Sage Journals: Discover world-class research

Abstract

Purpose

Adaptive eating refers to internally regulated eating that is enjoyable and gently guided by the nutritional quality of foods. Despite growing interest in this construct, there is currently no validated scale to measure adaptive eating among US college students aged ≥18 years. This study aimed to develop and evaluate the psychometric properties of the Adaptive Eating Scale (AES) in a diverse sample of US college students.

Design

Thirty-seven items were developed to comprise the AES. Students completed AES and measures of intuitive eating, mental health, diet, and anthropometrics.

Sample

Eight hundred and forty-nine university students aged ≥18 years, from racially and ethnically diverse backgrounds.

Analysis

The sample was divided randomly into two. Exploratory Factor Analysis (EFA) was completed on subsample A (n = 424). Confirmatory Factor Analysis (CFA) was conducted on subsample B (n = 425) to confirm factor structure from subsample A.

Results

EFA showed 17 items representing gentle nutrition (GN), unpreoccupied by food cravings (UBFC), enjoyment of food (EOF), and honoring hunger (HH) explained 63.1% of the variance. CFA showed a bifactor model comprising one global factor and four orthogonal factors (GN, UBFC, EOF, and HH) had excellent fit [CFI = 0.972, RMSEA = 0.051, SRMR = 0.048]. GN, EOF, and HH were reliable, and significantly associated with each other, intuitive eating, emotional eating, weight status, and mental health.

Conclusion

Results showed that AES is best conceptualized as a multidimensional measure with one general factor and specific factors representing GN, EOF, and HH.

Keywords

eating behavior cross-sectional validity college students US

Introduction

Behavioral patterns concerning what, when, and how much to eat are influenced by physiological, genetic, psychological, and social factors.^1,2 According to behavioral susceptibility theory genetic predispositions toward food approach or avoidance traits are shaped over time by environmental factors such as family dynamics and food insecurity.³ Eating behaviors are considered maladaptive when they are associated with adverse physical and mental health outcomes.^4-9 Examples of maladaptive eating behaviors include binge eating, emotional eating, compulsive eating, hedonic eating, and excessive restriction.^10-12 In contrast to the large body of research on maladaptive eating, relatively little is known about adaptive eating, eating behaviors that promote physical and psychological health and well-being.^5,13

Wellness is defined as optimal mental and physical functioning despite one’s current environment and circumstances.¹⁴ More than the absence of disease, it focuses on thriving and resilience to life challenges. Based on positive psychology, wellbeing encompasses positive emotions, meaningful activities, positive relationships, mindfulness, and goal achievement.¹⁵ Adaptive eating may support wellbeing when it is pleasurable (ie, positive emotions), supports long-term health (ie, meaningful), reflects a positive relationship with food and body image (ie, positive relationships), is internally regulated (ie, mindful), and serves the goals of satisfying hunger and satiety (ie, goal achievement). While there are measures of adaptive eating that capture some of these constructs, no validated measure currently captures adaptive eating from a wellness perspective. For example, the Intuitive Eating Scale, which measures internally regulated eating, is the most popular measure of adaptive eating, but it was developed based on a theory of eating recovery for chronic dieters, some versions of the scale were found to have inconsistent factor structures across various populations,^16-20 and the practical application of the current version (3) is limited by the bifactorial structure of the items.²¹ A measure of adaptive eating based on a theory of wellness for a diverse population would allow for research on adaptive eating, its determinants, consequences, and mechanisms of action, which may inform health promotion interventions.

The college years represent a period of transition and growth. College students experience several emotions and stress due to changes in their home and academic environment, societal expectations, and peers.²² These changes affect dietary habits, which track throughout the lifespan.²³ A validated instrument to measure adaptive eating is the first step to finding ways to support college students in achieving wellness. This study aimed to develop and validate the Adaptive Eating Scale (AES) among racially/ethnically diverse college students. It was hypothesized that the AES would capture (1) the habit of eating when hungry (honoring hunger, HH) and terminating meals and snacks when satiated (respecting satiety, RS), (2) the ability to derive pleasure from food (enjoyment of food, EOF), (3) moderate interest in nutrition when making food choices (gentle nutrition, GN), and (4) freedom from impulses to eat energy dense foods without hunger (unpreoccupied by food craving, UBFC).

Methods

Development of the AES included a literature review, item generation, expert consultation, and measurement. A definition of each step is presented in Text S1. The initial item pool (including subscales of HH, RS, EOF, and UBFC) included 37 items rated on a 5-point Likert scale. For 19 items, response options ranged from “1 = never” to “5 = always” and for 18-reverse-coded items, response options ranged from “5 = never” to “1 = always.” Response options were summed so that higher scores indicate more adaptive eating.

Data Collection

The Research on Eating, Activity, and Community Health (REACH) study evaluated eating behaviors among undergraduate students. In spring 2022, students ≥18 years old from an introductory science course at the University of Houston were eligible. In exchange for course credit, students completed the survey via the web, and anthropometric measurements were collected in the lab. The study was approved by the Institutional Review Board. Out of 974 participants, 125 were removed because they had missing covariate data (n = 98), or missing eating behavior data (n = 27), resulting in a final sample of 849 (87.2%).

Demographic Data

Demographic data comprising age, sex, race, and ethnicity were gathered through a questionnaire.

Internally Regulated Eating

A 6-item subscale (Reliance on Hunger and Satiety Cues, RHSC) from the IES-2 assessed internally regulated eating.²⁴ Participants responded to items using a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree) (Cronbach’s α = 0.83) (Text S2).

