Available online at www.sciencedirect.com
Eating Behaviors 8 (2007) 457 – 463
About Your Child's Eating: Factor structure and psychometric properties of a feeding relationship measure W. Hobart Davies a,b,⁎, Lynn K. Ackerman c , Cheryl M. Davies d , Kathryn Vannatta e , Robert B. Noll f a
Department of Psychology, University of Wisconsin — Milwaukee, PO Box 413, Milwaukee, WI 53201-0413, United States b Children’s Hospital of Wisconsin, PO Box 1997, Milwaukee, WI 53201-1997, United States c Ackerman Research, 1418 S. Birch Drive, Mt. Prospect, IL 60056, United States d WIC Program, MLK Heritage Health Center, 2555 N. Dr. Martin Luther King Dr., Milwaukee, WI 53212, United States e Center for Biobehavioral Health, Columbus Children’s Research Institute and the Ohio State University, 700 Children’s Drive, Room G362, Columbus, OH 43205, United States f Children’s Hospital of Pittsburgh, University of Pittsburgh Medical Center, 3705 Fifth Avenue, Pittsburgh, PA 15213, United States Received 28 July 2006; received in revised form 13 December 2006; accepted 15 January 2007
Abstract Conducted exploratory and confirmatory factor analysis of the About Your Child’s Eating (AYCE) inventory with 763 parents. Parent subjects were drawn from a community study of families with physically healthy and chronically ill children between 8 and 16 years of age. Three correlated factors were identified: Child Resistance to Eating, Positive Mealtime Environment, and Parent Aversion to Mealtime. The internal consistency of the factors remained satisfactory across all examined demographic subgroups. Evidence for convergent validity was found by comparing the AYCE factors to higher order factors of the Family Environment Scale. Empirically derived clinical range cut-off scores are presented. Results support the AYCE as a psychometrically sound measure of the parent–child feeding relationship for school-aged children. © 2007 Published by Elsevier Ltd. Keywords: Mealtime interactions; Measurement; Children; Parents; Reliability; Validity
Concerns about eating and weight are a common reason for referral to professionals working with children. The complexity of interactions across the cellular, individual, and family systems make these cases difficult to evaluate (Davies et al., 2006). The complexity of interactions has been demonstrated in families of children with failure-tothrive (Kessler & Dawson, 1999), cystic fibrosis (Stark & Powers, 2005), cancer (Gerhardt et al., 2006), and physically healthy children (Stanek, Abbott, & Cramer, 1990).
⁎ Corresponding author. Department of Psychology, University of Wisconsin—Milwaukee, PO Box 413, Milwaukee, WI 53201-0413, United States. Tel.: +1 414 229 6594; fax: +1 414 229 5219. E-mail address:
[email protected] (W.H. Davies). 1471-0153/$ - see front matter © 2007 Published by Elsevier Ltd. doi:10.1016/j.eatbeh.2007.01.001
458
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
These cases often benefit from the expertise available in a multidisciplinary evaluation. However, professionals have to make decisions about when such an intensive and expensive step is needed. A psychometrically robust scale to document behavioral and interactional difficulties in the feeding relationship would be valuable. Satter (1999) defines the feeding relationship as, “the complex of interactions that take place between parent (or primary caregiver) and child as they engage in food selection, ingestion, and regulation.” (p. 122). Disruptions in the feeding relationship are associated with a range of behavioral and emotional difficulties for the relationship between parent and child and/or for each of them individually. We have few psychometrically stable instruments dealing with mealtime behaviors, and most of the available instruments are focused explicitly on the child’s behavior. The parent-report questionnaire “About Your Child’s Eating” (AYCE; Davies, Noll, Davies, & Bukowski, 1993) included questions asking about the