Exploratory factor analysis of The Comprehensive Feeding Practices Questionnaire (CFPQ) in a low-income hispanic sample of preschool aged children

Abstract

The Comprehensive Feeding Practices Questionnaire (CFPQ) is an important measure to assess parent feeding practices as it encompasses a broad range of feeding behaviors, not just behaviors negatively associated with child weight outcomes. However, parent feeding practices have been shown to differ across ethnicities and the CFPQ has not been tested among low-income, Hispanic-American parents with preschool-aged children, a group at elevated risk for developing obesity. A confirmatory factor analysis was performed with the present Hispanic-American sample of Head Start mothers with preschoolers to confirm the original 12-factor, 49-item structure of the CFPQ. Because the original factor structure was not confirmed in the present Hispanic-American sample, an exploratory factor analysis was conducted to examine the psychometric properties of the CFPQ in this sample (n = 187). Among this sample, a five-factor model with 34 items was found to more appropriately assess parent feeding practices than the original 12-factor, 49-item CFPQ. This study provides preliminary validation of the CFPQ for use among low income, Hispanic-American families. Although future research is needed to replicate findings among a larger sample, this study takes an important first step toward more accurately assessing parent feeding practices among this high-risk population to inform tailored interventions that aim to reduce economic and ethnic disparities in child obesity.

1. Introduction

Hispanic-American children, two to five years of age, have approximately three times the rate of obesity compared to their Caucasian peers (Ogden et al., 2016), and compared to other ethnicities, the inequality in the prevalence of childhood obesity by household income is greater among Hispanic-American families (Ogden et al., 2018). Factors contributing to ethnic and income disparities in the development of childhood overweight and obesity are poorly understood, which limits the development of effective interventions for the prevention and treatment of obesity in high-risk children. Due to parents’ direct influence on child eating behaviors and, consequently, weight status (Spruijt-Metz, Lindquist, Birch, Fisher, & Goran, 2002), the study of parent feeding practices is a growing area of research. Parent feeding practices are specific goal-oriented behaviors or directives that parents use to influence what, when, and how much children eat (Blissett, 2011). Given associations between feeding practices and child weight outcomes, parent feeding practices may be an important target for successful child obesity interventions. For example, restrictive and controlling feeding practices are generally associated with higher weight, and pressure to eat has traditionally been associated with lower child weight status (Shloim, Edelson, Martin, & Hetherington, 2015).

Relatively few studies have specifically focused on the feeding practices of low-income Hispanic-American parents (Anderson, Hughes, Fisher, & Nicklas, 2005; Cardel et al., 2012; Hughes et al., 2006; Tschann et al., 2013; Tschann et al., 2015). However, feeding practices of parents in this population do not meet recommendations for responsive feeding practices (Power, Hughes, et al., 2015). Given the high risk for childhood obesity among this population, measures that accurately assess parent feeding practices among low-income, Hispanic parents are needed. The Comprehensive Child Feeding Practices Questionnaire (CFPQ) (Musher-Eizenman & Holub, 2007) is a broad measure of feeding practices that extends beyond controlling feeding practices (Vaughn, Tabak, Bryant, & Ward, 2013), to include scales associated with the promotion of child health, such as modeling of healthy eating and making healthy foods available in the home (Kroller & Warschburger, 2009). The CFPQ has been highlighted as a well-developed instrument due to the researchers’ systematic approach to creating and refining the item pool, and the use of theory to conceptualize constructs (Vaughn et al., 2013). However, like almost all parent feeding measures, the CFPQ was created with predominantly Caucasian, middle-to upper-income families, which limits its generalizability and use across other populations. Items and response options on parent feeding questionnaires may be interpreted differently across varying income levels and ethnic groups (Anderson et al., 2005; Lindsay et al., 2012). Parent feeding practices vary by income and ethnicity (Evans et al., 2011; Hughes, Power, Orlet Fisher, Mueller, & Nicklas, 2005; Power, Hughes, et al., 2015; Sherry et al., 2004; Ventura, Gromis, & Lohse, 2010), and existing questionnaire items may not adequately capture feeding behaviors specific to certain economic and ethnic groups (Jain, Sherman, Chamberlin, & Whitaker, 2004). Further investigation of feeding practices among low-income Hispanic American parents is important to better understand and ultimately diminish the disproportionately higher prevalence of obesity among this population in the United States. A lack of validated measures appropriate for use among this population is a critical barrier to conducting research to reduce obesity and obesity-related health disparities.

