Predictors of dropout among personality disorders in a specialist outpatients psychosocial treatment: A preliminary study


Francesca Martino, D Psychol, Institute of Psychiatry, Bologna University, Viale C. Pepoli 5, 40123 Bologna, Italy. Email:


Aim:  The aim of this study was to identify factors that may affect treatment retention in a 1-year psychosocial program for adult personality disorders.

Methods:  The sample consists of patients admitted to the Adult Personality Disorder Outpatient Programme of the Bologna Community Mental Health Centre in the period 2003–2008. At the beginning of the program, patients were evaluated through a comprehensive assessment including sociodemographic form, diagnostic interviews and self-report questionnaires. Patients who dropped out from treatment were retrospectively compared with patients who completed the program.

Results:  Out of 39 patients enrolled in the program, 20 (51.3%) dropped out and 19 (48.7%) completed the treatment. Out of 20 patients who dropped out, 14 terminated the treatment within the first 2 months. The dropout group and the group which remained showed significant differences in diagnosis (borderline personality disorder [BPD]), demographic data (age, time from first contact with psychiatric services), clinical variable (impulsiveness) and subjective experience (motivation, treatment expectation, therapeutic relation perception and barriers to access). BPD and subjective evaluation were found to be predictors of premature termination in the sample. In detail, BPD patients who experienced a less satisfactory therapeutic relationship and reported many external problems were more likely to drop out of the program.

Conclusion:  Important factors contributing to dropout were identified, with potential implication for clinical practice. Further efforts need to be made to find ways to retain BPD patients who find the first subjective experience of the service more problematic.

SATISFACTORY THERAPEUTIC COMPLIANCE among patients with personality disorder (PD) is challenging and notoriously difficult to achieve in all treatment settings and especially when a borderline personality disorder (BPD) diagnosis is present.1–3 In particular, dropout, ranging from 50% to 60%, mainly occurs in the early phase of treatment.4–6 Despite the importance of this problem, very little research on predictors of premature termination of treatment for BPD has been carried out. Previous studies have investigated clinical and demographic variables and they have shown that factors such as young age,7,8 high levels of hostility and anxiety,9 anger and impulsiveness10,11 may be important for predicting dropout.

In addition, some authors have supported the relevance of considering patients' subjective experience for compliance, especially with PD. Gunderson et al.4 evaluated reasons reported from patients who dropped out from treatment in a specialist program for PD. Dropouts used to have a less satisfactory therapeutic relationship; they felt more criticized by the staff, they did not perceive that they were given enough support by their family and they appeared to be less motivated from the beginning of treatment. Recently Chiesa et al.12 used a semi-structured interview to assess qualitative data in patients with PD who prematurely left a community treatment program. In detail, they highlighted several significant areas in dropouts, who tended to have a less satisfactory relationship with the staff, more difficulty regarding treatment organization and delivery, or problems with living together in the community. Other authors13 have taken a qualitative approach, asking ex-residents themselves about their reasons for leaving the therapeutic community and about their experiences with clinicians and therapy. They used a semi-structured interview composed of some fixed dichotomic answer options and of a series of open questions to investigate new areas. Observing the results relative to fixed responses, they highlighted the most relevant subjective factors implicated in dropout, such as difficulties in changing dysfunctional coping mechanisms, external unfavorable circumstances, like pressure from their family and problems with the staff. Exploring the responses to open questions, they found that patients used to report a fear of stigma, problematic relationships with other residents and external barriers for attending an intensive treatment program.

According to these studies, interpersonal problems represent an obstacle for the development of a satisfactory therapeutic alliance.14 Group therapy has relevant implications in enhancing compliance or improvements. On one hand, a group setting seems to favor changes better than an individual setting through the mirroring process occurring between members. The mirroring process provides ‘hard-wired’ support for the group therapists and it could be powerful in leading clinical improvements and favoring work alliance among members. On the other hand, a group may elicit dysfunctional mechanisms, such as splitting and fragmentation, or behavior, such as manipulation among patients and staff. Teamwork is central to prevent splitting and to keep patients in treatment, by providing meaning, consistency and cohesion within the group.15,16

The aim of the present work is to evaluate the demographic and clinical variables, the patients’ subjective experience and to understand better the reasons that could impact on dropout in patients with a diagnosis of PD beginning a group therapy. In accord with the previous studies we consider as subjective data: relationships with the staff, motivation and external problems. In addition, we examine two more variables that seem relevant on a clinical ground: patients’ expectations of treatment and stress perceived during the assessment.