Body Mass Index

Using standard protocols, trained staff collected weight with a scale (Tanita Corporation, Chicago, IL, USA) and height with a stadiometer (Seca 222, WA, USA). Body mass index (BMI) was calculated according to Quetelet’s index (weight (kg)/Height² (m²).²⁵

Depression

The Center for Epidemiological Studies Depression Scale (20-items) is a validated self-report measure of current depression symptoms.²⁶ Responses were scored as 0 = “Not At All” to 3 = ” A Lot” with scores ranging from 0-60 (Text S2).

Emotional Eating Scale

The emotional eating scale is a subscale of the Motivations to Eat measure and examines the frequency of eating in response to negative emotions.²⁷ It comprised 5 items with response options “never or almost never” (1) to “almost always or always” (5). Higher scores indicate more emotional eating (Cronbach’s α = 0.88).

Starting the Conversation

Starting The Conversation assesses disease risk due to poor quality diet. Participants indicated how often they consumed 8 types of foods: fast food, fruit, vegetables, sodas, beans, chicken, fish, chips, desserts, and margarine²⁸ (Text S2). Frequent consumption of fruit, vegetables, chicken, fish, and beans lowers the risk of chronic disease, so these items were reverse-coded.

Rosenburg Self-Esteem Scale

Rosenburg Self Esteem Scale (RSES) is a validated scale that measures self-esteem.²⁹ Response options were 1 = “Strongly agree” and 4 = “Strongly disagree”. Higher scores indicated higher self-esteem (Text S2).

Veggie Meter

Skin carotenoids were assessed using the Veggie Meter (Longevity Link Corp), a pressure-mediated reflection spectroscopy device. The device was calibrated daily and after every ten participants. All participants had the non-dominant ring finger cleaned with an alcohol swab before placing it on the sensor. The device produced a Skin Carotenoid Score (0-800), with higher scores indicating higher carotenoid levels. Three readings were taken using the standard triplicate protocol, and the average was automatically generated.^30,31

Data Analysis

Descriptive analyses were conducted to calculate means ± standard deviations (SDs) or frequencies (%). The sample was randomly split into two subsamples A and B, and a cross-validation approach was pursued. All subsample A analyses were conducted using SPSS V29.

Exploratory Factor Analysis

Subsample A was used to investigate the AES item distribution and factor structure via principal axis factoring with varimax rotation. Cronbach’s alpha was used to assess internal consistency (≥0.7 acceptable). Of the 37 original AES items, only items with factor loadings of ≥0.60^32,33 were retained. Factors were retained based on eigenvalues >1, scree test, and parallel analysis.^34,35 Factors with <3 items loading at ≥0.60 were considered unreliable and excluded.³⁶

Convergent Validity

The relation between AES total and subscale scores (GN, UBFC, EOF, HH) with the correlates was examined using Spearman’s correlation coefficients for subsample A.

Confirmatory Factor Analysis

Subsample B was used to cross-validate findings from subsample A within a confirmatory factor analysis (CFA) framework. Specifically, four alternative CFA models were estimated in Mplus vs 7 (Muthén & Muthén, 1998-2012) using maximum likelihood estimation. Global model fit was evaluated using a non-significant (P > 0.05) model chi-square test, Comparative Fit Index (CFI) > 0.90 for acceptable and >0.95 for good fit, Root Mean Square Error of Approximation (RMSEA) < 0.05 for good and <0.08 for acceptable fit, a non-significant (P > 0.05) test of the close-fit RMSEA hypothesis (H₀: RMSEA ≤0.05), and Standardized Root Mean Square Residual (SRMR) < 0.08.^37,38 Given the sample size, less weight was placed on the chi-square test when evaluating the global model fit compared to the other fit indices (ie, CFI, RMSEA, SRMR).³⁹ Local model misfit, as indicated by Heywood cases and significant standardized residuals, was evaluated. Parameter estimates were also examined for interpretability.

Results

Participants

The participant characteristics are indicated in Table 1.

Table 1.

General Characteristics of participants Included in our Study

Variables	Total (n = 849)
Age, years	20.8 (2.62)
Sex, n (%)
Women	399 (47%)
Men	329 (38.8%)
Height, cm	168 (9.94)
Race, n (%)
White	308 (36.3%)
Black or African American	125 (14.7%)
Asian	307 (36.2%)
American Indian or Alaska native	11 (1.3%)
Native Hawaiian or pacific islander	3 (0.4%)
Other races	95 (11.2%)
Ethnicity, n (%)
Latino or Hispanic^a	279 (33%)
Non-Hispanic^b	570 (67.1%)

Data are presented as mean ± standard deviation (SD) or percent.

^aA person who identifies with or is of Mexican, Puerto Rican, Cuban, Central or South American, or other Spanish culture or origin.

^bAny possible options not covered in the above category.

Descriptive Statistics and EFA (Subsample A)

Table 2 shows descriptive statistics for the 17 AES items. Item distributions indicated normality (skewness: −0.78 to 0.06; kurtosis: −0.76 to 0.55) without floor/ceiling effects. The Kaiser-Meyer-Olkin value of 0.83 and Bartlett’s test (χ² (136) = 4284.48, P < 0.001), supported the application of EFA. Principal axis factoring analysis with varimax rotation suggested a four-factor model as best capturing the factorial structure of the 17 items. The AES items, rotated factor loadings, communalities, eigenvalues, and percentages of explained variance for the four-factor model are reported in Table 2. The 4 factors accounted for 63.10% of the variance. Item communalities met the ≥0.50 threshold (except for item 4 (GN4): 0.43). All factors were reliable with ≥ 3 rotated factor loadings ≥0.60 (range: 0.64 to 0.84) and no cross-loadings ≥0.30. During EFA, 20 items (many positively phrased; see Table 2) were removed due to low factor loadings. The remaining 17 items loaded on 4 factors: (HH, 3 items), (EOF, 4 items), (UBFC, 5 items), and (GN, 5 items), reflecting the potent adaptive eating components. (Table 2).

