Psychometric properties and factor structure of the Diabetes Eating Problem Survey Revised in Danish adolescents with type 1 diabetes

Abstract

Aim

The aim of this study was to examine the psychometric properties and factor structure of the Danish translation of Diabetes Eating Problem Survey – Revised (DEPS‐R) among adolescents with type 1 diabetes (T1D).

Methods

A total of 131 adolescents (53% women) with T1D aged 11–19 years completed DEPS‐R and Youth Eating Disorder Examination Questionnaire (YEDE‐Q). Additional anthropometric, biochemical, and medical data were obtained from medical records. Exploratory Factor Analysis was performed to examine the factor structure of DEPS‐R.

Results

The DEPS‐R demonstrated good internal consistency (Cronbach α = 0.87) and was significantly correlated with YEDE‐Q (r = 0.83; p < 0.01), HbA1c (r = 0.35; p < 0.01), zBMI (r = 0.35; p < 0.01), and age (r = 0.20; p < 0.05), indicating high convergent validity. The median DEPS‐R score was 10 [5–18] for the entire sample; however, it was significantly higher among women (15 [9–21]) than men (6 [4–10]). Factor analysis revealed a three‐factor structure accounting for 58% of the variance and indicated that item 4 contributed less strongly to the overall score.

Conclusions

This study supports the validity of DEPS‐R as a screening tool for detecting the risk of disordered eating in adolescents with T1D but challenges the relevance of item 4. Future research is needed on updating the original DEPS‐R, including a sensitivity analysis of DEPS‐R in detecting the risk of serious eating disorder‐related behaviour in T1D.

What's new?

• The DEPS‐R is a widely used, diabetes‐specific screening tool for identifying the risk of disordered eating, with evidence supporting its reliability and structural validity in various populations.

• This study identified a three‐factor structure in an adolescent sample and found that item 4 may not contribute meaningfully to the scale.

• Findings suggest that the DEPS‐R may benefit from flexible, context‐specific interpretation and question the relevance of item 4 in an adolescent population with type 1 diabetes.

1. INTRODUCTION

People living with type 1 diabetes (T1D) have a 2–3‐fold increased risk of disordered eating (DE) and eating disorders (ED), ¹ , ² , ³ associated with suboptimal metabolic regulation and diabetes‐related complications. ⁴ , ⁵ Proposed contributing factors are the necessary focus on carbohydrate intake, increased body weight, and the repeated weighing at clinic visits. ⁶ , ⁷ , ⁸

The severe medical complications of ED in people with T1D emphasize the importance of systematic screening for ED symptoms. ⁹ However, generic screening tools might not be suitable for people with T1D as these do not include insulin restriction, nor account for the dietary focus that is an intrinsic part of diabetes self‐management. Hence, international guidelines recommend assessing the risk of DE with diabetes‐specific tools such as the Diabetes Eating Problem Survey – Revised (DEPS‐R). ¹⁰ , ¹¹

Although the reliability and validity of the DEPS‐R have been examined in several countries with acceptable results, ¹² , ¹³ , ¹⁴ findings on its structural validity have been inconsistent. The original DEPS‐R was primarily developed as a single‐factor scale when revised from DEPS to DEPS‐R. ¹⁵ Studies including factor analysis, such as the exploratory factor analysis (EFA) by Wisting et al., ¹⁶ and subsequent confirmatory factor analysis (CFA) in the Greek, Italian, Chinese, and Arab versions, ¹² , ¹⁷ , ¹⁸ , ¹⁹ have organized DEPS‐R into three factors referred to as ‘Maladaptive eating habits’, ‘Preoccupation with thinness’ and ‘Maintaining high blood glucose to lose weight/Concerned with desire to maintain high blood glucose’. A French and Spanish validation study, including an adult T1D population, has not been able to confirm the factor structure reported by Wisting et al. ¹⁴ , ²⁰ The French study reported a four‐factor structure including ‘Self‐control’, ‘Weight loss concerns’, ‘Diabetes medication’, and ‘Compensatory behaviors’, and the Spanish version yielded five factors including ‘Food attitudes’, ‘Bulimic behaviors’, ‘Weight control’, ‘Avoidance’, and ‘Restriction’. ¹⁴ , ²⁰

The present study aims to explore the psychometric properties of the DEPS‐R, focusing on the structural validity, but also including internal reliability and convergent validity, compared with a widely used generic ED questionnaire in Danish adolescents with T1D.

