Table 1 Mean scores of the EDI subscales for anorexia nervosa, bulimia nervosa, EDNOS and female students
EDI subscales
Anorexia Nervosa,
restricting
Anorexia Nervosa,
binge/purging
Bulimia Nervoas ED, not otherwise
specified
Female students
(n = 35)
(n = 76)
(n = 349)
(n = 91)
(n = 751)
Drive for Thinness
∗∗∗7.4 ± 5.9
a11.8 ± 6.0
b13.6 ± 4.4
b11.9 ± 4.6
b6.3 ± 5.0
cBulimia
∗∗∗2.2 ± 3.6
a12.2 ± 7.5
b14.7 ± 4.8
c10.7 ± 6.1
b2.2 ± 3.5
aBody Dissatisfaction
∗∗∗10.5 ± 6.6
a12.5 ± 5.8
a19.0 ± 6.7
b18.5 ± 6.8
b13.1 ± 7.1
aIneffectiveness
∗∗∗10.7 ± 7.3
a14.9 ± 6.9
b14.3 ± 7.3
b12.7 ± 8.0
b5.2 ± 4.9
cPerfectionism
∗∗∗6.3 ± 4.3
a,b8.7 ± 5.1
a8.3 ± 4.6
a7.4 ± 5.3
a4.0
± 3.6
bInterpersonal Distrust
∗∗∗4.2 ± 4.2
a,b6.1 ± 4.0
a4.9 ± 3.4
a5.4 ± 4.3
a3.7
± 3.3
bIntroceptive Awareness
∗∗∗7.0 ± 6.7
a13.0 ± 6.6
b13.5 ± 7.0
b11.1 ± 6.6
b,c7.4
± 7.2
aMaturity Fear
∗∗∗9.0 ± 5.3
a,b12.2 ± 5.7
a9.5 ± 5.3
b,c8.8 ± 5.6
c,d8.5
± 5.2
dEDI Total
∗∗∗57.3 ± 31.5
a91.3 ± 29.0
b97.8 ± 28.1
b86.4 ± 30.7
b64.9
± 5.8
c
IntroductionThe Eating Disorder Inventory (EDI-1) is a widely used self-reported
measure of symptoms associated with anorexia nervosa (AN) and bulimia nervosa (BN). Previous factor analytic work on the EDI supported the
first-order 8-factor measurement structure among clinical patients, but it appeared less consistent in nonclinical samples. In this study, we aimed to examine the psychometric properties of the Mandarin Chinese version of the EDI (C-EDI) among both patient and non-patient groups.
Methods
The EDI-1 consists of three subscales, Drive for Thinness (DT), Bulimia (B), and Body Dissatisfaction (BD), assessing attitudes and behaviors about eating, weight, and shape and five subscales,
Perfectionism (P), Ineffectiveness (IE), Interpersonal Distrust (ID),
Interoceptive Awareness (IA), and Maturity fear (MF), for measuring psychological functioning. A total of 551 female patients with eating
disorders and a group of 751 nursing college students completed the C-EDI. Patients were diagnosed as having AN (n = 111), BN (n = 349), or ED not otherwise specified (EDNOS) (n = 91). Both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were conducted to examine the construct validity of the C-EDI in both groups. EFA was performed by a forced eight-factor solution using principal axis
factoring with promax rotation. CFA was examined by structural
equation modeling with maximum likelihood method using LISREL 8.71 (SSI Inc., Chicago, IL, USA).
Results
The values of the alpha coefficients were larger than 0.80 for B, BD, IE, and IA subscales and near or above 0.70 for the remaining subscales in the patient group, indicating acceptable level of internal consistency in patient group. Using 0.3 as the cutoff, the item-subscale correlations were
acceptable for the eight subscales with the exception of a few items (1, 6, 11, and 57). The values of the alpha coefficient in the non-patient group were comparable to the patient group except ID (Cronbach’s α = 0.56), and more items with item-subscale correlations lower than 0.3 (1, 6, 26, 30, 34, 53, 54, and 57) were found. One-way ANOVA showed significant differences of 8 subscale scores between 3 patient groups and female
students. Post-hoc comparison revealed that subscale scores of patient
group were significantly higher than those of non-patient group except IA of AN-restricting patients, BD of AN patients, and MF of EDNOS patients (Table 1). With a few exceptions, the original clinically-derived eight EDI subscales were clearly identified in EFA in both groups and explained
50.8% and 47.3% of the variance in the patient and non-patient group,
respectively. The model-fit indices of CFA indicated that the original 1st-order 8-factor structure was barely acceptable for Taiwanese patients with ED. Moreover, we grouped the 3 eating and weight-related subscales and the 5 general psychological functioning subscales to derive two 2nd-order factors, the model-fit indices for this 2nd-order factor structure were
comparable to those of 1st-order structure in both the patient and non-patient groups (Table 2).
Conclusions
The C-EDI is a psychometrically sound questionnaire for both clinical and non-clinical subjects.
Table 2 Goodness-of-Fit indicators for 1st-order and 2nd- order confirmatory factor analyses for patients and female students
Mode
χ
2χ
2/dfs
CFI
SRMSR
RMSEA (90% CI)
Eating disorder patients
1
storder
Original eight factor model
5970.662
3.10
a0.916
0.0914
0.0713 (0.0695, 0.0730)
2
ndorder
Two 2
ndfactor model
7537.614
3.88
b0.914
0.0971
0.0724 (0.0706, 0.0741)
Female college students
1
storder
Original eight factor model
4479.893
2.33
c0.929
0.0591
0.0421 (0.0405, 0.0437)
2
ndorder
Two 2
ndfactor model
4898.725
2.52
d0.923
0.0611
0.0450 (0.0435, 0.0466)
CFI = comparative fit index; SRMSR = standardized root-mean-square residual; RMSEA = root-mean-square error of approximation
adf = 1924, bdf = 1943, cdf = 1924, ddf =1943
Different lower case letter indicates differences among groups (P < 0.01 for each comparison)
Internal consistency and factor structure of the Eating Disorder Inventory
among clinical and non-clinical subjects in Taiwan
Mei-Chih Meg Tseng1, Grace Yao2, Fu-Chang Hu3, Kwan-Yu Chen4, David Fang5
1 Department of Psychiatry, National Taiwan University Hospital Yun-Lin Branch and National Taiwan University College of Medicine, Taipei, Taiwan
2 Department of Psychology, National Taiwan University, Taipei, Taiwan
3Department of Medical Research and National Center of Excellence for General Clinical Trial and Research, National Taiwan University Hospital,
Taipei, Taiwan
4Department of Psychiatry, Taipei City Psychiatric Center, Taipei City Hospital, Taipei, Taiwan