frequency of 15 positive, negative, and neutral parent–child interactions at mealtimes. Scales of Positive and Negative Interactions were constructed which yielded adequate internal consistency. The present study presents improvements to the AYCE and more intensive exploration of its psychometric properties. The number of interaction items was increased from 15 to 31, and the items were subjected to exploratory and confirmatory factor analyses on separate samples to develop a more refined factor structure. The internal consistency of the new factors is evaluated across a range of child and family variables. 1. Method 1.1. Participants The sample included 763 mothers and co-parents of children (ages 8–16) with chronic illness and physically healthy children taking part in the University of Cincinnati Friendship Study (Noll et al., 1999). The selection procedure consisted of (1) selection of children with chronic illness from rosters at their treatment center; and (2) a matched-group classroom nomination procedure. Children with chronic illness and their parents who agreed to participate gave us permission to contact the child’s school where physically healthy children matched on gender, race, and age and their parents were asked to participate. In the resulting sample, 49.6% of the children had a chronic illness; 51.4% were male; 64% were white, with the remaining 36% black. The mean child age was 11.3 years (SD = 1.99 years). Parents of children with chronic illness did not differ significantly from parents with physically healthy children in age, race, education, and socioeconomic status. 1.2. Measures 1.2.1. About Your Child’s Eating (AYCE) The AYCE consists of 31 Likert scale items rated from 1 (never) to 5 (nearly every time) asking parents about the frequency of child eating behaviors, their mealtime interactions with the child, and their feelings about mealtime. Items were developed by querying dietitians and psychologists familiar with Feeding Relationship theory (Satter, 1999) about common scenarios. Previous research employing the AYCE utilized a factor structure that grouped the negative and positive items into subscales. This factor structure has not been verified through factor analysis. 1.2.2. Family Environment Scale (FES) The Family Environment Scale (Moos & Moos, 1981) is a 90-item true/false questionnaire tapping various family characteristics. The measure contains nine subscales that were empirically combined into three higher order factors: Supportive, Conflicted, and Controlling (Kronenberger & Thompson, 1990). The Supportive factor is scored by summing the Cohesion, Expressiveness, Independence, Active–Recreational Orientation, and Intellectual–Cultural Orientation scales. For the Conflicted factor, the Cohesion and Organization scales are subtracted from the Conflict scale. The Controlling factor is the sum of the Control, Achievement Orientation, and Moral–Religious Emphasis scales minus the Independence scale. A fourth FES construct is the Family Relationships Index (FRI), conceptually derived by Moos and Moos (1981) as the sum of the Cohesion and Expressiveness scales minus the Conflict scale. Studies have found the FES to be a valid instrument for examining family issues, including the family mealtime environment (Davies et al., 1993; Stanek et al., 1990).
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
459
