Cluster B personality disorders (PDs) (borderline, narcissistic, antisocial, and histrionic) represent around 13% of the clinical psychiatric population and roughly 2.5% to 3% of the general population (1, 2). Within this group, borderline personality disorder (BPD) is the most prevalent diagnosis in clinical settings. Cluster B PDs are not only common in both general and clinical populations, but they are also associated with a significantly reduced life expectancy, by approximately 9 years for women and 13 years for men at age 20, primarily due to elevated suicide rates and comorbid physical illnesses (1). People with cluster B PDs are also high health care service users: in one year, 78% of them have consulted a family physician, 62% of them visited a psychiatrist, 44% were admitted into the emergency room, and 22% were hospitalized (1). This vital health care use translates into high treatment costs, ranging from $15,000 USD to $50,000 USD per patient every year (3, 4).

Furthermore, many studies have shown that PDs are often comorbid, particularly with borderline personality disorder (BPD) (2, 5). While studying BPD in isolation through randomized controlled trials helps optimize sample homogeneity and design targeted psychotherapeutic interventions, clinicians in real-world settings must frequently manage significant comorbidity, especially among Cluster B PDs. Furthermore, patients with comorbid PDs tend to have a poorer prognosis and are less likely to achieve complete symptomatic remission (6). Therefore, given the high mortality rates, healthcare service utilization, and unfavorable prognosis for patients with comorbid cluster B PDs, it is essential to study this population as a whole.

Mentalization-based therapy (MBT) and dialectical behavior therapy (DBT) are two evidence-based treatments designed for individuals with borderline PD (7, 8). MBT is grounded in psychodynamic theory and attachment theory, with a specific focus on mentalization (9). DBT is a type of cognitive-behavioral therapy that integrates Eastern mindfulness practices to enhance a person’s ability to regulate distress, accept experiences, and manage interpersonal emotions (10). While meta-analyses have established that these treatments have a moderate effect size (Setkowski et al., 2023), their implementation in real-world clinical settings presents several challenges. Patient populations are more heterogeneous, comorbidities are frequent, and therapists differ in their level of experience and training, especially in managing the clinical complexity of Cluster B PDs. Thus, although their efficacy is well established—and even explored in real-world settings for patients with borderline PD (11)—their efficiency in a broader PD population remains underexplored. Similarly, beyond the documented efficacy of specialized psychotherapies in research settings, treatment dropout has been the subject of numerous studies for borderline PDs (12). However, none have specifically focused on the population of individuals with Cluster B PDs.

In this retrospective, naturalistic study, we examine the outcomes of MBT and DBT as delivered in routine clinical practice for adult patients with predominantly BPD along with other Cluster B PDs. Specifically, we aim to describe pre-post differences of these treatments in a real-world setting by analyzing two key clinical indicators during the first year of therapy: the number of emergency department (ED) visits and psychiatric hospitalizations. These ‘hard outcomes’ are frequently used in the literature to evaluate clinical trajectories, with reductions typically reflecting a more favorable prognosis (11, 13, 14). Additionally, we investigate treatment completion rates by comparing dropout rates before and after the start of therapy. Given the naturalistic design, absence of a control group, and lack of a priori hypotheses, all findings are interpreted as associative and exploratory rather than as evidence of treatment effects or comparative effectiveness.

A total of 288 participants began treatment and were included in the descriptive outcome analyses: 130 patients in MBT and 158 in DBT. Table 1 summarizes the baseline sociodemographic and clinical characteristics of the two groups, as well as the 75 patients who dropped out before treatment began. Notably, most of the participants in both treatment groups were female, with a female-male sex ratio of 1.9 in the MBT group and 10.3 in the DBT group. In the MBT group, 29.2% of participants have a narcissistic PD, as opposed to 4.4% in DBT. Most participants in both groups were single (69.3% in MBT and 71.2% in DBT) and of White origin (90.3% in MBT and 79.5% in DBT). Notable descriptive group differences included age (p = .01), sex ratio (p <.001), and the prevalence of borderline (p = .007) and narcissistic (p <.001) PDs. These differences reflect the non-randomized, clinically assigned nature of the treatment allocation.

Table 2 summarizes the results concerning pre-post differences in ED visits and hospitalizations, as well as dropout trends for the MBT and DBT groups.

To our knowledge, this study is among the first to descriptively report real-world outcome patterns on patients with predominantly BPD and other cluster B PDs referred to either MBT or DBT in an outpatient setting. Several descriptive observations emerge from the data. Differences in diagnostic profiles at treatment entry were observed: clinicians tend to refer more patients with narcissistic PD to MBT and those with borderline PD to DBT. These patterns likely reflect clinician judgment within a non-randomized allocation process rather than differential treatment indications. Overall, we observed a decrease in ED utilization in both groups during the first year of treatment. Observed reductions in psychiatric hospitalizations were limited to the MBT group, while no clear change was observed in the DBT group. Finally, while the overall dropout rate during treatment was similar between the two groups, the dropout trajectories differed: it was steady over time in the MBT group but concentrated in the early stages of treatment in the DBT group. Additionally, a substantial proportion of patients dropped out before initiating psychotherapy, particularly those from low-income backgrounds.

