Results of a randomized trial show that at-risk women who interacted with the chatbot showed lower concern about their weight and body shape compared to a wait-list control group.
“Chatbots are widely used in industry and have begun to be used in medical settings, although few studies have examined their effectiveness for mental health issues and none address EDs or ED prevention,” senior investigator C. Barr Taylor, MD, a research faculty member at Palo Alto (Calif.) University, said in a press release.
“We found that the group with access to the chatbot had a greater reduction in weight and shape concerns, both right after using it at 3 months and at the 6-month follow-up. The effects had sustainability over time, and we also found indication that the chatbot may reduce ED onset more so than the control group, where there was a greater incidence of EDs,” Dr. Taylor told this news organization.
The study was published online Dec. 28, 2021, in the International Journal of Eating Disorders.
Deadly disorders
“EDs are a common problem with huge risk factors; and, given how widespread they are, we need scalable tools that can reach a lot of people at low cost, reduce risk factors for developing an ED – which is the second most deadly of all psychiatric illnesses – so prevention is of the utmost importance,” Dr. Taylor said.
The investigators developed a targeted Internet-based preventive program called StudentBodies that utilizes cognitive-behavioral therapy approaches. The program was successful in reducing weight/shape concerns in women at high risk for the onset of an ED, and it reduced ED onset in the highest-risk women.
However, it required trained moderators who spent over 45 minutes with participants. Given the large number of people at risk for an ED who might benefit, the researchers noted that it is unlikely that a human-moderated version would be widely disseminated.
A chatbot may represent a “possible solution to reducing delivery costs” because it mimics aspects of human moderation in simulating conversations, the investigators noted.
“We wanted to take the earlier program we developed into this century and program it for delivery in this new format that would allow for bite-size pieces of information for the chatbot to communicate to the user,” lead author Ellen Fitzsimmons-Craft, PhD, assistant professor of psychiatry, Washington University, St. Louis, told this news organization.
“Our ED prevention online version was more effective when there was guidance from a human moderator who could provide feedback on progress, encourage you to go on, and apply the skills in daily life. But that’s not the most scalable. So we thought that a chatbot, in addition to providing content in this perhaps more engaging format, could also provide some aspect of human moderation, although the person is chatting with a robot,” added Dr. Fitzsimmons-Craft, associate director of the Center for Healthy Weight and Wellness.
Tessa will speak to you now
Participants (n = 700 women; mean [SD] age, 21.08 [3.09] years; 84.6% White; 53.8% heterosexual; 31.08% bisexual), were randomized to an intervention group or a wait-list control group (n = 352 and 348, respectively). There were no significant differences between groups in age, race, ethnicity, education, or sexual orientation.
The StudentBodies program was adapted for delivery via a chatbot named Tessa “while retaining the core intervention principles” and referred to as “Body Positive.”
It consisted of several components programmed into the chatbot, which initiated each conversation in a predetermined order. Participants were encouraged to engage in two conversations weekly. The program included an introduction and eight sessions as well as a crisis module that provided users with a referral to a crisis hotline in case of emergency. Referral was triggered on the basis of “recognized keywords,” such as “hurting myself.”
The researchers used the Weight Concerns Scale questionnaire to assess weight and shape concerns and the Internalization: Thin/Low Body Fat subscale of the Sociocultural Attitudes Toward Appearance Questionnaire–4 to “assess the cognitive aspect of thin-ideal internalization.”
Secondary outcomes tested the hypothesis that the chatbot would be more likely to reduce clinical outcomes (ED psychopathology, depression, and anxiety) and prevent ED onset, compared to the control condition.
Ready for prime time
At 3- and 6-month follow-up, there was significantly greater reduction in the intervention group compared with the control group in weight/shape concerns (d = -.20, P = .03 and d = -.19, P = .04, respectively), although there were no differences in thin-ideal internalization change.
The chatbot intervention was associated with significantly greater reductions in overall ED psychopathology at 3 months (d = -.29, P = .003) compared to the control condition, but not at 6 months.
Notably, the intervention group had significantly higher odds than the control group of remaining nonclinical for EDs at 3- and 6-month follow-up (OR, 2.37 [95% confidence interval, 1.37-4.11] and OR, 2.13 [95% CI,1.26-3.59], respectively).
“We were very excited about the study, and frankly, I was surprised by the effectiveness [of the chatbot intervention] because I didn’t think it would have as much of an impact as it did,” said Dr. Taylor. “Prevention gets short shrift everywhere, and I think we succeeded very well.”
Dr. Fitzsimmons-Craft added that the National Eating Disorders Association (NEDA) has agreed to make the chatbot available on its website for people who screen positive for having an ED or for being at high risk, and so their group is working with their industry partner, a company called X2AI, which developed the chatbot, to make this happen.
“This is definitely the fastest research-to-practice translation I’ve ever seen, where we can so quickly show that it works and make it available to tens of thousands almost immediately.”
Dr. Fitzsimmons-Craft is optimistic that it will be available to launch the week of Feb. 21, which is National Eating Disorders Week.
Innovative, creative research
Commenting on the research, Evelyn Attia, MD, professor of psychiatry, Columbia University Medical Center, and director of the Columbia Center for Eating Disorders New York–Presbyterian Hospital, New York, described the study as “innovative and creative.”
New York-Presbyterian Hospital
Dr. Evelyn Attia
Dr. Attia, a member of the Research Advisory Council of the NEDA, noted that the structure of the study is “very preliminary” and that the comparison to a wait-list control makes it hard to know whether this is an effective intervention compared with other types of interventions, rather than compared with no intervention.
“But I’m sure that when the researchers are set up and primed to study this more robustly, they will consider a more active control intervention to see whether this preliminary finding holds up,” she said.
Also commenting on the study, Deborah R. Glasofer, PhD, associate professor of clinical medical psychology (in psychiatry), Columbia Center for Eating Disorders, said, “Higher-than-average concern about appearance – body shape, size, or weight – and a tightly held belief that it is ideal to be thin are known risk factors for the development of an eating disorder.
“This study offers an indication that technology can be leveraged to fill a gap and help folks before unhelpful and sometimes misguided thoughts about food, eating, and appearance evolve into a full-blown eating disorder,” said Dr. Glasofer, who was not involved with the study.
The study was supported by the NEDA Feeding Hope Fund, the National Institute of Mental Health, the National Heart, Lung, and Blood Institute, and the Swedish Research Council. The authors and Dr. Glasofer have disclosed no relevant financial relationships. Dr. Attia is on the board and the Research Advisory Council of NEDA.
A version of this article first appeared on Medscape.com.