1.3. Procedure As part of the Cincinnati Friendship Study (Noll et al., 1999), parents completed several individual and family measures. Of interest to this study were the demographic survey, the AYCE, and the FES. The sample was split randomly into two groups, half for the exploratory factor analysis and the other half for the confirmatory factor analysis. 2. Results 2.1. Exploratory factor analysis The raw data for the 31 items were submitted to an exploratory factor analysis with maximum likelihood extraction and promax rotation. The following criteria for retaining factors were established a priori (Cattell, 1966; Gorsuch, 1983): (a) eigenvalues greater than 1.0, (b) a scree test, (c) interpretability of the rotated factor solution, and (d) item loadings of .45 or greater. Of the original 31 items, 20 were retained, creating three correlated factors accounting for 42% of the total variance. Eleven items loaded substantially on the Child Resistance to Eating factor (25% of variance), five items on the Positive Mealtime Environment factor (10% of variance), and four items on the Parent Aversion to Mealtime factor (6% of variance). Cronbach alpha values for the Child Resistance to Eating, Positive Mealtime Environment, and Parent Aversion to Mealtime factor scales were .89, .80, and .72, respectively. 2.2. Confirmatory factor analysis Two competing models were evaluated for goodness of fit using confirmatory factor analysis. For model one, we created a two-factor structure by forcing the 15 items of the Child Resistance to Eating and Parent Aversion to Eating Table 1 Full sample descriptives and internal consistency for AYCE factors and items (N = 763) Item
M
SD
Item-total correlation
Factor 1: Child Resistance to Eating 19.90 1 1.71 2 1.94 4 1.66 5 1.59 6 2.10 7 2.57 9 1.95 16 1.66 17 1.67 19 1.64 20 1.39
6.60 .87 .99 .91 .79 1.10 1.14 .87 .76 .81 .78 .69
.61 .50 .77 .67 .53 .65 .51 .53 .59 .54 .66
Factor 2: Positive Mealtime Environment 18.55 3 3.48 8 3.66 10 3.83 11 3.97 13 3.62
3.37 .87 .92 .88 .93 .91
.56 .65 .73 .38 .62
Factor 3: Parent Aversion to Mealtime 7.41 12 1.71 14 1.58 15 1.25 16 1.66 18 1.20
2.32 .86 .73 .56 .76 .48
.44 .60 .49 .42 .40
Cronbach alpha .88
.80
.70
460
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
factors to load onto a single negative factor and the five items of the Positive Mealtime Environment factor onto a single positive factor. If adequate, this model would confirm that the AYCE is comprised simply of negative and positive dimensions. For model two, items were forced onto the three-factor structure identified in the exploratory analysis. A good fit to the data would confirm this three-factor representation of the AYCE. For both models, the factors had variances set at 1.0 and were allowed to correlate. The normalized estimate of multivariate kurtosis was 40.78, which indicates positive kurtosis and is highly suggestive of nonnormality in the population. We used the robust maximum likelihood (ML Robust) estimation procedure in EQS for Windows 5.1 to test each model. This estimation procedure is robust to violation of normality. Because the traditional χ2 statistic is sensitive to sample size, the overall fit of each model was evaluated using five preestablished criteria: (a) a normed fit index (NFI) of .90 or higher, (b) a nonnormed fit index (NNFI) of .80 or higher, (c) a robust comparative fit index (R-CFI) of .90 or higher, (d) a root mean square residual (RMSR) of .05 or lower, and
Table 2 Internal consistency of the AYCE factors for various demographic subgroups Demographic subgroup
Race Caucasian African-American Sex of child Male Female Age of childa Seven years Eight years Nine years Ten years Eleven years Twelve years Thirteen years Fourteen years Fifteen years Child chronic illness Yes No Parental status Mother Co-parent Parent education High school degree Some college College degree Family SESb Lower 50% Upper 50% Age of parentc Lower 50% Upper 50% Family size 1 child 2 children >2 children Full sample a
Cronbach alpha CRE
PAM
PME
.88 .87
.82 .75
.69 .70
.88 .87
.81 .79
.70 .69
.91 .87 .89 .83 .89 .90 .90 .83 .88
.70 .78 .78 .78 .85 .74 .85 .75 .85
.68 .58 .77 .65 .78 .57 .68 .67 .64
.90 .83
.79 .81
.75 .62
.87 .87
.79 .81
.64 .78
.86 .89 .89
.78 .80 .84
.66 .67 .78
.86 .90
.82 .81
.60 .76
.89 .87
.79 .82
.66 .74
.79 .90 .88 .88
.85 .81 .78 .80
.65 .76 .65 .70
Median = 37.2 on Duncan TSEI2 index of occupational status; bmedian = 38.83 years. CRE = Child Resistance to Eating, PAM = Parent Aversion to Mealtime, PME = Positive Mealtime Environment.
