Facilitation of Psychiatric Advance Directives by Peers and Clinicians on Assertive Community Treatment Teams

When a person with mental illness is in crisis and lacks capacity to give consent to treatment, he or she often must be treated involuntarily, which can be distressing and may make future voluntary care feel more coercive (1). Psychiatric advance directives (PADs) provide a legal mechanism for competent adults with serious mental illness to document their future crisis care preferences, consent to or refuse treatment during a crisis, and authorize a trusted surrogate to make treatment decisions for them during periods of decisional incapacity. A PAD may consist of an advance instruction (AI) for future health care, designation of a health care power of attorney (HCPA), or both. PADs are designed to empower people to direct their future treatment during times of crisis and give consent to treatment, thus avoiding involuntary treatment. Every state provides some legal mechanism for advance treatment planning in the event of an incapacitating mental health crisis; 26 states have enacted legislation specifically authorizing advance mental health care planning.

Previous research suggests very high consumer interest in PADs, but studies have found low rates of actual completion (2–4). PADs are challenging to complete without assistance, and there are misunderstandings of processes involved, concerns that directives may not be followed, and feelings of uncertainty about directing future decisions in unknown circumstances. Clinicians in routine fee-for-service treatment encounters may be unable to devote the time required to work with individuals to overcome these barriers and to complete PADs. Other assistance is often lacking.

An evidence-based intervention, the facilitated psychiatric advance directive (FPAD), was found to overcome most barriers to PAD completion when it was delivered by trained research associates in a controlled research setting (5). Findings also suggested that PADs can serve as effective recovery tools: PAD completion has been shown to yield significant benefits to consumers, including fewer coercive interventions (6), a better working alliance with care providers (4), and higher odds that individuals will receive and adhere to their preferred medications (7). Yet implementation of PADs lags far behind consumer interest, despite provider requirements to implement PADs as a condition of participation in Medicare and Medicaid (8), dissemination of research findings via online resources (http://www.nrc-pad.org), and awareness raising by advocacy groups (9,10).

In this study, we implemented the FPAD intervention in usual care settings as delivered by two kinds of community-based staff on assertive community treatment (ACT) teams: certified peer support specialists and nonpeer clinicians (referred to below as peers and clinicians). People served by ACT teams are high-priority candidates for PADs because they have serious mental illnesses (11) and are at high risk of experiencing repeated crises and potential loss of capacity to make health care decisions. PAD facilitation is likewise well aligned with recovery-oriented ACT teams and staff roles (12).

Successful implementation of FPAD by ACT team peers or clinicians would create a pool of potential facilitators. Peers and clinicians have different skill sets and work responsibilities that could affect PAD facilitation rates and quality. The peer’s role emphasizing empowerment and recovery may be a natural fit with PAD facilitation, because peers may be particularly able to help consumers achieve some recovery outcomes (13,14), serve as role models in recovery, and foster a person-centered process in successful PAD completion. On the other hand, clinicians’ training may make it easier to formulate treatment preferences into clinically feasible and effective PADs. If PAD facilitation can be conducted comparably well by a peer or a clinician, it may be more advantageous to deploy peers as PAD facilitators, given other clinical demands on clinicians (15).

The goal of this randomized controlled study was to implement PAD facilitation in one type of usual care setting and to compare the results for two different types of facilitators: peers and clinicians. We asked the following research questions: Is the rate of PAD completion in a community mental health setting not inferior to the rate established by research staff in the prior controlled experimental trial (4)? Is the rate of PAD completion different across peer versus clinician facilitators? Is the quality of PAD documents completed in the usual care setting not inferior to that of the prior trial? Does the quality of PADs differ between peer and clinician facilitators? What baseline consumer characteristics predict completion of a facilitated PAD?