2. METHODS AND MATERIALS

2.1. Study design

A cross‐sectional survey design was applied.

2.2. Participants

Inclusion criteria were a diagnosis of T1D with a duration of >1 year, age 11–19 years, and treated in Denmark Central Region. A total of 580 adolescents were identified and contacted by secure email, including information about the project and a possibility for digital consent between June 2021 and October 2022. Questionnaires were distributed by and stored in a secure RedCap.org database hosted by Aarhus University.

2.3. Measures

DEPS‐R is a 16‐item diabetes‐specific tool designed to assess risk of DE behaviours among individuals living with T1D. ¹⁵ Answers are scored on a Likert scale from 0 to 5 points, with higher scores indicating a stronger risk of ED pathology. Authors of the DEPS‐R have suggested a cut‐off of ≥20 points, implying the necessity for further investigation of ED symptoms. ¹⁵

The Danish translation of DEPS‐R followed the forward–backward procedure for translation of questionnaires. ²¹ DEPS‐R was initially translated into Danish by two different health professionals, experienced in both diabetes and psychology, separately. Translations were compared and a consensus version was agreed upon, which was subsequently back translated by an independent bilingual third party. Based on semi‐structured interviews, using a cognitive interviewing technique, the Danish version of the DEPS‐R has previously showed an appropriate level of understanding. ²² Permission for the Danish translation and its use was obtained from the developer of the English version of DEPS‐R, original author Prof. Dr. Lisa Volkening, Associate Director of Pediatric Research Programmes at Joslin Diabetes Centre.

The Youth Eating Disorder Examination Questionnaire (YEDE‐Q) is a youth adapted version of the Eating Disorder Examination Questionnaire (EDE‐Q). ²³ YEDE‐Q consist of four subscales (Restraint, Eating Concern, Shape Concern, Weight Concern) and a global total score, enabling comparison across countries and providing a common language for assessing ED psychopathology. The adaptation to a youth format consists of a modified language, an illustrative introduction to the concept of ‘overeating’, and a recurrent reminder of the timeframe. The Danish version of YEDE‐Q was based on the Danish validated EDE‐Q and the English YEDE‐Q. The translational process was similar to the aforementioned translation of the DEPS‐R and included a native English adolescent psychiatrist in the back‐translation. For this study, the YEDE‐Q was adapted for diabetes by adding one additional question involving insulin restriction with inspiration from d'Emden and colleagues ²⁴ and Meltzer and colleagues ²⁵ ; ‘I take less insulin than I should to influence my shape or weight’.

Body mass index (BMI) is calculated based on weight and height (kg/m²) and standardized to a z‐score according to age and sex (zBMI) according to the World Health Organization ²⁶ using WHO AntroPlus App version 1.0.4. All participants were divided into three groups with the following cut‐offs: Underweight, <−1 standard deviation (SD), Normal weight, ≥−1 SD ≤ +1 SD, and Overweight/obesity, >+1 SD.

Haemoglobin A1c (HbA1c) refers to glycated haemoglobin to measure glycaemic management. ²⁷

Additionally, data such as age at diabetes onset, diabetes treatment (pen or pump), additional diagnoses, both somatic (e.g., autoimmune disease) and psychiatric (e.g., depression, anxiety), were retrieved from medical records.

2.4. Statistical analysis

Statistical analyses were conducted using Stata ver.18.0 (StataCorp LLC, USA). Continuous variables were expressed as mean ± standard deviation (SD) or as median with interquartile range (IQR). Categorical or binary data were reported as frequencies or percentages. Group differences were explored using Student's t‐test for continuous parametric data, Chi‐square for binary data, and Wilcoxon rank‐sum test for non‐parametric data. p‐values below or equal to 0.05 were interpreted as significant.