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
461
(e) a Satorra–Bentler scaled statistic (SB-χ2/df) ratio of 2.0 or lower (Bentler, 1990; Byrne, 1994; Satorra & Bentler, 1988). The results indicated that the two-factor oblique model demonstrated a poor fit to the data as NFI = .75, NNFI = .77, R-CFI = .80, RMSR = .08, and SB-χ2/df = 3.15. The three-factor oblique model provided a good fit to the data as NFI = .87, NNFI = .91, R-CFI = .93, RMSR = .04, and SB-χ2/df = 1.75. The drop in the Satorra–Bentler χ2 from the two-factor model to the three-factor model was statistically significant (χ2 (7) = 248.81, p < .001) suggesting an improved fit to the data. The alpha reliability indices for the three factor scales in the confirmatory sample were .86 for Child Resistance to Eating, .80 for Positive Mealtime Environment, and .66 for Parent Aversion to Mealtime. 2.3. Combined sample statistics Table 1 provides a descriptive summary and reliability analysis of the three-factor AYCE for the combined samples (N = 763). Table 2 displays reliability coefficients for various demographic subgroups. The internal consistency of the three AYCE factors were satisfactory across all subgroups. Correlations among the factors were significant at r = − .24 for Child Resistance to Eating and Positive Mealtime Environment, r = .55 for Child Resistance to Eating and Parent Aversion to Mealtime, and r = − .37 for Positive Mealtime Environment and Parent Aversion to Mealtime. 2.4. Convergent validity Pearson correlations were computed to evaluate relationships between the AYCE factors and the Supportive and Conflicted factors, and Family Relationship Index of the Family Environment Scale (see Table 3). As expected, the AYCE factors correlated significantly and in expected directions with the FES factors. 2.5. Preliminary clinical cutoffs We elected to use a conservative approach of labeling a two standard deviation elevation as a clinical cut-off. We found that 7% of parents with chronically ill children scored above the clinical cutoff of 33 for Child Resistance to Eating versus 1.3% of parents with healthy children. For Parent Aversion to Mealtime, the chronic illness parents again had a greater percentage of scores (4.8%) above the two standard deviation cutoff of 12 than the healthy group (1.0%). For Positive Mealtime Environment, the pattern was reversed with 4.4% of parents with healthy children versus 2.4% of parents with chronically ill children scoring below the two standard deviation cutoff of 12. 2.6. Final version of instrument A final version of the instrument is shown in the Appendix. The 20 items that loaded on factors are included. The items load on the factors as shown in Table 1. Five additional items from the 31-item AYCE were maintained because they were judged to be of independent clinical interest.
Table 3 Correlations between AYCE factors and family environment factorsa
Child Resistance to Eating Positive Mealtime Environment Parent Aversion to Mealtime
FESb
FESb
Supportive
Conflicted
− .09⁎ .34⁎⁎ − .21⁎⁎
.20⁎⁎ − .39⁎⁎ .40⁎⁎
FRIc − .14⁎⁎ .37⁎⁎ − .34⁎⁎
⁎p < .05; ⁎⁎p < .001; N = 763; ap-values corrected for multiple comparisons using Holm's procedure; bFamily Environment Scale; cFamily Relationships Index Holm, 1979.
462
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
3. Discussion We found that the AYCE consists of three robust factors: Child Resistance to Eating, Positive Mealtime Environment, and Parent Aversion to Mealtime. The AYCE factors had good internal consistency across race, sex of child, age of child, child health status, parent status, parent education, family SES, age of parent, and family size. Evidence of convergent validity was provided by the significant correlation of AYCE factors with measures of family functioning. Data are drawn from a community-based sample of children with chronic illnesses and physically healthy children drawn from the same classrooms. Participation rates exceeded 90% for both groups. This eliminates concerns about generalizability due to volunteer samples or other biases in subject selection. This is a strength of the study. Research is needed to support the utility of the AYCE. Replication of our findings in clinical samples of children with feeding disorders and other age groups would be useful, as would comparison of AYCE factor scores to actual observed mealtime interactions. Additional psychometric analyses such as test–retest reliability and other evidence of validity are also needed. These results provide evidence that the AYCE is a robust measure of important aspects of the feeding relationship for school-aged children. Acknowledgement This project was supported in part by grants to R. B. Noll from the Arthritis Foundation and the American Cancer Society, Ohio Branch. Appendix A. About Your Child’s Eating Child’s Name: _________________ Date: _________ Age: ______ Filled Out By: _________________ A variety of situations take place in families around children’s eating. Please indicate how often each of the following occur between you and your child or in your family 1 Never
