Feasibility of Peer-Delivered Suicide Safety Planning in the Emergency Department: Results From a Pilot Trial

With suicide rates having risen >30% in the past two decades before the COVID-19 pandemic (1–3), suicide is now the 11th leading cause of death in the United States (4). Many patients who end their lives had previously received treatment at an emergency department (ED) (5–8), and patients who present to the ED because of suicidal ideation or suicide attempt are at increased risk for future suicidal behavior (9, 10), Even a single visit to the ED for overdose, suicidal ideation, or self-harm is associated with an increased and persistent (months to years) risk for suicide relative to that of other ED patients (9, 11).

Given that EDs are open 24 hours per day and provide nearly half of all medical care in the United States (12), it is reasonable to assume that many, if not most, patients who experience suicidal thoughts will eventually present to an ED. Thus, the implementation of suicide prevention efforts in the ED is important. Unfortunately, general medical and mental health staff in the ED have limited time and training to maintain the fidelity of suicide interventions, and therefore brief, low-threshold evidence-based efforts (7, 13) are likely most feasible for implementation.

The safety planning intervention (SPI) (14) is a brief evidence-based intervention that has been shown to be acceptable and to effectively reduce suicidal behavior (14, 15) after ED discharge. The SPI provides a personalized list (i.e., safety plan) of coping skills and social support to patients at increased risk for suicide that can be used if suicidal thoughts reemerge. The SPI has been consistently identified as a best practice by multiple suicide prevention experts (16–20) and reduces return visits to the ED for suicidal behavior after discharge (14). A safety plan is usually developed by the patient in collaboration with a trained medical or mental health provider and takes approximately 20–45 minutes to complete (21).

Although higher-quality safety plans are associated with better outcomes (22, 23), and treatment of suicidal patients is included in the 2019 model of the clinical practice of emergency medicine (24), few ED physicians feel prepared to assess risk or to create a safety plan for patients with suicidal behavior (25, 26). Consequently, trained mental health workers, typically nurses or social workers, often take this role. However, most EDs do not have continuous availability of these trained workers (27, 28), which limits the ability of the typical ED to perform SPI.

In the context of suicide prevention, peers are individuals with lived suicide experience (i.e., recovering from suicidal thoughts or suicide attempt). Although the SPI is typically conducted by medical or mental health providers and is not overwhelmingly considered to be within the scope of peer practice, many of the skills used in the SPI (i.e., person-centered services, trauma-informed care, and collaborative relationships) are similar to the core competencies of peer specialists as defined by the U.S. Substance Abuse and Mental Health Services Administration (29) and others (30). Additionally, peers likely have more time to spend with patients than do medical or mental health providers.

Thus, a peer-delivered SPI in the ED might allow for more empathetic and compassionate care than general medical or mental health providers in EDs have the time to provide (31). Patients with suicidal ideation may perceive a peer to be more of a friend than a provider (32), and so they may be inspired to engage more meaningfully in the safety planning process. Peer-delivered SPI in the ED could be more acceptable to patients and result in more complete and higher-quality safety plans than an SPI administered by a busy ED clinician. Furthermore, studies from both outpatient and inpatient settings have shown that the use of peers to deliver suicide prevention efforts has been associated with reductions in emergency services use (33), reduced numbers of readmissions to a psychiatric unit (34), and increased scores on various recovery assessment scales (33, 35). However, no studies to date have prospectively evaluated peer-delivered SPI in the ED setting.

