Reliability and Validity of the Mandarin Chinese Version of Eating Disorder Inventory

Mei-Chih Tseng, Huei-Chen Ko, Ming Been Lee

Research output: Contribution to journalArticlepeer-review


To establish the Mandarin Chinese Version of Eating Disorder Inventory (EDI), proceed it’s reliability and validity test and to address the psychometric properties of EDI in Taiwanese eating disorders patients and university students. Methods: After getting the permission agreement from the copyright owner of EDI (Psychological Assessment Resources. Inc), the EDI was translated into Mandarin Chinese via two-step procedure. The study subjects included 4 groups of persons: university students (N=341, 183 women and 154 men), clinical patients fulfilled DSM-IV eating disorders diagnoses (N=69, 19 AN and 50 BN), recovered eating disorders patients (N=20), and obese patients (N=63). Reliability (internal consistency) and construct validity were established by Cronbach’s α method and item-total correlation. The scores between groups were compared by Student’s t-test or One way ANOVA. Correlation analyses were used between all EDI subscales and other self-rating scales, e.g. Bulimic Investigatory Test, Edinburgh (BITE), Visual Analogue Scale (VAS) for body weight/ shape and general appearance dissatisfaction, and Brief Symptom Rating Scale (BSRS). Results: Cronbach’s alpha values were above or near 0.7 on all subscales and EDI for both female students and clinical patients. The item-subtotal correlation coefficients were below 0.3 for 4 items (item 1, 6, 22 and 43). To delete item 1 may markedly improve internal consistency and construct validity of Drive for Thinness (DT) subscale. Eating disordered patients as well as BN patients were differentiated from female students on all EDI subscales. However, Body Dissatisfaction (BD) subscale can’t differentiate effectively between AN patients and female college students. Clinically recovered patients scored similarly to female students on all subscales, and scored lower than eating disordered patients on all subscales except Interpersonal Distrust (ID) and Maturity Fears (MF), which was compatible with the characteristics of recovered patients in the follow-up studies of eating disorders. Convergent validity and discriminate validity were established for all subscales. Group comparison showed BN patients scored higher than AN patients on DT, B and BD subscales but not on the subscales measuring general psychopathology. Obese patients can be well differentiated with AN and BN patients except BD subscale. As expected, the binge-eating/purging AN patients scored higher than the restrictive AN patients only on Bulimia subscale. But the binge-eating/purging AN patients scored lower than BN patients on BD subscale, but higher on MF subscale. The above findings were partially explained by the positive correlation between BD score and percentage of standard body weight (γ=0.40, p<0.001) as well as the heterogeneity of AN and BN groups (evidenced by the large standard deviations compared to means on all subscales). The other possible reasons were denial of illness in AN patients and cultural difference in phenomenology of AN. We defined high DT group as the female students who scored at or above the eating disordered patients’ mean score of 12 on the DT subscale of the EDI (N=26, 14.2%). The high DT group scored higher than eating disorder patients on DT (p<0.01) and BD (p=0.05) subscale, but lower on B. I. IA (p<0.001) and ID (p<0.01) subscales. Conclusions: The Mandarin Chinese Version of Eating Disorder Inventory had satisfactory internal reliability and validity. It can be applied to measure the multidimensional psychopathology of Taiwanese eating disordered patients. However, it did not differentiate the AN patients and female students effectively. We need to be aware the multidetermined nature of EDI-I when apply it to both clinical and general population.
Original languageChinese (Traditional)
Pages (from-to)379-388
Issue number4
Publication statusPublished - Jul 1 2001

Cite this