Internal consistency of DEPS‐R was determined using Cronbach alpha (α) coefficient. Levels above 0.5 were interpreted as acceptable and levels above 0.8 were considered as very good. ²⁸ Correlation analyses between DEPS‐R and HbA1c, sex, zBMI, and age were conducted using the Spearman correlation coefficient as DEPS‐R represents an ordinal category. Correlations between 0.10 and 0.29 were interpreted as small, 0.30–0.49 as moderate, and >0.50 as large. ²⁹

Sampling adequacy for factor analysis was evaluated using the Kayser–Meyer–Olkin (KMO) measure, with values 0.80–0.89 interpreted as meritorious. Bartlett's test of specificity was performed with statistical significance (<0.01), indicating sufficient inter‐item correlations to proceed. Structural validity was examined in two stages. First, a CFA was conducted to test the three‐factor structure found by Wisting et al. ³⁰ Goodness of fit statistical tests were conducted in R version 4.1.1 as lavaan 0.614 test. Model fit was evaluated using the following commonly accepted thresholds: a root mean square error of approximation (RMSEA) less than 0.05, a comparative fit index (CFI) greater than or equal to 0.95, a standardized root mean square residual (SRMR) close to 0.08, and a Tucker–Lewis index (TLI) greater than or equal to 0.95. ³¹ Our fit statistics showed that only CFI was fulfilled, and therefore, the CFA was considered inappropriate (RMSEA = 0.081, CFI = 0.95, SRMR = 0.112, TLI = 0.946). An EFA was then performed, using principal axis factoring (PAF) as the extraction method, which is appropriate for identifying latent constructs based on shared variance among items. Oblique quartimin rotation was applied, allowing factors to correlate as this delivers a more accurate solution compared with the orthogonal rotation. ³² Factors with eigenvalues >1 were retained based on a scree plot. ³³ A solid factor includes ≥3 items, preferably ≥5 items, and items with loadings >0.50 were considered strong, while loadings ≤0.32 were considered weak. Likewise, cross‐loadings should be accounted for if an item loads at ≥0.32 or higher on two or more factors. ³⁴

Convergent validity was assessed in an exploratory analysis using Spearman's rank correlations between the DEPS‐R total score and theoretically related variables: YEDE‐Q global score, HbA1c, age, and zBMI.

2.5. Ethical considerations

Participants were provided written information about the study in preparation. The consent included parental consent for those <15 years. Participants were given time to consider participation. All data were handled in accordance with the General Data Protection Regulation (GDPR). The study was registered at The Science and Legal Office, Central Denmark Regional Office (case number: 1‐45‐70‐8‐21). The Committee on Health Research Ethics evaluated the project as not needing additional specific approval according to Danish legislation.

3. RESULTS

3.1. Participant characteristics

The final sample consisted of 131 adolescents (69 women, 62 men) with T1D (response rate 22.6%) with a mean age of 15.3 (±2.2) years. Baseline characteristics of the sample have been partially reported previously. ³⁵ In this article, we present an abridged version of these characteristics for completeness and context, with emphasis on variables relevant to the current research question (Table 1).

TABLE 1.

Participant characteristics.

	All N = 131 (%)	Males N = 62	Women N = 69	p‐Values
DEPS‐R median score, (range 0–80) [IQR]	10 [5–18]	6 [4–10]	15 [9–21]	<0.001
DEPS‐R ≥ 20	28 (21%)	6 (10%)	22 (32%)	<0.001
Age (years)	15.3 ± 2.2	15.1 ± 2.2	15.5 ± 2.2	0.357
HbA1c (N = 128*)	54.4 ± 9.7	53.3 ± 9.3	55.4 ± 10	0.230
zBMI (N = 127*)	0.5 ± 1.1	0.3 ± 1.2	0.6 ± 1.0	0.083
Diabetes duration (years) (N = 129*)	6.1 ± 3.9	5.5 ± 3.8	6.6 ± 4.0	0.110
Insulinpump (N = 126*), N (%)	104 (83%)	49 (82%)	55 (83%)	0.806
Psychiatric co‐morbidity (N = 126*), N (%)	20 (16%)	5 (8%)	15 (23%)	<0.001
Somatic co‐morbidity (N = 126*), N (%)	20 (16%)	7 (12%)	13 (20%)	0.166
Other medication** (N = 124*), N (%)	53 (43%)	20 (33%)	33 (50%)	0.042