The Adult Personality Disorder Outpatient Programme (APDOP), developed in the Bologna Community Mental Health Centre (CMHC),15,16 was composed by a multidisciplinary team to provide specialist treatment for patients with PD. The program consists of: (i) 1 year of weekly group-based psychodynamic psychotherapy; (ii) regular contact (every 1–2 months) with the senior psychiatrist who prescribes psychotropic medications if necessary; (iii) regular contact with a psychologist and a nurse who provides individual psychological or social support depending on needs; and (iv) a psychometric assessment at baseline, regular test reviews every 6 months, and a 1-year follow up. Patients who are engaged in the program agree to participate for 1 year.

According to some authors,17–21 a long-term outpatient treatment provides significant improvements on a number of clinical outcome measures. It is believed that purely community-based approaches to the treatment of personality disorders may avoid an earlier admission to a hospital milieu with no detriment to effectiveness. Therefore, we intend to verify the efficacy of an adult outpatients program for personality disorders (APDOP) measuring clinical outcomes through an established assessment. So far we have observed some preliminary outcomes relative to the clinical program, but further studies are needed to generalize our data.

Every new patient with dysfunctional personality traits who requested a consultation in the Bologna Community Mental Health Centre in the period 2003–2008 was referred to the psychiatrist in charge of the program for a first evaluation. Patients are asked to fill out a written consent form before the first evaluation for the protection of their privacy. Patients with the following conditions were considered not to be suitable for the program: mental retardation, severe psychotic symptoms, moderate to severe mood disorder, moderate to severe suicidal ideation, and substance dependence. On the contrary, patients who were supposed to be suitable and available were asked to proceed into the baseline assessment and then into the program. Information about the group psychotherapy was provided, including the frame of scheduled appointments, the setting and the length of treatment.

Data of patients participating in the program were evaluated to investigate the impact of dropout and the reasons that may have affected it. The term ‘dropout’ is used in relation to patients who have left the program before the end; ‘early dropout’ refers to the patients who left within the first 2 months. The retrospective data collection and the study publication were approved by the Ethical Committee of the Bologna Local Health Unit on 30 June 2011 (EC code: 19OS11).

Baseline assessment

Patients who were found to be suitable for the program were asked to complete a sociodemographic and clinical form and to proceed to the psychometric assessment which comprised: (i) Structured Clinical Diagnostic Interview for DSM-IV – Axis II (SCID-II);22 (ii) Symptom Check-list-90-R (SCL-90-R);23 (iii) Aggression Questionnaire (AQ);24 (iv) Barratt Impulsiveness Scale (BIS-11);25 (v) Social Disability Scale (DISS);26 and (vi) Patient's Assessment Evaluation Questionnaire (PAEQ),16 which was appositely developed by the staff to evaluate subjective initial experience in service users.