Table 2.

Descriptive Statistics for Adaptive Eating Scale (AES) Items, Cronbach Alpha of AES Subscales, and Results of Principal Axis Factoring Analysis With Varimax Rotation for Random Split Subsample a (N = 424)

How often do you	M (SD)	Factor 1 GN (α = 0.88)	Factor 2 UBFC (α = 0.89)	Factor 3 EOF (α = 0.88)	Factor 4 HH (α = 0.80)	h ²
GN1. Feel preoccupied by nutritional content of foods you eat^a	3.54 (1.07)	0.84	0.26			0.78
GN2. Think too much about the nutritional quality of foods you eat^a	3.48 (1.10)	0.83	0.15			0.72
GN3. Feel distracted by thoughts of nutrition^a	3.50 (1.09)	0.81	0.27	0.13	0.11	0.75
GN4. Concentrate a lot on nutritional quality of your food^a	2.89 (1.01)	0.64		−0.12		0.43
GN5. Spend too much time planning meals^a	3.77 (0.99)	0.65	0.28			0.50
UBFC1. Feel preoccupied by thoughts of food^a	3.67 (1.00)	0.19	0.84			0.74
UBFC2. Think too much about food^a	3.77 (1.05)	0.20	0.82			0.72
UBFC3. Feel distracted by thoughts of food^a	3.41 (0.99)	0.11	0.79			0.64
UBFC4. Feel burdened by food cravings^a	3.72 (1.16)	0.21	0.69		0.24	0.59
UBFC5. Struggle with food cravings^a	3.72 (1.20)	0.15	0.66		0.25	0.52
EOF1. Savor the flavors of the food you are eating	3.97 (0.77)			0.79		0.62
EOF2. Enjoy the food you are eating	3.99 (0.69)	0.12		0.79	0.13	0.67
EOF3. Like the smell of the food you are eating	3.96 (0.76)			0.80		0.65
EOF4. Like the textures of the food you are eating	3.93 (0.73)			0.80		0.65
HH1. Ignore your hunger^a	3.47 (1.00)				0.84	0.71
HH2. Wait until you are very hungry to eat^a	3.17 (1.01)		0.11		0.72	0.54
HH3. Avoid eating when you are hungry^a	3.68 (0.93)		0.21	0.15	0.66	0.51
Eigenvalue		2.45	5.05	1.85	1.38
% Explained variance		14.40	29.72	10.88	8.10

Removed items
UBFC6: Experience food cravings?^a
EOF5: Eat what you are in the mood to eat?
EOF6: Go out of your way to find something pleasurable to eat?
RS1. Stop eating when you are comfortably full?
RS2. Leave food on your plate?
RS3. Feel comfortably full when you stop eating?
RS4. Feel satisfied at the end of a meal or snack?
RS5. Stop eating before you feel full?^a
RS6. Feel unsatisfied when you stop eating?^a
RS7. Eat beyond the point of comfortable fullness?^a
RS8. Feel too full when you stop eating?^a
HH4. Start eating when you feel hunger?
HH5. Feel hunger when you begin eating?
HH6. Feel an empty sensation in your stomach when you begin to eat?
HH7. Experience a little gnawing or rumbling in your stomach when you begin to eat?
HH8. Begin eating when you feel emptiness in your stomach?
HH9. Begin eating when you feel a little gnawing or rumbling in your stomach?
HH10. Eat without feeling any hunger sensations?^a
HH11. Feel no hunger at all when you begin to eat?^a
HH12. Begin eating without feeling hungry?^a

Note. Rotated factor loadings ≥0.60 are depicted in bold font. Factor loadings below 0.10 are not shown for visual clarity.

Abbreviations: GN, Gentle Nutrition; UBFC, Unpreoccupied By Food Cravings; RS, respecting satiety; EOF, Enjoyment Of Food; HH, Honoring Hunger; h², item commonality.

^aReverse scored items.

Internal Consistency and Convergent Validation (Subsample A)

Cronbach α (0.80 to 0.89) for 4 factors (Table 2) indicated internal consistency. All subscales were significantly (P < 0.05) positively (r = 0.11 to 0.38) associated with RSES and RHSC and negatively (r = −0.46 to −0.08) associated with depression and emotional eating. Associations with BMI and diet were mixed (Table 3).

Table 3.

