Predictors of outcome during inpatient psychotherapy for posttraumatic stress disorder: a single-treatment, multi-site, practice-based study.

Objective: The aims of this study were to determine the effectiveness of a routine clinical care treatment and to identify predictors of treatment outcome in PTSD inpatients. Methods: A routinely collected data set of 612 PTSD inpatients (M = 42.3 years [SD = 11.6], 75.7% female) having received trauma-focused psychotherapy was analyzed. Primary outcome was the clinical symptom severity change score, secondary outcomes were assessed using functional, anxiety, and depression change scores. Hedges g-corrected pre-post effect sizes (ES) were computed for all outcomes. Elastic net regulation as a data-driven, stability-based machine-learning approach was used to build stable clinical prediction models. Results: Hedges g ES indicated medium to large effects on all outcomes. The results of the predictor analyses suggested that a combined predictor model with sociodemographic, clinical, and psychometric variables contribute to predicting different treatment outcomes. Across the clinical and functional outcome, psychoticism, total number of diagnoses, and bronchial asthma consistently showed a stable negative predictive relationship to treatment outcome. Conclusion: Trauma-focused psychotherapy could effectively be implemented in a routine inpatient setting. Some important prognostic variables could be identified. If the proposed models of predictors are replicated, they may help personalize treatment for patients receiving routine clinical care.

Inpatient psychotherapy for depression in a large routine clinical care sample: A Bayesian approach to examining clinical outcomes and predictors of change.

BACKGROUND: A routinely collected dataset was analyzed (1) to determine the naturalistic effectiveness of inpatient psychotherapy for depression in routine psychotherapeutic care, and (2) to identify potential predictors of change. METHODS: In a sample of 22,681 inpatients with depression, pre-post and pre-follow-up effect sizes were computed for various outcome variables. To build a probabilistic model of predictors of change, an independent component analysis generated components from demographic and clinical data, and Bayesian EFA extracted factors from the available pre-test, post-test and follow-up questionnaires in a subsample (N = 6377). To select the best-fitted model, the BIC of different path models were compared. A Bayesian path analysis was performed to identify the most important factors to predict changes. RESULTS: Effect sizes were large for the primary outcome and moderate for various secondary outcomes. Almost all pretreatment factors exerted significant influences on different baseline factors. Several factors were found to be resistant to change during treatment: suicidality, agoraphobia, life dissatisfaction, physical disability and pain. The strongest cross-loadings were observed from suicidality on negative cognitions, from agoraphobia on anxiety, and from physical disability on perceived disability. LIMITATIONS: No causal conclusions can be drawn directly from our results as we only used cross-lagged panel data without control group. CONCLUSIONS: The results indicate large effects of inpatient psychotherapy for depression in routine clinical care. The direct influence of pretreatment factors decreased over the course of treatment. However, some factors appeared stable and difficult to treat, which might hinder treatment outcome. Findings of different predictors of change are discussed.

Cognitive Change and Relaxation as Key Mechanisms of Treatment Outcome in Chronic Pain: Evidence From Routine Care.

Despite effective treatment approaches within the cognitive behavioral framework general treatment effects for chronic pain are rather small to very small. Translation from efficacy trials to naturalistic settings is questionable. There is an urgent need to improve the effectiveness of well-established treatments, such as cognitive-behavior therapy (CBT) and the investigation of mechanisms of change is a promising opportunity. We performed secondary data analysis from routine data of 1,440 chronic pain patients. Patients received CBT in a multidisciplinary setting in two inpatient clinics. Effect sizes and reliable change indices were computed for pain-related disability and depression. The associations between changes in the use of different pain coping skills (cognitive restructuring, activity despite pain, relaxation techniques and mental distraction) and changes in clinical outcomes were analyzed in structural equation models. Pre-post effect sizes range from g = 0.47 (disability) to g = 0.89 (depression). Changes in the use of cognitive restructuring, relaxation and to a lesser degree mental distraction were associated with changes in disability and depression. Effects from randomized trials can be translated to naturalistic settings. The results complement experimental research on mechanisms of change in the treatment of chronic pain and indicate an important role of cognitive change and relaxation as mechanisms of change. Our findings cautiously suggest that clinicians should optimize these processes in chronic pain patients to reduce their physical and emotional disability.

Drawing the borderline: Predicting treatment outcomes in patients with borderline personality disorder.

BACKGROUND: A routinely collected big data set was analyzed to determine the effectiveness of naturalistic inpatient treatment and to identify predictors of treatment outcome and discontinuation. METHODS: The sample included 878 patients with borderline personality disorder who received non-manualized dialectic behavioral therapy in a psychosomatic clinic. Effect sizes (Hedge's g) were calculated to determine effectiveness. A bootstrap-enhanced regularized regression with 91 potential predictors was used to identify stable predictors of residualized symptom- and functional change and treatment discontinuation. Results were validated in a holdout sample and repeated cross validation. RESULTS: Effect sizes were small to medium (g = 0.28-0.51). Positive symptom-related outcome was predicted by low affect regulation skills and no previous outpatient psychotherapy. Lower age, absence of work disability, high emotional and physical role limitations and low bodily pain were associated with greater improvement in functional outcome. Higher education and comorbid recurrent depressive disorder were the main predictors of treatment completion. The predictive quality of the models varied, with the best being found for symptom-related outcome (R2 = 18%). CONCLUSION: While the exploratory process of variable selection replicates previous findings, the validation results suggest that tailoring treatment to the individual patient might not be based solely on sociodemographic, clinical and psychological baseline data.