The PAEQ is a 10-item self-report measure that evaluates five dimensions relevant to the assessment of the subjective initial approach of patients who are engaged in the program. It has already been used in a previous study16 to underline reasons which have an impact on treatment uptake. Service users are asked to self-rate their perception and their personal experience regarding: (i) relationships with professionals; (ii) patients' expectations regarding the effectiveness of the proposed treatment; (iii) their motivation to engage in long-term therapy and perceived need for change; (iv) the degree of distress elicited during the assessment process; and (v) the presence of possible adverse life circumstances and barriers to access into the treatment (for example, external life problems or logistic factors, such as timetable clashes, working or family engagements, etc.). The PAEQ is a five-point self-rated questionnaire which provides quantitative values. In contrast to other studies which adopted a categorical approach exploring the presence or the absence of subjective factors, we support a dimensional approach for all variables for the achievement of a better understanding of individual cases, for enhancing strategies and for maximizing adherence when patients' scores tend to be lower. Although the PAEQ has not yet been formally validated with regard to content validity and reliability, in common with other measures of satisfaction used to elicit opinions in surveys, it relies on face validity as a primary feature of usability. During the administration of the test, all subjects had no difficulty in clearly understanding what the individual items were addressing. Internal consistency for the PAEQ total mean score and individual dimensions were calculated from the raw scores obtained from the study sample. In a previous study16 the PAEQ showed a satisfactory level of internal consistency for the whole test (α = 0.82). Two out of the five dimensions (Therapeutic relationship = 7.42 ± 2.67, α = 0.82; and Expectation = 6.92 ± 2.72, α = 0.72) had a good level of Cronbach's alpha; another two (Perceived stress = 4.47 ± 2.33, α = 0.59; and Barriers to access = 6.47 ± 2.80, α = 0.53) showed a lower internal consistency and only one had an unacceptable value (Motivation = 7.45 ± 2.02, α = 0.30).

Statistical analysis

In order to ascertain the significance of differences between dropouts and completers, χ2-tests for categorical variables and independent sample t-tests for Likert-scale variables were used. Mann–Whitney tests were used when the distribution was not normal. In order to test the predictors of dropout, most significant variables were entered into stepwise logistic regression models. A backward procedure was conducted with ‘completers’ (yes vs no) as dependent variable and BPD, age and patients' subjective evaluation (total score) as independent variables. A second backward procedure was conducted to explore the PAEQ subscales, which had a stronger impact on the dropouts. ‘Completers’ (yes vs no) has been inserted as a dependent variable, while ‘motivation’, ‘expectation’, ‘therapeutic relation’, ‘stress’ and ‘barriers to access’ have been added as independent variables. All analyses were performed using the spss, Version 15 (Chicago, IL, USA).


Sixty patients were referred to the program, 13 (21.7%) subjects did not attend the assessment consultation and eight (13.3%) cases were found not to be suitable for the program, leaving a final sample of 39 patients. The majority of the patients were female (84.6%), young adults (mean age = 37.2; ±9.9 years) and single (74.4%). On average, patients had a relatively long history of psychiatric disorder (time from the first contact with the psychiatric service = 11.3; ±8.8 years).

With regard to DSM-IV Axis II diagnosis, 11 (28.2%) patients met criteria for a Cluster A PD (six schizotypal, eight paranoid and one schizoid); 29 (74.4%) for a Cluster B PD (26 borderline, seven narcissistic and five histrionic), 21 (53.8%) for a Cluster C PD (16 depressive, 11 avoidant, six dependent and five obsessive–compulsive). The majority of patients met criteria for two or more PD (62%). In detail, each patient showed a mean of 2.6 PD disorders. At the admission in the program neither mood disorders nor substance dependence were present.

Out of 39 patients with PD, 20 (51,3%) dropped out while 19 (48,7%) completed the program. Most of the dropouts (n = 14; 36% of the sample) were ‘early’. Because of the program abandonment, we are not able to provide a primary clinical outcome for patients who dropped out. Differences between dropouts and completers are reported in Table 1.