While the CFPQ has not yet been validated among Hispanic-American populations, the psychometric properties of the CFPQ have been examined among diverse populations outside the United States (Al-Qerem, Ling, & AlBawab, 2017; Haszard, Williams, Dawson, Skidmore, & Taylor, 2013; Mais, Warkentin, Latorre Mdo, Carnell, & Taddei, 2015; Shohaimi, Wei, & Shariff, 2014).The majority of these factor analyses have shown poor fit to the original model (Al-Qerem et al., 2017; Haszard et al., 2013; Mais et al., 2015; Shohaimi et al., 2014; Warkentin, Mais, Latorre Mdo, Carnell, & Taddei, 2016). In addition to ethnicity and nationality, socioeconomic status and child age may be reasons for the varying structures found across these studies. Lower income status has been associated with unfavorable parent feeding practices, such as greater restriction and pressure to eat (Cardel et al., 2012). A lack of consistent reporting of income level and lack of subgroup analyses among diverse income samples prohibit conclusions regarding how income influences the psychometric properties of the CFPQ. Conversely, consistent reporting of child age enables general comparison across studies. The factor analysis of the Portuguese version of the CFPQ did not vary between the age ranges of two to five and five to nine years (Mais et al., 2015; Warkentin et al., 2016). However, longitudinal analysis of the CFPQ among American Caucasians indicates that parent feeding practices may differ by the age of the child, even within the preschool age years (from 37 to 57 months old) (Saltzman, Balantekin, Musaad, Bost, & Fiese, 2018). Additionally, the CFPQ factor analysis among New Zealand children ages five to eight found parental pressure to eat to be less prevalent the older the child (Haszard et al., 2013).

The purpose of this study was to examine the psychometric properties of the CFPQ in a low-income sample of Hispanic-American parents with preschool-aged children in the United States. Although Brazilian culture and language differs from that of Hispanic-Americans, a series of studies among Brazilian children (Mais et al., 2015; Warkentin et al., 2016) may be most comparable to the present study because of the shared Latin American roots between populations. Factor analysis of the Portuguese translation of the CFPQ among Brazilian children resulted in a six-factor model. Based on this, it was hypothe-sized that changes to the CFPQ would be necessary for use in low-income, Hispanic-American populations. Specifically, the original twelve-factor model would not be a best fit; instead, a model with fewer factors would be more appropriate.

2. Methods

2.1. Participants and procedures

Recruitment and data collection for this study took place August 2011 to May 2013. Recruitment for this study was focused on Head Start parents who had a three-five year old child whom they identified as being Hispanic. Recruitment procedures included presentations at parent meetings, flyers placed at Head Start centers and sent home with the child, and active recruitment by Head Start teachers. Recruitment was conducted across three head start districts in a large urban south-western city in the United States. Of these districts, 26 centers were approached about the study. Approximately 440 families expressed interest in the study, from which 187 parent-child dyads participated. All parents recruited into the study were mothers except for two grandmothers (here in referred to as mothers). Mothers reviewed and signed consent forms in either English or Spanish and received a total of $90 for participating in the study. The study was reviewed and approved by the Institutional Review Board at Baylor College of Medicine.

The recruited mother-child dyads came into the study laboratory on two days to participate in observational tasks not relevant to the present paper. Questionnaires were completed by the mother on the second day while the child was participating in tasks not involving the mother. Seventy-nine percent of the questionnaires were completed in Spanish. The translation process, similar to that used for a separate feeding measure (Hughes et al., 2005), included several Spanish speaking research staff. Initial translation was performed, and a second member of the research staff back-translated, taking note of differences between the English and Spanish versions. Differences were discussed between the two translators plus a third Spanish-speaker who was naïve to the translation and author. The final questionnaire was the product of the consensus reached between all three individuals.

2.2. Measures

Comprehensive Feeding Practices Questionnaire (CFPQ). The CFPQ was used to assess maternal feeding practices (Musher-Eizenman & Holub, 2007). This parent-report questionnaire consists of 49 items measuring a broad range of parental feeding practices with preschool and elementary school age children. The twelve subscales include child control, emotion regulation, encourage balance and variety, environment, food as reward, involvement, modeling, monitoring, pressure to eat, restriction for health, restriction for weight control, and teaching about nutrition. Tables 2 and 3 list each questionnaire item with its corresponding subscale. The CFPQ is a well-known measure that has been extensively used (Vaughn et al., 2013), including low-income samples (Gross, Velazco, Briggs, & Racine, 2013). Reliability and validity of the measure was originally established among middle class, Caucasian families with preschool aged children (Musher-Eizenman & Holub, 2007).