Note: Data are mean ± SD, except as noted. p‐values refer to the significance of the difference between men and women. N varies due to missing records.

As shown in Table 1, 21% of participants scored ≥20 on DEPS‐R, with more women than men scoring above this cutoff (32% vs. 10%) (p < 0.001). The median DEPS‐R score was significantly higher among women (15 [9–21]) than men (6 [4–10]) (p < 0.001). We found no significant differences between men and women regarding age (p > 0.357), diabetes regulation (p > 0.230), zBMI (p > 0.083), or diabetes duration (p > 0.110). However, more women than men had a psychiatric disorder (p < 0.001), such as hyperactivity, anxiety, depression, obsessive‐compulsive disorder, and post‐traumatic stress disorder, based on their medical chart.

3.2. Internal consistency

In the total sample, Cronbach α for the DEPS‐R was 0.87 (α = 0.84 for women and α = 0.88 for men). Stratifying the sample by age showed decreased internal reliability in younger adolescents (<15 years) compared with older (≥15 years) (α = 0.75 vs. 0.87). The Cronbach α of the YEDE‐Q global score was 0.96 across the sample and revealed minor variation within subgroups (women α = 0.96; men α = 0.91; ≥15 years α = 0.97; <15 years α = 0.94).

3.3. Normative data

As presented in Table 2, 60% of the sample presented with a body weight within the normal range, 32% with overweight/obesity, and 7% with underweight. DEPS‐R scores were higher in women than in men across all BMI groups. DEPS‐R scores increased with age, especially in women, and differences between sexes became more pronounced and statistically significant from age 14 onwards. The median DEPS‐R score was 19.5 [16–24] (n = 12) among women with overweight/obesity between 17 and 19 years.

TABLE 2.

DEPS‐R scores according to sex, age, and zBMI.

	N*	All	Males	Women	p‐Values
zBMI	127
Underweight	9	8 [4–10]	7 [3–8]	16 [12–20]	0.0396
Normal weight	77	8 [4–17]	5.5 [2–8.5]	14 [7–20]	<0.001
Overweight/obesity	41	13 [7–21]	8.5 [5–11]	17 [13–23]	<0.001
Age (years)	131
11–13	23	8 [2–18]	7.5 [2–18]	12 [7–15]	0.4110
14–16	53	9 [5–15]	5 [1–8]	13 [7–17]	<0.001
17–19	55	11 [6–21]	6.5 [4–10]	18 [12–25]	<0.001

Note: Data are median [IQR] of DEPS‐R score categorized by sex, age, and BMI group classification (underweight = zBMI < −1 SD, normal weight = zBMI ≥ −1 < 1 SD, overweight/obese ≥1 SD). p‐values reported using the Wilcoxon rank‐sum test. N varies due to missing records.

3.4. Convergent validity

There was a significant correlation between DEPS‐R total score and YEDE‐Q global score, HbA1c, zBMI, and age for the total sample. Overall, correlations were stronger for women compared with men (Table 3).

TABLE 3.

Correlation between DEPS‐R and clinical variables.

	All	Women	Males
YEDE‐Q global score (N = 131)	0.83 (p < 0.01)	0.85 (p < 0.01)	0.69 (p < 0.01)
HbA1c (N = 126)	0.35 (p < 0.01)	0.40 (p < 0.01)	0.26 (p = 0.04)
zBMI (N = 127)	0.35 (p < 0.01)	0.24 (p = 0.05)	0.40 (p < 0.01)
Age	0.20 (p = 0.02)	0.34 (p < 0.01)	0.07 (p = NS)

Note: Spearman correlation coefficient.

