Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Treatment Response in Body Dysmorphic Disorder with Interpretable Machine Learning

Omar Costilla-Reyes, Morgan Talbot

TL;DR

This study demonstrates that in predicting treatment outcomes for Body Dysmorphic Disorder, interpretable machine learning models—especially linear ones—perform comparably to more complex algorithms and highlight treatment credibility as a key, modifiable predictor. Using forward feature selection across 100 imputed datasets and 5-fold cross-validation, credibility consistently emerged as the first feature across outcomes, with decision-tree analyses revealing practical thresholds on a 27-point credibility scale. These thresholds (around 16 and 22) offer actionable guidance for clinicians to tailor interventions and resource allocation, while simpler models maintain robust performance. The findings advocate for integrating credibility-enhancing strategies, including digital, technology-assisted interventions, into personalized treatment plans, though they acknowledge limits due to small samples and self-report data, calling for larger, multi-site studies to establish causal effects. $30 ext%$ reductions and $16$-point thresholds are used as representative benchmarks to illustrate the approach, highlighting the potential impact of harnessing patient beliefs to improve engagement and outcomes in BDD care.

Abstract

Body Dysmorphic Disorder (BDD) is a highly prevalent and frequently underdiagnosed condition characterized by persistent, intrusive preoccupations with perceived defects in physical appearance. In this extended analysis, we employ multiple machine learning approaches to predict treatment outcomes -- specifically treatment response and remission -- with an emphasis on interpretability to ensure clinical relevance and utility. Across the various models investigated, treatment credibility emerged as the most potent predictor, surpassing traditional markers such as baseline symptom severity or comorbid conditions. Notably, while simpler models (e.g., logistic regression and support vector machines) achieved competitive predictive performance, decision tree analyses provided unique insights by revealing clinically interpretable threshold values in credibility scores. These thresholds can serve as practical guideposts for clinicians when tailoring interventions or allocating treatment resources. We further contextualize our findings within the broader literature on BDD, addressing technology-based therapeutics, digital interventions, and the psychosocial determinants of treatment engagement. An extensive array of references situates our results within current research on BDD prevalence, suicidality risks, and digital innovation. Our work underscores the potential of integrating rigorous statistical methodologies with transparent machine learning models. By systematically identifying modifiable predictors -- such as treatment credibility -- we propose a pathway toward more targeted, personalized, and ultimately efficacious interventions for individuals with BDD.

Predicting Treatment Response in Body Dysmorphic Disorder with Interpretable Machine Learning

TL;DR

This study demonstrates that in predicting treatment outcomes for Body Dysmorphic Disorder, interpretable machine learning models—especially linear ones—perform comparably to more complex algorithms and highlight treatment credibility as a key, modifiable predictor. Using forward feature selection across 100 imputed datasets and 5-fold cross-validation, credibility consistently emerged as the first feature across outcomes, with decision-tree analyses revealing practical thresholds on a 27-point credibility scale. These thresholds (around 16 and 22) offer actionable guidance for clinicians to tailor interventions and resource allocation, while simpler models maintain robust performance. The findings advocate for integrating credibility-enhancing strategies, including digital, technology-assisted interventions, into personalized treatment plans, though they acknowledge limits due to small samples and self-report data, calling for larger, multi-site studies to establish causal effects. reductions and -point thresholds are used as representative benchmarks to illustrate the approach, highlighting the potential impact of harnessing patient beliefs to improve engagement and outcomes in BDD care.

Abstract

Body Dysmorphic Disorder (BDD) is a highly prevalent and frequently underdiagnosed condition characterized by persistent, intrusive preoccupations with perceived defects in physical appearance. In this extended analysis, we employ multiple machine learning approaches to predict treatment outcomes -- specifically treatment response and remission -- with an emphasis on interpretability to ensure clinical relevance and utility. Across the various models investigated, treatment credibility emerged as the most potent predictor, surpassing traditional markers such as baseline symptom severity or comorbid conditions. Notably, while simpler models (e.g., logistic regression and support vector machines) achieved competitive predictive performance, decision tree analyses provided unique insights by revealing clinically interpretable threshold values in credibility scores. These thresholds can serve as practical guideposts for clinicians when tailoring interventions or allocating treatment resources. We further contextualize our findings within the broader literature on BDD, addressing technology-based therapeutics, digital interventions, and the psychosocial determinants of treatment engagement. An extensive array of references situates our results within current research on BDD prevalence, suicidality risks, and digital innovation. Our work underscores the potential of integrating rigorous statistical methodologies with transparent machine learning models. By systematically identifying modifiable predictors -- such as treatment credibility -- we propose a pathway toward more targeted, personalized, and ultimately efficacious interventions for individuals with BDD.

Paper Structure

This paper contains 13 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Outcome Measures.
  4. Results
  5. Discussion
  6. A Simple Heuristic Based on Credibility.
  7. Integration with Standard Measures.
  8. Future Directions for Practice.
  9. Sample Size and External Validity.
  10. Reliance on Self-Reported Data.
  11. Causal Ambiguities.
  12. Role of Digital Interventions and Technology Acceptance.
  13. Conclusions

Figures (3)

  • Figure 1: Illustration of the forward feature selection procedure used in our study. In (a), candidate features (e.g., Age, Credibility, BDD-YBOCS) are initially considered. In (b), the addition of BDD duration fails to meet the 0.05 AUC-improvement threshold (indicated by the red cross). In (c), the final selected combination of features (shown in pink) includes Credibility and BDD-YBOCS.
  • Figure 2: A representative decision tree for predicting treatment response, highlighting how treatment credibility interacts with gender identity. Patients scoring above 22 on credibility are likely to respond; in contrast, female-identifying patients with credibility scores below 16 are at higher risk for non-response.
  • Figure 3: Scatter plot of BDD patients by baseline treatment credibility (x-axis) and BDD-YBOCS (y-axis). Blue markers denote treatment responders; red markers denote non-responders; star-like markers indicate remission. Patients with credibility scores $\leq 16$ rarely responded or remitted, whereas scores $> 22$ were strongly predictive of positive outcomes.