Factor analysis of the Comprehensive Feeding Practices Questionnaire in a large sample of children

Appetite 62 (2013) 110–118

Contents lists available at SciVerse ScienceDirect

Appetite journal homepage: www.elsevier.com/locate/appet

Research report

Factor analysis of the Comprehensive Feeding Practices Questionnaire in a large sample of children Jillian J. Haszard a, Sheila M. Williams b, Anna M. Dawson c, Paula M.L. Skidmore a, Rachael W. Taylor d,⇑ a

Department of Human Nutrition, University of Otago, PO Box 56, Dunedin 9054, New Zealand Department of Preventive and Social Medicine, University of Otago, PO Box 913, Dunedin 9054, New Zealand c Department of Women’s and Children’s Health, University of Otago, PO Box 913, Dunedin 9054, New Zealand d Department of Medicine, University of Otago, PO Box 913, Dunedin 9054, New Zealand b

a r t i c l e

i n f o

Article history: Received 5 April 2012 Received in revised form 4 November 2012 Accepted 22 November 2012 Available online 30 November 2012 Keywords: Feeding practices Children Factor analysis Diet Parenting

a b s t r a c t How parents feed their children may impact on their weight and eating behaviours, both now and in the future. The Comprehensive Feeding Practices Questionnaire (CFPQ) proposes to measure parental feeding practices and was originally developed using 12 factors in relatively small, homogenous samples. In contrast the present study used a large, diverse sample (n = 1013) of children aged 4–8 years. A confirmatory factor analysis showed that the original 12-factor model was not a good fit and that several factors were strongly inter-correlated. A subsequent exploratory factor analysis yielded five scales of interest: Healthy Eating Guidance, Monitoring, Parent Pressure, Restriction and Child Control. These scales were largely supported by further analyses in these data. Parents who were concerned about their child being overweight reported more Healthy Eating Guidance and Restriction and less Parent Pressure, whereas parents concerned about their child being underweight used more Parent Pressure and less Healthy Eating Guidance. Parents who rated a healthy diet for their child as very important undertook more Healthy Eating Guidance and Monitoring of food intake and less Child Control. These five factors from the CFPQ provide a well-supported and useful set of feeding practices that could be applicable to a wide variety of population groups. Ó 2012 Elsevier Ltd. All rights reserved.

Introduction Although parental feeding practices have been shown to influence eating patterns, dietary intake and body weight of children, how different feeding practices contribute to the development of obesity remains uncertain (Stang & Loth, 2011). Restricting or pressuring children to eat, or manipulating their behaviour using food, may interfere with a child’s ability to have an appropriate response to, and relationship with, food (Birch & Fisher, 1998; Blissett, Haycraft, & Farrow, 2010; Fisher & Birch, 2002; Galloway, Fiorito, Francis, & Birch, 2006). Furthermore, parents who teach and model healthy eating have the ability to markedly affect their child’s diet in a positive way (Fisher, Mitchell, Smiciklas-Wright, & Birch, 2002; Kroller & Warschburger, 2009; Vereecken, Keukelier, & Maes, 2004). A greater understanding of how feeding practices influence the development of child obesity is warranted. A variety of measures have been used to assess a range of different child feeding practices, mostly using questionnaires. Interest arose in the 1980s when Costanzo and Woody found that greater parental restriction of girls’ eating was associated with overweight ⇑ Corresponding author. E-mail address: [email protected] (R.W. Taylor). 0195-6663/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.appet.2012.11.017

(Costanzo & Woody, 1984, 1985). Since 2001, the most widely used instrument to measure feeding practices has been the Child Feeding Questionnaire (CFQ), which incorporated assessment of parental restriction of a child’s diet with parental monitoring of child food intake and parental pressure on the child to eat (Birch et al., 2001). Additional child feeding practices of interest have included parental control (containing aspects of Restriction and Pressure to Eat) (Arredondo et al., 2006; Johannsen, Johannsen, & Specker, 2006; Wardle, Carnell, & Cooke, 2005; Wardle, Sanderson, Guthrie, Rapoport, & Plomin, 2002), emotion regulation (when food is used to control the child’s moods) (Wardle et al., 2002), prompting and encouragement (Arredondo et al., 2006; Vereecken et al., 2004; Wardle et al., 2002), rewarding (when food is given or withheld in response to good or bad behaviour) (Kroller & Warschburger, 2008; Wardle et al., 2002), parental modelling (Campbell, Crawford, & Ball, 2006; Kroller & Warschburger, 2008) and how much the child controls his/her own eating (Kroller & Warschburger, 2008; Rhee et al., 2009; Vereecken et al., 2004). With the introduction of the Comprehensive Feeding Practices Questionnaire (CFPQ) in 2007 (Musher-Eizenman & Holub, 2007), the measurement of feeding practices was broadened. This questionnaire was designed to measure 12 different feeding practices that parents were thought to employ, thereby creating a tool that

