Using the Minnesota Multiphasic Personality Inventory–2 Restructured Form to Predict Functioning After Treatment for Borderline Personality Disorder: A Machine Learning Approach.

Insight into predictors of functioning after treatment for borderline personality disorder (BPD) is limited, despite growing recognition that more focus on other aspects of recovery, especially psychosocial functioning, is warranted. The present study explored the utility of a widely used omnibus as...

Descripción completa

Detalles Bibliográficos
Publicado en:Psychological Assessment Vol. 37; no. 6/7; pp. 261 - 273
Autores principales: Wibbelink, Carlijn J. M., Sellbom, Martin, Grasman, Raoul P. P. P., Arntz, Arnoud, Sinnaeve, Roland, Kamphuis, Jan H.
Formato: Artículo
Publicado: American Psychological Association Jul2025
Materias:
Acceso en línea:Ver este registro en EBSCOhost
Descripción
Sumario:Insight into predictors of functioning after treatment for borderline personality disorder (BPD) is limited, despite growing recognition that more focus on other aspects of recovery, especially psychosocial functioning, is warranted. The present study explored the utility of a widely used omnibus assessment instrument, the Minnesota Multiphasic Personality Inventory–2 Restructured Form (MMPI-2-RF), to predict change in functioning during treatment for BPD. Data were obtained from a randomized clinical trial into the effectiveness of 2-year evidence-based treatment for BPD (dialectical behavior therapy or schema therapy) among 130 participants diagnosed with BPD. Different machine learning algorithms, including elastic net regression (ENR), random forest, gradient boosting machine, and extreme gradient boosting, were implemented using nested cross-validation. The ENR model had an average explained variance of 42%. A combination of baseline functioning and four MMPI-2-RF scales emerged as key predictors of change in functioning. Baseline functioning was the most important predictor, with lower initial functioning levels related to more improvement. Higher scores on ideas of persecution, somatic complaints, family problems, and disconstraint were associated with less improvement in functioning. Given the risk of overfitting and the lack of an independent data set, future research should focus on the replicability and generalizability of the findings, as well as clarifying the underlying mechanisms. Our study serves as a first step in identifying patients at risk of poor functional outcome after treatment for BPD. Public Significance Statement: Given the increasing emphasis on improving psychosocial functioning alongside symptom recovery in treatment for borderline personality disorder, the present study employed machine learning algorithms to identify predictors of functional outcome after treatment using the Minnesota Multiphasic Personality Inventory–2 Restructured Form. Baseline functioning was the strongest predictor, while ideas of persecution, somatic complaints, family problems, and disconstraint were also associated with change in functioning, tentatively suggesting the importance of the therapeutic relationship.