A Comprehensive Assessment of Food Parenting Practices: Psychometric Properties of the Portuguese Version of the HomeSTEAD Family Food Practices Survey and Associations with Children’s Weight and Food Intake

The Home Self-Administered Tool for Environmental Assessment of Activity and Diet (HomeSTEAD) survey evaluates a broad spectrum of food parenting practices related to parental use of control, autonomy support, and structure. This study aims to test the psychometric properties of the Portuguese version of the HomeSTEAD family food practices survey in parents of 3–12 year old children. Data were collected from 184 parents/caregivers. We performed an exploratory factor analysis (EFA), calculated the internal consistency coefficients of each subscale, and tested for associations with children’s food intake and weight. Based on the EFA, 61 items were included in the Portuguese version of the HomeSTEAD family food practices survey, and were distributed among four Coercive Control Practices (16 items); five Autonomy Support Practices (17 items); and nine Structure Practices (28 items). All scales demonstrated an acceptable level of internal consistency. A higher body mass index (BMI) SD score in children was associated with higher levels of restriction and weight talk by parents and distractions during meals. Higher levels of distractions during meals were also associated with higher sweets intake in children. Additionally, higher levels of parental modeling and the establishment of rules and limits were associated with lower intake of sugar-sweetened beverages. These associations provide preliminary evidence of the HomeSTEAD family food practices survey’s construct validity and attest to its potential to assess parental strategies and provide useful information to improve children’s eating.


Introduction
Parents are the main mediators between children and food. They guide their dietary intake and eating behavior through strategies called food parenting practices. The study of food parenting practices has traditionally focused on practices of control, namely restriction and pressure to eat [1]. However, as interest in the field grew due to its potential to prevent obesity, instruments were created to study alternative approaches to control [1]. For instance, those food parenting practices that are concerned with the organization of the food environment in terms of availability of healthy foods, and positive strategies that parents can apply to promote healthy eating, such as nutrition education, have more recently been taken into consideration [2]. However, inconsistencies remained in the terminology and definitions and a comprehensive measure was lacking [3].
The HomeSTEAD family food practices survey is a promising measure for the comprehensive evaluation of food parenting practices. It includes a broad spectrum of parenting practices in addition to the focus on 'Restriction', 'Pressure to eat', and 'Monitoring' as addressed in the Child Feeding Questionnaire [22]. Another well-known comprehensive measure of food parenting practices, the Comprehensive Feeding Practices Questionnaire [23], can already be used to evaluate how parents model, teach, or encourage healthy eating. However, the HomeSTEAD family food practices survey integrates a broader evaluation of the concepts of 'Structure' (with subscales such as distractions and rules and limits) and 'Support of Autonomy' (with subscales such as guided choices). However, attempts to develop this more comprehensive measure require further research on construct validity and reliability in varying linguistic and cultural contexts [21]. Therefore, we sought to investigate the psychometric properties of a Portuguese translation of the HomeSTEAD family food practices survey in parents of [3][4][5][6][7][8][9][10][11][12] year old children. We tested the factorial structure of each group of practices independently, following the same procedure that was adopted in the study with the original version. Additionally, we tested the construct validity by analyzing the association with children's weight (BMI SD score and weight status) and food intake. Based on the described findings, we expected a positive correlation between children's weight and the 'Restriction' and 'Weight talk' subscales, and a negative correlation with 'Pressure to eat' [16][17][18]. Regarding children's intake, a poor-quality diet (high-fat and high-sugar food and drink intake) was expected to be negatively associated with the 'Rules and limits' [10] and 'Guided choices' [3] subscales and positively associated with the 'Distractions' subscale [13,14]. On the other hand, a better-quality diet (higher fruit and vegetables intake) was predicted to be negatively associated with the 'Guided choices' subscale [11] and positively associated with the 'Availability and accessibility of fruit and vegetables' [10], 'Nutritional education', 'Encouragement', and 'Modeling' subscales [8]. Given the lack of empirical data regarding the other subscales of the HomeSTEAD family food practices survey, no significant correlations were expected, with the exception of two subscales. Based on theoretical reasoning, we expected to find an association between the 'Planning and preparation of healthy meals' and 'Praise' subscales and higher fruits and vegetables intake [24].