J.J. Haszard et al. / Appetite 62 (2013) 110–118

incorporated and expanded the current measures, allowing for a better description of child feeding practices. The CFPQ included the three factors from the CFQ, (while distinguishing between Restriction for Health and Restriction for Weight Control) with additional factors that assessed a variety of other practices, including how much the child controls his/her own eating, the extent to which a parent uses food to regulate their child’s emotions or behaviour and how much the parent models, teaches and encourages healthy eating (Musher-Eizenman & Holub, 2007). The 12 factors were developed from existing feeding measures and from surveys with parents, resulting in a 49-item questionnaire on which a confirmatory factor analysis was undertaken. This was then validated using correlations between the factors and parents’ concerns for overweight and underweight in their child and their feelings of responsibility. As this original analysis occurred in relatively small, homogenous subgroups (n = 269 mothers, n = 248 fathers, n = 152 mothers), and because child-feeding practices can be influenced by many factors including ethnicity, maternal weight status, maternal education, concern for child’s weight, and socio-economic status (SES) (Costa, Del Pino, & Friedman, 2011; Hupkens, Knibbe, van Otterloo, & Drop, 1998; Kroller & Warschburger, 2008; OrrellValente et al., 2007; Powers, Chamberlin, van Schaick, Sherman, & Whitaker, 2006; Robinson, Kiernan, Matheson, & Haydel, 2001; Spruijt-Metz, Li, Cohen, Birch, & Goran, 2006; Spruijt-Metz, Lindquist, Birch, Fisher, & Goran, 2002; Vereecken et al., 2004; Wardle et al., 2002; Webber, Hill, Cooke, Carnell, & Wardle, 2010), further investigation of the CFPQ is warranted. The aim of this study was to conduct both confirmatory and exploratory factor analyses on the CFPQ in a large, diverse sample with an investigation of the relationships between the resulting factors and attitude measures. Methods Participants Phase 1 of the MInT (Motivational Interviewing and Treatment) Study (Taylor et al., 2010) recruited 1093 children aged 4–8 years from primary and secondary care clinics in Dunedin, New Zealand to screen for overweight. The second phase of the MInT Study was a trial of a 2-year intervention to treat overweight in those children. The study gained ethical approved by the Lower South Regional Ethics Committee (LRS/09/09/039) and a parent or guardian of each participant gave informed consent before inclusion. Exclusion criteria included medical conditions or medication that affected BMI, severe disabilities and if the family did not expect to be living locally for the next 2 years.

111

The CFPQ contained 49 items answered using a 5-point Likert response scale. These were either questions (‘‘If this child does not like what is being served, do you make something else?’’) answered with ‘‘never, rarely, sometimes, mostly or always’’, or statements (‘‘My child should always eat all of the food on his/ her plate’’) answered with ‘‘disagree, slightly disagree, neutral, slightly agree or agree’’. Missing data in the CFPQ were excluded list-wise resulting in a final sample size of 1013. Questions about parent’s concern and perception of their child’s weight were also asked: ‘‘How concerned are you about your child’s weight?’’ with a response scale ‘‘not at all concerned, not concerned, a little concerned, quite concerned, very concerned’’ and ‘‘Compared to other children of the same age and sex, how would you rate your child’s weight?’’ with response scale ‘‘underweight, a little underweight, about right, a little overweight, overweight’’. Concern for overweight was assumed if the parent indicated concern for the child’s weight combined with a rating of ‘‘a little overweight’’ or ‘‘overweight’’. A similar process was used to identify those expressing concern for underweight. Parents also rated how much they agreed with the statement ‘‘It is important that my child eats a healthy diet’’, using a scale of 0–10 (with 0 being ‘‘definitely not’’ and 10 being ‘‘definitely’’). Confirmatory factor analysis Confirmatory factor analysis was undertaken on the 12-factor model proposed by Musher-Eizenman and Holub (2007) using the mean and variance-adjusted weighted least-squares method in Mplus version 5.21 (Muthen & Muthen, Los Angeles). In this analysis the items were treated as ordinal and were restrained to load only on the specified factor. Tests of model fit used included Comparative Fit Index (CFI), Tucker–Lewis Index (TLI) and Root Mean Square Error of Approximation (RMSEA). Modification indices (which indicate an improvement of fit if items were permitted to load on more than one factor) were calculated and a correlation matrix for the 12 factors was generated. Exploratory factor analysis Exploratory factor analysis was undertaken using STATA version 12.0 (StataCorp, Texas). A scree plot was used to indicate the number of factors that should be extracted and a Horn’s Parallel Analysis was also undertaken to verify this (Costello & Osborne, 2005). As factors were hypothesised to correlate, an oblique rotation was employed (Promax). Items with loadings greater than 0.4 were initially included in a factor. This cut-off was chosen from possible cut-offs of 0.3, 0.4 and 0.5 to give optimal Cronbach’s alpha scores and no complex items (when items load above the cut-off on more than one factor). Items were removed if their removal improved the Cronbach’s alpha (up to 0.9).

Measures Confirmatory factor analysis of new factors The parent completed a questionnaire that included demographic questions and the CFPQ while the children’s height, weight, waist circumference and blood pressure were measured. Most questionnaires were completed online at the appointment. Socioeconomic status (SES) was determined by home address and where this placed them on the New Zealand deprivation index 2006 (NZDep2006), which is a scale of 1–10 establishing deprivation in areas by factors such as income, housing, employment and qualifications (White, Gunston, Salmond, Atkinson, & Crampton, 2008). This was then split into tertiles representing low SES (high deprivation, 8–10 NZDep2006), medium SES (medium deprivation, 4–7 NZDep2006) and high SES (low deprivation, 1–3 NZDep2006). Maternal BMI was calculated by measures taken at the appointment or using self-reported data.

A confirmatory factor analysis was then undertaken on the new factors extracted from the exploratory analysis, using methods as for the initial confirmatory analysis. A correlation matrix was also generated. Split sample factor analyses As the sample was large, it was then split into two samples with similar distributions for sex, ethnicity, SES, weight status and maternal education. An exploratory analysis in Group 1 was undertaken, resulting in the proposal of factors with items loading greater than 0.4. A confirmatory analysis on this proposed structure was then carried out in Group 2 using structural equation modelling in Stata

112

J.J. Haszard et al. / Appetite 62 (2013) 110–118

12.0. How these analyses supported or contradicted the newly proposed model from the full sample was noted. Face validity Data was analysed using STATA version 12.0 (StataCorp, Texas). Factor scores were calculated by averaging the contributing item scores. Pair-wise correlations (and p-values) were calculated between the factor scores and age, concern for overweight, concern for underweight and importance of a healthy diet. Results Participant characteristics The final sample used in this analysis, with complete CFPQ data, consisted of 1013 children (92.7% of the total sample) with an average age of 6.5 years (SD = 1.4). They were diverse in weight status, SES, ethnicity, maternal BMI and maternal education (Table 1) and did not differ from those excluded from this analysis (n = 80, data not shown). Compared to national figures, the sample slightly under-represented Ma¯ori, Pacific and Asian populations. However, compared to the local population from which this sample was drawn, this group considerably over-represented these ethnicities. In the main, this sample was of higher SES and education than New Zealand as a whole. Confirmatory factor analysis The confirmatory factor analysis for the 12-factor model showed that all loadings were greater than 0.4 on the designated factor. However, the goodness of fit statistics (CFI = 0.84, TLI = 0.90), which ideally have values greater than 0.9, and the RMSEA (0.09), which should be less than 0.05, indicated that the model was not a good fit (Marsh, Balla, & Hau, 1996). Furthermore

