Machine learning v. traditional regression models predicting treatment outcomes for binge-eating disorder from a randomized controlled trial

Abstract

Background While effective treatments exist for binge-eating disorder (BED), prediction of treatment outcomes has proven difficult, and few reliable predictors have been identified. Machine learning is a promising method for improving the accuracy of difficult-to-predict outcomes. We compared the accuracy of traditional and machine-learning approaches for predicting BED treatment outcomes. Methods Participants were 191 adults with BED in a randomized controlled trial testing 6-month behavioral and stepped-care treatments. Outcomes, determined by independent assessors, were binge-eating (% reduction, abstinence), eating-disorder psychopathology, and weight loss (% loss, ⩾5% loss). Predictors included treatment condition, demographic information, and baseline clinical characteristics. Traditional models were logistic/linear regressions. Machine-learning models were elastic net regressions and random forests. Predictive accuracy was indicated by the area under receiver operator characteristic curve (AUC), root mean square error (RMSE), and R2. Confidence intervals were used to compare accuracy across models. Results Across outcomes, AUC ranged from very poor to fair (0.49–0.73) for logistic regressions, elastic nets, and random forests, with few significant differences across model types. RMSE was significantly lower for elastic nets and random forests v. linear regressions but R2 values were low (0.01–0.23). Conclusions Different analytic approaches revealed some predictors of key treatment outcomes, but accuracy was limited. Machine-learning models with unbiased resampling methods provided a minimal advantage over traditional models in predictive accuracy for treatment outcomes.

Predictors of neurofeedback treatment outcome in binge‐eating disorder: An exploratory study

“…Overall, the findings highlighted the relevance of BED‐ versus general psychopathology‐related symptoms for NF treatment success, consistent with predictors for diverse psychological treatments in BED (Forrest et al, 2021) and meta‐analytical predictors for diverse treatment options across EDs (Vall & Wade, 2015). Differential predictors emerged for OBE frequency and EDP, composed of restraint, eating, weight, and shape concern, indicating that the improvement of core symptoms versus EDP was associated with different baseline eating disorder‐ and general psychopathology‐related characteristics of patients with BED.…”

“…The interaction between anxiety symptoms and treatment group did not reach significance, but anxiety symptoms were surprisingly linked to lower EDP in rtfNIRS-NF in follow-up analyses. This vague finding resembled the inconclusive evidence on the predictive role of anxiety symptoms for outcomes of diverse BED treatments (Forrest et al, 2021;Lydecker & Grilo, 2021). Specifically, the presence of comorbid anxiety disorders was linked to higher EDP after psychological and pharmacological treatment for BED (Lydecker & Grilo, 2021).…”

Exploration of the individual and combined effects of predictors of engagement, dropout, and change from digital interventions for recurrent binge eating

“…While novel in the context of eating disorder research, our multivariate D approach is not the only means available for predictive modeling. If multivariate D may be considered a ‘brute force’ strategy for combining a large set of predictors to optimize prediction, it would be logical to then consider in relation to other methodological advances such as machine learning‐based approaches (Forrest et al, 2023; Krug et al, 2021). Approaches such as random forests and neural networks may combine predictors in ways that differ from a standard regression‐based approach, and with potential to enhance prediction of outcomes of interest.…”

Machine learning models for diagnosis and risk prediction in eating disorders, depression, and alcohol use disorder

“…In ED research, these have been used in cross-sectional diagnostic classi cation models derived from distinct datasets, such as questionnaires 16,17 , social media 18 , or neuroimaging data 19,20 . Longitudinal models have also been built to predict illness course 21 and treatment outcomes 22 . Yet, to our knowledge, no ED study to date has combined a wide range of data domains to build models for diagnostic classi cation or risk prediction.…”