Table 2.

Characteristics of deleted items.

Item		Mean (SD)	Original CFPQ Factor
5	Do you let your child eat whatever s/he wants?	2.30 (0.98)	Child Control
6	At dinner, do you let this child choose the foods s/he wants from what is served?	2.74 (1.12)	Child Control
11	Do you allow this child to eat snacks whenever s/he wants?	2.09 (0.97)	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?	2.99 (1.15)	Child Control
16	I keep a lot of snack food (potato chips, Doritos, cheese puffs) in my house.	2.21 (1.10)	Environment
18	I have to be sure that my child does not eat too many high-fat foods.	4.05 (1.19)	Restriction for Weight Control
21	If I did not guide or regulate my child’s eating, s/he would eat too much of his/her favorite foods.	3.63 (1.34)	Restriction for Health
28	If I did not guide or regulate my child’s eating, s/he would eat too many junk foods.	3.75 (1.38)	Restriction for Health
30	If my child says, “I’m not hungry,” I try to get him/her to eat anyway	3.58 (1.37)	Pressure
36	I withhold sweets/dessert from my child in response to bad behavior.	3.20 (1.60)	Food as reward
37	I keep a lot of sweets (candy, ice cream, cake, pies, pastries) in my house.	1.90 (1.08)	Environment
42	I tell my child what to eat and what not to eat without explanation.	2.17 (1.36)	Teaching about Nutrition
43	I have to be sure that my child does not eat too many sweets (candy, ice cream, cake, or pastries).	4.35 (1.04)	Restriction for Health
45	I often put my child on a diet to control his/her weight.	1.42 (0.96)	Restriction for Weight Control
49	When s/he says s/he is finished eating I try to get my child to eat one more (two more, etc.) bites of food.	3.18 (1.43)	Pressure

Table 3.

Final model of the CFSQ and item characteristics: Mean (standard deviation), Corrected Item-Total Correlation (CITC), and Factor Loadings.