the high modification indices suggested that a large number of items could have had substantial loading on several factors, with 24% (12/49) of the items having loadings more than 0.4 on at least one other factor. For instance, statement 18 ‘‘I have to be sure that my child does not eat too many high-fat foods’’ (from Restriction for Weight Control) had loadings greater than 0.4 on eight other factors, including Teaching about Nutrition and Emotion Regulation. Item number 45, ‘‘I often put my child on a diet to control his/her weight’’ loaded on nine other factors. This cross-loading of items further indicated a poor model fit. The correlation matrix for the 12 factors (Table 2) shows that some factors were strongly correlated with one or more of the other factors. Of particular note are the strong correlations between Modelling, Teaching about Nutrition, Encourage Balance and Variety and Environment (r = 0.54–0.81). Restriction for Health and Restriction for Weight Control were strongly correlated (r = 0.52) and also Food as a Reward with Emotion Regulation (r = 0.48) and Pressure (r = 0.44). The results of the confirmatory factor analysis, the many high modification indices and the strong correlations between factors indicated that this 12-factor model did not adequately represent the proposed underlying concepts. We therefore carried out an exploratory factor analysis to see if the data could be reduced in a more suitable way. Exploratory factor analysis Both the scree plot and the Horn’s Parallel Analysis indicated that five factors should be extracted and these explained 84% of the variance. Fifteen items with loadings of less than 0.4 were excluded (items 7, 8, 12, 13, 15–18, 20–22, 32, 37, 42, 43) and two further items excluded because the Cronbach’s alpha improved if they were not included in the relevant factor (items 28 and 44). This resulted in 32 items contributing to the five factors. The items, loadings (from the pattern matrix) and Cronbach’s alpha for the

Table 1 Demographics of the study population. Variable

Category

Participants with full CFPQ data (n = 1013) n (%)

Sex

Female Male

513 (50.6) 500 (49.4)

Socio-economic status

High Medium Low Unknown

398 (39.3) 383 (37.8) 206 (20.3) 26 (2.6)

Ethnicity

New Zealand European Ma¯ori Pacific Asian Other

767 (75.8) 135 (13.3) 39 (3.9) 45 (4.4) 26 (2.6)

Weight status (BMI percentile)

Underweight (BMI < 3rd) Normal weight (3rd 6 BMI < 85th) Overweight (85th 6 BMI < 95th) Obese (BMI P 95th)

11 (1.1) 754 (74.4) 150 (14.8) 98 (9.7)

Maternal BMI

Underweight (BMI < 18) Normal (18 6 BMI < 25) Overweight (25 6 BMI < 30) Obese (BMI P 30) Unknown

7 (0.7) 417 (41.2) 297 (29.3) 250 (24.7) 42 (4.1)

Maternal education

Some secondary Completed secondary Tertiary qualification (not University degree) University degree Other Unknown

275 (27.1) 68 (6.7) 206 (20.3) 410 (40.5) 43 (4.2) 11 (1.1)

113

J.J. Haszard et al. / Appetite 62 (2013) 110–118 Table 2 Correlations in the 12-factor model from the confirmatory analysis. 1 1. Monitoring 2. Emotion Regulation 3. Food as Reward 4. Pressure 5. Child Control 6. Teaching about Nutrition 7. Environment 8. Restriction for Weight Control 9. Restriction for Health 10. Modelling 11. Involvement 12. Balance and Variety * **

2 – 0.29* 0.18* 0.05 0.38* 0.34* 0.48* 0.13* 0.06 0.36* 0.20* 0.44*

3 – 0.48* 0.25* 0.37* 0.16* 0.30* 0.09** 0.19* 0.11* 0.15* 0.28*

4

– 0.44* 0.18* 0.03 0.29* 0.15* 0.32* 0.05 0.18* 0.10**

5

– 0.14* 0.02 0.07 0.09** 0.19* 0.06 0.16* 0.09**

6

– 0.04 0.35* 0.15* 0.09** 0.17* 0.00 0.22*

– 0.54* 0.10* 0.17* 0.65* 0.46* 0.81*

7

8

– 0.03 0.07** 0.58* 0.34* 0.65*

9

– 0.52* 0.12* 0.05 0.11**

– 0.22* 0.13* 0.09**

10

11

– 0.32* 0.78*

– 0.50*

p < 0.01. p < 0.05.

five factors are shown in Table 3, along with the excluded items and which factors they originated from. The factors are as follows: Healthy Eating Guidance (9 items): this factor indicates how much a parent models, teaches and encourages healthy eating for their child. It contains items from the original subscales of Environment, Encourage Balance and Variety, Teaching about Nutrition and Modelling. Monitoring (4 items): this factor assesses how much a parent keeps track of the unhealthy foods that their child consumes and replicates the original Monitoring factor of the CFPQ. Parent Pressure (7 items): this factor investigates how much a parent pressures the child to eat or uses food to control their behaviour and includes some items from three of the original subscales (Emotion Regulation, Food as a Reward and Pressure.) Restriction (8 items): this factor assesses how much a parent restricts or controls their child’s eating and brings together most of the items from Restriction for Weight Control with one item from Restriction for Health. Child Control (4 items): this factor determines how much the parent allows the child to make decisions around what and when they eat and uses most of the questions from the original Child Control subscale. The only item not fitting this model was question 12 (‘‘Do you allow this child to leave the table when s/he is full, even if your family is not done eating?’’) which loaded at 0.33 (<0.4) and was subsequently removed.

proposed by this analysis were the same as those proposed by the full sample. Confirmatory analysis on the model indicated by the exploratory analysis in Group 1 was then undertaken using Group 2 (n = 506) and loadings are shown in Table 3. The tests of model fit indicated poor model fit in this group (CFI = 0.75, TLI = 0.73, RMSEA = 0.07). Most of the new items suggested by the Group 1 analysis, that were not part of the factors from the full sample, loaded below 0.4 in this confirmatory analysis, supporting the full sample model. Face validity The five hypothesised factors from the full sample were correlated against four measures that may influence feeding practices to test face validity (Table 5). Parents reporting more concern for overweight tended to use more Healthy Eating Guidance with higher levels of Restriction and lower levels of Parent Pressure. Conversely, parents indicating concern for underweight showed more Parent Pressure to eat and also had lower scores for Healthy Eating Guidance. As age increased, Parent Pressure decreased and parents who strongly agreed that a healthy diet for their child is important used more Healthy Eating Guidance and Monitoring with less Child Control. These associations mostly acted in expected ways and support the five-factor model.

