This review aims to complement quantitative reviews and meta-analysis (17–19) by synthesizing the existing qualitative, peer-reviewed evidence regarding the experiences of diverse stakeholders (consumers, support persons, and clinicians) involved in the SPI. These stakeholder experiences include but are not limited to: what is perceived as helpful and unhelpful about safety planning; what processes facilitate positive effects; the collaborative process regarding how the safety plan is developed, used, accessed, and revised; as well as the perceived impact of the safety plan on suicide-related outcomes and other well-being indicators.

The search strategy was developed by MF, based on a previous safety planning systematic review (17), and refined in consultation with an academic librarian. We conducted searches on 28 November 2023 in seven databases: Embase, Emcare, MEDLINE and PsycInfo, in the Ovid platform; as well as CINAHL, Scopus and Web of Science. The final search strategy was broad, including terms for safety planning and suicide. Additional terms were trialed (e.g., for participant groups and study designs), however these restricted results and were excluded from the final strategy. We limited results to English language and a publication date range of 2000 to present. See Supplementary Data Sheet 1 for the search strategies used in each database. Reference lists of included articles were pearled in duplicate (MF, EO, KR) for potentially relevant studies.

Search results were imported to EndNote 21 (Clarivate, Philadelphia, USA) to manually identify and remove duplicates (MF). We screened the remaining results using Covidence (Veritas Health Innovation, Melbourne, Australia) in two stages, in duplicate: 1) title and abstract screening (MF, KR); 2) full-text screening (MF, EO, KR). Reviewers discussed any disagreements until 100% consensus was reached.

Eligibility criteria included: published in English language; qualitative in design (or mixed-methods, but where qualitative data were able to be extracted); participants of any age who had direct involvement in safety planning (including consumers, support persons, service providers, clinicians, etc.) in any setting (e.g. emergency department, inpatient, outpatient, community, online, school, etc.); and where it was clear that safety planning was based on the Stanley and Brown (7) version. Studies could include the SPI as a standalone intervention, or as part of a wider intervention approach. Studies were excluded if they: were not published in English; were not primary research; were not qualitative in design (either purely quantitative or where qualitative method and data could not be extracted); participants had no direct involvement in safety planning; or where the type of safety planning intervention was irrelevant or unclear (i.e., no reference to Stanley and Brown, and/or no definition or description of safety planning procedures).

We custom-built an electronic survey (LimeSurvey, Hamburg, Germany) to extract key information from the included studies, including: aim; study location and setting; study design; participant characteristics (sample size, population description, age, sex); SPI details (delivery modality, format, other intervention components if relevant); methods of data collection and analysis. Reviewers (MF, EO, KR) extracted data independently, in duplicate. Where necessary, we discussed and consulted the original papers until consensus was reached.

As part of the data extraction phase, and to facilitate the meta-aggregation process, we read and re-read included studies in duplicate (MF, EO, KR) to extract individual findings (i.e., authors’ analytic interpretative statements of qualitative data) and accompanying illustrations (i.e., verbatim participant quotation that exemplifies the finding). Any verbatim analytic statement was eligible to be extracted as a finding, provided an accompanying illustration was available. Where an accompanying illustration was not available, the finding was not included in this review. As per JBI guidelines (24), we (independently and in duplicate) assigned finding and illustration pairings a credibility rating: unequivocal (i.e., illustration supports the finding beyond reasonable doubt and therefore not open to challenge), credible (i.e., illustration lacks clear association with the finding and is therefore open to challenge) or not supported (i.e., illustration does not support the finding).

Risk of bias assessment was conducted for each eligible study independently by three reviewers (MF, EO, KR) using the JBI Checklist for Qualitative Research (25). In this 10-item tool, each item is rated as: yes, no, unclear, or not applicable. We resolved discrepancies via discussion, re-checking the papers together, and discussion with a fourth author (NP) as required. As per recent guidelines for ensuring review results represent the best available evidence (26) eligible studies were included if they satisfied at least six criteria on the appraisal tool.