Item		Description	M (SD)	CITC	Original CFPQ Factor	Factor Monitoring	Restriction for Weight	Promotion of Overconsumption	Healthy Eating Guidance	Healthy Eating Variety
Monitoring (α = 0.86)			4.20 (0.77)
2		How much do you keep track of the snack food (potato chips, Doritos, cheese puffs) that your child eats?	4.29 (0.79)	0.75	Monitoring	0.7894	−0.0088	−0.0021	−0.0108	0.0358
1		How much do you keep track of the sweets (candy, ice cream, cake, pies, pastries) that your child eats?	4.29 (0.85)	0.71	Monitoring	0.7889	−0.0848	0.0306	−0.0273	−0.0029
4		How much do you keep track of the sugary drinks (soda/pop, kool-aid) this child drinks?	4.23 (0.94)	0.77	Monitoring	0.7594	0.0123	−0.0493	0.0593	−0.0192
3		How much do you keep track of the high-fat foods that your child eats?	3.97 (1.05)	0.62	Monitoring	0.6487	0.0993	0.0125	0.0218	0.0190
Restriction for Weight (α = 0.87)			2.76 (1.09)
34		I restrict the foods my child eats that might make him/her fat.	3.06 (1.46)	0.73	Restriction for weight control	−0.0412	0.7908	−0.1334	0.0454	−0.0248
35		There are certain foods my child shouldn’t eat because they will make him/her fat.	2.81 (1.56)	0.69	Restriction for weight control	−0.0408	0.7490	−0.0207	0.0053	−0.0441
29		I give my child small helpings at meals to control his/her weight.	2.62 (1.50)	0.70	Restriction for weight control	0.0138	0.7166	−0.0821	−0.0343	0.0314
27		I encourage my child to eat less so s/he won’t get fat.	2.54 (1.56)	0.64	Restriction for weight control	0.0220	0.6589	−0.0181	−0.0489	0.0344
33		If my child eats more than usual at one meal, I try to restrict his/her eating at the next meal.	2.61 (1.37)	0.62	Restriction for weight control	0.0480	0.6485	0.0900	0.0258	−0.0372
41		I don’t allow my child to eat between meals because I don’t want him/her to get fat.	2.29 (1.35)	0.60	Restriction for weight control	−0.0880	0.5782	0.0698	−0.0358	−0.0219
40		I have to be sure that my child does not eat too much of his/her favorite foods.	3.35 (1.32)	0.54	Restriction for health	0.1193	0.5348	0.1476	−0.0034	0.0118
Promotion of Overconsumption (α = 0.73)			2.45 (0.68)
8		Do you give this child something to eat or drink if s/he is bored even if you think s/he is not hungry?	1.65 (0.84)	0.56	Emotion regulation	0.0309	−0.0476	0.8001	0.0008	−0.1029
9		Do you give this child something to eat or drink if s/he is upset even if you think s/he is not hungry?	1.45 (0.77)	0.56	Emotion regulation	−0.0226	0.0149	0.7781	−0.0094	−0.0713
7		When this child gets fussy, is giving him/her something to eat or drink the first thing you do?	1.73 (0.90)	0.46	Emotion regulation	0.0344	0.0291	0.5817	−0.0610	−0.0979
19		I offer my child his/her favorite foods in exchange for good behavior.	2.77 (1.47)	0.52	Food as reward	−0.1441	0.0951	0.4478	0.0470	0.1629
39		If my child eats only a small helping, I try to get him/her to eat more.	3.58 (1.29)	0.36	Pressure	−0.0012	−0.0729	0.3969	0.0528	0.0619
10		If this child does not like what is being served, do you make something else?	2.56 (1.09)	0.36	Child control	−0.0373	−0.0031	0.3725	0.0904	−0.0931
23		I offer sweets (candy, ice cream, cake, pastries) to my child as a reward for good behavior.	2.58 (1.36)	0.37	Food as reward	0.1161	0.0310	0.3635	−0.0487	0.1594
17		My child should always eat all of the food on his/her plate.	3.27 (1.32)	0.35	Pressure	−0.0428	0.0192	0.3523	−0.0446	0.1670
Healthy Eating Guidance (α = 0.86)			4.20 (0.63)
48	I show my child how much I enjoy eating healthy foods.		4.54 (0.81)	0.73	Modeling	−0.0972	−0.0144	−0.0303	0.7027	0.0457
47	I try to show enthusiasm about eating healthy foods.		4.49 (0.87)	0.65	Modeling	−0.0357	−0.0195	−0.0230	0.6358	0.0242
15	I involve my child in planning family meals.		3.71 (1.15)	0.62	Involvement	−0.0245	−0.0896	0.0168	0.5961	−0.0954
44	I model healthy eating for my child by eating healthy foods myself.		4.37 (0.97)	0.61	Modeling	0.0645	0.0441	0.1110	0.5319	0.0430
22	A variety of healthy foods are available to my child at each meal served at home.		4.25 (0.93)	0.60	Environment	0.1599	0.0466	−0.0617	0.4992	0.0035
32	I encourage my child to participate in grocery shopping.		4.37 (0.99)	0.54	Involvement	−0.0580	−0.0335	0.0703	0.4936	−0.0021
31	I discuss with my child the nutritional value of foods.		4.26 (1.04)	0.59	Teaching about nutrition	−0.0379	0.1176	0.0821	0.4812	0.1274
46	I try to eat healthy foods in front of my child, even if they are not my favorite.		4.15 (1.16)	0.43	Modeling	−0.0931	0.0932	−0.0276	0.4796	−0.1003
38	I encourage my child to eat a variety of foods.		4.51 (0.78)	0.50	Encourage balance & variety	−0.0351	−0.0669	−0.0747	0.4565	0.1238
14	Most of the food I keep in the house is healthy.		4.06 (0.92)	0.55	Environment	0.1512	0.0801	0.0483	0.4398	0.0510
13	Do you encourage this child to eat healthy foods before unhealthy ones?		4.33 (0.86)	0.48	Encourage balance & variety	0.1589	−0.0480	−0.0191	0.4093	0.0214
20	I allow my child to help prepare family meals.		3.39 (1.32)	0.41	Involvement	0.0346	−0.1038	−0.0571	0.3919	−0.0930
Healthy Eating Variety (α = 0.73)			4.68 (0.55)
26	I tell my child that healthy food tastes good.		4.78 (0.61)	0.70	Encourage balance & variety	−0.0927	−0.0521	0.0458	−0.0368	0.9027
25	I discuss with my child why it’s important to eat healthy foods.		4.70 (0.61)	0.57	Teaching about nutrition	0.0808	0.0346	0.0981	0.1564	0.5550
24	I encourage my child to try new foods.		4.55 (0.77)	0.43	Encourage balance & variety	0.1238	−0.0098	−0.1310	0.0542	0.3832

Demographics.