Confirmatory factor analysis of new factors The tests of model fit for this five factor model showed acceptable fit with CFI = 0.90 and TLI = 0.93, however the RMSEA was 0.10. The loadings from the confirmatory analysis (Table 3) showed that all items loaded greater than 0.4 on the designated factor. The correlations between factors (Table 4) revealed only two relatively high correlations (r > 0.2): Monitoring correlated positively with Healthy Eating Guidance and negatively with Child Control. There were only two modification indices that showed that one item would have loaded greater than 0.4 on other factors if permitted to do so. This was a considerable improvement from the 33 alternative loadings suggested by the 12-factor confirmatory analysis. Split sample factor analyses To further investigate the robustness of the new factor structure of the CFPQ, the sample was split into two matched groups. Exploratory factor analysis was undertaken in Group 1 (n = 507) with five factors specified. Loadings are shown in Table 3 and items that loaded greater than 0.4 were included in the factor. Healthy Eating Guidance had three items different in this group, Parent Pressure had two new items, and Restriction and Child Control had only 1 item different from the full sample factors. Otherwise the factors

Discussion This study showed, in a large sample of New Zealanders, that there were five feeding practices measured by the CFPQ, namely: Monitoring, Child Control, Restriction, Parent Pressure and Healthy Eating Guidance. The original CFPQ proposed 12 child feeding scales using a 49-item questionnaire and was factor analysed and validated in an inadequately sized sample (<5 participants per question) by Musher-Eizenman and Holub (2007). These samples were predominantly Caucasian, with high education and good socio-economic status and may have been subject to chain-sampling bias, when acquaintances of the researchers were recruited and then asked to recruit their acquaintances (Musher-Eizenman & Holub, 2007). The current study, using a confirmatory factor analysis in a much larger, more diverse sample, showed that the observations did not fit the proposed theoretical structure very well and that several of the factors were strongly correlated with other factors. An exploratory factor analysis indicated five parental feeding scales based on 32 of the original items all with loadings of more than 0.4. A confirmatory factor analysis based on these 32 items showed that five factors had a better model fit than the 12-factor model as well as lower correlations between factors. Fewer factors provide a more parsimonious solution for further

Factors and items

Original factor

Mean (S.D.)

114

Table 3 Factors, items and loadings from exploratory and confirmatory analyses in the full sample and split sample. Factor loadings in full sample (n = 1013) Confirmatory analysis

Exploratory analysis in Group 1 (n = 507)

Confirmatory analysis in Group 2 (n = 506)

0.40

0.58

0.37

–

0.47

0.59

0.49

0.37

0.70

0.78

0.68

0.64

0.63

0.62

0.54

0.65

0.62

0.64

0.57

0.54

0.52

0.68

0.48

0.48

0.56

0.71

0.58

0.59

0.74

0.84

0.74

0.75

Exploratory analysis/internal consistency coefficient (a) Healthy Eating Guidance

Factor loadings in split sample

a = 0.82

Environment

24. I encourage my child to try new foods

44. I model healthy eating for my child by eating healthy foods myself

Encourage Balance and Variety Teaching About Nutrition Encourage Balance and Variety Teaching About Nutrition Encourage Balance and Variety Modelling

47. I try to show enthusiasm about eating healthy foods

Modelling

48. I show my child how much I enjoy eating healthy foods

Modelling

22. A variety of healthy foods are available to my child at each meal served at homea 46. I try to eat healthy foods in front of my child, even if they are not my favouritea

Environment Modelling

3.7 (1.3)

0.47

4.2 (0.7) 4.2 (0.8) 4.2 (0.8) 4.1 (0.9) 4.4 (0.8)

a = 0.90

a = 0.72

0.53

0.55

0.56

0.32

Emotion Regulation

2.7 (0.8) 1.7 (0.7) 2.2 (1.2) 2.7 (1.3) 3.1 (1.3) 2.6 (1.4) 3.4 (1.3) 3.3 (1.3) 1.7 (0.7)

0.39

–

0.47

0.31

Restriction for Health

3.5 (1.3)

0.43

–

0.46

0.29

25. I discuss with my child why it’s important to eat healthy foods 26. I tell my child that healthy food tastes good 31. I discuss with my child the nutritional value of foods 38. I encourage my child to eat a variety of foods

Monitoring 1. How much do you keep track of the sweets that your child eats?

Monitoring

2. How much do you keep track of the snack food that your child eats?

Monitoring

3. How much do you keep track of the high-fat foods that your child eats?

Monitoring

4. How much do you keep track of the sugary drinks that your child drinks?

Monitoring

Parent Pressure 9. Do you give this child something to eat or drink if s/he is upset even if you thing s/he is not hungry? 19. I offer my child his/her favourite foods in exchange for good behaviour 23. I offer sweets to my child as a reward for good behaviour

Emotion Regulation

Food as a Reward

30. If my child says ‘‘I’m not hungry,’’ I try to get him/her to eat anyway

Pressure

36. I withhold sweets/dessert from my child in response to bad behaviour

Food as a Reward

39. If my child eats only a small helping, I try to get him/her to eat more

Pressure

49. When he/she is finished eating, I try to get my child to eat one more (two more, etc.) bites of food 8. Do you give this child something to eat of drink if s/he is bored even if you think s/he is not hungry?a 28. If I did not guide or regulate my child’s eating, s/he would eat too many junk foodsa