Accordingly, we conducted a pilot randomized controlled trial (RCT) with patients who presented to an urban general ED with suicidal ideation or after a suicide attempt, to document the feasibility, acceptability, and preliminary effects (defined as the ability of safety planning to prevent crises necessitating an ED return) of peer-delivered SPI compared with provider-delivered SPI. We hypothesized that participants in the peer-delivered SPI condition would find the peer-delivered intervention acceptable, given the peers’ focus on empathetic and compassionate care; would have safety plans that were both similarly complete and of similarly high quality as plans in the provider-delivered SPI condition, if peer training was appropriate; and might have longer safety plan completion times compared with participants receiving provider-delivered SPI but that this increase would neither increase ED lengths of stay (LOS) nor times to disposition because peers have no other clinical duties. Given previous conflicting results (14, 36) regarding the ability of ED mental health interventions to increase outpatient treatment engagement, we hypothesized that peer-delivered SPI either would have no effect on ED returns or would result in decreased ED use compared with provider-delivered SPI.

Methods

Clinical Setting and Participants

The study was performed in the Department of Emergency Medicine at the University of Arkansas for Medical Sciences in Little Rock, which annually serves approximately 65,000 patients and >1,200 patients with suicidal ideation. Potential study participants presenting to the ED were assessed for suicidal risk with the Patient Safety Screener–3, a validated instrument used for triage purposes (37). If a patient was determined to be at risk of suicide, this was flagged in the electronic medical record and the patient then approached by the peer after initial evaluation by an ED physician. Inclusion criteria for the study included presentation to the ED because of suicidal ideation or after suicide attempt; age 18–89 years, English speaking (because translators were not available and study materials have not been validated in other languages), willingness to provide informed consent, and no safety plan at the current ED visit. Patients were excluded from the study if they were incarcerated, unwilling or unable to complete a safety plan with either a peer or a provider, unwilling or unable to show the safety plan to a mental health provider, too psychiatrically ill to be approached safely, or if clinical personnel objected to study enrollment for any reason. The race-ethnicity of participants was not collected.

Research Design and Procedures

An RCT among eligible patients who presented to the ED because of suicidal ideation or suicide attempt was conducted to document the feasibility (assessed as quality and completeness scores of the safety plans, LOS in the ED or time to disposition, and time to develop safety plan), acceptability to patients, and preliminary effects (number of ED return visits poststudy) of peer-delivered SPI compared with mental health provider–delivered SPI. Random allocation of the participants in a 1:1 fashion to peer-delivered SPI or provider-delivered SPI was conducted by using Research Electronic Data Capture (REDCap) software (38). The randomization sequence was generated by the software, and allocation therefore was concealed until time of random selection.

Consecutive patients with suicidal ideation or after a suicide attempt were approached on Mondays and Tuesdays (November 1, 2019, to September 1, 2020) during study hours. Study enrollment was conducted on these days because available intramural funding limited peer time and because these days are among the busiest times for mental health presentations at the study site. All study procedures were approved by the University of Arkansas for Medical Sciences Institutional Review Board before data collection and were limited to a maximum of 37 participants because of the pilot nature of this work in a vulnerable population. Reporting of this study followed CONSORT guidelines (39).

SPI

The Stanley and Brown Patient Safety Plan Template was used (14, 21, 40). The template includes identification of warning signs for possible impending suicidal crises, internal coping strategies, people and social settings that can provide distraction, people to ask for help when in crisis, professionals or agencies to contact during a crisis, and ways to make one’s environment safe. Individuals are also prompted to list one thing most important to them that is worth living for. Participants were required to provide written responses themselves, and peers were trained to discuss only safety planning with participants. All completed safety plans were approved by mental health providers in the ED before being entered into the electronic medical record. If providers did not approve the safety plan, the peer was asked to work with the participant again until the plan was acceptable. Any additional amount of time was added to the safety planning process.

Safety Plan Grading

A research coordinator removed identifying information from all safety plans, and the plans were then graded blindly by two investigators (A.W., R.G.T.) with expertise in safety planning by using materials developed by Gamarra et al. (22). Safety plans were graded individually, then resolved by consensus, for completion (0, complete; 1, partially complete; and 2, complete; range of total scores 0–16) and quality (0, blank; 1, boilerplate; 2, some evidence of personalization; and 3, highly personalized; range of total scores 0–24).