Demographic information collected included child age and gender, maternal age, marital status, employment status, level of education, nativity, and the number of children in the household.

3. Statistical analysis

The frequencies, percentages, means and standard deviations (SD) were calculated for all demographic and study variables. Means, SDs, and corrected item-total correlations (CITC) were calculated for all 49 items on the CFPQ. The internal consistency of each subscale was examined using Cronbach’s alpha. Values greater than 0.3 and 0.7 were considered acceptable for CITC and internal consistency, respectively (Nunnally, 2010).

A confirmatory factor analysis (CFA) was performed using Mplus version 7.31 (Múthen & Múthen, 1998–2015) to examine the goodness of fit of the original 12-factor, 49-item structure of the original CFPQ among a Hispanic sample. The performance of model fit was evaluated using Hu and Bentler’s two-index presentation strategy. The combinational rule of Comparative Fit Index (CFI) > 0.95 and Standardized Root Mean Square Residual (SRMR) < 0.09 were recommended for a sample size smaller than 250 (Hu & Bentler, 1999). Should the original factor structure not be confirmed, an exploratory factor analysis (EFA) using maximum likelihood with squared correlation as initial communality estimates will be conducted to identify an alternative underlying factor structure. The optimal number of components will be selected based on the Scree plot (Cattell, 1966) and the variance explained by the factors (> 0.7; Stevens, 2002). The final solution will be rotated using an oblique promax rotation. The Pearson correlations among the identified factors will also be provided. The EFA will be conducted using SAS PROC FACTOR (9.4,Cary, NC).

4. Results

4.1. Sample descriptives

Table 1 presents the sample characteristics. On average children were 4.78 (± 0.43) years old and 52% were male. Mothers were on average 31.98 years (± 6.47), married (59.68%), unemployed (76.47%), had a high school diploma or some college education (52.94%), from Mexico (77%), and had 2.66 children living at home.

Table 1.

Sample characteristics (n = 187).

Characteristic	Mean (SD) or % (n)
Child Characteristics
Age (years)	4.78 (0.43)
Gender
Female	47.59 (89)
Caregiver & Household Characteristics
Age	31.98 (6.47)
Gender
Female	100.00 (187)
Marital Status
Single	40.10 (75)
Married	59.89 (112)
Employment Status
Employed	23.53 (44)
Level of Education
Less than high school	40.10 (75)
High school graduate or some college	52.94 (99)
College graduate	6.95 (13)
Nativity
Mexico	77.01 (144)
Central America	20.86 (39)
Other	2.14 (4)
Number of Children in the Household	2.66 (1.33)

4.2. Confirmatory factor analysis

The 12-factor, 49-item structure was not confirmed by the CFA. Specifically, the Root Mean Square Error of Approximation (RMSEA) was 0.07, 90% CI [0.061, 0.07]; the CFI was 0.74; the Tuker Lewis Index (TLI) was 0.71; and the SRMR was 0.09. However, the parameter estimates and standard error may not be trustworthy due to having more parameters than the sample size.

4.3. Exploratory factor analysis

An EFA was performed to identify the simplest structure. The number of factors retained was determined by the Scree plot and the variance explained by the retained factors. The Scree plot indicated a clear break between five and six components. The eigenvalues for the first five factors were 14.21, 10.33, 6.04, 5.74, and 2.92. The next three factors had the eigenvalues of 1.92, 1.31, and 1.14. The five-factor model was also supported by the variance explained by the five retained factors. Specifically, the variances explained by the first four, five, and six factors combined were 68.38%, 74.16%, and 79.02%, respectively.

Fifteen items either did not load on a factor (had a factor loading < 0.35) or loaded on more than one factor, and were deleted during the structure development process (Table 2). The cut-point of 0.35 to balance the cutoff of 0.40 generally used and the cutoff of 0.32 suggested by Tabachnick and Fidell (2007) as a good rule of thumb for the minimum loading of an item. The analysis was then conducted with the reduced set of items and yielded a final set of 34 items.The rotated loadings between each item and each component derived from an oblique rotation are presented in Table 3.

The EFA resulted in a five-factor model, with factors similar to those in the original CFPQ. The first factor comprised four high-loading items and was labeled “monitoring.” These items are identical to the items in the monitoring factor in the original CFPQ validation study (Musher-Eizenman & Holub, 2007). Monitoring represents how much parents keep track of child consumption of unhealthy foods. The variance explained by the monitoring factor was 2.45 (18.92%).

