Borderline personality disorder (BPD) is associated with an elevated treatment dropout rate irrespective of the therapeutic approach [3, 5, 10, 39]. Dialectical behavior therapy (DBT) has most often been proven to be an effective treatment for patients with borderline pathology [26]. However, even for DBT, a mean dropout rate of 27.3% was found in a meta-analysis [26]. Despite the prevalence and high clinical relevance, the attention drawn to this phenomenon has increased only in the last years. Samples in different therapeutic settings and therapeutic approaches have mostly been examined with respect to “patient variables” such as demographic and clinical characteristics.

Only a few studies have investigated the predictive value of therapeutic processes on treatment completion in psychotherapy for BPD [5]. An important process factor is the therapeutic alliance, that can be defined as the overall bond between therapist and patient evolving during the process of therapy [24]. Sharf et al. [36] found an effect size of Cohen’s d = .55 to describe the association between the therapeutic alliance and therapy dropout in adult individual therapy across all mental disorders. Interpersonal difficulties represent core difficulties in patients with BPD and may affect the therapeutic alliance [1]. Patients show frantic efforts to avoid real or imagined abandonment and a pattern of unstable and intense interpersonal relationships characterized by alternating between extremes of idealization and devaluation. Therefore, a predictive value of the therapeutic alliance on treatment retention rates seems specifically likely in this patient group. However, results are heterogeneous. Whilst most studies found that a poor therapeutic alliance predicts dropout in BPD [5, 39, 40], a recent study of Barnicot et al. [4] showed that frequent use of skills in DBT was even more important than therapeutic alliance regarding treatment retention rates. A recent study investigating treatment characteristics and first-session relationship variables as predictors of dropout in traumatized youth, found that the level of approval of therapy after the first-session significantly influenced treatment completion rates [33]. We assume that a change of therapist after the first therapeutic contact may be experienced as a rupture in the therapeutic process that hypothetically harms the therapeutic alliance and thus influences treatment dropout.

Furthermore, comorbid disorders, such as a generally higher burden of axis I disorders [39], alcohol and substance abuse, and anorexia nervosa, have been found to be associated with dropout [27, 28]. It is known that patients with BPD and comorbid posttraumatic stress disorder (PTSD) are a particularly burdened population with an increased level of general distress, elevated numbers of suicide attempts and non-suicidal self-injury [21, 34]. DBT is increasingly recommended for PTSD, however, patients with BPD and comorbid PTSD benefit less from DBT as compared to patients with BPD alone [22, 41]. Accordingly, patients with BPD and comorbid PTSD are affected by many of the risk factors associated with dropout described above and may therefore be at risk for dropping out of treatment. Nevertheless, the relationship between comorbid PTSD and therapy dropout is poorly understood. Arntz et al. [3] found that childhood physical abuse predicted treatment discontinuation. To the best of our knowledge, there is no study targeting the effect of a comorbid PTSD on (DBT) dropout rates.

Results are also heterogeneous in terms of the impact demographic factors and clinical characteristics. Younger age has been associated with higher dropout rates [27, 29]. However, other studies did not find evidence for an influence of demographic factors [5]. In a recent study, Landes et al. [29] found an increased pretreatment level of general distress to be a significant predictor of dropout, whereas other studies did not find an association between dropout and pretreatment symptom load [8, 35].

To date there are no studies investigating the effect of a therapist change on treatment process of BPD. The purpose of this study was to investigate factors related to dropout from inpatient treatment in a large sample of patients with Borderline pathology. More specific, we hypothesized that (i) the change of the therapist between DBT-briefing and start of DBT-treatment as well as (ii) comorbid PTSD and childhood trauma history are associated with elevated dropout. In addition, demographic and clinical factors cited above that were reported to predict dropout in adult populations, should be controlled for.

This study investigated baseline variables and one therapeutic process variable related to dropout from 10-week inpatient DBT among patients with borderline pathology. Our hypotheses were partially confirmed. High childhood emotional abuse was associated with premature termination of treatment. However, higher CTQ-values of physical neglect during childhood were associated with a protective effect on treatment dropout. This was also true for comorbid PTSD. In addition, a change of therapist between DBT-briefing and treatment was associated with a significantly elevated risk of drop-out.

The dropout rate (24.7%) found in this naturalistic sample of treatment seeking patients with borderline pathology is comparable to dropout rates found in previous inpatient-DBT-studies [10, 25, 26, 35]. Reasons for dropout in around half of all cases was discharge on the patients’ own request (e.g. because of ambivalence with regard to the treatment, conflicts with fellow patients); the other half was discharged on part of the staff because of repeated dysfunctional behaviors (e.g. substance use, aggressive behaviors, negative therapy-interfering behaviors). In sum, reasons for dropout were balanced out. No patient explicitly mentioned a change of therapist being a reason for dropout.