2 Once in a while
3 Sometimes
1. My child hates eating. 1 2 3 2. I feel like a short-order cook because I have to make special meals for my child. 1 2 3 3. Meal times are among the most pleasant in the day. 1 2 3 4. I feel that it is a struggle or fight to get my child to eat. 1 2 3 5. My child refuses to eat. 1 2 3 6. I worry that my child will not eat right unless closely supervised. 1 2 3 7. My child is a picky eater. 1 2 3 8. The family looks forward to meals together. 1 2 3 9. My child enjoys eating. 1 2 3 10. Mealtime is a pleasant, family time. 1 2 3 11. I get pleasure from watching my child eating well and enjoying his/her food. 1 2 3 12. I dread meal times. 1 2 3
4 Often
5 Nearly every time
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
W.H. Davies et al. / Eating Behaviors 8 (2007) 457–463
463
13. We have nice conversations during meals. 1 2 3 4 5 14. Meal times are the pits. 1 2 3 4 5 15. It is hard for me to eat dinner with my child because of how he/she behaves. 1 2 3 4 5 16. There are arguments between me and my child over eating. 1 2 3 4 5 17. My child seems to have no appetite. 1 2 3 4 5 18. My child has mealtime tantrums. 1 2 3 4 5 19. My child refuses to eat a planned meal. 1 2 3 4 5 20. I have to force my child to eat. 1 2 3 4 5 21. I use preferred foods (such as dessert) as rewards or bribes to get my child to eat “good” foods. 1 2 3 4 5 22. We watch television during meals. 1 2 3 4 5 23. There are house rules about how much kids have to eat (for example, the “Clean Plate Club”; No dessert until you eat what’s on your plate). 1 2 3 4 5 24. I have thought about putting my child on a diet. 1 2 3 4 5 25. We end up grabbing meals whenever we can with no time for planning. 1 2 3 4 5
References Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107, 238−246. Byrne, B. M. (1994). Structural equation modeling with EQS and EQS/Windows. Thousand Oaks, CA: Sage. Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 245−276. Davies, C. M., Noll, R. B., Davies, W. H., & Bukowski, W. M. (1993). Mealtime interactions and family relationships for families with children who have cancer in long-term remission and controls. Journal of the American Dietetic Association, 93, 773−776. Davies, W. H., Satter, E., Berlin, K. S., Sato, A. F., Silverman, A. H., Fischer, E. A., et al. (2006). Reconceptualizing feeding and feeding disorders in interpersonal context: The case for a relational disorder. Journal of Family Psychology, 20, 409−417. Gerhardt, C. A., Baughcom, A. E., Johnston, A., Noll, R. B., Davies, W. H., Mackner, L. N., et al. (2006). Parent perceptions of nutritional issues during their child’s treatment for cancer. Journal of Pediatric Hematology/Oncology, 28, 454−460. Gorsuch, R. (1983). Factor analysis, 2nd ed. Hillsdale, NJ: Erlbaum. Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6, 65−70. Kessler, D. B., & Dawson, P. (Eds.). (1999). Failure to thrive and pediatric undernutrition: A transdisciplinary approach. Baltimore, MD: Brookes. Kronenberger, W. G., & Thompson, R. J. (1990). Dimensions of family functioning in families with chronically ill children: A higher order factor analysis of the Family Environment Scale. Journal of Clinical Child Psychology, 19, 380−388. Moos, R. H., & Moos, B. S. (1981). Family environment scale manual. Palo Alto, CA: Consulting Psychologists Press. Noll, R. B., Gartstein, M. A., Vannatta, K., Correll, J., Bukowski, W. M., & Davies, W. H. (1999). Social, emotional, and behavioral functioning of children with cancer. Pediatrics, 103, 71−78. Satorra, A., & Bentler, P. M. (1988). Scaling corrections for chi-square statistics in covariance structure analysis. Proceedings of the American Statistical Association (pp. 308−313). Satter, E. (1999). The feeding relationship. In D. B. Kessler, & P. Dawson (Eds.), Failure to thrive and pediatric undernutrition: A transdisciplinary approach (pp. 121−149). Baltimore, MD: Brookes. Stanek, K., Abbott, D., & Cramer, S. (1990). Diet quality and the eating environment of preschool children. Journal of the American Dietetic Association, 90, 1582−1584. Stark, L. J., & Powers, S. W. (2005). Behavioral aspects of nutrition in children with cystic fibrosis. Current Opinion in Pulmonary Medicine, 11, 539−542.