Participants
The participants included 184 parents or caregivers of children aged 3-12 years. Participants were recruited from December 2017 to September 2018 in two different contexts. One context was a health care center, where parents were invited sequentially to participate by family doctors and nurses during child health care visits (n = 76). The other context was 11 parents' associations of preschools and primary schools that invited parents to answer the questionnaire (n = 108). Parents or caregivers were eligible to participate if they were ≥18 years old and if they were involved in the management of a child's eating habits (an involvement score of ≥5 on a 0-10 rating scale, from 'Not at all' (0) to 'Extremely involved' (10), regarding food acquisition and meal planning and preparation). Furthermore, parents or caregivers were not eligible to participate if the child whose eating habits were under management had any disability that required regular medical care or specific food habits or had had nutritional counseling in the previous 6 months.

Procedures
This study complied with the Ethical Principles of the Declaration of Helsinki [25] and was approved by the Ethics Committee of the Faculty of Psychology and Educational Sciences of the University of Porto (reference number 2017/10-4).
To adapt the HomeSTEAD family food practices survey to Portuguese, we followed the recommended guidelines for cross-cultural adaptation of instruments [26]. Permission for translation, adaptation, and validation was obtained from the authors. In the translation process, some cultural adjustments were needed, mainly in relation to foods and cooking methods (e.g., in the item 'When you serve potatoes, how often are they fried, like French fries or hash browns?', we dropped the examples, because they are not common in Portuguese culture). We applied the back-translation technique to find a preliminary version and tested it with five volunteers. These volunteers satisfied the eligibility criteria of our sample and were from different sociodemographic backgrounds. We asked each of them what the meaning of each item was and made adjustments when needed to improve comprehension.
Questionnaires were self-administered in an electronic format. The exception was the Food Frequency Questionnaire, which was only answered at the health care center with the support of a professional dietician. Therefore, data on food intake were not collected from parents that were not evaluated face-to-face (n = 75). Additionally, we assumed that data were missing in anthropometrics terms if the data relative to the last measurement in a child were missing or if they had been collected more than three months ago (n = 40).

The HomeSTEAD Family Food Practices Survey
The HomeSTEAD family food practices survey is an 86-item questionnaire that evaluates 24 food parenting practices and is organized into three main domains: coercive control, structure, and autonomy support. Coercive practices, which reflect parents' attempts to impose their will, are measured with 16 items that are grouped into five subscales: restriction, threats and bribes, soothing with food, a "clean plate" policy, and pressure to eat. Structure practices, which reflect parents' organization of a child's environment, are measured with 46 items that are grouped into 12 subscales: monitoring of unhealthy foods, rules and limits around unhealthy foods, components of the child's eating environment (i.e., meal setting, family eating, atmosphere of meals, eating area, and distractions), planning and preparation of healthy meals, attractive presentation of healthy foods, availability and accessibility of healthy foods, modeling, and weight talk. Autonomy support practices, which reflect the support of child competence and self-endorsement of behaviors, are measured with 24 items that are grouped into seven subscales: encouragement, reasoning, praise, nutrition education, and guided choices around when, what, and how much food is eaten. Parents answered each item using a five-point Likert scale (with the exception of one item that was answered in a seven-point scale) [5].

Food Intake
Parents/caregivers answered a Food Frequency Questionnaire [27], which rated the frequency of their children's intake of a list of foods and drinks within a nine-point scale that ranged from "never" to "three or more times a day". Portions of a food group's intake (Fruit and Vegetables, Sugar-Sweetened Beverages, and Snacks and Sweets) resulted from the sum of frequencies of the items that composed each group.

Child's Anthropometric Data
Parents/caregivers reported data contained in their child's health record booklet related to the last anthropometric measurement (date, weight, and height). For those recruited in child health care visits (n = 76), the child's anthropometrics were evaluated just before the interview. We computed the BMI as the value of weight (kg) over the squared height (m), and the age-and sex-specific BMI SD score (BMIz) according to World Health Organization (WHO) Growth References [28]. We defined a child's weight status (underweight, normal weight, overweight, and obesity) based on the cut-offs of the WHO [28].