The second factor included seven items and was labeled “restriction for weight.” Six of the items included in this factor were from the “restriction for weight” factor and one item was from the “restriction for health” factor in the original CFPQ (Musher-Eizenman & Holub, 2007).These items examined how much and conditions for when a mother restricted her child’s food intake. The variance explained by the “restriction for weight” factor was 3.28 (25.33%).

The third factor included eight items indicative of practices that may override children’s hunger and satiety cues and was labeled “promotion of overconsumption.” Specifically, this factor included the following from the original CFPQ: emotion regulation (all three items), child control (one of the five items), food as a reward (two of four items), and pressure (two of four items) (Musher-Eizenman & Holub, 2007). Specifically, these items represent whether parents use food as a way to regulate the child’s emotional state or behaviors, whether parents pressure the child to consume more food at meals, and whether parents make something else if the child doesn’t like what is being served (child control item). The variance explained by this factor was 2.47 (19.11%).

The fourth factor was labeled “healthy eating guidance.” This factor was labeled “healthy eating guidance” to be consistent with prior validation studies in which a similar factor emerged (Haszard et al., 2013; Mais et al., 2015; Saltzman et al., 2018; Warkentin et al., 2016). It included items related to modeling, food availability in the home, and parental encouragement of healthy food intake. Specifically, this factor consisted of all of the items from the original CFPQ’s modeling (four items) and involvement (three items) factors, one of the three items from the teaching nutrition factor, two of the four items from the environment factor, and two of the four items from the encourage balance and variety factor (Musher-Eizenman & Holub, 2007). The variance explained for healthy eating guidance was 3.28 (25.33%).

The final factor was labeled “healthy food variety” and was comprised of three items. These items came from the original teaching about nutrition (one item) and encourage balance and variety (two items) factors (Musher-Eizenman & Holub, 2007). The variance explained for this healthy food variety factor was 1.46 (11.31%).

The Pearson correlations among the five identified factors are presented in Table 4. Monitoring was significantly positively related to healthy eating guidance (r = 0.210, p = 0.004) and nutrition education (r = 0.147, p = 0.045); while monitoring was marginally negatively associated with restriction for weight (r = −0.132, p = 0.072) and promotion of overconsumption (r = −0.135, p = 0.065). Additionally, healthy eating guidance was positively correlated with healthy food variety (r = 0.459, p < 0.001).

Table 4.

Pearson Correlation among the five identified factors.

	2	3	4	5
1. Monitoring	− 0.132 (0.072)	0.210 (0.004)	− 0.135 (0.065)	0.147 (0.045)
2. Restriction for Weight		0.121 (0.098)	0.136 (0.064)	0.075 (0.306)
3. Health Eating Guideline			0.079 (0.282)	0.459 (< 0.001)
4. Promotion of Overeating				0.016 (0.827)
5. Healthy Food Variety

4.4. Internal consistency

Cronbach’s alpha, which measures internal consistency, was calculated for each of the five factors. The results are presented in Table 2 along with mean, SD, and CITC. Item means ranged from 1.73 (SD = 0.9) to 4.78 (SD = 0.61). CITC acceptably ranged from 0.35 to 0.77. Internal consistencies were considered good for monitoring (α = 0.86), healthy eating guidance (α = 0.86), and restriction for weight (α = 0.87); and were considered acceptable for promotion of overconsumption (α = 0.73) and healthy food variety (α = 0.73).

5. Discussion

This was the first study to investigate the psychometric properties of the CFPQ among low-income, Hispanic-American mothers with preschool aged children. Consistent with our hypothesis, a modified version of the CFPQ with five factors and 34 items was found to better assess maternal feeding practices in this sample than the original 12-factor, 49-item CFPQ (Musher-Eizenman & Holub, 2007). The lack of fit to the original model is consistent with other studies examining the appropriateness of using the CFPQ among other ethnicities outside the United States (Al-Qerem et al., 2017; Doaei, Kalantari, Gholamalizadeh, & Rashidkhani, 2013; Haszard et al., 2013; Mais et al., 2015; Melbye, Ogaard, & Overby, 2011; Shohaimi et al., 2014; Warkentin et al., 2016). As with the present study, many of these studies also required an EFA with substantial modification to achieve an acceptable fit. Correlation patterns between factors in this study were generally consistent with prior factor analyses (Mais et al., 2015; Warkentin et al., 2016) and the original validation study (Musher-Eizenman & Holub, 2007). The factors represent favorable parent feeding practices such as healthy eating guidance, healthy food variety, and monitoring being positively correlated with one another and being negatively correlated with unfavorable parent feeding practices such as restriction and promotion of overconsumption.