Correlation Table for AES Total and Subscale Scores [Gentle Nutrition (GN), Unpreoccupied With Food Cravings (UBFC), Honoring Hunger (HH), and Enjoyment of Food (EOF)] With Intuitive Eating, Emotional Eating, Diet Quality, Depression, Veggie Meter, Self-Esteem, and Body Mass Index (BMI) for a Random Split Subsample a (n = 424)

	AES	Gentle nutrition	Unpreoccupied by food cravings	Honoring hunger	Enjoyment of food
Mean (SD)	61.5 (9.24)	17.2 (4.36)	18.3 (4.49)	10.2 (2.47)	15.8 (2.53)
Gentle nutrition	0.75***	—	0.41***	0.14**	0.13**
Unpreoccupied by food cravings	0.80***	0.41***	—	0.29***	0.14**
Honoring hunger	0.52***	0.14**	0.29***	—	0.15**
Enjoyment of food	0.45***	0.13**	0.14**	0.15**	—
Reliance on hunger and satiety (IES)	0.38***	0.19***	0.35***	0.20***	0.23***
Emotional eating	−0.37***	−0.08	−0.46***	−0.23***	−0.17***
CESD	−0.42***	−0.15**	−0.33***	−0.44***	−0.25***
RSES	0.35***	0.11*	0.33***	0.30***	0.22***
STC^a	−0.002	0.19***	−0.011*	−0.02	−0.11*
Veggie meter	−0.06	−0.14**	−0.06	0.04	0.08
Body mass index	−0.07	−0.03	−0.06	−0.11*	0.02

Abbreviations: CESD, Center for Epidemiological Studies Depression Scale; RSES, Rosenburg Self Esteem Scale; STC, Starting The Conversation; SD, standard deviation; IES, Intuitive Eating Scale.

^*P < 0.05, ^**P < 0.01, ***P < .001.

^ahigher STC indicates poorer diet quality and vice versa.

CFA (Subsample B)

Informed by EFA results, four confirmatory factor analysis (CFA) models were estimated using subsample B (Figure 1 & Table 4). First, a single-factor CFA model (M1): Poor fit (CFI = 0.382, RMSEA = 0.224, SRMR = 0.181), rejected as expected; Second, a four-correlated first-order factor CFA model (M2): Acceptable fit (CFI = 0.931, RMSEA = 0.076, SRMR = 0.056), though RMSEA close-fit hypothesis was rejected. Its completely standardized factor loadings were similar to EFA results from subsample A: GN: 0.63-0.95; UBFC: 0.66-0.90; EOF: 0.74-0.77; HH: 0.75-0.83). The intercorrelations among the 4 first-order latent factors were small-to-moderate (r = 0.00 to 0.40), similar to the EFA (Table 3); Third, a second-order CFA model (M3): This model showed acceptable fit to the data. However, the close-fit RMSEA hypothesis was again rejected (CFI = 0.927, RMSEA = 0.078, SRMR = 0.066). The UBFC first-order latent factor was strongly associated with the general second-order latent factor (completely standardized second-order factor loading: 0.90), whereas the factor loadings of the other first-order latent factors were much smaller in magnitude (EOF = 0.02, HH = 0.33, GN = 0.44); Fourth, a symmetrical bifactor CFA model (M4): Best overall fit (CFI = 0.972, RMSEA = 0.051, SRMR = 0.048), meeting all criteria including RMSEA close-fit. This was selected as the final model. Alternative bifactor models showed, at best, only acceptable fit.

Figure 1.

Four alternative confirmatory factor analysis models estimated for subsample B.

Table 4.

Fit of Confirmatory Factor Analysis Models for Random Split Subsample B (N = 425)

Model	χ² (df)	CFI	RMSEA (90% CI) PCLOSE-FIT	SRMR
M1: Single-factor CFA model	2646.75 (119) P = 0.000	0.382	0.224 (0.216, 0.231) P = 0.000	0.181
M2: Four correlated first-order factor CFA model	398.96 (115) P = 0.000	0.931	0.076 (0.068, 0.084) P = 0.000	0.056
M3: Second-order CFA model with four first-order factors	413.91 (115) P = 0.000	0.927	0.078 (0.070, 0.086) P = 0.000	0.066
M4: Symmetrical bifactor CFA model (one general and four orthogonal specific factors)	215.22 (102) P = 0.000	0.972	0.051 (0.042, 0.061) P = 0.411	0.048

Note. Based on maximum likelihood estimation.

Abbreviations: CFA, Confirmatory Factor Analysis; CFI, Comparative Fit Index; RMSEA, Root Mean Square Error of Approximation; PCLOSE-FIT, close-fit RMSEA hypothesis; SRMR, Standardized Root Mean Square Residual; CI, confidence interval.

Symmetrical Bifactor CFA Model (Subsample B)

Bifactor measurement models are based on the assumption that all items in an instrument load on a single general latent factor and that each item additionally loads on one specific latent factor. Each specific factor was uncorrelated with each other and captured unique aspects of the construct beyond the general latent factor. The results for the symmetrical bifactor CFA model are reported in Table 5.

Table 5.

Maximum Likelihood Estimates for Symmetrical Bifactor CFA Model With 17 Adaptive Eating Survey Items for Random Split Subsample B (N = 425)