Statistical Anaysis
The assumption of normality was confirmed by a Kolmogorov-Smirnov test and by analyses of each item distribution by the skewness and kurtosis values [29]. We excluded items with high values of skewness and kurtosis (|Sk| > 3 and |K| > 7). We then performed an exploratory factor analysis (EFA) to test the dimensionality of items. We ran separate EFAs for each group of practices (control, structure, and autonomy support), similarly to the procedure that was adopted in the original study, assuming the distinctiveness of the constructs [5]. We confirmed the factorability of data with Bartlett's test of sphericity (p < 0.05) and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (≥0.6).
To obtain a factor's loading structure, we applied an orthogonal rotation (VARIMAX). To determine the number of dimensions, the criterion of an eigenvalue greater than 1, the scree test, and a parallel analysis were used [30]. We identified and excluded items with low factor loadings (<0.40) and then repeated the analysis.
Internal consistency coefficients (Cronbach's Alpha) for each subscale were held to reflect an acceptable level of internal consistency at values over 0.60, considering the reduced number of items in each factor [29]. Inter-item correlations were expected to be at values over 0.30, and we calculated the mean inter-item correlation (MIIC), which was expected to be between 0.15 and 0.50 [31]. For subscales with an MIIC higher than 0.50, multicollinearity between items was tested by variance inflation factors (<4.0) and tolerance (>0.2) [29].
The final score of each subscale was calculated by the mean of the items that composed it, with higher punctuations reflecting higher levels of each practice. The exception was 'Guided choices', in which higher levels reflected a lower level of guidance by parents and a higher degree of autonomy for the child concerning when, what, and how much to eat. Therefore, the associations to be expected with this subscale should be read inversely.
We tested the construct validity by analyzing the Pearson's correlations of the score of each parental practice with children's BMIz and food intake (Fruit and Vegetables, Sugar-Sweetened Beverages, and Snacks and Sweets). Additionally, we tested differences in scores between weight status and specific groups with one-way ANOVA.
All analyses were performed with the IBM ® SPSS ® Software, version 24.0 [32].

Coercive Control Scales
All items had satisfactory factor loadings. The EFA identified four factors, which were confirmed by the parallel analysis to explain 61.3% of the variance. Three factors ('Restriction', 'Soothing with food', and '"Clean plate" policy') remained the same as in the original scale (see Table 2). Items from the other two original subscales, 'Pressure to eat' and 'Threats and bribes', were punctuated together in the same factor that we renamed to 'Threats, bribes, and pressure to eat'. All subscales of Coercive Control had a good level of internal consistency, with acceptable coefficients (between 0.67 and 0.81) and MIIC (between 0.39 and 0.50). Table 2. The factor loadings and internal consistency of Coercive Control Practices of the Home Self-Administered Tool for Environmental Assessment of Activity and Diet (HomeSTEAD) family food practices survey for parents of 3-12 year old children (n = 184).

Portuguese Version of Coercive Control Scales
Original Factor Factor Loadings How often do you restrict (or try to restrict) your child's food intake so that she or he will not gain weight?

Autonomy Support Scales
We excluded seven items, and the reasons for exclusion per item are provided in Appendix A. Afterwards, we repeated an EFA with the remaining 17 items. All items had sufficient factor loadings. We found five factors, explaining 61.7% of the total variance. Two factors ('Praise' and 'Guided choices: amount of food eaten') replicated the original subscales of the HomeSTEAD family food practices survey (see Table 3). The factor 'Nutrition education' was very similar to that in the original subscale, with only one excluded item. We excluded one item of the original subscale 'Guided choices: when food is eaten' and another of the original subscale 'Guided choices: what food is eaten'. The remaining items of those subscales were punctuated together in the same factor, which was therefore called 'Guided choices: when and what food is eaten'. We excluded three items of 'Reasoning' and one of 'Encouragement'. The remaining items of those subscales were punctuated together in the same factor that we renamed 'Encouragement and reasoning'. All final subscales had an acceptable level of internal consistency (between 0.62 and 0.73) and an acceptable MIIC (between 0.31 and 0.50).

Structure Scales
We excluded 18 items; the reasons for exclusion per item are provided in Appendix A. We repeated the EFA with the remaining 28 items, which revealed the presence of nine factors, explaining 66.9% of the total variance. The scree plot and the parallel analysis confirmed this nine-factor structure. Three ('Monitoring of unhealthy foods', 'Weight talk', and 'Distractions') of the nine factors replicated the original scale (see Table 4). The factors 'Atmosphere of meals' and 'Availability and accessibility of healthy foods' were very similar to the original subscales (with only one excluded item per subscale). Five of the excluded items represented the factors 'Meal setting' and 'Family eating' that, therefore, did not emerge. Only one factor of 'Meal setting' persisted and was punctuated with 'Rules and limits'. As this was theoretically related to the factor 'Rules and limits', it was included in this subscale, which in turn remained similar to the original one (with only one excluded item). Compared to the original subscales, the factors 'Eating area/physical space' and 'Modeling' had two and three excluded items, respectively. Major differences were found in the subscale 'Planning and preparation of healthy meals', with four deleted items. The other two items were punctuated together with the remaining item of 'Attractive presentation of healthy foods' (the other two were excluded) and the remaining item of 'Availability and accessibility of healthy foods'. We considered these items to be theoretically related and we renamed this factor 'Planning, preparation, and attractive presentation of foods and meals'.  The structure subscales of the Portuguese version of the HomeSTEAD survey had an acceptable level of internal consistency (between 0.61 and 0.94) and MIIC (between 0.26 and 0.85). We confirmed that there was no multicollinearity between items of subscales with an MIIC of over 0.50.
In total, the Portuguese version of the HomeSTEAD family food practices survey ended up with 61 items: 16 items representing four Coercive Control Practices, 17 items representing five Autonomy Support Practices, and 28 items representing nine Structure Practices.

