Participants were 113 adult patients from a psychotherapy facility that is part of a large mental health care institute in the Netherlands. The facility offers specialized in- and outpatient treatments for patients with a PD as their primary classification, for a time-limited period of 36 weeks (nine months). After this 36-week period, patients receive follow-up aftercare in groups (‘resocialization phase’) for an extra couple of months, before terminating the treatment completely. Patients are only admitted to the facility when previous treatment attempts had been made and proved unsuccessful. Diagnostic classification was made by means of structured diagnostic interviews (i.e., the SCID-II, based on the DSM-IV Axis II classifications of PDs), administered by an experienced clinician. All patients received a standard program including group psychotherapy based on either cognitive behavioral therapy when mainly experiencing externalizing problems (i.e., group schema-focused therapy according to Farrell & Shaw, 2012), or psychodynamic psychotherapy when experiencing internalizing problems (according to Lemma et al., 2011). Both group psychotherapies were combined with other therapies, such as emotion-regulation training, sociotherapy, creative therapy, and pharmacotherapy whenever appropriate.

The current sample consisted of 88 females (77.9%) and 25 males (22.1%). Their mean age was 33.6 years old (SD = 9.7; range 18–65). The most frequent PD within the current sample was a PD not otherwise specified (PD-NOS; n = 56; 49.6%), followed by an avoidant PD (n = 30; 26.5%), borderline PD (n = 11; 9.7%), obsessive compulsive PD (n = 11; 9.7%); dependent PD (n = 4; 3.5%); and histrionic PD (n = 1; 0.9%). The prevalence of PD-NOS in the current sample is relatively high but corresponds with other studies (e.g., Verheul et al., 2007). Most patients (n = 89, 78.8%) received an intense treatment consisting of 4 to 5 days of treatment a week, which could be both inpatient or outpatient (daycare). The remaining patients (n = 24, 21.2%) received a less intense treatment of 2 days a week. There was no significant association between treatment intensity (four or five days a week versus two days a week of treatment) and treatment cluster (schema-focused therapy versus psychodynamic and mentalization Based Therapy), χ2 = 0.060, p = 0.807, nor between treatment cluster (Schema-focused therapy versus Psychodynamic and Mentalization Based Therapy) and diagnoses (Fisher’s exact = 7.43, p = 0.156), nor between diagnoses and treatment intensity (four or five days a week versus two days a week), Fisher’s exact = 6.81, p = 0.200. Also, there was no association between gender and diagnoses (Fisher’s exact = 3.36, p = 0.649).

The current study was approved by the ethical committee of the Faculty of Social Sciences of the Radboud University. Patients who started a new treatment in that particular psychotherapy facility between 29-02-2016 and 29-06-2017 were asked by the care manager to participate in the current study. During the intake phase, the diagnostician explained the content and goal of the current study. Patients received a written information letter and informed consent form. All patients were informed that they could continue treatment without any negative consequence, should they decide not to participating in the study.

Within the inclusion period, 204 new patients started their treatment within the center, of which 174 patients were actually asked to participate (85.3%) (some patients were not informed about the study due to absence of the relevant care manager, or because patients started treatment with urgency leading to a divergent intake procedure). Of these 174 patients, 133 patients agreed to participate (76.4% of 174 patients). Of the 133 initially included patients, 20 patients discontinued their treatment within the facility prematurely; three patients who were officially logged as drop-outs (i.e., initiative to quit treatment by the patient and not by the therapist), while the treatment of the remaining 17 patients ended for other reasons (e.g., by mutual agreement with therapist, or because of re-assessment and reallocation to another department within or outside Pro Persona). As most of these 20 patients quit treatment in an early phase of treatment, and none of them remained in treatment for a longer period than 6 out of 9 months, we chose to remove all these 20 patients from our dataset. This resulted in the final sample of 113 patients participating in our study.

Participants were asked to complete a set of paper questionnaires, including the Care Dependency Questionnaire (CDQ) and their wish for treatment continuation, multiple times during their treatment: After the first day of the introduction phase (T0; on average completed 2.05 weeks before the start of the treatment, SD = 2.40), as well as immediately after the start of the treatment (T1; M = 0.63 ‘weeks’ after the start of treatment, SD = 0.69), in the first phase of treatment (T2; M = 5.47 weeks, SD = 1.96), after three to four months of treatment (T3; M = 14.15 weeks, SD = 1.88), after almost 6 months of treatment (T4; M = 23.27 weeks, SD = 1.98); and in the final phase of treatment, between eight and nine months of treatment (T5, M = 32.21 weeks, SD = 1.96). The responses on these questionnaires were combined with patient information derived from the electronic patients files (e.g., DSM classifications), as well as with patients’ responses to other questionnaires that were part of the digitally Routine Outcome Monitoring (i.e., Outcome Questionnaire-45 [OQ-45] and the Severity Indices of Personality Problems-118 [SIPP-118]). Regarding the OQ-45 and the SIPP-118, scores were only used when the questionnaires were completed eight weeks before the treatment or within the first four weeks after the start of treatment for the pre-measurement, and between 28 and 44 weeks after the start of treatment for the post-measurement.

Patients’ dependency on their treatment or therapists was measured with the Care Dependency Questionnaire (CDQ; Geurtzen et al., 2018). The CDQ consists of 18 items, measuring three dimensions of patients’ care dependency. All items were rated on a 7-point Likert scale, ranging from 1 (totally disagree) to 7 (totally agree). Both the CDQ total score and the CDQ subscales have shown to have adequate reliability and validity (Geurtzen et al., 2018). Since earlier research hinted towards potential different associations between the different subscales and treatment outcome (Geurtzen et al., 2019), we decided to focus primarily on the three subscales separately.

