Determinants of Completion and Use of Psychiatric Advance Statements in Mental Health Care in the Netherlands

Psychiatric advance statements (PASs) are important instruments for patients and clinicians who wish to prevent crisis situations and to handle them in a mutually agreeable way (1,2). PASs may also have a positive effect on the therapeutic relationship and patient satisfaction (3). The term “psychiatric advance statement” encompasses a variety of statements used in psychiatric care, such as crisis plans (CPs), joint CPs, and psychiatric advance directives. They can vary in form, content, and judicial context (4).

PASs are a promising practice, with equivocal evidence as to their effectiveness. Because the process of implementing evidence-based interventions may take more than a decade, it is not surprising that PAS implementation in the mental health system has been problematic (3,5,6). A U.S. study of outpatients at five public mental health services showed that only 4% to 13% of patients who wanted a PAS had completed one (7). Reasons for lack of PAS implementation include clinicians’ doubts about the need for them and clinicians’ reluctance to share decisions with patients (8,9).

We found only two studies that investigated determinants of PAS completion (3,7). One organizational factor that seems to be important for completion is the use of an independent facilitator. In one study, individuals were randomly assigned to two conditions—one with and one without a facilitator (3). In the group with the facilitator, 61% completed a PAS, compared with 3% in the group without the facilitator (3). Patient-level factors associated with PAS completion included an understanding of key PAS components, greater insight into the illness and the need for treatment, and adverse medication experiences (3,7). More evidence about patient- and clinician-related factors may help inform interventions to increase PAS completion rates.

Another aspect of PAS implementation is whether they are used during a crisis. Although PAS use seems logical from a clinical standpoint, many factors can interfere. In times of crisis, the plan may be overlooked by clinicians working in the crisis service because time is lacking or because no one is aware that a PAS exists. To our knowledge, this issue has not been studied previously.

Therefore, the purposes of this study were to identify factors associated with PAS completion and to determine how often PASs are used when a patient is in crisis. We expected that PAS completion would be more likely among patients with better psychosocial functioning (that is, greater illness insight, fewer symptoms, better social functioning, and fewer behavioral problems); those with more positive attitudes toward mental health services, the clinician, and the usefulness of a PAS; and those whose clinician had more positive attitudes toward the patient and the usefulness of a PAS.

Methods

Design

Data were derived from a randomized controlled trial (RCT) conducted in the Netherlands on the effects of a CP—a type of PAS—for patients with psychotic and bipolar disorders. This study, which has been described in detail (2,10), investigated the effects of CPs on the number of voluntary and involuntary admissions and outpatient emergency visits. The trial distinguished two intervention conditions and one control condition in which patients did not have a CP (no CP). In the first intervention condition, the CP was created solely by the patient and the clinician (CCP). In the second intervention condition, a patient advocate (PA) facilitated the process of CP completion (PACP) by preparing the draft and by supporting the patient in negotiating the specifics of the plan with the clinician. The study was approved by the Medical Ethics Trial Committee for Mental Health Organizations. Informed consent was obtained from all participants.

CPs

CPs comprise two aspects of a PAS: crisis prevention (for example, how to recognize early warning signs) and provision of practical information for future psychiatric care. Practical information is summarized on a small card—the “crisis card”—which users carry with them. The format of the crisis card distinguishes four domains: indicators of relapse, the patient’s care preferences in the event of a crisis, medical information, and contact information of relevant others. In the Netherlands, CPs are not legally binding and can be viewed as an advance agreement between patient and clinician.

Recruitment

We included adult outpatients who had had at least one crisis contact with mental health services or had been admitted to a hospital—voluntarily or involuntarily—in the previous two years. Participants were recruited from 12 community mental health teams in three mental health institutions in Rotterdam, the Netherlands. Eligible participants were selected on the basis of case notes (N=537); 151 refused to participate, and 174 could not be contacted. A total of 212 patients were included in the RCT (no CP, N=73; PACP, N=69; and CCP, N=70). All patients randomly assigned to the intervention conditions (PACP and CCP) were included in our analyses (N=139). Data were collected between 2008 and 2011.

Procedures

A researcher and two participating PAs trained the clinicians (N=93) in creating CPs during a two-hour session. Participants were psychiatric nurses, who were the patients’ primary clinicians. PAs were social workers with extensive professional experience in mental health care. Both worked for a patient advocacy organization, where their main task was to draw up CPs.

The following steps were taken to create a CP. At the first meeting with the patient, the procedure was discussed, information—for example, on crisis-precipitating factors—was collected, and strategies for preventing crises were developed. Over the next few meetings, a draft of the CP was prepared. When the CP was complete, it was signed by the patient’s psychiatrist, the clinician, and any others involved in the CP, such as partners or family. The final step was to summarize the CP on a card that was given to the patient.

Measures and Instruments

Outcome variables.