Construct Validity
The differences between food parenting practices in terms of children's weight status categories are reported in Table 5. We also present correlations between the score of each food parenting practice of the HomeSTEAD family food practices survey and children's BMIz.
Children's BMIz was strongly correlated with 'Restriction' (r = 0.48; p < 0.001). When evaluating the differences between weight statuses, parents reported higher levels of restriction when their Correlations between the score of each food parenting practice of the HomeSTEAD family food practices survey and food intake are reported in Table 6, as a means to analyze construct validity. Table 5. Scores of food parenting practices for the full sample and according to children's sex, age and weight status groups and correlations with body mass index (BMI) SD score (N = 184).   1 Missing data for 75 participants. Note: We did not find the expected significant correlations between children's intake of fruit and vegetables with 'guided choices', 'availability and accessibility of fruit and vegetables', 'nutritional education', 'encouragement', 'modeling', 'planning and preparation of healthy meals', and 'praise'. * p < 0.05; *** p < 0.001.

Discussion
This study aimed to investigate the factor structure and psychometric properties of a Portuguese version of the HomeSTEAD family food practices survey in a sample of parents of children aged 3-12 years old.
The Portuguese version of the HomeSTEAD family food practices survey is composed of 18 scales with an acceptable level of internal consistency. A total of 25 items were removed from the Portuguese version and, consequently, two factors were eliminated: 'Family eating' and 'Meal setting'. The majority of the excluded items were highly dependent on culture. For instance, to the question 'what do family dinners on weekdays look like at your home', almost all parents answered 'family members sit down and eat together'. The scale with the most excluded items was 'Planning and preparation of healthy meals', and we should acknowledge that it was also the scale we made major cultural adjustments to during the adaptation to Portuguese, mainly in relation to foods and cooking methods.
Regarding construct validity and unlike the original study, we found significant correlations between some food parenting practices and children's BMIz. Parents of children with a higher BMIz reported higher levels of restriction. Comparing the scores of the practices by children's weight statuses, parents of children with overweight and obesity revealed higher scores than parents of children with a normal weight for their age. This finding is in line with those of previous studies on restriction [16][17][18]. We should also note that the 'Restriction' questions of the HomeSTEAD family food practices survey are relevant to weight control (e.g., "How often do you restrict your child's food intake so that she or he will not gain weight?") compared to the 'Restriction' questions from other existing instruments, such as the Child Feeding Questionnaire (e.g., "If I did not guide or regulate my child's eating, she would eat too much of her favorite foods"). Therefore, this positive association between restriction and weight is even more likely using the HomeSTEAD family food practices survey and this measure may be more useful for evaluating restrictions related to weight control. We also found that 'Weight talk' was higher in parents of children with obesity compared with parents of children with overweight or children with a normal weight for their age. This finding was also in the predicted direction [16][17][18]. We did not find a significant correlation between BMIz and the parent's use of pressure to eat, which might be explained by the fact that this subscale in the Portuguese version also integrates items related to threats and bribes.
Regarding the practices of coercive control, the only correlation found with children's intake was between the use of soothing with food and children's snacks intake. Previous studies have already observed an association between emotional feeding and children's snacking [33]. To date, this specific practice of 'Soothing with food' has mostly been studied in infants, and a longitudinal association with a higher weight has been reported [34]. This scale of the HomeSTEAD family food practices survey adds the possibility to evaluate this specific practice in parents of older children.
Concerning the construct validity of autonomy support practices, the only practice that correlated with children's BMIz and food intake was guided choices. Lower levels of parental guidance in relation to what and when to eat were associated with a lower BMIz in children. This was not in the direction initially predicted, as some studies have found that a more permissive parenting style (with less guidance by parents on children's choices) is associated with a higher weight status [3]. However, this negative association may be related to a reverse causality, i.e., parents may be giving a higher degree of autonomy to children in relation to food intake as they have a lower BMIz. In fact, some studies have suggested that parents tend to apply strategies to monitor and control eating in response to children's weight [35][36][37], and our findings point in the same direction with less guidance associated with a lower BMIz. Additionally, concern about a child with overweight was found to be positively associated with healthy eating guidance, as evaluated by the Comprehensive Feeding Practices Questionnaire in a previous study [38]. On the other hand, a lower number of parent-guided choices was associated with sugar-sweetened beverages intake in our sample [3]. However, we did not find any association with snacks, sweets, or fruit and vegetables intake.