The first subscale, patients’ Submissive Stance, consists of five items measuring patients’ compliant and docile stance in treatment, and patients need for advice of their therapists (e.g., “I present all my decisions to my therapist(s)”). Cronbach’s alphas in the current sample ranged from 0.71 (T4) to 0.80 (T2). The second subscale, patients’ Need for Contact with the therapists, consists of four items measuring patients’ wish to stay in touch with the therapists (e.g., “I dread ending the contact with my therapist(s) at the end of the treatment”). Cronbach’s alphas ranged from 0.77 (T1) to 0.87 (T4). The third subscale, patients’ Lack of Perceived Alternatives, consists of nine items measuring patients’ believe that there are no other options besides the current treatment to get rid of their symptoms, as well as patients’ feelings that only their therapist(s) can help them to keep going (e.g., “Only my therapist(s) can help me with my problems”). Cronbach’s alphas ranged from 0.82 (T1) to 0.89 (T5).

Next to the CDQ we also measured patients’ perceived dependency on their treatment at a very explicit level, by means of the single item “I am dependent on my treatment”, answered on a Likert scale ranging from 1 (totally disagree) to 7 (totally agree).

To measure the severity of personality psychopathology, the Severity Indices of Personality Problems was used (SIPP-118; Verheul et al., 2008). The SIPP-118 is an elaborate questionnaire that is standardly administered as the main outcome measure in the specialized treatment center. It’s concurrent, convergent, and discriminant validity in the Dutch language are good and the SIPP-188 is sensitive to change over a mid- to long-term treatment periods (Verheul et al., 2008). The SIPP consists of 118 items (e.g., “I can cope very well with disappointments”, “I know exactly who I am and what I am worth”, “It is hard for me to show affection to other people”), all rated on a 4-point Likert scale, ranging from 1 = fully disagree, to 4 = fully agree. The 118 items of the SIPP are part of 16 different facets of personality problems, clustered in five higher-order domains (Verheul et al., 2008). However, since we were mainly interested in one overall score reflecting the severity of personality problems, we used the sum score as our main treatment outcome variable. We pooled the items so that higher scores on the SIPP-118 reflect more personality pathology. Cronbach’s alphas based on all 118 items were 0.77 (T0) and 0.89 (T5).

General symptom severity was measured by means of the Outcome Questionnaire (OQ-45; Lambert et al., 1996; Dutch adaptation by de Jong et al., 2009). The OQ-45 consists of 45 items covering three different domains; symptomatic distress (25 items, e.g. “I feel no interest in things”); complaints or dissatisfaction regarding interpersonal relations (11 items, e.g., “I am concerned about family troubles”); and difficulties in social role (9 items, e.g., “I work/study too much”). All items were rated on a 5-point Likert scale ranging from 0 = never, 1 = rarely, 2 = sometimes, 3 = often, to 4 = almost always. Positively formulated items were pooled so that higher scores indicated higher levels of symptoms, more dissatisfaction with interpersonal relationships, and more difficulties in functioning, with a potential range of 0 to 180. Sum scores above 55 indicate a clinically significant impairment, and a decrease of 14 points or more over time should be considered as a reliable change (de Jong et al., 2009). The Dutch translation has shown to have adequate to good psychometric properties (de Jong & Nugter, 2004). Cronbach’s alphas in the current sample were 0.88 (T0), and 0.96 (T5).

Patients’ wish for treatment continuation was based on three additional self-construed items: (1) “Despite the fact that the current treatment is (almost) finished, I would like to continue my current treatment”; (2) “I am very motivated to continue my current treatment”; (3) “If I had the opportunity to continue my current treatment, I would do so”. All items were rated on a 7-point Likert scale, ranging from 1 (completely disagree) to 7 (completely agree). Cronbach’s alpha based on these three items was 0.83.

To validate the assumption that care dependency is a contextual effect, we determined whether care dependency scores varied over time during the treatment by means of a repeated measures MANOVA with Time (T0 to T5) as within variable, and the three subscales of the CDQ, i.e., submissive stance, need for contact, and the lack of perceived alternatives, as dependent variables. Next, a repeated measures ANOVA with Time (T1 to T5) as within variable, and patients’ scores on their self-perceived dependency item as dependent variable was conducted. As nonlinear patterns might be possible, we also explored any significant time-effects on multiple polynomial contrasts.

The second aim of the current study was to examine whether increases or decreases in care dependency during treatment were associated with patients’ symptom reduction and patients’ wish for treatment continuation. To do that, we first tested whether there were significant reductions in symptoms by means of a repeated measures MANOVA with time (T0 and T5) as independent variable and SIPP-118 and OQ-45 as dependent variables. Next, difference scores of the three subscales of care dependency, self-perceived dependency, OQ-45, and the SIPP-118 were calculated (T5—T0). Subsequently, two GLM-Multivariate tests were performed, both with the OQ-45 and SIPP-118 differences scores as dependent variables, and either the difference scores of the three care dependency subscales scores, or the difference score of the self-perceived dependency item as predictor variables.

Regarding patients’ wish for treatment continuation, we first explored the Spearman’s rho correlations (since not all variables were normally distributed) between the CDQ subscales, the self-perceived dependency item, and patients’ wish for treatment continuation (all at T5), and we explored the partial correlations between patients’ care dependency and their wish for treatment continuation, while controlling for the OQ-45 and the SIPP-118 scores at T5.