Pressure

Food as a Reward

0.79

0.88

0.73

0.83

0.37

–

0.40

0.34

–

0.46

0.50

0.97

0.96

0.94

0.91

0.94

0.95

0.89

0.90

0.80

0.86

0.72

0.84

0.74

0.80

0.74

0.71

0.50

0.48

0.53

0.44

0.60

0.80

0.60

0.75

0.62

0.79

0.60

0.78

0.45

0.47

0.45

0.29

0.44

0.50

0.47

0.48

0.49

0.56

0.51

0.32

J.J. Haszard et al. / Appetite 62 (2013) 110–118

14. Most of the food I keep in the house is healthy

4.4 (0.6) 4.0 (0.9) 4.7 (0.5) 4.6 (0.7) 4.4 (0.9) 3.9 (1.2) 4.8 (0.5) 4.1 (1.0) 4.4 (0.9) 4.3 (0.9) 4.6 (0.6)

Restriction 27. I encourage my child to eat less so he/she won’t get fat 29. I give my child small helpings at meals to control his/her weight 33. If my child eats more than usual at one meal, I try to restrict his/her eating at the next meal 34. I restrict the food my child eats that might make him/her fat 35. There are certain foods my child shouldn’t eat because they will make him/her fat 40. I have to be sure that my child does not eat too much of his/her favourite foods 41. I don’t allow my child to eat between meals because I don’t want him/ her to get fat 45. I often put my child on a diet to control his/her weight

Restriction for Weight Control Restriction for Weight Control Restriction for Weight Control Restriction for Weight Control Restriction for Weight Control Restriction for Health Restriction for Weight Control Restriction for Weight Control

Child Control Child Control

6. At dinner, do you let this child choose the foods s/he wants from what is being served? 10. If this child does not like what is being served, do you make something else? 11. Do you allow this child to eat snacks whenever s/he wants?

Child Control

12. Do you allow this child to leave the table when s/he is full, even if your family is not done eating?a Excluded items 7. When this child gets fussy, is giving him/her something to eat or drink the first thing you do? 13. Do you encourage this child to eat healthy foods before unhealthy ones? 15. I involve my child in planning family meals 16. I keep a lot of snack food in my house 17. My child should always eat all of the food on his/her plate 18. I have to be sure that my child does not eat too many high-fat foods 20. I allow my child to help prepare family meals 21. If I did not guide or regulate my child’s eating, s/he would eat too much of his/her favourite foods 32. I encourage my child to participate in grocery shopping 37. I keep a lot of sweets in my house 42. I tell my child what to eat and what not to eat without explanation 43. I have to be sure that my child does not eat too many sweets a

Child Control Child Control Child Control

a = 0.75

2.3 (0.6) 2.5 (0.8) 2.4 (1.0) 1.8 (0.8) 2.4 (0.9) 2.7 (1.1)

a = 0.63

0.63

0.75

0.61

0.69

0.66

0.78

0.61

0.77

0.64

0.81

0.68

0.64

0.53

0.66

0.54

0.45

0.51

0.66

0.51

0.46

0.43

0.48

0.39

–

0.51

0.66

0.47

0.57

0.45

0.75

0.47

0.46

0.49

0.74

0.52

0.65

0.59

0.45

0.58

0.44

0.50

0.53

0.49

0.44

0.42

0.69

0.50

0.58

0.33

–

0.41

0.25

Emotion Regulation Encourage Balance and Variety Involvement Environment Pressure Restriction for Weight Control Involvement Restriction for Health

J.J. Haszard et al. / Appetite 62 (2013) 110–118

5. Do you let your child eat whatever s/he wants?

1.9 (0.6) 1.6 (1.0) 1.5 (0.9) 1.3 (0.7) 2.8 (1.4) 2.5 (1.4) 3.0 (1.2) 1.5 (0.8) 1.1 (0.4)

Involvement Environment Teaching About Nutrition Restriction for Health

Items in italics were excluded from the final model but were included in the split-sample confirmatory analysis.

115

116

J.J. Haszard et al. / Appetite 62 (2013) 110–118

Table 4 Correlations in the five-factor model from the exploratory analysis. 1 1. 2. 3. 4. 5. * **

Healthy Eating Guidance Monitoring Parent Pressure Restriction Child Control

2 – 0.41* 0.01 0.01 0.15*

3 – 0.18* 0.05 0.37*

– 0.09** 0.12*

4

5

– 0.15*

–

p < 0.01. p < 0.05.

statistical analysis and may make the results more interpretable (Tabachnick & Fidell, 2001). Fewer items also reduces participant burden. Two of the five factors were very similar to factors in MusherEizenman and Holubs’ (2007) original model – Monitoring and Child Control. Monitoring is the practice of keeping track of what your child eats and comes almost directly from the CFQ (Birch et al., 2001). It has been associated with healthier outcomes for children (Arredondo et al., 2006; Faith et al., 2004; Klesges, Stein, Eck, Isbell, & Klesges, 1991). Child Control has been less explored and results are not all in agreement (Kroller & Warschburger, 2008; Rhee et al., 2009) but it is a practice of increasing interest as the ‘you provide, they decide’ model of feeding gains popularity, in which the parent is responsible for providing appropriate nutritious foods (you provide) but lets the children control what and how much they eat (they decide) (American Academy of Pediatrics Committee on Nutrition, 2003). Allowing children greater control over their eating may encourage better regulation of appetite and the body’s responses to food (American Academy of Pediatrics Committee on Nutrition, 2003; Birch & Fisher, 1997, 1998; Birch, Johnson, Andresen, Peters, & Schulte, 1991). Although a decreased ability to regulate intake may result in a higher BMI in children (Fisher & Birch, 2002), increased child control could result in unhealthy or over eating if the ability to regulate intake is already compromised or if the available food is not healthful (Hill & Peters, 1998). This analysis resulted in the loss of the Environment subscale meaning that now this questionnaire alone cannot specifically examine Child Control in relation to availability. However, the original Environment subscale was strongly correlated with other factors, and was not validated as a good measure of food environment. One Environment question is still represented in Healthy Eating Guidance. As more research is needed on Child Control, its place in this questionnaire is justified. The most studied feeding practice is that of Restriction. Studies have shown that Restriction has been associated with less healthy diet and higher weight in children (Costa et al., 2011; Faith et al., 2004; Fisher & Birch, 1999a, 1999b, 2002; Jansen, Mulkens, & Jansen, 2007; Spruijt-Metz et al., 2002, 2006; Van Strien, van Niekerk, & Ouwens, 2009; Webber et al., 2010). However, concern for overweight might mediate this relationship (Webber et al., 2010) and Restriction might be a useful practice in the treatment of childhood overweight (Campbell et al., 2010). Restriction is therefore an important practice to include in this questionnaire as there is a need to fully understand its impact.