Abbreviations: DEPS‐R, Diabetes Eating Problem Survey – Revised; YEDE‐Q, Youth Eating Disorder Examination Questionnaire.

3.5. Structural validity

The overall KMO value was 0.81, indicating meritorious sampling adequacy. All individual item KMO values exceeded 0.70, supporting retention of all items in the factor analysis. PAF revealed four factors with an eigenvalue of >1 explaining 64% of the variation. This was confirmed in a scree plot of the data. These four factors were retained in the preliminary analyses. Table 4 presents factor loadings after rotation, percentages of accumulated variance, uniqueness, Cronbach α for each subscale, eigenvalues, and interpreted concepts of factors. For clarity, the full wording of each DEPS‐R item is presented in Table 4, where items are listed according to their primary factor loadings rather than numerical order.

TABLE 4.

Exploratory factor loading matrix.

	Exploratory factor analysis
	1	2	3	4
Cronbach α	0.83	0.80	0.77	–
Eigenvalues	5.99	2.20	1.29	1.29
Item					Uniqueness	Acc. variance	Concepts
1: Losing weight is an important goal to me	0.86	−0.01	−0.01	−0.0	0.27	–
6: I feel that it's difficult to lose weight and control my diabetes at the same time	0.81	−0.15	0.05	−0.06	0.34	–
16: I would rather be thin than have good control of my diabetes	0.71	0.11	0.15	−0.25	0.37	–
2: I skip meals and/or snacks	0.55	−0.07	(0.35)	0.05	0.45	–
5: I eat more when I am alone than when I am with others	0.50	0.17	−0.17	(0.47)	0.41	–
15: I alternate between eating very little and eating huge amounts	0.49	0.07	0.06	(0.43)	0.41	32%	Eating and weight concerns
10: I try to eat to the point of spilling ketones in my urine	−0.13	0.91	−0.06	−0.02	0.25	–
9: I try to keep my blood sugar high so that I will lose weight	0.15	0.88	−0.07	−0.06	0.20	–
13: After I overeat, I skip my next insulin dose	−0.19	0.79	0.20	0.15	0.25	–
8: I make myself vomit	0.20	0.59	0.24	(−0.37)	0.32	50%	Purging behaviours
7: I avoid checking my blood sugar when I feel like it is out of range	0.11	−0.1	0.79	−0.11	0.36	–
12: Other people tell me to take better care of my diabetes	−0.09	0.10	0.75	0.21	0.33	–
11: I feel fat when I take all my insulin	0.22	0.21	0.50	0.03	0.47	–
3: Other people have told me that my eating is out of control	0.08	(0.36)	0.43	0.23	0.41	58%	Diabetes care and concerns
4: After I overeat, I do not take enough insulin to cover the food	−0.04	−0.09	0.16	0.77	0.39	–	–
14: I feel that my eating is out of control	(0.43)	(0.37)	0.00	0.48	0.23	64%	–

Note: Factor loadings ≥0.50 are displayed in italic. Cross‐loadings ≥0.32 are shown in parentheses. Items are ordered according to their highest primary loading. Variance values listed represent the cumulative variance explained after adding each factor to the preceding factors.

Reported uniqueness is relatively low (<0.6), which is interpreted as the variables being well explained by the corresponding factor (Table 4). With reference to Costello et al. ³² a factor requires >3 items; thus, we decided to omit factor 4, which only contains two items (4 and 14), still maintaining 58% of accumulated variance. Item 4 is omitted in the final three‐factor model after including item‐total correlation analysis, which revealed a low range moderate correlation to the total DEPS‐R score for item 4 (r = 0.34) and small correlations to the YEDE‐Q global score (r = 0.27). Additionally, item 4 has a very similar phrasing to item 13. Item 14 strongly cross‐loads on both factor 1, 2, and 4; thus, we have decided to include item 14 in factor 1 due to its connection with the concept of ‘eating concern’. Items 2, 5, 15, 8, 3, and 14 had cross‐loadings ≥0.32, as indicated by parentheses. Internal reliability was re‐established on the remaining 15 items (α = 0.88).