Methods

Sample Selection, Screening, Recruitment, and Randomization

The sample consisted of 145 people with serious mental illness receiving care from one of six ACT teams in central North Carolina during 2013–2015. Study enrollment criteria included current receipt of ACT services; age 18–70; chart diagnosis of schizophrenia, schizoaffective disorder, bipolar disorder, other psychotic disorder, or mood disorder with psychotic features; ability to give informed consent to participate in a research study; not in crisis, jail, or a hospital at the time of study enrollment; likely ability to complete a research interview in English; no existing PAD; and interest in learning more about PADs.

Informed consent was obtained in person. Interviews averaged 45 minutes, ranging from 30 to 60 minutes, and the interviewer was blinded with regard to random assignment. Respondents received $25 at completion of a baseline interview and another $25 at completion of the follow-up interview. PAD completion was not a requirement of participation. Selected medical chart information was provided to the researchers by the ACT team.

Each ACT team designated a peer and a clinician to receive training in PAD facilitation from the research team. Consumers were randomly assigned, within teams, to facilitation from either a peer or a clinician. Facilitators were given four months to complete PAD facilitation with each randomly assigned consumer, but this time limit was extended during the study to accommodate facilitators and consumers.

This study was approved by research ethics committees at Duke University School of Medicine and entities associated with participating ACT teams.

Intervention and Training

The FPAD intervention is a manualized structured conversation to identify a person’s instructions for treatment in advance of a future period of incapacity (5). The facilitators are trained to listen carefully and ensure that the PAD document reflects the individual’s wishes. Peers and other clinicians were trained by a doctoral-level psychologist (LLM) according to procedures in the prior study (4). Consumers who did not wish to complete or notarize a PAD were still eligible for the follow-up interview.

Measures

PAD completion.

To measure the rate of PAD completion, we counted PADs that were completed, witnessed, and notarized according to North Carolina law.

Descriptive and evaluative content analysis.

A psychiatrist (MSS), blinded to the random assignment, coded PADs to categorize their structure (AI, HCPA, or both); information content; and quality with regard to feasibility, consistency with community standards, and usefulness and relevance of the information provided, taking into account diagnosis, local practice standards, and local health care resources. Four domains were assessed: medication preferences; hospital preferences; clinical information; and a summary assessment of overall consistency, feasibility, and usefulness. A 4-point scale (for example, very consistent, mostly consistent, somewhat inconsistent, and very inconsistent) was applied to the summary assessment and each of the three domains. We dichotomized each of the four variables for analysis (for example, 1, very or mostly consistent). The same rating scale was used in prior PAD facilitation research (4).

Other variables.

From participants, we collected information on age, gender, racial background, Latino ethnicity, years of education, marital status, residential status, working for pay in the last 6 months, problems associated with alcohol or substance use in last 30 days, arrest history, ACT team utilization, involuntary treatment, and the perception of ever having needed hospitalization but not being admitted. The baseline interview included the Insight and Treatment Attitudes Questionnaire (ITAQ) (16) to assess understanding of illness and the Colorado Symptom Index (CSI) (17) to measure psychiatric symptomatology. Perceived barriers to completing a PAD (for example, poor understanding of PADs, lack of trust in doctors, and no one to serve as a health care agent) were also evaluated (18). From the medical record, we collected information on diagnosis.

Statistical Analysis

The rate of PAD completion was calculated for two samples: the number of people assigned to facilitation at baseline (the intent-to-treat [ITT] sample) and the number who completed a follow-up interview (the retained sample). Unlike the ITT sample, by definition all individuals in the retained sample continued to be available and eligible for the study during the study window; thus, the retained sample was an arguably more stable subset than the ITT sample. PAD completion rates were compared with the previous study rate of 61% (4) by using one-tailed chi-square tests to test for inferiority of the community-delivered intervention to the prior standard, a directional hypothesis. Two-tailed chi-square tests were used to test the difference between facilitator types, because there was no hypothesized direction of effect. Comparisons of completion rates and quality were conducted by using survey methods to account for clustering of facilitators and consumers into six ACT teams (SAS PROC SURVEYFREQ using a Rao-Scott–modified chi-square test).

To assess the results of randomization across the peer and clinician study arms and the potential differences between the ITT and retained samples, we compared demographic and clinical variables by using chi-square tests for percentages and t tests for means.