Parent Pressure is a new construct that combines the well-studied practice of Pressure to Eat, from the CFQ, with the relatively new practice of Food as a Reward (whether parents offer or withhold treat food in response to good or bad behaviour) and one question from Emotion Regulation, together describing a situation where parents encourage eating in a controlling manner. Associations with less desirable eating patterns in children have previously been made (Campbell et al., 2006; Carnell & Wardle, 2007; Fisher et al., 2002; Galloway et al., 2006; Kroller & Warschburger, 2008, 2009; Powers et al., 2006). Further investigation of this feeding practice is required to establish its relationship with children’s diets as it captures an important aspect of feeding, that of overt parental control. Healthy Eating Guidance is a new measure consisting of the familiar concepts of Modelling, Teaching about Nutrition, Environment and Encouraging Balance and Variety, which were all strongly inter-correlated in the confirmatory analysis. Modelling and encouragement have been associated with healthier eating (Arredondo et al., 2006; Campbell et al., 2006; Fisher et al., 2002; Kroller & Warschburger, 2009; Vereecken et al., 2004) and this offers a positive feeding practice that could be used as a recommendation to parents. The two new factors emerging from this analysis, Parent Pressure and Healthy Eating Guidance, are supported by data from the original CFPQ analysis, which showed significant correlations between the subscales Pressure and Food as a Reward (now Parent Pressure), and Modelling, Encourage Balance and Variety and Teaching about Nutrition (now Healthy Eating Guidance) (Musher-Eizenman & Holub, 2007). As these correlations indicate considerable concordance between these subscales, which is replicated in our dataset, combining them as new factors in this analysis was satisfactory. The combination of factors has meant the loss of some subscales (for example Environment, Emotion Regulation, Encourage Balance and Variety, Food as a Reward) and with this the ability to detect more specifically which behaviours are most effectual. However, the fact that they are so strongly correlated in both datasets suggests that these behaviours do not occur in isolation and are highly related therefore the scales are more robust as a collection. Furthermore, factors that are excluded by an exploratory analysis indicate that they account for less of the variance and are less reliable. Extracting too many factors can result in a less replicable structure (Zwick & Velicer, 1986). As this sample was large, many statistically significant correlations between the five subscales were present, but were substantially fewer (and lower) than those observed with the 12-factor model, which had 28 correlations of greater than 0.2 (up to 0.81). The correlation in the new model between Monitoring and Healthy Eating Guidance comes as no surprise as parents who model, teach and encourage healthy eating are more likely keep track of their child’s diet. Furthermore those parents who keep better track of what their child eats would feasibly reduce the amount of control and freedom that the child has over what they eat, illustrated by the negative association between Monitoring and Child Control. Thus these correlations support the five-factor model, rather than undermine the structure.

Table 5 Correlations between the five factors from the exploratory analysis and measures that might influence feeding practices. N Age Concern for underweight Concern for overweight Importance of a healthy diet *

p < 0.05.

1013 159 118 1012

Healthy Eating Guidance 0.04 0.17* 0.28* 0.35*

Monitoring 0.05 0.08 0.08 0.22*

Parent Pressure 0.18* 0.35* 0.24* 0.02

Restriction 0.03 0.02 0.27* 0.02

Child Control 0.02 0.04 0.11 0.07*

J.J. Haszard et al. / Appetite 62 (2013) 110–118

Additional analyses were undertaken by dividing the sample into two groups, running an exploratory analysis in one group, followed by a confirmatory analysis in the other. Both groups had at least 10 participants per variable. The exploratory analysis led to factors similar to those proposed in the full sample analysis. Most of the small number of extra items suggested for the factors loaded weakly in the subsequent confirmatory analysis. However, the majority of the loadings for the items that were included in the full sample factors loaded strongly in this confirmatory analysis. The confirmatory analysis in Group 2 showed overall poor model fit. It is generally agreed that larger sample sizes are better for factor analysis. Indeed it has been shown that a 20:1 ratio of participants to questions more accurately represents the data when conducting an exploratory analysis compared to lower ratios (Costello & Osborne, 2005). The full sample used here exhibited a ratio of nearly 21:1, suggesting that the full sample analysis gave the most accurate structure. As the five-factor structure for the whole sample was, for the most part, supported by the split sample analyses, these results suggest a relatively robust model with potential to be applicable in other samples. Both the CFPQ and the CFQ original analyses reported correlations between feeding practices and concern for child weight. Results concur with this study, that concern for overweight is positively associated with restrictive feeding practices and that concern for underweight is positively associated with pressuring feeding practices (Birch et al., 2001; Musher-Eizenman & Holub, 2007). Furthermore, in this study concern for overweight was positively associated with Healthy Eating Guidance and negatively associated with Parent Pressure – outcomes which seem logical and attest to the subscales. Although our results demonstrated that those parents concerned that their child was underweight exhibited less Healthy Eating Guidance, which might seem counter-intuitive, it is perhaps feasible given that parents worried about their child’s relative thinness may want them to eat anything, regardless of nutritional content. Moreover, Musher-Eizenman and Holub (2007) showed that concern for underweight had negative correlations with Environment and Modelling (two factors that were partially captured by the Healthy Eating Guidance factor), supporting this result. Face validity was further evaluated with the importance of a healthy diet scale and age of the child, which again gave logical associations and endorsed the five-factor model. A major strength of this analysis was the large, diverse sample (n = 1013) resulting in nearly 21 participants per variable. Sample size recommendations for factor analyses vary but at least 5–10 participants per variable is advised, with 20 participants per variable suggested as optimal (Costello & Osborne, 2005; Floyd & Widaman, 1995). Furthermore, as feeding practices are influenced by maternal education (Kroller & Warschburger, 2008; Vereecken et al., 2004), socio-economic status (Hupkens et al., 1998; OrrellValente et al., 2007), ethnicity (Spruijt-Metz et al., 2002, 2006) and maternal weight status (Powers et al., 2006; Robinson et al., 2001; Wardle et al., 2002) these results have greater applicability because of the sample diversity (exhibited in Table 1). Reducing the questionnaire, from 49 items to 32, by removing 17 items that did not contribute to any one feeding practice, has the potential to reduce the response burden. The data were represented by five feeding practices of interest, which need further investigation to determine how they might influence child preferences, eating patterns, dietary intake and weight status. Unlike the original analysis on the 12-factor model, these five practices were not pre-specified and were allowed to emerge from the data, supported by face validation, correlation results and split sample analyses. Our study was limited by its lack of information about the person who completed the questionnaire and who did the feeding in the household. While anecdotally it was mostly mothers who

