Predicting treatment outcome for complex posttraumatic stress disorder using the personalized advantage index

<b>Background:</b> Ample studies have demonstrated the effectiveness of psychotherapy for posttraumatic stress disorder (PTSD). However, large individual variation in treatment outcome remains unsolved and treatment options for complex posttraumatic stress disorder (CPTSD) are debated. There is a need for exploring methods for matching patients with treatment they will most likely benefit from. <b>Objective:</b> To develop a personalized advantage index (PAI) based on relevant clinical and demographic predictors of outcome from exposure therapy and skills-training for CPTSD. <b>Method:</b> Data from a previous randomized controlled trial (RCT) in 92 patients with a CPTSD diagnosis was used to compare Prolonged Exposure (PE; <i>n</i> = 32) and Skills Training in Affective and Interpersonal Regulation (STAIR; <i>n</i> = 60). Outcome measures were clinician-assessed and self-reported PTSD symptoms. Predictors of outcome in PE and STAIR were identified separately from sixty-one candidate variables using random forest and bootstrap procedures. Relevant predictors were then used to calculate PAI and retrospectively identify optimal versus suboptimal treatment in a leave-one-out cross-validation approach. <b>Results:</b> In PE, somatoform dissociation, depression, suicidal ideation, and reduced physical health predicted worse outcome. In STAIR, interpersonal problems, total PTSD symptom severity, intrusions, elevated guilt, and psychoticism predicted worse outcome, while being a witness to trauma predicted better outcome. Allocation to optimal treatment according to the PAI was associated with large improvements in clinician-assessed (Cohen’s <i>d</i> = 0.83) and moderate improvement in self-rated (Cohen’s <i>d</i> = 0.60) PTSD symptoms as compared to allocation to suboptimal teatment. <b>Conclusions:</b> Using the PAI in personalizing psychological treatment for CPTSD is a promising approach to improve treatment benefits. Further research on larger samples and external validation of the PAI is needed. The present study aimed to assess the potential for improving psychotherapy outcome by applying machine learning methods in treatment selection.Different baseline variables predicted the outcome of exposure treatment and skills training for patients with complex posttraumatic stress disorder.The study demonstrated the utility of the machine learning method personalized advantage index in predicting differential treatment outcomes. The present study aimed to assess the potential for improving psychotherapy outcome by applying machine learning methods in treatment selection. Different baseline variables predicted the outcome of exposure treatment and skills training for patients with complex posttraumatic stress disorder. The study demonstrated the utility of the machine learning method personalized advantage index in predicting differential treatment outcomes.