The final three‐factor structure of the Danish DEPS‐R includes factor 1 ‘Eating and Weight Concerns’, factor 2 ‘Purging Behaviors’ and factor 3 ‘Diabetes Care and Concerns’.

Associations between all factors and YEDE‐Q global score were significant (Table 5). The strongest associations were found between factor 1 in DEPS‐R and YEDE‐Q global score; oppositely, the weakest associations were found between factor 2 in DEPS‐R and the YEDE‐Q global score.

TABLE 5.

Associations between the three‐factor structure of the Danish DEPS‐R, clinical variables, and DE behaviour.

DEPS‐R	Factor 1 eating and weight concerns	Factor 2 purging behaviour	Factor 3 diabetes care and concerns
HbA1c (n = 126)	0.28 (p < 0.01)	0.17 (p < 0.05)	0.39 (p < 0.01)
zBMI (n = 127)	0.43 (p < 0.01)	0.08 (p = NS)	0.15 (p < 0.05)
Age	0.21 (p = 0.01)	0.24 (p < 0.01)	0.09 (p = NS)
Diabetes duration (n = 129)	0.21 (p = 0.02)	0.22 (p = 0.01)	0.17 (p = 0.05)
YEDE‐Q
YEDE‐Q global score	0.87 (p < 0.01)	0.50 (p < 0.01)	0.60 (p < 0.01)
YEDE‐Q subscales
Eating concern	0.78 (p < 0.01)	0.55 (p < 0.01)	0.58 (p < 0.01)
Weight concern	0.83 (p < 0.01)	0.48 (p < 0.01)	0.55 (p < 0.01)
Shape concern	0.84 (p < 0.01)	0.47 (p < 0.01)	0.58 (p < 0.01)
Restraint	0.73 (p < 0.01)	0.46 (p < 0.01)	0.47 (p < 0.01)
ED characteristics
Purging behaviour a	0.19 (p = 0.03)	0.31 (p < 0.01)	0.16 (p = NS)
Binge eating b	0.43 (p < 0.01)	0.40 (p < 0.01)	0.35 (p < 0.01)
Restraint c	0.41 (p < 0.01)	0.36 (p < 0.01)	0.31 (p < 0.01)
Excessive exercise	0.12 (p = NS)	0.13 (p = NS)	0.07 (p = NS)
ED behaviour items d	0.60 (p < 0.01)	0.41 (p < 0.01)	0.37 (p < 0.01)

Note: Spearman correlation coefficient.

^a Purging behaviour includes vomiting, IR, and laxative misuse.

^b Binge Eating includes both objective and subjective binge eating.

^c Restraint only includes item 2 in YEDE‐Q.

^d ≥1 ED behaviour: >8 h fasting, vomiting, excessive exercise, insulin misuse, laxative misuse.

All three factors in the Danish version of DEPS‐R correlated with the four subscales of YEDE‐Q; however, the strongest correlations were with factor 1. Associations were also seen with subcategories from YEDE‐Q such as binge eating, restraint eating, and ED behaviours, but not with excessive exercise. Factor 1 and 2, but not 3 in DEPS‐R, also correlated with purging behaviour measured by YEDE‐Q. Further, the three‐factor structure in DEPS‐R showed a moderate to strong correlation to HbA1c and diabetes duration. Factor 1 was positively associated with zBMI as the sole factor, whereas factor 2 was the sole factor to significantly correlate with age. Diabetes duration also showed an association, however weak, towards all three factors in DEPS‐R (Table 5).

4. DISCUSSION

This is the first study to explore the psychometric properties and structural validity of the Danish version of DEPS‐R, extending the validation of DEPS‐R to include Danish adolescents with T1D.