Provider SPI Training

All medical and mental health providers at the study site participated in the ED-SAFE (Emergency Department Safety Assessment and Follow-up Evaluation) study (15, 41), which involved extensive training on the treatment of suicidal patients in the ED, including SPI delivery. No additional training was provided, because SPI competency was an assured standard of care in the studied ED.

Peer SPI Training

For this study, the term “peer” referred to an individual who had experienced severe suicidal ideation or had survived a suicide attempt in the past and who was, at the time of the study, state certified as a peer recovery support specialist. Peers were recruited specifically for this project and were provided approximately 12 hours of initial training (see an online supplement to this article).

Peer Supervision and Safety

Because peers had lived through personal trauma and had never held a similar role within the ED, debriefing and supervision were paramount. Peers received biweekly feedback by the study team on the quality and completeness of the plans and adherence to study protocol. No revisions to safety plans were allowed during these feedback sessions. The peers also received a weekly debriefing by a licensed clinical counselor throughout the study.

Measures

After creating a safety plan, each participant completed a brief survey containing demographic information (age, gender, lifetime suicide attempts, past-year attempts, and whether the participant had ever made a safety plan before the current visit), questions about acceptability (see below), and questions from the Columbia–Suicide Severity Rating Scale (C-SSRS) focused on lifetime and past-month suicidal ideation and behavior (42). The peers documented the number of plans that were initially unacceptable to providers, as well as the time it took to make a plan plus any revisions. Data collected from the electronic medical record included the patient’s chief complaint, time to disposition (defined as an order for admission or discharge minus the triage time), total ED LOS (defined as the time the participant left the ED minus the triage time), and the number of ED visits during the 3 months before and after study enrollment. Abstraction of data followed best practices (43, 44), including comparison of all data for accuracy.

Feasibility measures included duration of time spent on safety planning (including any revisions), time to disposition (defined as the time the disposition order was placed minus the triage time), total LOS in the ED, safety plan quality, and safety plan completion. Measures of acceptability included patient satisfaction, which was measured by having patients rate their safety planning process on a 7-point Likert scale (i.e., 1, strongly disagree; 2, disagree; 3, moderately disagree; 4, neutral; 5, moderately agree; 6, agree; or 7, strongly agree). Questions included “Did you like making this safety plan today?”; “Did you find completing a safety plan helpful today?”; and “Would you recommend completing a safety plan with a peer?” In the provider group, participants were additionally asked “Do you think working with a nurse, social worker, or doctor helped you more than working with a peer?” In the peer group, participants were asked, “Do you think working with a peer support specialist helped you more than working with the nurses, doctors, or other clinical staff?” Preliminary effects were measured by the change in the number of repeat ED visits made in the 3 months before the study and in the 3 months after the study and by the number of study participants whose deaths were recorded in the electronic medical record within the 3 months after study enrollment.

Data Analyses

We used t tests for continuous variables such as age; number of suicide attempts; quality and completeness scores of safety plans; ED LOS and time to disposition; the frequency of ED visits during the 3 months before an ED visit, when participants had not yet been enrolled or randomly assigned to an intervention; the frequency of ED visits during the 3 months after enrollment; and time to make the safety plan. Because ED times are typically nonnormally distributed (45), these data were log-transformed before the analysis. Chi-square tests were used for questions containing proportions, such as gender or percentage answering “yes” to a particular question. Changes in ED visits from 3 months before to 3 months after enrollment were analyzed with a Wald-type statistic, by using an F1-LD-F1 design (46). Acceptability measures assessed with the 7-point Likert scale were recoded to a 5-point scale for statistical analysis only, by collapsing “strongly disagree” and “disagree” into one data point and “strongly agree” and “agree” into another data point. The resulting 5-point Likert scales were then compared between groups by using chi-square tests. Given the limited sample size mandated by the internal review board, subanalyses of variables such as gender were not performed. All statistical analyses were performed with RStudio, version 2021.09.2.