Monitoring was the only factor in the present study that was identical to that of the original CFPQ. This is consistent with studies examining the psychometric properties of the CFPQ in countries besides the United States, which have consistently included all monitoring items from the original CFPQ (Al-Qerem et al., 2017; Doaei et al., 2013; Haszard et al., 2013; Mais et al., 2015; Melbye et al., 2011; Shohaimi et al., 2014). This indicates that these items are well understood and that monitoring is likely a relevant construct across varying cultures and ethnicities.

Due to the possibility that the relationship between restriction and children’s weight depends on mothers’ motivation for restriction, the original CFPQ sought to distinguish between restriction for purposes of weight and for health (Musher-Eizenman & Holub, 2007). While the motivation behind restrictive practices is likely important to elucidate, the removal of all but one of the items in the original restriction for health factor indicates that this may not have been achieved in the present model among Hispanic-American mothers. As in the present study, the removal of all or almost all of the items from the restriction for health factor also occurred in two prior factor analyses of the CFPQ (Saltzman et al., 2018; Haszard et al., 2013). One of them occurred among Caucasian children aged 37 months (but not among these children at 57 months of age) in the United States (Saltzman et al., 2018). Another example occurred among Caucasian children aged five to eight years old from New Zealand (Haszard et al., 2013). In a CPFQ factor analysis among Malaysian school children (mean age of eight years old), restriction for health was kept as a separate factor; however, only two items remained in the factor (Shohaimi et al., 2014). Conversely, in the factor analyses among Brazilian preschool and school-age children, all restriction for health items remained as a distinct factor from restriction for weight (Mais et al., 2015; Warkentin et al., 2016). A detailed description and comparison of CFPQ subscales across studies examining the measure’s psychometric properties amongst diverse populations has recently been conducted, and provides a reference for further comparison (Saltzman et al., 2018).

Mixed results across studies regarding the delineation of restriction for weight and restriction for health make it difficult to disentangle the potential influences of ethnicity, nationality, income level, and age on the appropriateness of the two separate factors for restriction. Regarding ethnicity, in many Hispanic culturesweight status is not always indicative of health, and a healthy child is often considered one that is loved, happy, and able to play (Crawford et al., 2004; Reifsnider et al., 2006). Thinness, which may be indicative of illness, is considered a greater health risk than overweight status, and a moderately overweight body size is considered to be most preferable (Crawford et al., 2004; Reifsnider et al., 2006). Furthermore, regarding income level, having a heavier child is a sign of being a good parent and able to appropriately provide for the child, particularly among lower income families (Baughcum, Burklow, Deeks, Powers, & Whitaker, 1998). Prior research has suggested low income to be associated with greater parental restriction (Cardel et al., 2012), and for maternal concern for children’s weight to mediate the relationship between restriction and child weight (Gray, Janicke, Wistedt, & Dumont-Driscoll, 2010). Therefore, it will be important for future work to clarify how motivations for restriction and perceptions of weight status are related to weight, and how these factors can best be assessed among populations of varying income levels and ethnicities.

In the present study, the promotion of overconsumption factor that emerged contrasts with prior CFPQ factor analyses, which all retained a distinct subscale for pressure. The collapse of these items into one factor in the present study may be due to an underlying responsiveness of Hispanic mothers to their children’s happiness during meals. The indulgent parent feeding style, characterized by low levels of demandingness and high levels of responsiveness (Hughes et al., 2005), is common among low-income, Hispanic families in the United States (Hughes et al., 2011; Hughes, Shewchuk, Baskin, Nicklas, & Qu, 2008; Power, O’Connor, Orlet Fisher, & Hughes, 2015). Indulgent feeders are not demanding of their children and are likely to be highly responsive to their child’s emotional state, prioritizing child happiness at mealtimes (Hughes et al., 2011). This is consistent with a child’s happiness being considered an important determinant of health among Hispanic families. The only item from the child control factor of the original CFPQ in the present model (item 10: “If this child does not like what is being served, do you make something else?”) is in line with ensuring a child’s happiness at mealtime. Similarly, the two food as reward items that remained in the model (item 23 and 19) reflect rewarding good behavior; whereas the food as a reward item that was dropped (item 36) reflects withholding food in response to poor behavior, which is unlikely to make the child happy. Although the inclusion of items relating to pressuring a child to eat may be counterintuitive to the goal of making a child happy at mealtimes, the idea that “food is love” in Hispanic culture is particularly prevalent (Gomel & Zamora, 2007). What is interpreted as pressuring a child to eat may simply be a demonstration of love and is construed in a positive manner. Similarly, in the original CFPQ the food as reward and pressure constructs were highly correlated (Musher-Eizenman & Holub, 2007), indicating that “pressuring” children to eat may be viewed as a positive reinforcement for a particular behavior or a sign of love. The grouping of items conceptually related to responding to children’s emotions into one factor may indicate an underlying reason for why some “controlling” parent feeding practices differ between Caucasian and Hispanic-American mothers, with practices being more emotionally driven among Hispanic-American compared to Caucasian mothers.