AES Item	General factor G	Specific factor GN	Specific factor UBFC	Specific factor EOF	Specific factor HH	Proportion explained variance
GN1.	1.49 (0.42)	1.35 (0.85)				0.90
GN2.	1.35 (0.37)	1.32 (0.80)				0.77
GN3.	1.52 (0.43)	1.14 (0.72)				0.71
GN4.	0.43 (0.13)	1.00 (0.65)				0.44
GN5.	0.98 (0.31)	0.81 (0.57)				0.42
UBFC1.	2.24 (0.71)		0.99 (0.59)			0.84
UBFC2.	2.29 (0.70)		0.79 (0.45)			0.69
UBFC3.	2.11 (0.65)		1.00 (0.58)			0.75
UBFC4.	2.76 (0.81)		0.05 (0.03)			0.66
UBFC5.	3.09 (0.88)		−0.27 (−0.15)			0.79
EOF1.	−0.02 (−0.01)			1.17 (0.78)		0.61
EOF2.	0.16 (0.07)			1.00 (0.76)		0.58
EOF3.	0.02 (0.01)			1.08 (0.74)		0.55
EOF4.	0.10 (0.04)			1.06 (0.77)		0.59
HH1.	0.73 (0.23)				1.15 (0.73)	0.59
HH2.	0.98 (0.29)				1.33 (0.79)	0.70
HH3.	1.00 (0.34)				1.00 (0.67)	0.57
Factor variance (SE)	0.10 (0.03)	0.48 (0.07)	0.34 (0.09)	0.27 (0.03)	0.39 (0.05)
ω	0.93	0.90	0.93	0.85	0.83
ω _H	0.57	0.74	0.12	0.85	0.72
ECV	0.37
PUC ^a	0.79
FD ^a	0.95	0.95	0.86	0.92	0.90
H ^a	0.90	0.87	0.57	0.85	0.78

Note. Based on maximum likelihood estimation in Mplus vs 7.

Significant factor loadings are shown in bold font (for item GN4 at P < 0.05; for all other items at P < 0.001). Unstandardized factor loadings for items GN4, UBFC3, EOF2, and HH3 are fixed to value 1 for identification purposes. Completely standardized factor loadings are reported in parentheses.

All estimated factor variances are significant at P ≤ 0.001.

Abbreviations: GN, Gentle Nutrition; UBFC, Unpreoccupied By Food Cravings; EOF, Enjoyment Of Food; HH, Honoring Hunger; ω (omega), proportion of scale score variance that can be attributed to all involved common factors (i.e., general and specific factors); ω_H (omega hierarchical) = proportion of scale score variance that can be attributed to a single common factor; ECV (explained common variance), proportion of the common variance that can be attributed to the single general factor; PUC, percent of uncontaminated correlations; FD, Factor Determinacy; H = Construct replicability H-index.

^aBased on the calculator developed by Dueber (2017).

Omega Reliability Coefficients

Model-based omega reliability coefficients assessed the proportion of total scale score variance that can be attributed to (1) all (both general and specific) latent factors that underlie a scale score (ω coefficients) and (2) a targeted latent factor (general or specific only) that underlies the scale score at hand (omega hierarchical (ω_H) coefficients). ω_H should exceed 0.50 (preferably>0.75).⁴⁰ Findings for ω coefficients indicated a highly reliable multidimensional factor structure (general factor ω = 93% explained variance, the 4 specific factors ω = 83%-93%).

Findings for ω_H indicated that the general AES factor was moderately well-defined (57% of the total scale score variance could be attributed solely to the general latent factor). None of the EOF items had significant factor loadings on the general factor, indicating that they had little in common with the items from the other 3 subscales. Three specific factors, GN, EOF, and HH captured unique and reliable information beyond the general factor (ω_H = 72-85% of their subscale score variance could be attributed to their specific factor after adjustment for the general factor). However, the UBFC specific factor showed poor reliability (ω_H = 12%); most of its subscale score variance was explained by the general factor. Thus, the UBFC-specific factor should be considered negligible and unreliable.

Factor Determinacy and Construct Replicability

The Factor Determinacy (FD) index provides an estimate of the correlation between factor scores and the factors. The range of FD is between 0 and 1, with higher values indicating better determinacy. It is advisable to generate factor scores only when FD is greater than 0.90.⁴¹ With one exception (FD = 0.86 for UBFC), the cut-off was met for the general factor and the other 3 specific factors. This finding suggested that factor scores may be generated for the AES general and GN, EOF, and HH specific factors (but not for UBFC).

The Construct Replicability (H-index) estimates the degree to which a group of items represents the underlying latent factor (cut off >0.7).⁴² With one exception (H = 0.57 for UBFC), the AES general and GN, EOF, and HH-specific factors met this criterion. These results suggest that all latent factors (except the UBFC) can be considered stable and replicable across studies.

Dimensionality of AES

A high Explained Common Variance (ECV)-Index value >0.70-0.80 indicates the instrument has a strong general factor compared to specific factors.⁴² The observed value (ECV = 0.37) suggested the 17-item instrument cannot be considered sufficiently unidimensional to justify using a one-factor measurement model. Jointly with the ECV-index, the Percentage of Uncontaminated Correlations (PUC)-Index influences the parameter bias resulting from imposing a unidimensional measurement model on a multidimensional bifactor data structure.⁴² The value for the AES instrument (PUC = 0.79) stayed below the recommended cut-off. In sum, the ECV- and PUC-index findings are consistent with the poor model fit for the single-factor CFA model (M1).

Discussion

The purpose of this study was to develop and validate the AES. Of the original 37 items, 17 were retained, representing HH, EOF, UBFC, and GN. No items representing RS were retained. The AES is best conceptualized as a multidimensional measure with one general factor and specific factors representing GN, EOF, and HH. Multiple indices suggested that UBFC subscale scores are not reliable.

Although not expected, it is interesting that no RS items were retained in the final model. As with other eating behavior measures the satiety items assessed “fullness” but what made this measure unique was that the items differentiated between “comfortably full” and “too full”. It is possible these subjective terms were interpreted inconsistently by participants, as revealed in previous study on satiety measurement.⁴³ Moreover, satiety cues are influenced by context and interoceptive awareness, making them difficult to capture reliably through brief self-report items.⁴⁴

Implications for Researchers

The findings of this study have important implications for research purposes. First, the hypothesis that the AES represents a unidimensional construct was rejected. Future research aiming to replicate AES factor structure should include bifactor CFA models. Results are consistent with Swami et al (2022)⁴⁵ and Tylka et al (2024)²¹ who identified a bifactor measurement model for the Intuitive Eating Scale. In their case, a bifactor exploratory structural equation model was chosen. The findings from both studies suggest that bifactor specifications should be included in future measurement model tests of eating behavior scales.