Qualitative findings were pooled via meta-aggregation (24). Findings, illustrations, and credibility data were exported and printed for repeated reviewing in hard copy and for discussion in duplicate by two authors (EO, KR). Using butchers paper, we manually grouped the printed findings into categories based on our discussions. We first placed findings into categories based on similarity of meaning. Second, we combined similar categories into ‘synthesized findings’, referring to indicatory statements that convey the whole, inclusive meaning of a collection of categories, and which can be used to develop policy and practice recommendations. We then transferred these hard copy synthesized findings back to an Excel spreadsheet for discussion with the wider team. Following team discussion, we prioritized these synthesized findings into conceptual order for presentation in the manuscript.

As per JBI guidelines, we used the ConQual approach (27) to establish confidence in each synthesized finding. ConQual argues that confidence in a meta-synthesized finding is determined by the dependability and credibility of the studies and individual findings that comprise it. Confidence ratings range from high, moderate, low, to very low. By default, qualitative studies are initially given a ‘high’ confidence rating, which can be downgraded based on dependability and credibility. Dependability is determined based on performance of each study on items 2-4 and 6-7 of the JBI Checklist for Qualitative Research, with the overall confidence level unchanging if the majority of individual findings are from studies with 4-5 ‘yes’ responses, downgraded one level for majority 2-3 ‘yes’ responses, and downgraded two levels for majority 0-1 ‘yes’ responses. For credibility, where a synthesized finding contains only unequivocal individual findings, no downgrading penalty is applied; however, confidence is downgraded one level if the synthesized finding comprises a mix of unequivocal and credible individual findings.

The overarching qualitative methodology guiding this review was an interpretivist approach, which recognizes subjectivity and reflexivity (28). This approach makes the perspectives and positioning of the authors explicit, ensuring that the impact of researcher lenses on the synthesis and examination of results is transparent. While the components of the SPI should be universal, we acknowledge our positioning in the Australian context, which is associated with a unique set of cultural factors and policy frameworks that influence SPI practices and implementation. It is also important to acknowledge the authors’ backgrounds. Collectively, the research team brings expertise across lived experience, clinical practice, and research. EO is a postdoctoral researcher with expertise in behavioral science and mental health. KR is an experienced mental health nurse and doctoral level health psychologist working in research and education. NP is a professorial level mental health nurse expert and leader in suicide prevention research and education. ML is a Lived Experience academic. AP is a PhD researcher in health and medical sciences and Expert by Experience with the SPI. J-AR is a mental health nurse expert in clinical and senior management. SP is an experienced mental health nurse. MF is a senior suicide prevention researcher.

Database searching yielded 1862 results, reduced to 588 after removal of duplicates. Results were screened at the title/abstract level, leaving 60 eligible for full-text screening. One additional article was identified via a correction that appeared in the search results. No further articles were identified through reference list pearling. Twelve eligible studies were critically appraised; two (15, 29) were excluded by the minimum risk of bias threshold, leaving ten studies for inclusion. See

Figure 1 for the full screening process, and Supplementary Data Sheet 2 for a list of all ineligible full-text results.

Included studies were published between 2015 and 2023 and primarily conducted in the United States (n =7). Results for this review are based on data from n = 243 participants (note: this relates to the total number of participants from eligible phases of the included studies). The mean sample size was 24 (range, n=12-50). Across all studies, participants included n = 113 clinicians/staff (n=5 studies), n = 103 adults (including 95 veterans, n=4 studies; and 8 general population, n=1 study), n = 20 adolescents (n=2 studies), and n = 7 support persons (n=2 studies). Eight studies included both female and male participants, two did not report any gender data, and none reported data on other gender identities. Study settings included combined inpatient and outpatient (n=4), outpatient only (n=4), emergency department (n=1), and community services (n=1), with six studies relating to the context of veterans.