117

participated, the large sample would have had some fathers who completed the questionnaire, adding to the diversity. Although the sample was diverse, there may also be country-specific issues that affect the applicability of the final CFPQ. The new 32-item CFPQ should be tested in different samples to determine if the subscales are robust in other populations. While the RMSEA test indicated that the new five-factor model was not a good fit, both the Tucker–Lewis and Comparative Fit indices showed that it was a good fit, which was not the case for the 12-factor model. In addition, considerable improvement in modification indices and correlations also supported the new model. It is noted that the mean scores for Healthy Eating Guidance and Monitoring were high, exhibiting ceiling effects, which may be indicative of participants answering in a socially desirable way. Additionally, given the design of the CFPQ other potential factors of interest were omitted, such as food as a reward for food (‘‘If you eat your peas, you can have dessert’’), which has been shown to increase a preference for the reward food, and repeated exposure of new foods, which increases acceptance and liking of those foods (Wardle & Cooke, 2008). In conclusion, with modifications suggested by this analysis, the Comprehensive Feeding Practices Questionnaire has been enhanced, resulting in a tool that could be used in many settings and contribute to the understanding of a child’s eating environment.

References American Academy of Pediatrics Committee on Nutrition (2003). Prevention of pediatric overweight and obesity. Pediatrics, 112, 424–430. Arredondo, E. M., Elder, J. P., Ayala, G. X., Campbell, N., Baquero, B., & Duerksen, S. (2006). Is parenting style related to children’s healthy eating and physical activity in Latino families? Health Education Research, 21, 862–871. Birch, L. L., & Fisher, J. O. (1997). Food intake regulation in children. Annals of the New York Academy of Sciences, 819, 194–220. Birch, L. L., & Fisher, J. O. (1998). Development of eating behaviors among children and adolescents. Pediatrics, 101, 539–549. Birch, L. L., Fisher, J. O., Grimm-Thomas, K., Markey, C. N., Sawyer, R., & Johnson, S. L. (2001). Confirmatory factor analysis of the Child Feeding Questionnaire. A measure of parental attitudes, beliefs and practices about child feeding and obesity proneness. Appetite, 36, 201–210. Birch, L. L., Johnson, S. L., Andresen, G., Peters, J. C., & Schulte, M. C. (1991). The variability of young children’s energy intake. New England Journal of Medicine, 324, 232–235. Blissett, J., Haycraft, E., & Farrow, C. (2010). Inducing preschool children’s emotional eating. Relations with parental feeding practices. American Journal of Clinical Nutrition, 92, 359–365. Campbell, K., Andrianopoulos, N., Hesketh, K., Ball, K., Crawford, D., Brennan, L., et al. (2010). Parental use of restrictive feeding practices and child BMI z-score. A 3year prospective cohort study. Appetite, 55, 84–88. Campbell, K., Crawford, D., & Ball, K. (2006). Family food environment and dietary behaviors likely to promote fatness in 5–6 year-old children. International Journal of Obesity, 30, 1272–1280. Carnell, S., & Wardle, J. (2007). Associations between multiple measures of parental feeding and children’s adiposity in United Kingdom preschoolers. Obesity, 15, 137–144. Costa, F. S., Del Pino, D. L., & Friedman, R. (2011). Caregivers’ attitudes and practices. Influence on childhood body weight. Journal of Biosocial Science, 43, 369–378. Costanzo, P. R., & Woody, E. Z. (1984). Parental perspectives on obesity in children. The importance of sex differences. Journal of Social and Clinical Psychology, 2, 305–313. Costanzo, P. R., & Woody, E. Z. (1985). Domain-specific parenting styles and their impact on the child’s development of particular deviance. The example of obesity proneness. Journal of Social and Clinical Psychology, 3, 425–445. Costello, A., & Osborne, J. (2005). Best practices in exploratory factor analysis. Four recommendations for getting the most from your analysis. Practical Assessment, Research and Evaluation, 10(7). Faith, M. S., Berkowitz, R. I., Stallings, V. A., Kerns, J., Storey, M., & Stunkard, A. J. (2004). Parental feeding attitudes and styles and child body mass index. Prospective analysis of a gene-environment interaction. Pediatrics, 114, e429–e436. Fisher, J. O., & Birch, L. L. (1999a). Restricting access to foods and children’s eating. Appetite, 32, 405–419. Fisher, J. O., & Birch, L. L. (1999b). Restricting access to palatable foods affects children’s behavioral response, food selection, and intake. American Journal of Clinical Nutrition, 69, 1264–1272. Fisher, J. O., & Birch, L. L. (2002). Eating in the absence of hunger and overweight in girls from 5 to 7 y of age. American Journal of Clinical Nutrition, 76, 226–231.