Regarding the structure practices, we found positive correlations between being distracted during meals and children's BMIz. As expected [13,14], those distractions were also correlated with a worse-quality diet by a higher snacks and sweets intake. However, we did not find any correlations with sugar-sweetened beverages intake, in line with previous studies [13][14][15]. Recent studies have shown positive associations between daily television watching and children's weight [39], and between children watching advertising during a meal and higher intake of poor-quality food [40]. What this scale adds is the possibility to evaluate the specific practice of parents leaving the television turned on during meals, thus allowing for this practice to be assessed. Concerning correlations between structure practices with children's diet, we found that a higher degree of establishment of rules and limits by parents was associated with a lower intake of snacks, sweets, and sugar-sweetened beverages. These were the expected directions of associations [10]. Additionally, a higher degree of reported modeling by parents was associated with lower sugar-sweetened beverages intake, which is also supported by the literature [8].
We did not observe the expected correlations between 'Availability and accessibility of healthy foods', 'Nutrition education' 'Encouragement', 'Planning and preparation of healthy meals', and 'Praise' with fruit and vegetables intake. Some associations between practices of parents and children's weight and food intake may not have been found because the questionnaire targeted a large age range. The age ranges of 3-6 and 6-12 years are characterized by different levels of autonomy and parental and peer influence, which affect children's behaviors differently [10,41].
The strength of this study is our attempt to make the HomeSTEAD family food practices survey available for use in research on food parenting practices, considering that it covers a broad spectrum of practices. Additionally, it adds knowledge on the psychometric properties of the HomeSTEAD family food practices survey. This study has also some limitations that deserve attention. First, we used the minimum acceptable sample size to analyze the factor structure. Second, the food intake assessment was based on a self-report by parents, without an observational assessment of the home's food environment (as performed in the original study) or of meals. Despite this, the Food Frequency Questionnaire has been shown to be a good instrument to evaluate dietary intake in children, with strong correlations with data from food diaries and serum biomarkers [27]. Regarding anthropometrics, we asked parents to report data collected by clinicians as an assurance of quality. However, it would have been useful to measure objectively the height and weight of each child in order to avoid excluding those with less-current measurements or parents without access to their child's health record booklet. Third, the majority of the participating parents were mothers; despite this, we attempted to obtain answers from both mothers and fathers. Whenever we had both parents present, the mother was mostly asked to answer because of her greater involvement in the management of infant feeding. This problem is common to other studies on food parenting practices [42], and limited our ability to evaluate precisely the food parenting practices that were exerted on the child, as meal time observations have indicated that fathers' practices differ from mothers' [43].
In the future, the factor structure of the Portuguese version of the HomeSTEAD family food practices survey should be tested in a larger sample. There is also a need to test for invariance in children's age and gender [42]. It would also be important to confirm the temporal stability with a test-retest reliability analysis.

Conclusions
In conclusion, the HomeSTEAD family food practices survey, which evaluates food parenting practices of Coercive Control, Structure, and Autonomy Support, was adapted to be used with Portuguese parents of children aged 3-12 years old. This study supports the view that some items need to be reformulated during the adaptation of this survey to countries with different cultural habits, namely those related to foods and cooking methods and meal settings. We found significant correlations between food parenting practices, identified in the literature as closely linked to obesity and children's BMIz and food intake, which provide preliminary evidence of the construct validity of the Portuguese version of the HomeSTEAD family food practices survey. Acknowledgments: The authors gratefully acknowledge the families that participated in this study and the health care center's professionals and parents' associations involved in their recruitment. They also acknowledge the Maia-Valongo Health Centre Group and particularly Emília Mendes and Isabel Cortijo for their support in testing the system within the Family Health Units Viver Mais and Íris, respectively.

Conflicts of Interest:
The authors declare no conflicts of interest. Guided choices: When food is eaten Low Cronbach's α As a parent, I decide the kinds of food my child eats. 2 Guided choices: What food is eaten

Structure Scales
What do family dinners on weekdays look like at your home? 3 Family eating Asymmetry/ kurtosis What do family dinners on weekends look like at your home? 3 Family eating