CP completion was operationalized as the handover of the crisis card to the patient. CP use was assessed during follow-up (at nine and 18 months). We interviewed the patient’s clinician, asking the following questions: “Has the CP been used over the past nine months?” (yes/no); “Who has consulted the CP?” (patient, clinician, family, or other stakeholder); “Which parts of the CP have been consulted?” (parts 1–8); “The instructions in the CP were followed/ignored during the crisis, because . . . ” (open question). Data from the nine- and 18-month follow-ups were combined. The occurrence and type of crises were assessed over 18 months. Service use data were collected from patients’ files and checked against the Rotterdam Region Psychiatric Case Register (11). Three categories were used for crises: 0, no crisis; 1, outpatient emergency visit or voluntary admission; and 2, involuntary admission (whether acute or court ordered).

Determinants.

Patient characteristics included chart diagnosis, gender, age, ethnicity, education, income, and marital status. Clinician characteristics included years of professional experience. The patient’s psychosocial functioning was assessed by an independent interviewer with the Health of the Nation Outcome Scale (12,13). The patient’s insight into illness was measured with the self-report Insight Scale (13). The patient’s level of service engagement as seen from the clinician’s perspective was measured by using the Service Engagement Scale (14). The quality of the therapeutic relationship was measured with the Working Alliance Inventory (15), which was scored from both the patient’s and the clinician’s perspectives. Expectations about the CP were measured by four questions, answered by both clinician and patient (10): “To what extent do you think the CP could be effective in preventing a crisis, recognizing a crisis, what to do in advance of a crisis, and what to do in case of a crisis?” The items were rated on a 5-point scale. Determinants were assessed at baseline, before randomization. Instruments used in this study have been described in detail elsewhere (10).

Statistical Analysis

Determinants of completion.

In the first step in the modeling procedure, univariate associations between CP completion and all determinants were analyzed with logistic regression analysis. For this purpose, nonnormally distributed variables were recoded into categories of approximately equal size. Determinants with a significance level of .25 were considered potentially relevant and retained for analysis (16). Next, all potentially relevant determinants and meaningful interactions between determinants were entered into a preliminary model and removed in a stepwise backward selection procedure on the basis of model fit. Collinearity between determinants was checked. The fit of the final model was evaluated by using diagnostic scatterplots of the standardized residuals and described using Nagelkerke’s R². The data were hierarchical—that is, patients were clustered within clinicians and treatment teams. However, sensitivity analysis using multilevel models produced no evidence that associations between determinants and CP completion varied across clinicians or treatment teams.

CP use.

Associations with CP use were explored by using logistic regression analysis. SPSS for Windows, version 21.0, was used for statistical procedures.

Results

Sample

Characteristics did not differ between the PACP and CCP conditions. Table 1 summarizes patient and clinician characteristics for the overall sample of 139 patients. Psychosocial functioning scores were consistent with average scores in populations of patients with psychosis (11). Illness insight was average to high. Service engagement scores indicated moderate engagement. Scores on the patient- and clinician-rated working alliance indicated positive therapeutic relationships. Scores on the expected usefulness of CPs indicated positive expectations. The patient and clinician working alliance scores were positively related (r=.33, p<.001). Furthermore, a positively rated alliance by the patient and by the clinician was related to better service engagement (r=–.28, p=.002, and r=–.61, p<.001, respectively).

CP Completion

Table 2 presents data on CP completion. The 93 trained clinicians were responsible for creating 139 CPs (median, one CP per clinician; range, one to three). Completion rates did not differ significantly between the PACP and CCP conditions, although clinicians in the CCP condition needed more reminders. Univariate analyses indicated the following potentially relevant (p<.25) determinants of CP completion: female gender, unemployment, psychotic disorder, a low education level, a higher level of service engagement, a positively rated working alliance by the patient and the clinician, fewer years of professional experience for the clinician, and positively rated expectations of the CP by the clinician.

Table 3 presents the final model. The probability of CP completion was lower when the working alliance was rated negatively by the clinician and when the clinician had more years of professional experience. Patients with a low education level were more likely to complete a CP than those with a moderate or high education level. We found no interaction effect between working alliance and education on CP completion.

Use of CPs

Of the 89 patients who completed a CP, 38 (43%) later experienced a crisis. Of these, 24 (27%) needed an outpatient emergency visit or were admitted voluntarily, and 14 (16%) were admitted involuntarily. Of the 50 patients whose CP was not completed, 32 (64%) experienced a crisis; 19 (38%) needed an outpatient emergency visit or were admitted voluntarily, and 13 (26%) were admitted involuntarily. The type of crisis did not differ between patients with complete and incomplete CPs.

The CP was consulted by the patient, clinician, family, or other stakeholder for a third (N=13, 34%) of the 38 patients who later experienced a crisis. For 24% (N=9) of the 38 patients, no information was available on whether the CP was consulted. CPs drafted by the clinician were consulted significantly more often than those drafted with the help of a PA (69% [N=9] versus 31% [N=4], p=.03). CPs were consulted equally by patients and clinicians. Of three patients whose CPs were consulted, no data were available on which part of the plan was consulted. For the other ten, the information consulted most often concerned the patient’s care preferences (N=6, 60%), followed by indicators of relapse (N=3, 30%) and medication (N=1, 10%).