Table 1.  Comparison between patients with personality disorders that completed the program and patients that dropped out
 Dropouts (n = 20)Completers (n = 19)Test of significance
χ2, T or ZP
Sex, female n (%)17 (51.5)16 (48.5)0.011.01
Age, years, mean (SD)31.5 (8.3)43.4 (7.7)4.66<0.001
Personality Disorders, DSM-IV criteria, mean (SD)2.6 (2.1)2.5 (2.3)−0.100.91
 Cluster A, n (%)6 (54.5)5 (45.5)0.061.01
 Cluster B, n (%)18 (62.1)11 (37.9)5.260.03
 Cluster C, n (%)10 (47.6)11 (52.4)0.240.75
Time from first contact with psychiatric services, years, mean (SD)6.7 (6.0)16.2 (8.9)3.91<0.001
Symptoms Checklist 90 Revised, global indices
 General Severity index, mean (SD)1.60 (0.79)1.73 (1.01)0.460.64
 Positive Symptom Distress Index, mean (SD)2.35 (0.60)2.30 (0.73)−0.220.83
 Positive Symptoms total score, mean (SD)57.9 (19.18)61.68 (23.38)0.550.58
Aggression Questionnaire
 Total score, mean (SD)3.07 (0.61)2.92 (0.82)−0.640.53
Barratt Impulsiveness Scale
 Cognitive Impulsiveness, average rank (sum)20.5 (410.5)19.4 (369.5)−0.290.77
 Motor Impulsiveness, average rank (sum)21.9 (438.5)17.9 (341.5)−1.080.28
 Planning ability, average rank (sum)23.4 (467.5)16.4 (312.5)−1.90.05
Social Disability Scale
 Total score79.4 (28.9)62.3 (31.5)−1.770.08
Patient's Assessment Evaluation Questionnaire    
 Total score, mean (SD)2.63 (0.74)3.91 (0.53)6.17<0.001
 Motivation for change, mean (SD)3.37 (1.16)4.08 (0.69)2.290.02
Therapeutic relationship, mean (SD)2.81 (1.29)4.60 (0.54)5.56<0.001
Treatment expectation, mean (SD)2.52 (1.28)4.39 (0.56)5.79<0.001
Barriers to access, mean (SD)2.47 (2.29)4 (1.05)3.98<0.001
Perceived stress, mean (SD)1.97 (0.99)2.5 (1.29)1.40.17

Dropouts appear to be prevalent in cluster B and particularly frequent in cases of BPD (n = 17 out of 20; χ2 = 6.2; d.f. = 1; P < 0.01). Patients who dropped out were significantly younger and had a markedly shorter psychiatric history since the first contact with psychiatric services than completers. In addition, they showed a higher score on a sub-scale of BIS (planning disability), which is particularly compromised in BPD (Z = −2.3; P < 0.02), and a lower total score for qualitative variables regarding: motivation for treatment, quality of therapeutic relationship, expectation for treatment effectiveness and external circumstances which may affect program participation.

No significant differences related to sociodemographic variables (such as education, civil status, and occupation) or to clinical variables (such as aggressiveness, severity of symptoms, or social disability) were found.

From the backward stepwise logistic regression, considering together the less satisfactory initial approach (B = 3.8; odds ratio [OR] = 47.2; 95% confidence interval [CI] 3.93–566.3) and the diagnosis of BPD (B = 2.5; OR = 12.8; 95%CI 0.95−171.5), a strong association has emerged with dropout (R2 = 0.56). On the contrary, the association between dropout and age disappeared.

In addition, a second stepwise logistic regression was conducted to explore the PAEQ subscales impact on the dropouts. From the backward elimination procedure we observed significant results for two of the five subscales of the PAEQ. In detail, the quality of therapeutic relationship (B = 2.4; OR = 11.6, 95%CI 1.84−73.4) and the external circumstances (B = 1.4; OR = 4.1, 95%CI 1.18−14.3) showed the strongest impact on premature termination (R2 = 0.57).


We examined the dropout rate and its predictors in a group of patients with a diagnosis of PD admitted in a specific outpatients program in a CMHC. We found that:

  • 1The dropout rate is very high (51.3%) and it mainly occurs within the first 2 months of the treatment (36%) and in subjects with a diagnosis of BPD;
  • 2Dropout seems to be associated with patients' first impression of the program and of the staff; moreover external circumstances may complicate the access, including life events, timetable clashes, working or family engagements.

The high dropout rate we found in an early phase of BPD is widely supported by previous studies and probably depends on dysfunctional personality traits of these patients that mainly affect interpersonal relationships.3–8 The present research confirms these data and shows that dropout is strongly associated with patients' subjective perception, which is assessed through PAEQ and includes: therapeutic relationship, patients' expectations regarding the effectiveness of the treatment, and their motivation for change. As authors have previously suggested,12,13 BPD patients may have an unsatisfactory perception of the therapeutic relationship because of their dysfunctional traits.

