Comparison of Interview to Questionnaire for Assessment of Eating Disorders after Bariatric Surgery

Abstract

The Eating Disorder Examination Interview Bariatric Surgery Version (EDE-BSV) assesses eating pathology after bariatric surgery but requires significant training and time to administer. Consequently, we developed a questionnaire format called the Eating Disorders After Bariatric Surgery Questionnaire (EDABS-Q). This study evaluates the consistency of responsiveness between the two formats. After surgery, 30 patients completed the EDE-BSV and EDABS-Q in a restricted randomized design. Patient reported behavior for each item which was converted to a score following the Eating Disorder Examination-Questionnaire (EDE-Q) scoring scheme. Responses fell into three distributions: (1) dichotomous, (2) ordinal, or (3) unimodal. Distributions of items were not different between the two formats and order did not influence response. Tests of agreement (normal approximation of the binomial test) and association (χ² analyses on binary data and spearman rank order correlations on ordinal items) were performed. Percent concordance was high across items (63-100%). Agreement was significant in 31 of 41 items (Bonferroni-P < 0.001). Association was significant in 10 of 21 in χ²-appropriate items (Bonferroni-P < 0.002), and the ordinal items had highly significant correlations between formats (Bonferroni-P < 0.0125). The EDABS-Q is an adequate substitute for the EDE-BSV and may be useful for research and clinical evaluation of eating pathology after bariatric surgery.

Keywords: bariatric surgery; binge eating disorders; eating disorders; eating pathology; questionnaire.

Figure 1

Graphical illustration of intermediate scores across items in the interview and questionnaire. Graph showing the number of participants (n = 30) intermediate scores (between 2 and 4) in the interview and questionnaire. The bold horizontal lines (at the 8 tick on the ordinate axis) are the cut offs between ordinal and not ordinal, with any items with bars breaching this line being considered ordinal. Items on the abscissa are in ascending order, for each format. The first three letters of each item correspond with a behavior/symptom category—RED, restraint for weight control; RPH, restraint to avoid physical discomfort; BPH, purging to avoid physical discomfort; PUR, purging for weight control; EAT, eating concern; WGT, weight concern; SHP, shape concern. See Appendix A for the full text of each item.

Figure 2

Panel of Pearson linear regression and Spearman rank order correlations plots. (A,C,E,G) are Pearson linear regression plots with scaled score response (0–6) from questionnaire (ordinate) regressed from interview (abscissa). Note: Since the distributions of response for these items are not normal, Pearson linear regression is not appropriate, and the presentation of these plots are for visual aid. (B) (R² = 0.72, p < 0.001), (D) (R2 = 0.69, p < 0.001), (F) (R2 = 0.54, p < 0.002), and (H) (R2 = 0.70, p < 0.001) are Spearman rank order correlation plots with the rank of questionnaire response (ordinate) and the rank of interview response (abscissa). Item number for (A–H) noted below the title of each plot. All items significant after Bonferroni correction (k = 4, α = 0.0125). See Appendix A for the full text of each item.