Second, evaluating the utility of subscales in bifactor models remains a methodological challenge.^40,42 Our study suggests that the GN, HH, and EOF specific factors were reliably measured and provide unique, meaningful information beyond the general factor. Therefore, it is appropriate to use GN, HH, and EOF subscale scores independently in future research. Conversely, the UBFC factor was not reliably measured and may lack replicability. Pending cross-validation, UBFC subscale scores should not be used independently.

Third, researchers examining how AES is related to external criteria should apply analytical approaches reflecting the multidimensionality of the measure. Chen et al (2012)⁴⁶ provided an example of how general and specific latent factors from a bifactor model can be linked to external criteria in a structural regression model. For the AES, the general and 3 specific factors (excluding the UBFC specific factor) should be simultaneously embedded into such models. Because the general and specific latent factors are specified as orthogonal (uncorrelated) in the symmetrical bifactor model, they should be interpreted as distinctive constructs that may be differentially associated with external criteria. Bifactor models may also be used to test differences in latent factor means of both the general and specific factors (eg, across gender groups after measurement invariance has been established).⁴⁶ We recommend using this approach in future research because evidence of differential associations with external criteria would further bolster the viability of the GN, HH, and EOF subscales.

Implications for Theory

Wellness is achieved when humans routinely experience positive emotions, participate in meaningful activities, have positive relations, practice mindfulness, and attain goals.¹⁵ Eating that is internally regulated, enjoyable, and gently guided by nutrition, should support wellness. An expert panel of dietitians and psychologists confirmed this definition was consistent with goals they have for their patients. While our definition of adaptive eating was based on research, theory, and clinical expert feedback, the psychometric properties of the AES suggest adaptive eating may not be a valid construct.

The factor structure of the final 17 items of the AES represents a construct different from that defined initially. The items clustered more into GN, EOF, and HH categories than around the general AES construct. Twenty items (including all satiety items) from the original pool did not load significantly on any factors. Three items that were originally conceptualized to represent UBFC did not reliably share variance with each other, although they contributed to the general factor. It is possible that adaptive eating as originally defined, is not a valid construct, meaning it is not a naturally occurring phenomenon. Alternatively, it may be possible but incredibly rare for people to have the constellation of behaviors that make up adaptive eating as we defined it. Additionally, item wording may not have captured the intended meaning among this population. Future qualitative research could help identify the language for assessing adaptive eating accurately.

As independent measures, HH, EOF, and GN showed factorial and concurrent validity. In addition to being positively correlated with each other, these subscales were associated with internally regulated eating and mental health, demonstrating congruence with wellness and positive psychology. Relationships with diet quality and BMI were more variable. HH was not associated with diet quality. GN was weakly associated with poorer diet quality, and UBFC and EOF were associated with better diet quality. In today’s obesogenic society, where low-quality foods are abundant, a strong, committed approach to nutrition rather than a gentle one may be necessary to achieve a higher-quality diet. Most people crave highly palatable, high-sugar, high-fat foods; thus, it is not surprising that those who are unpreoccupied with food cravings also enjoy a higher-quality diet. The positive association between EOF and diet quality aligns with evidence that relaxed, enjoyment-based eating attitudes support greater dietary variety, higher fruit and vegetable intake, and overall higher diet quality.^47,48 For BMI, only HH was inversely associated, while GN and EOF were not. Findings from previous studies have produced variable results on the relationship between intuitive eating and dietary quality,^49-53 and mindful eating and BMI.^54,55 It seems that HH, EOF, and GN may support general wellbeing, but this may not necessarily support healthier diet quality or body composition.

Strengths and Limitations

The results of the study should be considered within the context of the strengths and limitations of the study. This study was strengthened by a multi-step theory-based process for developing the AES, a large diverse sample, and use of validated instruments. Results are generalizable to college students from multiple racial/ethnic and socioeconomic groups at large Hispanic and Asian serving universities. The participants were not chosen randomly, and the sample was made up of students enrolled in an introductory science course, which may increase selection bias. Additionally, reliance on self-reported data raises the possibility of social desirability bias.⁵⁶ Future research should aim to replicate these findings using nationally representative samples.

Conclusion

The AES is best conceptualized as a multidimensional measure with a general factor and specific factors of GN, EOF, and HH. Subscale scores for GN, EOF, and HH may be useful in future research. Further studies are needed to assess the feasibility of adaptive eating among American college students, including internally regulated eating guided by enjoyment and moderate nutritional concern. Research should also explore the role of satiety in adaptive eating and evaluate the scale’s applicability in more diverse populations beyond college students.

So What?

Understanding adaptive eating behaviors may drive innovative interventions to promote well-being beyond simply decreasing disease risk.

What is Already Known on This Topic?

Maladaptive eating behaviors increase risk for chronic disease and mental health issues.

What Does This Article Add?

This study introduces and validates the AES, a reliable tool measuring adaptive eating as internally regulated, enjoyable, and nutrition guided.

What are the Implications for Health Promotion Practice or Research?

The AES offers a valid measure for assessing positive eating behaviors and can inform interventions promoting balanced, non-restrictive eating in young adults.