Results

In total, 37 (39%) of 96 potential participants assessed were eligible for this study. The first four participants were withdrawn from the study because their data were collected in a nonstandard manner; staff were subsequently retrained. One additional patient in the provider-delivered SPI condition withdrew consent in the ED after undergoing random assignment. One patient in the peer group denied suicidal ideation after undergoing random assignment, leaving 31 participants for the analysis (see the flow diagram in the online supplement). All peer-delivered safety plans created in the ED were approved by providers without revision.

The sample included 15 females (47%), and the participants had an average age of 41 years (range 20–64). The groups did not differ by demographic characteristics or by key outcomes at baseline (Table 1).

The quality and completeness of plans significantly varied between the two groups, with patients in the peer-delivered condition having more complete (t=−3.96, df=29, p<0.001) and higher-quality plans (t=−3.84, df=29, p<0.001; see Table 2). The time to make the safety plan increased in the peer group (t=−5.92, df=29, p<0.001), but this increase did not significantly affect time to disposition or total ED LOS (both p>0.05). Participants equally liked making safety plans with the peers and providers, and both groups found the planning to be helpful.

We found no significant differences in visits of patients in both groups during the 3 months before enrollment or during the 3 months after enrollment. However, when we compared the change in the number of visits from 3 months before enrollment to 3 months after enrollment, we noted that participants who made a safety plan with a peer (Wald χ²=7.75, p<0.01), but not those who made a plan with a provider (Wald χ²=0.50, p>0.48), had a significant decrease in ED return visits. No participant deaths were recorded in the electronic medical record within 3 months after enrollment.

Discussion

The findings of this pilot RCT provide preliminary support for the use of peer-delivered SPI in the ED. Creation of safety plans was found feasible and equally acceptable regardless of group assignment. Furthermore, all peer-delivered safety plans were approved by providers with no revisions. These findings are consistent with the rapidly growing popularity of using peers to deliver mental health and substance use interventions worldwide. Peer-delivered interventions are especially popular in the United Kingdom, where peers are used in many mental health facilities (47). In the United States, more than 30 states now have some level of Medicaid reimbursement for peer specialists (48).

Although peers spent more than twice as long with the participant during creation of the safety plan compared with medical and mental health providers, this additional time affected neither the ED time to disposition nor the total ED LOS, perhaps because peers had no other clinical duties. Of note, safety plan quality and completion were higher in the peer group, and participants who made a safety plan with a peer also showed a reduction in visits in the period from 3 months before to 3 months after the study. Additional research is required to determine whether this finding was due to more complete and higher-quality safety plans, delivery of the intervention by a peer, or a combination of both.

This study was strengthened by its rigorous research design (i.e., as an RCT) and because the random allocation resulted in two groups that did not differ significantly in baseline demographic characteristics or other key potential outcomes. At the request of the internal review board, the sample size was small. This small size limits the findings somewhat, although the study nonetheless found statistically significant results, suggesting that further studies on ED peer support are appropriate. A second limitation (and simultaneous strength) was that the peers received frequent feedback on the quality and completeness of the safety plans, whereas the providers did not. This was a limitation imposed by the clinical environment so as not to disrupt care. Between-group differences may be attenuated if feedback were provided to both groups.

Conclusions

To our knowledge, this study reports the first RCT documenting the feasibility and preliminary effects of peer-delivered SPI, compared with provider-delivered SPI, in the ED. Safety plans created with peers’ guidance were more complete and of higher quality than those made with providers. Furthermore, although these plans took longer to create when participants worked with a peer instead of a provider, the increased time did not affect time to disposition or ED LOS. Additionally, peer-delivered safety planning was associated with reduced repeat visits to the ED in the subsequent 3 months. An RCT with a larger sample size is needed to determine the efficacy and cost-effectiveness of peer-delivered SPI in the ED and to identify both barriers to and facilitators of SPI’s widespread implementation.