First, the observed longitudinal decrease in healthcare utilization aligns with findings from controlled studies in patients with BPD (13, 14). These patterns give rise to the hypothesis that participation in structured, specialized psychotherapy may be associated with reduced reliance on ED services during treatment, even in heterogeneous, real-world populations, by improving crisis anticipation, emotional regulation, and help-seeking behaviors. These therapies may also foster a stronger therapeutic alliance and a sense of containment, allowing acute distress to be managed within planned clinical contacts in therapy sessions rather than through unscheduled emergency visits. The absence of a clear reduction in hospitalizations may reflect multiple factors, including the relatively small number of hospitalization events, the inclusion of a less severely affected population, at least about hospitalization risk, compared to those enrolled in controlled trials, and/or a lack of statistical power. These findings should therefore be interpreted cautiously and as descriptive only. Furthermore, differences in service were only observed before and during the first of year therapy. Further research is needed into determining whether these differences persist after treatment concludes.

Secondly, dropout rates observed after one year were substantial and consistent with those reported in a recent meta-analysis by Arntz et al. (12). Regarding descriptive trends across three-month intervals, our findings similarly showed a steady attrition rate of approximately 7–8% per quarter for MBT, in line with Arntz et al.’s observations (12). Specifically, in the DBT group, the overall dropout rate (31.0%) was higher than in Linehan’s original trials—19% (1991) and 17% (2006)—but comparable to the 36% reported in a UK trial by Zimmerman et al. (15–17).

Notably, dropout in the DBT group was highest in the first quarter (16.5%), reflecting a pattern of early disengagement also reported by Arntz et al., who observed a similar rate of approximately 20% (12). Although the temporal dropout patterns qualitatively differed between treatments, with DBT showing early attrition and MBT a more evenly distributed dropout rate, it is hypothetically plausible that these trends likely reflect structural differences between the therapies, variations in diagnostic and sociodemographic characteristics of the patient groups, and the absence of specific program components in our setting, such as telephone coaching in DBT, which may have contributed to the higher early dropout rate. Future studies specifically designed to examine treatment processes would be required to test these hypotheses.

It is essential to highlight that approximately 20% of participants eligible for either therapy dropped out before beginning their first group session. This group appeared to have greater clinical complexity, with a higher proportion of patients having four or more concurrent mental disorder diagnoses and slightly more PD diagnoses than treatment starters. This pattern may indicate that more clinically complex patients may face greater barriers to engaging in intensive psychotherapy. Additionally, socioeconomic characteristics further distinguished this group. None of these individuals reported an annual income above $50,000 CAD, and approximately 70% earned $20,000 CAD or less. These findings give rise to the hypothesis that socioeconomic factors such as poverty and unemployment may pose significant barriers to engaging in psychotherapy. This is consistent with a systematic review by McMurran et al., which found that treatment non-completion in PDs is associated with unemployment and lower levels of education and occupational status (18). In addition, it is plausible that clinician-level and systemic factors, including implicit biases and service-level selection processes, may contribute to unequal access to care for individuals with low socioeconomic status (SES), further exacerbating mental health disparities. These early dropouts likely introduce a selection bias into real-world effectiveness studies, since the most clinically complex and socioeconomically disadvantaged individuals are systematically underrepresented in analyses. Although the observed differences in our study were relatively modest, increasing awareness of the link between SES and therapy engagement may encourage clinicians to reflect critically on their own referral and selection practices (19).

A key strength of this study is its naturalistic design, which enhances ecological validity by capturing the realities of routine clinical practice.

Nonetheless, several limitations should be noted. The retrospective, non-randomized design limits causal inference and introduces potential confounders such as therapist bias and variability in clinical background. MBT was delivered by clinicians with extensive psychotherapy experience, while a team with less general experience but formal DBT certification implemented DBT. The DBT protocol used was also adapted rather than standardized, reducing comparability with previous trials. Treatment delivery was further affected by disruptions related to the COVID-19 pandemic, particularly for patients who began therapy in late 2019.

Additionally, the reliance on hospital records may have led to underreporting of ED visits and hospitalizations occurring outside our institution. However, the sectorized structure of mental health care in our region likely limited this impact. The sample was primarily composed of individuals with borderline and narcissistic PDs, restricting generalizability to other Cluster B subtypes. The exclusion of non-service users from the ED visits and hospitalization differences may also bias estimates and limits generalizability to a heterogeneous general adult psychiatric population referred for either type of therapy. Finally, although dropout was clearly defined, the absence of systematic documentation of reasons for discontinuation limits our ability to interpret adherence and treatment engagement.

This naturalistic study highlights that both MBT and DBT, when integrated into routine care for adults with Cluster B personality disorders, are descriptively associated with reduced emergency service use during the first year of treatment despite high dropout rates and differences in patient profiles. These observations underscore the real-world complexities of structured therapies delivered by diverse clinical teams. Future research should prioritize identifying patient-, therapist- and systemic-level predictors of engagement and to advance personalized care for patients with Cluster B personality disorders. Understanding the factors influencing treatment allocation is essential, as is exploring potential sources of stigma, such as socioeconomic disadvantage, that may hinder access to care.