Six studies were purely qualitative (10, 30–34), one was mixed methods (35), and while a further three identified as qualitative they also included some minor quantitative aspects (e.g., quantitative measures to collect participant clinical information, 36, 37; or quantification of time spent creating safety plans, 38) but were not considered mixed methods. Most studies (n=8) collected qualitative data via semi-structured individual interviews but focus groups (n=1) and open-ended survey items (n=1) were also used. Studies analyzed qualitative data using thematic analysis (n=4), content analysis (n=2), interpretive phenomenological analysis (n=1), and matrix analysis (n=1). Two studies did not clearly report an analytic method.

There was substantial variability across studies in SPI features, and its role in suicide prevention and mental health care. Studies discussed versions of the SPI including additional components such as text-message and/or telephone follow-up support (31, 35), and the inclusion of support persons (36). Most studies (n=9) used or discussed the SPI as one component of care, alongside other psychological interventions (e.g., individualized, outpatient psychotherapy). The specific format of initial construction, ongoing access, or both, was often unclear. Only three studies described a specific SPI format, including a traditional hard-copy format (33), a mobile phone app-based version (30), and either hard copy or electronic versions (38). There was also a lack of detailed reporting regarding delivery modality, with four studies (30, 31, 33, 38) clearly indicating in-person creation of the SPI, and one describing a group-based SPI delivered online via telehealth (37). Eight studies described who the SPI was co-created with – working with a clinician was the most frequent approach (10, 30, 31, 33, 34, 38), with one study describing construction with a study counselor (35), and another describing a collaborative creation process with other SPI users in a group format (37). See Table 1 for full characteristics of included studies.

Included studies performed well on critical appraisal items related to congruity between research methodology and study methods, as well as ethical research conduct and appropriateness of study conclusions. However, guiding philosophical perspectives were largely unreported, with only one study mentioning this (32), and studies did not consistently meet criteria for reflexivity, with only one study (32) locating the researchers culturally or theoretically, and two studies (10, 32) discussing the influence of the researcher on the research and vice-versa. See Table 2 for study-level critical appraisal results.

Ninety findings (82 unequivocal; 8 credible) related to stakeholders’ experiences of the SPI were extracted and aggregated into 14 unique categories according to similarity of meaning. Four synthesized findings (one moderate confidence and three low confidence) were developed via meta-aggregation. See Table 3 for a summary of the findings and categories used to create each synthesized finding, and Supplementary Table 1 for full ConQual results. Complete details of individual findings and illustrations are presented in Supplementary Table 2 . We provide a narrative description of each synthesized finding and associated categories below.

Consumers, clinicians, and support persons viewed the SPI as broadly acceptable and beneficial for reducing consumers’ suicide risk. These qualitative data concur with previous findings (39), wherein 95% of veterans endorsed the SPI as both acceptable and helpful. In addition, clinicians in the present review perceived SPI practices to be helpful in reducing suicide risk during consumers’ transition from inpatient to home or community settings. This is an important finding, as risk of suicide may be most acute following discharge from psychiatric hospitalization, particularly for those with active suicidal ideation, perceived hopelessness, and history of suicidal behavior (40). Overall, the efficacy of the SPI in helping consumers to reduce suicidal ideation and behavior is supported by both quantitative systematic reviews (17–19) and by the experiences and perspectives synthesized in the present review.