Supplemental Material

Supplemental Material - Development and Validation of the Adaptive Eating Scale: Exploratory and Confirmatory Factor Analyses

Supplemental Material for Development and Validation of the Adaptive Eating Scale: Exploratory and Confirmatory Factor Analyses by Sheida Zeraattalab-Motlagh, Margit Wiesner, Cynthia Yoon, Sarah Sasson, Randi Betts, Craig Johnston and Tracey Ledoux in American Journal of Health Promotion

Footnotes

Ethical Considerations

Consent was written and collected before individuals began the study. This survey was also affirmed by the University of Houston Human Subjects Review Committee.

Consent for Publication

Consent was obtained from all study participants for publication.

Author Contributions

Sheida Zeraattalab-Motlagh: Writing – original draft, Conceptualization.

Margit Wiesner: Conceptualization, Methodology, Formal analysis, Data curation, Writing – review & editing, Writing – original draft.

Sarah Sasson: Writing – original draft, Conceptualization.

Craig Johnston: Conceptualization, Investigation.

Randi Betts: Conceptualization. Investigation.

Cynthia Yoon: Conceptualization, Investigation, Writing – review & editing.

Tracey Ledoux: Writing – original draft, Writing – review & editing, Data analysis, Supervision, Investigation, Conceptualization.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The current study’s datasets are available from the corresponding author for use and analysis.*

Supplemental Material

Supplemental material for this article is available online.

References

Gahagan

. Development of eating behavior: biology and context. J Dev Behav Pediatr. 2012;33(3):261-271.

Grimm

Steinle

. Genetics of eating behavior: established and emerging concepts. Nutr Rev. 2011;69(1):52-60.

Llewellyn

Wardle

. Behavioral susceptibility to obesity: gene–environment interplay in the development of weight. Physiol Behav. 2015;152:494-501.

Mustelin

Bulik

Kaprio

Keski-Rahkonen

. Prevalence and correlates of binge eating disorder related features in the community. Appetite. 2017;109:165-171.

Yoon

Emery

Hazzard

Mason

Neumark-Sztainer

. Adverse experiences as predictors of maladaptive and adaptive eating: findings from EAT 2018. Appetite. 2022;168:105737.

Hampton-Anderson

Craighead

. Psychosociocultural contributors to maladaptive eating behaviors in African American youth: recommendations and future directions. Am J Lifestyle Med. 2021;15(6):621-633.

Benbaibeche

Saidi

Bounihi

Koceir

. Emotional and external eating styles associated with obesity. Journal of Eating Disorders. 2023;11(1):1-7.

Bryant

Rehman

Pepper

Walters

. Obesity and eating disturbance: the role of TFEQ restraint and disinhibition. Curr Obes Rep. 2019;8:363-372.

Dakanalis

Mentzelou

Papadopoulou

, et al. The association of emotional eating with overweight/obesity, depression, anxiety/stress, and dietary patterns: a review of the current clinical evidence. Nutrients. 2023;15(5):1173.

10.

Gearhardt

Schulte

Schiestl

. Chapter 2 - food addiction prevalence: development and validation of diagnostic tools. In: Cottone

Sabino

Moore

Koob

, eds. Compulsive Eating Behavior and Food Addiction. Academic Press; 2019:15-39.

11.

Moore

Sabino

Koob

Cottone

. Chapter 4 - habitual overeating. In: Cottone

Sabino

Moore

Koob

, eds. Compulsive Eating Behavior and Food Addiction. Academic Press; 2019:83-95.

12.

Espel-Huynh

Muratore

Lowe

. A narrative review of the construct of hedonic hunger and its measurement by the power of food scale. Obes Sci Pract. 2018;4(3):238-249.

13.

Camilleri

Méjean

Bellisle

, et al. Intuitive eating is inversely associated with body weight status in the general population‐based NutriNet‐Santé study. Obesity. 2016;24(5):1154-1161.

14.

Stoewen

. Dimensions of wellness: change your habits, change your life. Can Vet J. 2017;58(8):861-862.

15.

Seligman

. Positive psychology, positive prevention, and positive therapy. Handbook of positive psychology. 2002;2(2002):3-12.

16.

Khalsa

Stough

Garr

Copeland

Kharofa

Woo

. Factor structure of the intuitive eating scale-2 among a low-income and racial minority population. Appetite. 2019;142:104390.

17.

Madanat

Hawks

Gonzales

, et al. Assessing evidence of validity for the intuitive eating scale-2 among adult Latina women. Health. 2020;12:293-304.

18.

Saunders

Nichols-Lopez

Frazier

. Psychometric properties of the intuitive eating scale-2 (IES-2) in a culturally diverse Hispanic American sample. Eat Behav. 2018;28:1-7.

19.

Anastasiades

Maiano

Argyrides

Swami

. Psychometric properties of a Greek translation of the intuitive eating scale-2 (IES-2) in adults from Cyprus. Body Image. 2022;43:348-361.

20.

Vintila

Todd

Goian

Tudorel

Barbat

Swami

. The Romanian version of the intuitive eating scale-2: assessment of its psychometric properties and gender invariance in Romanian adults. Body Image. 2020;35:225-236.

21.

Tylka

Maiano

Fuller-Tyszkiewicz

, et al. The intuitive eating Scale-3: development and psychometric evaluation. Appetite. 2024;199:107407.

22.

Choi

. Impact of stress levels on eating behaviors among college students. Nutrients. 2020;12(5):1241.

23.