This study supports the importance of therapy process variables as predictors of therapy dropout in borderline pathology. Patients that experienced a change of therapist between DBT-briefing and treatment were more vulnerable for dropping out of treatment. In the dropout-group twice as much patients experienced a change of therapist as compared to the completer group. Assuming that a change of therapist is a burden for the therapeutic alliance, the results of this study are in line with studies finding that a poor patient or therapist-rated alliance can predict dropout [37]. The early point in time of therapist change may be important. The briefing is used to gain an agreement on tasks and goals accompanied by empathetic resonance. Patients who agree to these tasks/goals and feel comfortable in the therapeutic bond within the very first session will decide to participate in the DBT program. Despite the fact that patients are explained that a change of therapist is probable, patients might nevertheless be affected through therapist change in many ways. For example, patients might be disappointed or frightened by the idea that goals or agreements reached in the DBT-briefing are not valid anymore or that the therapeutic bond is not the same. Patients with borderline pathology experience relationships in extremes of idealization and devaluation [1]. The therapeutic alliance that might be idealized within or in the aftermath of the DBT-briefing might switch to devaluation of the therapeutic alliance or of the whole treatment by a change of therapist. Trust in the treatment and commitment to change, which have also been shown to be associated with dropout [5], might decrease. The importance of the first session relationship variables for dropout later on in treatment has been shown in traumatized youth [33], also suggesting that the therapeutic alliance is a particularly important component of therapy in individuals with histories of childhood abuse [14].

In line with Arntz et al. [3], we found childhood abuse to be predictive of dropout in DBT. In our study, emotional abuse was predictive of premature therapy termination which was by trend significant in Arntz et al. [3]. Childhood physical abuse, unlike in the latter study, was not significantly associated with dropout. Sexual and non-sexual abuse can cause a wide range of mental health consequences [32]. Particularly emotional abuse is related to emotional regulation and interpersonal difficulties in adulthood, moderated by maladaptive schemas such as mistrust, abandonment, and shame [13, 31], that may complicate treatment and therefore lead to an elevated dropout risk. Therefore, our findings may be moderated by emotion regulation deficits (specifically non-acceptance of emotions) that have recently been found to predict DBT dropout rates [29]. As an inpatient setting is highly accommodative, physical neglect might have a protective effect on dropout. Basic needs (e.g. food, medical care) that might have been deprived during childhood, are fulfilled during the inpatient stay, thereby increasing the likelihood of therapy completion.

Comorbid PTSD, against our prediction, was not associated with elevated dropout rates. Indeed, a comorbid PTSD was even protective with regard to DBT dropout. It is known that patients with BPD and PTSD benefit less from DBT [20]. This is, however, not solely due to premature dismissal from treatment (dropout rate in the comorbid PTSD subgroup: 20.6%). Patients were committed to the treatment and they were not discharged more frequently on part of the ward for repeatedly showing dysfunctional behaviors, despite of their likely increased symptom load. This finding encourages treating patients with BPD and PTSD with specialized treatments for both disorders. Our hypothesis that an elevated general symptom load which is elevated in patients with BPD and PTSD in comparison to BPD patients without this comorbidity [22] would lead to elevated dropout rate was not confirmed, even though the PTSD patients significantly differed from the patients without PTSD in terms of general symptom severity (GSI). Opposed to the findings of Landes et al. [29], but in line with previous studies [8, 35] we did not find an effect of the general symptom severity predicting dropout rates. The protective effect of PTSD might be due to an elevated psychological strain; however, we also did not find a protective effect of GSI. Irrespective of their elevated general symptom severity or emotion regulation deficits, our study suggests that patients with comorbid PTSD have the motivation and/or competence to maintain an inpatient treatment. Suffering from PTSD on the one hand and the hope to be helped may also be the background of our findings. Therefore, this severely suffering patient group should not be rejected from treatment assuming them to be too unstable for psychotherapy. However, this result needs to be replicated.

Based on the results of the current study, patients with borderline pathology who experienced a change of therapist between DBT-briefing and treatment and childhood emotional abuse treatment are more likely to dropout from DBT. Additional research is needed to replicate these results in other – including outpatient - samples. Especially, it is necessary to better understand the therapy process variables involved, such as structural therapeutic elements of the therapeutic setting as well as the course of the therapeutic alliance. Furthermore, research should include other variables that may affect dropout (satisfaction with treatment, expectations of treatment etc.).