People involved in the SPI also perceived a range of specific benefits that may help to explain the effectiveness of SPI practices. First, person-centered safety planning was seen to facilitate greater consumer autonomy, giving individuals a greater sense of ownership over their own health care. Consumers and clinicians also described how SPI practices helped to increase consumers’ sense of hope by internalizing and valuing their existing reasons for living. The amplification of reasons for living is an important protective mechanism, with reasons for living associated with reduced suicidal ideation and suicide attempts (41). In the present results, reasons for living often included loved ones such as children, partners, family, and friends. As such, greater identification of reasons for living appeared to intersect with an improved sense of connection with supportive others. This fundamental need for connection was maximized when support persons were involved in consumers’ safety planning. Similarly, ongoing engagement with SPI practices supported individuals’ self-efficacy in recognizing early warning signs and engaging self-regulatory coping strategies to interrupt the trajectory of escalating distress. This latter result aligns with recent evidence for growth in suicide-related coping as a key predictor of reduced suicidal ideation during an SPI intervention (16). In sum, the lived experience data synthesized in this review broadly align with some of the psychological mechanisms of effect for the SPI as theorized by Rogers et al. (20). Specifically, these findings add support to Rogers et al.’s (20) suggestions that the SPI promotes autonomy among users, both in initial plan creation and in their choices surrounding whether, when and how to use the plan to keep themselves safe; encourages connection with others (including healthcare services, and friends, family and community), which is a known protective factors against suicide; and builds competence through encouraging individuals to identify personalized support strategies and to practice using these to build confidence over time.

Clinicians and consumers strongly recommended a collaborative, person-centered approach to constructing and using the SPI over time. This approach refers to clinicians and consumers working together, sharing decision making and having a balance of power, to develop plans that address the consumer’s unique needs and circumstances (42). Unlike a crisis risk assessment process, which can imply a mechanistic and alienating experience of safety planning, collaborative and person-centered approaches allow a normalizing space for consumers to feel supported and to have voice in exploring suicide-related feelings. Recent quantitative evidence suggests that stronger therapeutic alliance established early in psychotherapy is a key predictor of reductions in suicidal ideation and behavior (43) and this review supports those findings from many consumers using safety plans. Collaborative and person-centered interactions were viewed as essential for helping people in distress to understand and process difficult emotional states, to find meaningful connection with others, and for using their strengths and supports to cope in the future.

Most mental health professionals would recognize the importance of person-centered therapeutic engagement. However, our results highlight a range of organizational barriers impairing clinicians’ ability to use the SPI according to these core principles. Time constraints were the primary barrier impacting clinicians’ perceived ability to conduct person-centered safety planning. Thus, without sufficient organizational support, the SPI may be more likely to be delivered instrumentally with a focus on risk mitigation, rather than in a person-centered and collaborative way.

Consumers reported experiences of ‘tunnel vision’ or an inability to consider SPI coping strategies, while enduring acute distress. This finding converges with the understanding that the ability to engage cognitive and/or behavioral self-regulatory coping strategies is diminished during heightened periods of crisis (44). This perceived limitation of SPI utilization further highlights the importance of appropriate and effective methods to work with consumers in deciding to restrict access to lethal means. At an individual level, clinicians and consumers can work collaboratively to make changes to living environments to restrict access to high lethality means should they experience acute and unbearable distress. This part of the planning process should focus on means identified by the consumer that feature in suicidal ideation. Appropriate involvement of support persons may be particularly beneficial in maintaining safe environments and reducing the help seeking burden placed on consumers.

In the present results, the SPI was disregarded as unhelpful by some consumers and clinicians. Similar uncertainty regarding the SPI has recently been documented in a quantitative study, with clinicians doubtful of the effectiveness of safety planning in reducing risk of suicidal behavior (45). As noted by an included study (31), this hesitancy suggests a need for prior education and training about the efficacy, usability, and acceptability of the SPI. Consumers’ fear of disclosure was another barrier to SPI engagement identified in the present results (10). Self-stigma and fear of stigmatized responses to disclosure can deter consumers from seeking help for suicide-related concerns (46), and consumers also report fears of disempowerment from treatment orders under mental health Acts (47). Similar worries may also deter individuals from engaging with interventions such as the SPI.