118

J.J. Haszard et al. / Appetite 62 (2013) 110–118

Fisher, J. O., Mitchell, D. C., Smiciklas-Wright, H., & Birch, L. L. (2002). Parental influences on young girls’ fruit and vegetable, micronutrient, and fat intakes. Journal of the American Dietetic Association, 102, 58–64. Floyd, J. F., & Widaman, K. F. (1995). Factor analysis in the development and refinement of clinical assessment instruments. Psychological Assessment, 7, 286–299. Galloway, A. T., Fiorito, L. M., Francis, L. A., & Birch, L. L. (2006). ‘Finish your soup’. Counterproductive effects of pressuring children to eat on intake and affect. Appetite, 46, 318–323. Hill, J. O., & Peters, J. C. (1998). Environmental contributions to the obesity epidemic. Science, 280, 1371–1374. Hupkens, C. L. H., Knibbe, R. A., van Otterloo, A. H., & Drop, M. J. (1998). Class differences in the food rules mothers impose on their children. A cross-national study. Social Science and Medicine, 47, 1331–1339. Jansen, E., Mulkens, S., & Jansen, A. (2007). Do not eat the red food! Prohibition of snacks leads to their relatively higher consumption in children. Appetite, 49, 572–577. Johannsen, D. L., Johannsen, N. M., & Specker, B. L. (2006). Influence of parents’ eating behaviors and child feeding practices on children’s weight status. Obesity, 14, 431–439. Klesges, R., Stein, R., Eck, L., Isbell, T., & Klesges, L. (1991). Parental influence on food selection in young children and its relationships to childhood obesity. American Journal of Clinical Nutrition, 53, 859–864. Kroller, K., & Warschburger, P. (2008). Associations between maternal feeding style and food intake of children with a higher risk for overweight. Appetite, 51, 166–172. Kroller, K., & Warschburger, P. (2009). Maternal feeding strategies and child’s food intake. Considering weight and demographic influences using structural equation modeling. International Journal of Behavioral Nutrition and Physical Activity, 6, 78–86. Marsh, H., Balla, J., & Hau, K. (1996). An evaluation of incremental fit indices. A clarification of mathematical and empirical properties. In G. Marcoulides & R. Schumaker (Eds.), Advanced structural equation modeling techniques (pp. 315–353). Mahwah, NJ: Lawrence Erlbaum Associates Inc.. Musher-Eizenman, D., & Holub, S. (2007). Comprehensive feeding practices questionnaire. Validation of a new measure of parental feeding practices. Journal of Pediatric Psychology, 32, 960–972. Orrell-Valente, J. K., Hill, L. G., Brechwald, W. A., Dodge, K. A., Pettit, G. S., & Bates, J. E. (2007). ‘‘Just three more bites’’. An observational analysis of parents’ socialization of children’s eating at mealtime. Appetite, 48, 37–45. Powers, S. W., Chamberlin, L. A., van Schaick, K. B., Sherman, S. N., & Whitaker, R. C. (2006). Maternal feeding strategies, child eating behaviors, and child BMI in low-income African–American preschoolers. Obesity, 14, 2026–2033.

Rhee, K. E., Coleman, S. M., Appugliese, D. P., Kaciroti, N. A., Corwyn, R. F., Davidson, N. S., et al. (2009). Maternal feeding practices become more controlling after and not before excessive rates of weight gain. Obesity, 17, 1724–1729. Robinson, T. N., Kiernan, M., Matheson, D. M., & Haydel, K. F. (2001). Is parental control over children’s eating associated with childhood obesity? Results from a population-based sample of third graders. Obesity Research, 9, 306–312. Spruijt-Metz, D., Li, C. Y., Cohen, E., Birch, L., & Goran, M. (2006). Longitudinal influence of mother’s child-feeding practices on adiposity in children. Journal of Pediatrics, 148, 314–320. Spruijt-Metz, D., Lindquist, C. H., Birch, L. L., Fisher, J. O., & Goran, M. I. (2002). Relation between mothers’ child-feeding practices and children’s adiposity. American Journal of Clinical Nutrition, 75, 581–586. Stang, J., & Loth, K. A. (2011). Parenting style and child feeding practices. Potential mitigating factors in the etiology of childhood obesity. Journal of the American Dietetic Association, 111, 1301–1305. Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics. Boston: Allyn and Bacon. Taylor, R., Brown, D., Dawson, A., Haszard, J., Cox, A., Rose, E., et al. (2010). Motivational interviewing for screening and feedback and encouraging lifestyle changes to reduce relative weight in 4–8 year old children. Design of the MInT study. BMC Public Health, 10, 271. Van Strien, T., van Niekerk, R., & Ouwens, M. A. (2009). Perceived parental food controlling practices are related to obesogenic or leptogenic child life style behaviors. Appetite, 53, 151–154. Vereecken, C. A., Keukelier, E., & Maes, L. (2004). Influence of mother’s educational level on food parenting practices and food habits of young children. Appetite, 43, 93–103. Wardle, J., Carnell, S., & Cooke, L. (2005). Parental control over feeding and children’s fruit and vegetable intake. How are they related? Journal of the American Dietetic Association, 105, 227–232. Wardle, J., & Cooke, L. (2008). Genetic and environmental determinants of children’s food preferences. British Journal of Nutrition, 99, S15–S21. Wardle, J., Sanderson, S., Guthrie, C. A., Rapoport, L., & Plomin, R. (2002). Parental feeding style and the inter-generational transmission of obesity risk. Obesity, 10, 453–462. Webber, L., Hill, C., Cooke, L., Carnell, S., & Wardle, J. (2010). Associations between child weight and maternal feeding styles are mediated by maternal perceptions and concerns. European Journal of Clinical Nutrition, 64, 259–265. White, P., Gunston, J., Salmond, C., Atkinson, J., & Crampton, P. (2008). Atlas of socioeconomic deprivation in New Zealand NZDep2006. Wellington, New Zealand: Ministry of Health. Zwick, W. R., & Velicer, W. F. (1986). Comparison of five rules for determining the number of components to retain. Psychological Bulletin, 99(3).