Logistic regression was used to model the likelihood of PAD completion (either AI or HCPA), including variables such as facilitator assignment, demographic and clinical characteristics, and treatment experience. Candidate independent variables were tested in bivariate models and selected for the multivariate model when p<.1. We used survey methods to account for clustering by ACT team (SAS PROC SURVEYLOGISTIC). SAS 9.4 or Excel was used for all statistical procedures.

Results

Screening and Enrollment

ACT staff reviewed records for 553 consumers on six ACT teams during the study period, screened 357 for study eligibility, and forwarded consent to contact for 190. A total of 145 eligible individuals gave consent to participate in the study, completed baseline interviews, and were randomly assigned to FPAD by either a peer (N=71) or a clinician (N=74). Twenty-six became ineligible after randomization (17 because they were no longer involved with the ACT team), and three declined follow-up. The remaining 116 individuals (80% of those who enrolled in the study) completed a follow-up interview.

Sample Characteristics

Demographic, clinical, and attitudinal characteristics are presented in Table 1. Random assignment of consumers to peers and clinicians was successful, yielding peer and clinician study arms with no statistically significant differences between the ITT (N=145) and retained (N=116) samples on any of the variables listed in Table 1 (results not shown). The retained sample resembled the ITT sample on the listed variables; however, individuals in the retained sample had been with the ACT team twice as long as individuals in the unretained sample (41.2 versus 20.2 months). Study participants were clustered by ACT teams, which varied in size and contributed a range of 14 to 36 study participants each for the ITT sample and 11 to 30 each for the retained sample.

Completion of PADs

All individuals who were randomly assigned to a peer or a clinician and who remained eligible were approached about meeting with a PAD facilitator. The mean length of time between the baseline interview and declining or completing the FPAD intervention was about five months (162±70 days, for N=127 reported by ACT teams; data missing for N=18). The length of time between declining or completing the intervention and participation in the follow-up interview was 51±35 days (N=116).

PAD completion rates were somewhat lower than the 61% rate achieved under controlled study conditions, but the difference was statistically significant only for the ITT sample, not for the retained sample. Of 145 individuals who participated in the baseline survey (the ITT sample), 72 completed and notarized a PAD with their assigned facilitators, for a completion rate of 50% (one-tailed test, p=.030) (Table 2). Of 116 individuals who also completed the follow-up interview (the retained sample), 67 completed and notarized a PAD with their assigned facilitators, for a completion rate of 58% (a one-tailed test was not significant). Separate comparisons of completion rates for each facilitator type and the 61% completion rate obtained in the prior research study indicated no statistically significant differences.

Completion rates for peers and clinicians were not significantly different from each other, for either the ITT or retained sample. In the ITT sample, completion rates were 42% for peers and 57% for clinicians (a two-tailed test was not significant). In the retained sample, the completion rate was 51% for peers and 63% for clinicians (a two-tailed test was not significant).

ACT staff completed 72 PADs with consumers. Most (63%) were stand-alone AIs, and the rest were joint AI and HCPA documents. No stand-alone HCPAs were completed. ACT teams varied in their rates of PAD completion (35%−67%), in the percentage of completed PADs that were facilitated by peers as opposed to clinicians (25%−67%), and in the percentage of completed PADs that included an HCPA (9%−67%).

Quality of Completed PADs

PADs facilitated by ACT teams received quality ratings similar to those in the prior research study. The percentage rated as high quality (for example, mostly or very consistent) ranged from 86% to 93% across four measures in this study, similar to the range reported for the four measures in the prior study (83%−94%) (4). The percentage facilitated by peers and rated as high quality ranged from 97% to 100% across the four measures. The percentage facilitated by clinicians and rated as high quality ranged from 76% to 90%. The quality of PADs did not differ significantly between the two facilitator groups.