In addition to ethnic differences, the collapse of items from the original emotion regulation, child control, food as a reward, and pressure factors into one factor may also be due to economic differences between samples. Low-income parents’ food provisioning is shaped by a desire to care and provide for children. Parents with limited finances may purchase relatively inexpensive unhealthy foods upon children’s request to compensate for the inability to purchase more expensive material desires (Fielding-Singh, 2017). Alternatively, these items may have loaded together for statistical rather than theoretical reasons. For example, it is possible that this sample of low income, Hispanic-American families had a higher correlation between pressure to eat and food as a reward factors or that the sample size precluded the possibility of greater variability in those items, and accordingly a different factor structure.

The healthy eating guidance factor identified in the current study, comprised of items from the original CFPQ modeling, involvement, environment, teaching about nutrition, and balance and variety factors, appears to represent parent feeding practices associated with healthy child weight status or eating behaviors (Vaughn et al., 2016). The final factor that emerged in the present model was healthy food variety. It contained three items, which were items from the teaching about nutrition and encourage balance and variety factors from the original CFPQ. These items are conceptually similar to those that loaded onto the healthy eating guidance factor. It is possible that the delineation of seemingly similar items is due to item misinterpretation. Qualitative research should be conducted to further examine how Hispanic parents interpret questionnaire items to ensure that items are assessing the intended constructs.

This study is the first to use the CFPQ among economically dis-advantaged Hispanic-American families, a population at high risk for poor eating behaviors and the development of obesity and obesity-related health problems. Importantly this study focused on mothers of preschool-aged children. Preschool is a developmental period marked by increased autonomy and continued introduction to novel foods. Thus, this life stage represents an important opportunity for parents to facilitate the establishment of healthy eating habits and behaviors to prevent excessive child weight gain (Cooke, 2007; Lagattuta, Nucci, & Bosacki, 2010). Adapting the CFPQ for low income, Hispanic-American families is the first step in being able to assess the role that parent feeding practices may play in the development of obesity in this population.

In addition to these strengths, this study also has some notable limitations. Although an important population to study, the specificity of the study population limits the ability to disentangle the influence of income and ethnicity on the results of the factor analysis. This differentiation is important to better understand the underpinnings of how items loaded in the exploratory analysis. For example, it is possible that the collapse of the restriction for health and for weight into one factor and the collapse of the pressure to eat, emotion regulation, child control, food as a reward, and pressure to the promotion of overconsumption factor would have remained separate factors if income and ethnicity could have been distinguished. Further, the level of acculturation of families in this study was not assessed but may be an important factor. Finally, due to the small sample size, findings from this study should be viewed as preliminary. While accurate results can be derived from smaller sample sizes (i.e. 3 cases per item) (Gorsuch, R.L., 1997), small sample sizes are more prone to issues with the correct factor structure, misclassifying items, error in eigenvalues and factors, and Heywood cases (Costello & Osborne, 2005; de Winter, Doduo, & Wieringa, 2009; Rouquette & Falissard, 2011). Therefore, it will be important for this study to be replicated among a similar population.

The results of this EFA indicate that cognitive interviewing or a similar qualitative method is needed to help ensure the validity of the instrument prior to the CFPQ being used with low-income Hispanic-American populations. Validation of commonly used questionnaires such as the CFPQ is necessary to accurately assess low income, Hispanic-American parent feeding practices. This will allow for formative work to contribute to the development of obesity interventions specifically tailored to this high-risk population. Meanwhile, the process of adapting the CFPQ to be appropriate for use among low-income, Hispanic populations may provide valuable insight into understanding how obesity disparities in the United States can be better addressed.