TY  - JOUR
AU  - Arend, Ann-Kathrin
AU  - Kaiser, Tim
AU  - Pannicke, Björn
AU  - Reichenberger, Julia
AU  - Naab, Silke
AU  - Voderholzer, Ulrich
AU  - Blechert, Jens
PY  - 2023
DA  - 2023/2/23
TI  - Toward Individualized Prediction of Binge-Eating Episodes Based on Ecological Momentary Assessment Data: Item Development and Pilot Study in Patients With Bulimia Nervosa and Binge-Eating Disorder
JO  - JMIR Med Inform
SP  - e41513
VL  - 11
KW  - idiographic
KW  - individualized
KW  - N of 1
KW  - Ecological Momentary Assessment (EMA)
KW  - Just-In-Time Adaptive Intervention (JITAI)
KW  - binge eating
KW  - literature research
KW  - focus group
KW  - prediction algorithm
KW  - machine learning
KW  - Best Items Scales that are Cross-validated, Unit-weighted, Informative and Transparent
KW  - BISCUIT
AB  - Background: Prevention of binge eating through just-in-time mobile interventions requires the prediction of respective high-risk times, for example, through preceding affective states or associated contexts. However, these factors and states are highly idiographic; thus, prediction models based on averages across individuals often fail. Objective: We developed an idiographic, within-individual binge-eating prediction approach based on ecological momentary assessment (EMA) data. Methods: We first derived a novel EMA-item set that covers a broad set of potential idiographic binge-eating antecedents from literature and an eating disorder focus group (n=11). The final EMA-item set (6 prompts per day for 14 days) was assessed in female patients with bulimia nervosa or binge-eating disorder. We used a correlation-based machine learning approach (Best Items Scale that is Cross-validated, Unit-weighted, Informative, and Transparent) to select parsimonious, idiographic item subsets and predict binge-eating occurrence from EMA data (32 items assessing antecedent contextual and affective states and 12 time-derived predictors). Results: On average 67.3 (SD 13.4; range 43-84) EMA observations were analyzed within participants (n=13). The derived item subsets predicted binge-eating episodes with high accuracy on average (mean area under the curve 0.80, SD 0.15; mean 95% CI 0.63-0.95; mean specificity 0.87, SD 0.08; mean sensitivity 0.79, SD 0.19; mean maximum reliability of rD 0.40, SD 0.13; and mean rCV 0.13, SD 0.31). Across patients, highly heterogeneous predictor sets of varying sizes (mean 7.31, SD 1.49; range 5-9 predictors) were chosen for the respective best prediction models. Conclusions: Predicting binge-eating episodes from psychological and contextual states seems feasible and accurate, but the predictor sets are highly idiographic. This has practical implications for mobile health and just-in-time adaptive interventions. Furthermore, current theories around binge eating need to account for this high between-person variability and broaden the scope of potential antecedent factors. Ultimately, a radical shift from purely nomothetic models to idiographic prediction models and theories is required. 
SN  - 2291-9694
UR  - https://medinform.jmir.org/2023/1/e41513
UR  - https://doi.org/10.2196/41513
UR  - http://www.ncbi.nlm.nih.gov/pubmed/36821359
DO  - 10.2196/41513
ID  - info:doi/10.2196/41513
ER  -