Aneesh

Roy

. Eating behavior and stress levels among college students. Journal of Mental Health and Human Behaviour. 2022;27(1):60-64.

24.

Tylka

Kroon Van Diest

. The intuitive eating scale-2: item refinement and psychometric evaluation with college women and men. J Counsel Psychol. 2013;60(1):137-153.

25.

Garrow

Webster

. Quetelet's index (W/H2) as a measure of fatness. Int J Obes. 1985;9(2):147-153.

26.

Weissman

Orvaschel

Padian

. Center for epidemiological studies depression scale for children. J Nerv Ment Dis.

27.

Jackson

Cooper

Mintz

Albino

. Motivations to eat: scale development and validation. J Res Pers. 2003;37(4):297-318.

28.

Paxton

Strycker

Toobert

Ammerman

Glasgow

. Starting the conversation: performance of a brief dietary assessment and intervention tool for health professionals. Am J Prev Med. 2011;40(1):67-71.

29.

Rosenberg

. Society and the Adolescent self-image. Princeton university press; 2015.

30.

Ermakov

Ermakova

Sharifzadeh

, et al. Optical assessment of skin carotenoid status as a biomarker of vegetable and fruit intake. Arch Biochem Biophys. 2018;646:46-54.

31.

Obana

Asaoka

Miura

Nozue

Takayanagi

Nakamura

. Improving skin carotenoid levels in young students through brief dietary education using the veggie meter. Antioxidants. 2022;11(8):1570.

32.

MacCallum

Widaman

Preacher

Hong

. Sample size in factor analysis: the role of model error. Multivariate Behav Res. 2001;36(4):611-637.

33.

MacCallum

Widaman

Zhang

Hong

. Sample size in factor analysis. Psychol Methods. 1999;4(1):84-99.

34.

Fabrigar

Wegener

. Exploratory Factor Analysis. Oxford University Press; 2011.

35.

O’connor

. SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behav Res Methods Instrum Comput. 2000;32(3):396-402.

36.

Velicer

Fava

. Affects of variable and subject sampling on factor pattern recovery. Psychol Methods. 1998;3(2):231-251.

37.

Bentler

. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Model: A Multidisciplinary Journal. 1999;6(1):1-55.

38.

Kline

. Principles and Practice of Structural Equation Modeling. 4th ed. The Guilford Press; 2016.

39.

Marsh

Balla

McDonald

. Goodness-of-fit indexes in confirmatory factor analysis: the effect of sample size. Psychol Bull. 1988;103(3):391-410.

40.

Reise

Bonifay

Haviland

. Scoring and modeling psychological measures in the presence of multidimensionality. J Pers Assess. 2013;95(2):129-140.

41.

Grice

. Computing and evaluating factor scores. Psychol Methods. 2001;6(4):430-450.

42.

Rodriguez

Reise

Haviland

. Applying bifactor statistical indices in the evaluation of psychological measures. J Pers Assess. 2016;98(3):223-237.

43.

Herbert

Blechert

Hautzinger

Matthias

Herbert

. Intuitive eating is associated with interoceptive sensitivity. Effects on body mass index. Appetite. 2013;70:22-30.

44.

Van Dyke

Drinkwater

. Relationships between intuitive eating and health indicators: literature review. Public Health Nutr. 2014;17(8):1757-1766.

45.

Swami

Maïano

Furnham

Robinson

. The intuitive eating scale-2: re-evaluating its factor structure using a bifactor exploratory structural equation modelling framework. Eat Weight Disord. 2022;27(4):1349-1357.

46.

Chen

Hayes

Carver

Laurenceau

Zhang

. Modeling general and specific variance in multifaceted constructs: a comparison of the bifactor model to other approaches. J Pers. 2012;80(1):219-251.

47.

Lohse

Bailey

Krall

Wall

Mitchell

. Diet quality is related to eating competence in cross-sectional sample of low-income females surveyed in Pennsylvania. Appetite. 2012;58(2):645-650.

48.

Tabatabai

Holland

Curtis

Morris

. The relationship between intuitive eating and diet quality in a college population. California Journal of Health Promotion. 2021;19(1):34-43.

49.

Banks

. Nutritional Analyses of Intuitive Eaters as Compared to Dieters. Utah State University; 2008.

50.

Cole

Horacek

. Effectiveness of the my body knows when intuitive-eating pilot program. Am J Health Behav. 2010;34(3):286-297.

51.

Hawley

Horwath

Gray

, et al. Sustainability of health and lifestyle improvements following a non-dieting randomised trial in overweight women. Prev Med. 2008;47(6):593-599.

52.

Madden

Leong

Gray

Horwath

. Eating in response to hunger and satiety signals is related to BMI in a nationwide sample of 1601 mid-age New Zealand women. Public Health Nutr. 2012;15(12):2272-2279.

53.

Lopez

Hernandez

Bode

Ledoux

. A complex relationship between intuitive eating and diet quality among university students. J Am Coll Health. 2021;71:1-7.

54.

Minari

Araújo-Filho

Tácito

LHB

, et al. Effects of mindful eating in patients with obesity and binge eating disorder. Nutrients. 2024;16(6):884.

55.

Moor

Scott

McIntosh

. Mindful eating and its relationship to body mass index and physical activity among university students. Mindfulness. 2013;4(3):269-274.

56.

Brewis

Brennhofer

van Woerden

Bruening

. Weight stigma and eating behaviors on a college campus: are students immune to stigma's effects? Prev Med Rep. 2016;4:578-584.