Among the 51 patients who did not later experience a crisis, CPs were consulted for ten (20%) patients. In two of these cases, CPs were consulted by family members or others involved in the patient’s care, rather than by the patient or clinician.

The proportion of patients who had their CP consulted was greater among the 24 patients who needed an emergency visit or were admitted voluntarily (N=12, 60%; N with missing data=4) than among the 14 patients who were involuntarily admitted (N=1, 11%; N with missing data=5) (p=.02).

Discussion

Almost two-thirds of the PASs were completed successfully. PASs were not completed for one-third of the patients, irrespective of the patient’s consent. Completion was associated with a better working alliance from the clinician’s perspective, a lower education level of the patient, and fewer years of professional experience of the clinician. We found no evidence that PAS completion was related to the patient’s level of psychosocial functioning or attitude toward health care or the usefulness of a PAS. However, the likelihood of PAS completion was higher if the clinician had a positive attitude toward the patient. Among patients with a completed PAS, the plan was used during a crisis for only one-third, usually in the context of an emergency visit or voluntary admission.

Fifty-seven percent of PASs were completed only after the clinicians were reminded to do so. This result is in line with earlier U.S. research (3), in which 61% of PASs were completed after efforts were made to facilitate the process. In our study, we assume that clinicians gave priority to other treatment tasks or were prevented by practical barriers, such as lack of experience drafting such plans or lack of time. A legal requirement to create a PAS might have disadvantages. The statutory nature of such a requirement would risk making PAS completion an obligatory administrative procedure rather than a mutual agreement.

This study found no association between PAS completion and psychosocial functioning. This finding is supported by previous research on the effects of patient characteristics on PAS completion (3,7). Our finding that a better working alliance (from the clinician’s perspective) was positively associated with PAS completion has not been previously reported; however, it is in line with previous findings that clinicians’ reluctance to complete PASs is an important barrier (8,9).

It is noteworthy that previous studies found that greater professional experience reduced perceived barriers regarding PAS completion (8,9), whereas our own study suggests the opposite. Also, our finding that patients with a lower level of education had higher completion rates contradicts previous studies (3,7). We speculate that clinicians who are more senior have more difficulty with change, which makes them less likely to include PASs in their practice (17). Patients with a higher level of education may have stronger opinions regarding their treatment, and reaching an agreement with the clinician on PAS content may thus be more difficult (18). Swanson and colleagues (3,7) reported that patients’ positive attitudes toward PASs and toward health care contributed to higher completion rates, a finding we could not replicate.

Our results suggest that PASs are completed if a clinician believes treatment to be progressing. It is known that some clinicians experience problems forming a collaborative relationship with patients who have severe mental illness (19). We also found that the PASs were consulted for a greater proportion of patients who subsequently had an emergency visit or voluntary admission, compared with patients who subsequently had an involuntary admission. This may indicate that a PAS can positively affect the way a crisis is handled or that patients who are less severely ill or more motivated for treatment are also more willing to consult the plan when in crisis. More research is needed to examine this phenomenon (2).

Our results do not suggest that PAS completion was moderated by PA facilitation. However, previous studies have suggested that PA facilitation results in qualitatively better PASs and in higher completion rates (3,20). PASs drafted by a clinician were consulted more often. Therefore, we recommend that clinicians and PAs work together to improve PAS completion, especially when the alliance between clinician and patient is disturbed.

To our knowledge this is the first study to assess PAS completion in relation to patient and clinician characteristics and attitudes toward PASs and health care services. In addition, no previous studies to our knowledge have examined associations between PAS completion and the use of PASs during crisis situations.

Because of our limited sample size, associations between patient and clinician characteristics and PAS completion and use during crises should be interpreted with caution. Research is needed to replicate these findings. Although CPs are similar to PASs described elsewhere (4), specific features of the format or context might limit the generalizability of our findings. Also, data on PAS completion were collected in the context of an RCT on the effects of completed PASs. Because considerable efforts were made to encourage clinicians to complete PASs, the results cannot be generalized to a naturalistic context in which clinicians are not similarly encouraged. The completion rate would probably have been significantly lower without such monitoring. Our sample included only 40% of eligible patients, and although analysis did not suggest the presence of selection bias (2), the 40% rate might affect generalizability. The sample size limited the number of determinants that could be included in our analysis. Therefore, determinants, such as working alliance, were operationalized by using overall scores. A larger sample is needed to examine the impact of specific elements of these determinants. Finally, because clinicians provided data on PAS use, consultation of the plan by others might have been missed. In addition, the patient’s perspective would have provided valuable information.

Conclusions

We found that successful PAS completion was not related to patient and clinician attitudes toward the usefulness of the instrument. Instead, a positive working relationship between the patient and clinician was a predictor of PAS completion. Clinicians should be more aware of the potential benefits of consulting a PAS. Future research should assess how information from PASs is used during crisis situations and determine how use of this information affects the way a crisis develops.