Developing an Optimized Conversational Agent or "Chatbot" to Facilitate Mental Health Services Use in Individuals with Eating Disorders

PROJECT SUMMARY/ABSTRACT

Eating disorders (EDs) are serious mental illnesses associated with high morbidity and mortality, clinical

impairment, and comorbid psychopathology. Although evidence-based treatments for EDs have been

established, the treatment gap is wide. Indeed, 200,000 times. Among screen respondents,

the vast majority screen positive for an ED, but of those, most are not in treatment, and only a small

percentage click to learn more about one of the options for seeking intervention presented in the screening

feedback. I propose a research agenda to design a conversational agent or “chatbot” that is optimized to

increase mental health services use among individuals with EDs through such features as: 1) providing a

personalized recommendation for seeking intervention; 2) engaging the user in motivational interviewing to

overcome barriers to care; and 3) repeated check-ins with the user to encourage follow-up with care. This

research agenda aligns with NIMH’s focus on services research and interest in technology-driven approaches

to promote engagement with care. I will conduct two studies. First, I will utilize a user-centered design

approach to create a prototype chatbot and conduct usability testing with adults with EDs to inform chatbot

refinements (Aim 1). Second, I will conduct a randomized optimization trial with adults who have completed

screening on the NEDA website and screen positive for an ED but are not in treatment to determine chatbot

feasibility and to generate data on the effect of the chatbot on motivation for treatment post-initial chatbot use

and motivation for treatment and mental health services use at 1- and 3-month follow-ups (Aims 2 & 3). This

trial will employ the Multiphase Optimization Strategy framework, using a 23 full factorial design, to randomly

assign participants to a combination of the three proposed intervention components (n=8 conditions) to isolate

the active ingredients. These aims support my training plan in which I will receive expert mentorship and

training in: services research and implementation science (Training Goal 1); user-centered design and

usability testing, as well as exposure to machine learning (Training Goal 2); novel trial designs (Training Goal

3); and advanced statistical techniques for analyzing longitudinal trial and digital innovation data (Training

Goal 4). My expert mentorship team, along with the environment of Washington University, will ensure my

success. Results from the proposed study will be used to optimize the chatbot, which I will then test in a

subsequent R01 randomized controlled trial. This K08 will equip me with the skills to become a leader in

addressing the research-practice and treatment gaps for EDs, with a focus on digital innovations.