The Danish version of DEPS‐R demonstrated satisfactory psychometric properties and good internal consistency in alignment with the original DEPS‐R validation and results obtained in several other validation studies. ¹² , ¹³ , ¹⁴ , ¹⁶ , ¹⁸ , ²⁰ , ³⁶ Internal reliability in this present study is high; however, it is lower for participants ≤15 years. Part of this may be explained by the fact that ED symptoms typically emerge and intensify during adolescence, and DEPS‐R scores have been shown to increase with age. ³⁷ However, certain items might be less easily interpreted by younger adolescents, which could lead to falsely lower or higher DEPS‐R scores in the younger age groups. ²²

In accordance with other studies, we found significantly higher DEPS‐R scores among women compared with men, and a positive association between DEPS‐R, HbA1c, age, and zBMI. ¹⁷ , ³⁰ The higher DEPS‐R scores among women, particularly older women with overweight or obesity, align with evidence linking sex and weight status to DE in adolescents with T1D. ³⁷ This pattern likely reflects cumulative societal and diabetes‐specific pressures, underscoring the need for targeted screening in this high‐risk group. ³⁰

In 21% of our sample, the DEPS‐R score was above cut‐off, which is comparable to a previous Danish prevalence study ³⁸ ; however, it is slightly higher than in other northern European countries. In a Norwegian study, 18.7% scored above cut‐off (mean age 14.6 years old), ³⁰ and in a similar German study, 15.4% (mean age 15.8 years old) scored above cut‐off. ³⁶ A strength of both our study and the German and Norwegian studies is the balanced distribution of male and female participants. Most studies also show a positive association between age and DEPS‐R score, which in this case does not solely appear to be the explanation for the different prevalences of adolescents above cut‐off. It is likely that both temporal factors (e.g., shifts in social norms or diabetes treatment modalities) and cultural context may influence how individuals respond to the DEPS‐R; however, these influences have not yet been thoroughly examined in the literature.

Our results support a multidimensional factor structure of the DEPS‐R, consistent with findings from previous studies. ¹² , ¹³ , ¹⁶ , ¹⁷ , ³⁹ We identified a three‐factor model including ‘Eating and Weight Concerns’, ‘Compensatory Behaviors’, and ‘Diabetes Care and Concerns’, all capturing distinct patterns of DE in adolescents with T1D. The first factor reflects body dissatisfaction and weight‐related preoccupations, the second involves intentional insulin omission and other compensatory strategies, while the third centres on diabetes management avoidance and external concern from others. To the best of our knowledge, no previous factor analysis has included the concept of ‘Diabetes Care and Concern’, which may lead to a potential overestimation of the prevalence of those scoring above the cut‐off. Interestingly, we found that this factor was significantly correlated with the YEDE‐Q global score, which could indicate a broader emotional burden related to diabetes, linked to DE behaviour driven through mechanisms such as diabetes distress or feelings of frustration around diabetes self‐management. This factor may not directly indicate ED behaviours but could contribute to the psychosocial context in which such behaviours develop. A previous study reported that diabetes distress significantly predicted DE in adolescents with T1D. ⁴⁰

The present study suggests carefully considering the relevance of item 4 in DEPS‐R due to low loadings and cross‐loadings. A potential elimination of item 4 should prompt an investigation of a new cut‐off, in alignment with the Dutch version of DEPS‐R, where item 10 is omitted. ⁴¹ The emergent factor structure in this study may hold relevance for subsequent psychometric studies exploring the subscale scoring, as this could guide the development of clinical intervention strategies, informed by the specific domains in which individuals exhibit elevated responses.

Originally DEPS‐R was validated among a sample (n = 113) of 13–19‐year‐olds with T1D, without a comparison tool. ¹⁵ Several other studies have since then investigated the psychometric properties of DEPS‐R ³⁶ ; however, none have administered DEPS‐R with a comparison questionnaire such as YEDE‐Q in the adolescent population. Our study found a strong positive correlation between the DEPS‐R and the YEDE‐Q, supporting the validity of DEPS‐R as a screening tool.

