PROJECT SUMMARY/ABSTRACT
Eating disorders (EDs) are severe mental illnesses with the highest mortality rate of any
psychiatric disorder. The most widely used empirically supported treatment for EDs (cognitive
behavior therapy) is only efficacious for ~50% of individuals. This low response rate is due to
the fact that EDs are heterogeneous conditions with diverse symptom trajectories that are not
adequately addressed in current “one-size-fits-all” psychotherapies. Until we can identify what
maintains or exacerbates individual symptoms, clinicians will continue to have difficulty
accurately predicting prognosis and will have no empirical guidance to develop targeted
treatment plans to promote recovery. Our scientific premise, developed from our past work, is
that the application of network theory will enable the identification of cognitive-behavioral
symptom networks that maintain and ‘trigger’ EDs both between and within individuals. Our
study goals are to (1) identify individual ED ‘trigger’ symptoms (cognitions, behaviors, affect,
and physiology) and (2) correlate trigger symptoms with real-time physiological data to create
an algorithm predicting onset of ED behaviors. These goals will ultimately identify symptoms
that prevent full remission and lead to relapse. We will use a multiple units of analysis approach
combined with novel, cutting-edge advances in network science. We will collect intensive real-
time data on cognitions, behavior, affect, and physiology using mobile and sensor-technology
from 120 individuals with a diagnosis of anorexia nervosa (AN), atypical AN, and bulimia
nervosa across 30 days. At 1-month and 6-month follow ups we will assess ED outcomes (e.g.,
remission status, ED behaviors) to test if ‘trigger’ symptoms predict ED outcomes. Network
science and state-of-the-art machine learning techniques will allow us, for the first time, to
discover whether certain trigger symptoms predict worse outcomes. Specific aims are to (1)
develop personalized networks to identify which cognitive, behavioral, affective, and
physiological symptoms maintain EDs and predict ED outcomes and (2) utilize sensor data to
identify physiological patterns both within and across people that correlate with core maintaining
symptoms and that predict ED behaviors. The proposed research uses highly innovative
methods, combining intensive longitudinal data collection methods, all remote procedures, novel
advances in network science and sensor-technology, and state-of-the-art machine learning
techniques to answer previously unresolvable questions pinpointing which personalized
symptoms trigger EDs. The proposed research has clinical impact. If we identify patterns that
contribute to symptom network variation within individuals, these data will provide a model of
personalized medicine for the entire field of psychiatry, as well as providing novel intervention
targets to prevent and treat EDs.