Investigating predictors of treatment response in Dialectical Behavior Therapy for borderline personality disorder using LASSO regression

Abstract

Objective

Prior studies of Dialectical Behavior Therapy (DBT) for borderline personality disorder (BPD) have yielded heterogeneous findings on what factors differentiate individuals with or without sufficient treatment response, highlighting the need for further research.

Method

We investigated a sample of 105 individuals with BPD receiving a 6-month course of DBT. Participants were categorized as sufficient or insufficient responders using clinical and statistical change indices (based on emotion dysregulation, BPD symptom severity, utilization of DBT skills, and functional impairment). Sociodemographic, clinical severity, and treatment process factors were tested as potential predictors of treatment response using a machine learning approach (LASSO regression).

Results

Two cross-validated LASSO regression models predicted treatment response (AUCs > .75). They suggested that higher homework completion rate, retention in treatment, and greater baseline severity were the most important predictors of DBT treatment response indicated by BPD symptom severity and utilization of DBT skills. Favorable effects of some aspects of therapeutic alliance during initial sessions were also found.

Conclusions

Future research may benefit from consolidating the criteria of treatment response, identifying clinically relevant variables, and testing the generalizability of findings to enhance knowledge of insufficient treatment response in DBT for BPD.

Keywords:

• borderline personality disorder
• Dialectical Behavior Therapy
• treatment response
• machine learning
• LASSO regression

Acknowledgements

The current study was not pre-registered. The Rutgers University Institutional Review Board approved this study (Study ID: Pro2019001864). Materials of the study can be accessed by contacting the authors.

Disclosure Statement

Shireen L. Rizvi provides training and consultation for Behavioral Tech, LLC. Qingqing Yin, Molly Stern, and Evan M. Kleiman do not have competing interests.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1 Note that Wise (2004) specifies a continuum of treatment response categories ranging from recovery to deterioration. Only categories that included at least 1 participant in the current study are described here.

2 Note that the baseline outcome rating was also included in model 2 resulting in a total of 11 predictors entered.