Predictors of PAD Completion

Six of the baseline variables were associated with completing a PAD (a p value of <.1 was used as a screen for inclusion in the multivariate model): having more education (12 years or more), being married or partnered, living in one’s own home or apartment, having a diagnosis of a substance use disorder or self-reported substance use problem, a greater length of time being served by an ACT team, and responding “no” to the question, “Have you ever felt that you needed to go into the hospital but were not admitted?” In multivariate analysis, four variables remained significant (Table 3). The odds of completing a PAD were more than two times greater for participants who lived in their own home or apartment compared with those with other types of residential status, such as living in a shelter or rooming house or with parents. The odds of completing a PAD were also greater (odds ratio=1.68) for those with a diagnosed substance use disorder or a self-reported substance use problem compared with those without such a diagnosis or problem. For participants who answered “no” to the question about having an unmet past need for hospitalization, the odds of completing a PAD were more than two times greater than for those who answered “yes” to that question.

Discussion

This study of PADs implementation showed that PADs can be facilitated, completed, and notarized in a community-based mental health setting by both peers and clinicians working on ACT teams. The quality of PADs was similar to that achieved under controlled conditions in a prior study (4). The PAD completion rate achieved in the ITT sample was somewhat inferior to the rate in the prior study of 61%; however, the rate for the arguably more stable retained sample (those who completed the follow-up interview) was not inferior to the rate in the prior study. The rates achieved by peers and clinicians were not statistically different from each other for the ITT and retained samples.

The modestly lower rate of PAD completion for the ITT sample in this study compared with the rate in the prior controlled research study is not unexpected given the more clinically challenging ACT population. The most conservative explanation for the somewhat inferior PAD completion rate in the ITT sample is the trend toward a lower completion rate with the peer facilitators, compared with the previous controlled study conditions (42% versus 61%).

Far fewer HCPA documents were created in this study than in the prior study. Of the total number of PADs facilitated by research staff in the prior study, 71% (103 of 146) were joint AI-HCPA documents and 8% (11 of 146) were stand-alone HCPAs. In the study reported here, 38% (27 of 72) were joint AI-HCPA documents and there were no stand-alone HCPAs completed. It is possible that more consumers on ACT teams lacked trusted social contacts that could serve as HCPAs, compared with the consumers in the prior study, or that the continuous availability of the ACT team made the HCPA function seem less compelling. It is also possible that ACT facilitators felt less comfortable facilitating HCPA documents because such documents seemed less clearly aligned with consumer self-determination, compared with AI documents. Such barriers could be overcome with additional training on the role and potential value of assigning a HCPA—for example, as an advocate for wishes expressed in the AI.

Individual characteristics at baseline associated with PAD completion in multivariate analysis included living in one’s own home or apartment and ACT team stability. These characteristics suggest higher functioning; however, scores on the ITAQ and CSI were not associated with PAD completion. The finding of a higher PAD completion rate among individuals with a diagnosis of a substance use disorder or a self-reported substance use problem is difficult to interpret. One possibility is that persons in treatment for substance use disorders have become more aware of their risk of future relapse and decisional incapacity and the need for a PAD. Compared with the group without substance use disorders, the group with such disorders reported higher rates of past arrest and unmet need for hospitalization, suggesting that the experience of arrest rather than treatment could motivate those with substance use disorders to create PADs.

The limitations of this study included a small sample and the distribution of individuals into six ACT team clusters, reducing the ability to detect group-level differences between facilitator types across clusters. We did not study the facilitation process, and this study had less rigorous fidelity assessment than in the controlled trial. Nevertheless, high quality ratings of completed PAD documents suggest that the intervention as delivered was faithful to the model.

Conclusions

This research demonstrated that PAD facilitation can be carried out by ACT team staff with individuals at high risk of losing decisional capacity. Peers and clinicians facilitated PAD completion at similar rates, and the rates compared well to rates in prior research (4). PAD facilitation may be considered as an expanded role for peer specialists in ACT and other community mental health settings, offering an additional workforce to complete PADs. For PADs to fulfill their promise, there must be enough PADs in health care settings so that providers implement them in mental health care, just as they have come to do in advance medical care planning. Peers and clinicians could play an important role in increasing the supply of PADs to achieve this goal.