiety responsiveness, Slowness in eating, and Fussiness. However, no correlation was found between the "Dieting factor" of the Turkish version of ChEAT-26 or any factor of the Turkish Children's Eating Behavior Questionnaire (Table 2).

Group differences
The median (min-max) ChEAT-26 score was 12.0 (0.0-51.0) in the study. When evaluated according to gender, it was determined as 13.0 (0.0-44.0) in girls and 11.0 (0.0-51.0) in boys (p < 0.05) ( Table 3). The scores for ChEAT-26 and "Social pressure to eat" were higher for girls (p < 0.05). The "Dieting" factor scores for the class examination were statistically significant (p = 0.008). Post-hoc tests showed differences between the 3rd-5th, 4th-5th, 3rd-8th, and 4th-8th years. In addition, the 'Preoccupation with thinness and food' score in mothers with ≤ 8 years of education and the 'Social pressure to eat' score in mothers who were unemployed were significantly higher (p < 0.05). When the BMI-for-age Z score (BAZ) classification of children was considered, "Preoccupation with thinness and food" was significantly higher in obese children (p < 0.05), while "Social pressure to eat" was lower (p < 0.001) ( Table 4).

Discussion
This study aimed to adapt the Turkish version of the ChEAT-26 and evaluate its validity and reliability, an internationally recognized scale for assessing disordered eating attitudes, for a representative sample of Turkish elementary, middle, and high school students. The present study has provided preliminary evidence for the validity and reliability of a Turkish version of the ChEAT.
While the Japanese, Belarusian, and Spanish versions of the ChEAT had five factors [11,18,19], the Portuguese and Finnish versions had four factors [17,33]. In this study, the goodness-of-fit statistics obtained from the three-factor structure were sufficient. This solution was also in line with clinical consideration. All 26 items loaded 0.30 or higher. The 'Social pressure to eat' subscale and its items were the same as that in the Japanese study [11]. The different factor structures may be due to cultural differences. A similar structure may not be  provided in every society and culture. In our study, three factors with 26 items were appropriate and compatible with clinical consideration. In our study, Cronbach's alpha coefficient was 0.75 for ChEAT-26, 0.67 for "Preoccupation with thinness and food", 0.63 for "Social pressure to eat", and 0.71 for "Dieting". These results show that the Turkish version of ChEAT has good internal consistency. In addition, our results revealed that the scale has a good test-retest reliability. In the validity and reliability study conducted with Finnish children, Cronbach's alpha value of the scale was found to be 0.84 [17]. Recently, in a Japanese study, Cronbach's alpha has been found to vary between 0.58 and 0.82 for subscales and 0.81 for the total scale [11]. In a similar study conducted in Spain, Cronbach's alpha coefficient was between 0.58 and 0.84 for subscales and 0.86 for the total scale [19]. Several researchers have investigated an appropriate age cut-off for children [11,19]. In age groups between 9 and 17, cut-offs ranging from 10 to 20 have been examined. However, there is no consensus on the appropriate cut-off [11,14,19]. Due to the lack of eating disorder diagnoses, we were unable to investigate a ChEAT cut-off for recognizing eating disorder symptoms.  The ChEAT is the most widely used standardized selfreport measure of symptoms and concerns characteristic of eating disorders worldwide [23]. While the CEBQ is a scale that evaluates eating behaviors reported primarily by parents, CheAT is important because it is a self-administered scale used to evaluate eating behavior disorders. This study examined correlations between ChEAT-26 and its subscales and CEBQ. Significant weak correlations were found. In particular, the Satiety responsiveness and Slowness in eating subscales of CEBQ may be closely related to EDs. In our study, it was determined that these subscales showed weak positive correlations with ChEAT scores. These results demonstrate that the Turkish version of the ChEAT-26 has moderate external construct validity with CEBQ.
The ChEAT-26 may enhance awareness of factors (sex, grade, socio-economic status, and parental education levels) which can influence eating attitudes in school children [19]. In this study, the ChEAT score was significantly higher in girls than in boys, indicating that the tendency in girls to have EDs may be higher than boys. Moreover, the fact that girls had higher Preoccupation with thinness and food scores than boys suggests that girls are more preoccupied with thinness than boys. These results are similar to previous research [19,34]. A meta-analysis reported that females had more body dissatisfaction compared to males [35]. Furthermore, typically, females face greater pressure to achieve an ideal body shape [36]. Therefore, promoting healthy eating attitudes and habits in addition to examining distorted attitudes about thinness is essential.
In a Japanese study, the mean ChEAT score decreased with increasing grades [11]. According to Suzuki et al., the prevalence of EDs among junior high school children increased linearly and peaked in ninth grade [37]. Similar results were also found for the Dieting subscale in our study. Previous studies indicate increased EDs, especially in adolescence [38,39]. Studies reveal that negative body image and EDs are related and affect adolescent health [40]. Media influence and peer interactions shape the body image of adolescents [36]. Body image is essential to adolescent health, and precautions should be taken to promote a healthy body image among adolescents to prevent EDs.
The "Preoccupation with thinness and food" score was significantly higher in the children compared to that of mothers whose education period was ≤ 8 years and the "Social pressure to eat" score was significantly higher in the children of mothers who were unemployed (p < 0.05). Families play a major role in the development of the eating behaviors of children. It should be emphasized that family members impact the eating habits of one another [41]. Literature indicates that parental education, the occupation of the mother along with parental and teacher health consciousness are positively associated with healthy eating behaviors in children [42]. The findings of our study also indicate that parental education and occupation may influence children's eating behaviors.
According to the BAZ classification of the children, Preoccupation with thinness and food subscale scores was significantly higher in obese children (p < 0.05). In contrast, Social pressure to eat scores were lower (p < 0.001). Additionally, ChEAT distinguishes between children with and with no overeating. Previous research indicates that overweight children are more prone to EDs [43,44]. These results show that the risk of EDs may increase in obese children.

Limitations and future research
This study had some limitations. Firstly, the inclusion criteria for this study was children aged 8-15 with no chronic or mental disorders, based on self-reports from parents. However, we did not conduct structured diagnostic and cognitive interviews with participants in this study. It is possible that children with eating disorders or other psychiatric disorders were included. Secondly, we had no ED diagnoses; we could not evaluate a ChEAT cut-off point for disordered eating symptoms. Since there is a need for studies on eating disorders in Turkish children, this scale in prevalence/clinical/cause-effect studies will significantly contribute to the literature.

Conclusion
In conclusion, the present study has provided preliminary evidence for the validity and reliability of a Turkish version of the ChEAT. Factor analysis identified three factors. Due to no cut-off score, higher scores indicate the severity of the eating disorder. We propose using the Turkish version of ChEAT-26 to evaluate students for EDs in schools for epidemiological studies. Sex, grade, BMI, parental education, and working status affected the ChEAT-26 scores. Therefore, further research is needed to investigate the relationship between these factors and EDs in children.