Importantly, our findings also reflect recent concerns regarding the application of a fixed DEPS‐R cut‐off score across DE subtypes. A recent study by Moskovich et al. highlighted that while a DEPS‐R score ≥ 20 demonstrated good overall accuracy (81.8%) and specificity (88.2%) in adults with T1D, sensitivity was lower (76.7%), especially among individuals presenting with binge eating disorder or purging disorder. ⁴² In contrast, individuals with bulimia nervosa spectrum disorders were identified with greater accuracy, likely due to their engagement in both bingeing and purging behaviours that elevate scores across multiple DEPS‐R items. This uneven representation of DE behaviours means that DEPS‐R may not fully capture the spectrum of DE presentations, and its limited focus on restriction may reduce its ability to identify certain profiles. Specifically, DEPS‐R includes only one item that directly assesses restrictive eating (item 2), which may limit its sensitivity for detecting individuals whose primary presentation involves restriction. The general debate regarding the appropriateness of the ≥20 cut‐off is particularly important when considering its application in routine clinical screening.

This study was not without limitations. The modest sample size and potential selection bias in the recruitment are challenges in the normative data, but also in the output of the factor analysis. To correctly estimate the proportions of adolescents with T1D scoring ≥20 in DEPS‐R, knowledge of the prevalence among the non‐responders is crucial. The relatively low response rate of ≈22% despite several reminders is a common challenge in questionnaire surveys. Also, the study did not include a diagnostic interview such as the EDE to assess the presence of EDs, so we could not estimate the sensitivity and specificity of the DEPS‐R; however, our findings are comparable to those of Moskovich et al. ⁴² Furthermore, the present study did not include individuals above the age of 19, which limits the generalizability of the results across age groups. This is particularly relevant since the prevalence of insulin restriction, omission, and overeating appears to increase with age. ⁴³ These findings underscore the need for nuanced interpretation of DEPS‐R scores, particularly in adolescent populations where symptom presentations may differ. Our findings similarly suggest that while the ≥20 threshold may be useful as a general indicator of elevated DE risk, it should not be used in isolation; instead, clinicians may benefit from reviewing specific item responses or patterns, such as frequent insulin omission or restrictive eating, to inform treatment. Although our EFA suggested a three‐factor structure, the DEPS‐R was originally developed and is most commonly used as a unidimensional screening tool, with the total score serving as the primary indicator of DE risk. Because the factor structure identified in our study is preliminary, we considered it premature to interpret or report factor‐level scores as distinct constructs. We therefore restricted our convergent validity analyses to the total DEPS‐R score, which allows for direct comparability with previous validation studies and avoids attributing substantive meaning to unconfirmed subscales. Future studies with larger samples and confirmatory approaches should examine convergent and discriminant validity at both the total‐score and factor‐score levels once a stable factor structure is established.

This study contributes to the growing body of evidence indicating that while the DEPS‐R is reliable and structurally sound, its interpretation may need to be flexible and individualized based on clinical presentation and population. Factor structures should be interpreted with caution, as no single structure appears consistent across countries or age groups. The study adds to the field by providing further evidence of the reliability of DEPS‐R, including a preliminary three‐factor structure of DEPS‐R in Danish adolescents with T1D.

AUTHOR CONTRIBUTIONS

CBA was lead researcher on this study and produced the first draft of the paper with LC. LC, DDP, JMB, KK, RKS, and ETV participated in the discussion of findings and revised the manuscript. All participated in the final draft.

FUNDING INFORMATION

This work was supported by a research grant from the Danish Diabetes Academy, which is funded by the Novo Nordisk Foundation, grant number NNF17SA0031406. Steno Diabetes Centre Aarhus and Danish National Centre for Obesity also financially supported this work. None of these had an influence on the results or the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors have no relevant financial interests to disclose.

Abild CB, Vestergaard ET, Pauls DD, et al. Psychometric properties and factor structure of the Diabetes Eating Problem Survey Revised in Danish adolescents with type 1 diabetes. Diabet Med. 2026;43:e70158. doi: 10.1111/dme.70158

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.