The four synthesized findings in this review suggest specific recommendations for practice, policy, and future research. For practice, it is recommended that the SPI is developed via a person-centered and compassionate collaboration, where clinicians are afforded sufficient time (minimum 30 minutes) to develop authentic therapeutic rapport for the person to express their suicidal experiences. Further, to address the transient nature of suicidal thoughts and maximize effectiveness of the safety plan, the SPI should be viewed as a living document that is shared with others (support persons, care providers) and revised regularly. Given that involving support persons appears to enhance the SPI, practitioners should genuinely explore this involvement during the initial safety plan co-construction and at review appointments. Supportive others should receive SPI education with assistance from the clinician and guidance from the consumer regarding how to best provide support.

Regarding policy recommendations, services that use the SPI should include mandatory training for all staff using the SPI, to ensure consistent, evidence-based skill sets and to address the ambivalence of some clinicians identified in this review. Further, there should be clear guidelines and policies for use of the SPI within and across services to ensure continuity of care. For example, the SPI could be proposed as the recommended safety planning instrument in a local context, to be completed before discharge from emergency/inpatient settings and communicated with follow-up care providers as standard practice. Given the diverse contexts in which safety planning is used, there should be flexibility to adapt the SPI to meet diverse consumer needs (e.g., versions in various languages).

Further research is required to address gaps in our understanding of the SPI and how best to support the people who use it. First, the specific processes which assist consumers to reduce suicidal ideation and behavior require further examination. Our findings indicate that SPI practices may enhance consumers’ connection, autonomy, and competence: three of the processes of SPI effect proposed by Rogers et al. (20). Further mixed-methods research is required to investigate causal pathways from specific SPI strategy-use to improved suicide and wellbeing-related outcomes via theorized processes of effect. Greater integration of diverse user experiences is required to inform future SPI adaptations that meet the needs of the specific consumer groups for whom they are designed. In the current review, over half of the included papers related to veterans, their support persons and/or people who work with them. There has been little to no focus on the experiences of safety planning from other priority groups known to experience high rates of suicidality, such as LGBTQIA+ communities (48). Finally, our results reveal a common perception whereby states of acute and severe distress temporarily impair peoples’ capacity to engage in safety planning behaviors. This perceived barrier should be explored in more depth using rigorous qualitative approaches. Research has begun to illuminate the temporal dynamics of suicidal states, often using digital technologies to monitor suicidal distress in real-time (49). Lived experience research will be crucial to develop a greater understanding of how consumers experience the fluctuating and dynamic nature of suicidal states, as well as the relationship between current distress severity and specific SPI strategy use. Such understandings may assist consumers, support persons, clinicians, and researchers to adapt SPI practices to mitigate the onset and worsening of distress, and to improve safety during peak distress.

Our search strategy, study selection procedures and meta-aggregation approach were systematic and thorough. In the JBI approach, findings can only be extracted if accompanied by an illustrative participant quotation. Whilst methodologically rigorous, this may have excluded relevant qualitative data if reported in a different format. There is also substantial scope for improvement in the methodological quality of studies in this area. In the present review, the dependability of included studies was limited due to inconsistent reporting of reflexivity details and guiding methodological frameworks. Three of the four synthesized findings were also downgraded due to a mix of unequivocal and credible findings, resulting in “low” overall confidence ratings. To enhance confidence in future qualitative findings, studies should follow best-practice guidelines for reporting qualitative research. Further, some studies lacked SPI details, such as format and delivery modality. We did not attempt to contact the authors of these papers to seek confirmation of these details. Doing so may have improved the generalizability of findings. However, we do not believe these details to be crucial to the results, as the findings relate more to overall experiences with the SPI, rather than specific features (with the exception that we had one finding category related to digital modalities).

Finally, although one included study indicated a mental health lived experience academic as part of the authorship team (10), none of the included studies explicitly indicate involvement or consultation with people with lived experience of suicidality and/or safety planning in designing or conducting the studies. More high-quality qualitative studies of consumer, support person and clinician perspectives, conceived and conducted collaboratively with people with lived experience of suicidality and safety planning, would advance our understanding of peoples’ experiences of using SPI practices.