In detail, the difficulty of BPD patients to connect with others in a functional and mature manner may impair a positive working relationship with treatment staff, undermining an ability to remain engaged in the program. This assumption has also been supported by other authors27 who have studied how the therapeutic relationship may activate specific dysfunctional attachment neural patterns in BPD patients, leading to a reduced mentalization capacity and favoring early dropout from treatment. On the basis of these findings, from the beginning, clinicians should adopt specific strategies to keep BPD patients in treatment. For enhancing compliance, we suggest a more systematic initial approach which should be focused on a specific component of the therapeutic relationship,28 such as (i) clear and common goals; (ii) agreement on tasks to achieve goals; and (iii) bond between therapist and client. The initial interactive process should start from the assessment phase. It is believed that a collaborative approach could enhance an early alliance and a better final outcome in such a difficult population.29–30

Furthermore, in our study patients' unfavorable external circumstances are significant predictors of dropout. In detail, problems such as logistic factors (work engagements that clash with session times) and external pressure by significant others (family/partner) are present in a significantly greater degree in the dropouts and they hinder regular attendance to treatment. These vulnerable and impaired patients may need active help and support by mental health clinicians in finding strategies to overcome obstacles to access treatment. We suggest that a deeper attention in external circumstances is required to provide psychosocial interventions for patients and for their families, when it is necessary.

In our sample, motivation for changing and for attending a therapeutic program is significantly lower in the dropouts group. In a previous study16 we also found that treatment uptake is strongly influenced by patients' subjective approach with clinicians, but we did not verify any differences in motivation between the two groups who refused/accepted the therapeutic offer. Conversely, patients who dropped out from treatment have significantly lower motivation. Because of the use of closed answers, we cannot really know which reasons lead patients to be less motivated for treatment. Further studies are required for a deeper understanding of that. However, we suggest that individual training of volitional competence might be helpful to increase patients' motivation and to improve their compliance. Previous researchers31 have described motivation techniques adaptable for patients with PD for improving volition competence, which could help to keep them in a treatment program.

Finally, expectations with regard to the usefulness and effectiveness of the proposed treatment were rated significantly lower by dropouts compared to patients who completed the program. It is necessary to explore patients' expectations from the beginning of the therapeutic program, through open questions, which need to be added for a deeper subjective evaluation, to clarify common goals and tasks which are aimed for and required. As we have already suggested,16 a better initial approach during the assessment is needed to highlight the patients' own needs and wishes on one hand, and to specify therapeutic aims in a clearer way on the other.

The present study has several limitations. First, the sample was relatively small and the results therefore need to be interpreted with caution. The small size is due to several reasons: the number of borderline patients who are usually referred to a psychiatric outpatients service is limited (15%) according to the scientific literature;32 many patients did not decide to attend the program due to its length and frequency; and the exclusion criteria were restrictive and did not allow all PD patients to be admitted to the program. In addition, even though the PAEQ shows a satisfactory level of internal consistency, the lack of information relative to its capacity to evaluate the construct that it was based on raises questions about the credibility of findings. To overcome this limitation, further investigations are needed, re-testing patients' subjective experience after improving the initial qualitative approach, as we have suggested. Finally, the PAEQ does not provide open questions to evaluate more deeply the relevant reasons which lead patients to drop out. Further investigations with free responses will be useful to better understand these problematic areas and to provide specific intervention to overcome these obstacles.


A better understanding of subjective factors that contribute to premature termination is required to improve the quality of the initial approach to BPD patients and develop strategies for increasing the adherence to the treatment. Among subjective factors which are found to be relevant in influencing dropouts, initial subjective impression of the relationship with clinicians and unfavorable external circumstances have a stronger impact in predicting premature termination. From this data, we highlight the possibility to operate on that, improving our service and providing better strategies to enhance compliance. An assessment procedure, which also considers relational and subjective aspects, would give patients an initial experience of what they might expect once in full treatment and to expose them to a therapeutic interaction with a professional that strives to understand their state of mind.


There are no conflicts of interest to disclose.