Refining the Diagnosis of Binge Eating Disorder

An APA 2014 Poster Brief

Bret S. Stetka, MD; Brandon K. Bellows, PharmD


May 09, 2014

Binge Eating Disorder Classification

Editor's Note: While on site at the American Psychiatric Association 167th Annual Meeting, held in New York, New York, May 3-7, 2014, Medscape interviewed Brandon K. Bellows, PharmD, Research Associate in the Department of Pharmacotherapy at University of Utah College of Pharmacy in Salt Lake City, Utah, about his study[1] assessing the diagnostic criteria for binge eating disorder (BED).

Medscape: Can you give our readers some background on your study?

Dr. Bellows: We were interested in examining the characteristics of patients with a diagnosis of BED and how well they are represented by a diagnosis of eating disorder not otherwise specified (EDNOS). BED doesn't have an International Classification of Diseases (ICD)-9 or ICD-10 code and falls under the ICD-9 code for EDNOS, so it's difficult to easily identify patients with BED from their medical records in order to study them.

Most studies of BED have been done using surveys or patient registries, or some other time-intensive process. So we developed a natural language processing (NLP) algorithm to identify patients with BED using the clinical notes in their medical records.

Medscape: This is a computer algorithm that scans medical records for specific terms or patient characteristics that correlate with BED?

Dr. Bellows: Yes; it's basically a very sophisticated keyword search. The NLP algorithm searched clinical text notes to identify instances where a provider mentioned BED. Then, on the basis of the context in which BED was mentioned, the algorithm classified it as confirming a diagnosis of BED, ruling out BED, or considering it as part of a differential diagnosis. We then had human annotators review random samples of these classifications to see how the NLP algorithm performed. We previously published a paper in the Journal of the American Medical Informatics Association that describes the methodology in detail.[2]

Medscape: What did you find using this approach in your new study?

Dr. Bellows: We set out to compare patients who had BED with those who had EDNOS but not BED (EDNOS only), and also with matched cohorts that did not have an eating disorder. Using the NLP algorithm, we were able to search the entire national medical record database for the Veterans Affairs system, which has a lot of patients -- 20 million or so.

We found nearly 1500 patients with BED, and about 600 of those patients met our inclusion criteria. Only about 11% of patients with BED actually had an EDNOS diagnosis, which is very low. We also saw that BED patients had a significantly higher body mass index (BMI) than patients with EDNOS only, and there were more men with BED.

In the year before diagnosis, we compared the prevalence of about 15-20 different comorbidities and prescription fills for around 10 different classes of medication between the groups. We found that overall, patients with BED and those with EDNOS only had similar rates of comorbidities, but when there were differences, more EDNOS-only patients had the comorbidity. Despite this, patients with EDNOS only generally had much higher use of medications than patients with BED.

When we compared the patients who had BED with those who did not have an eating disorder, significantly more patients with BED had comorbidities diagnosed and more were on medications.

One of the interesting findings when comparing BED patients with those who did not have an eating disorder involves obesity diagnoses. Although these patients were matched on BMI, patients with BED were far more likely to be diagnosed with obesity and morbid obesity. So physicians, for whatever reason, recognized and diagnosed obesity in patients with BED more than in patients without an eating disorder, even though they had the same BMIs. Along with the increase in comorbidities and medication use, this may indicate that patients with BED have a higher clinical burden than patients without an eating disorder and are being seen more by providers.

Medscape: BED was included as a specific diagnosis in the Diagnostic and Statistical Manual of Mental Disorders (DSM)-5, correct? And your study further suggests that it's unique enough from the EDNOS diagnosis that it should be its own diagnostic entity in the ICD?

Dr. Bellows: Yes, BED is in the DSM-5. And we've shown that EDNOS doesn't accurately represent patients with BED, so it needs its own identifier in the ICD system.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: