Suicide Prevention

Suicide constitutes a public health issue of major concern. Ongoing progress in the field of artificial intelligence, particularly in the domain of large language models, has played a significant role in the detection, risk assessment, and prevention of suicide. The purpose of this review was to explore the use of LLM tools in various aspects of suicide prevention. PubMed, Embase, Web of Science, Scopus, APA PsycNet, Cochrane Library, and IEEE Xplore-for studies published were systematically searched for articles published between January 1, 2018, until April 2024. The 29 reviewed studies utilized LLMs such as GPT, Llama, and BERT. We categorized the studies into three main tasks: detecting suicidal ideation or behaviors, assessing the risk of suicidal ideation, and preventing suicide by predicting attempts. Most of the studies demonstrated that these models are highly efficient, often outperforming mental health professionals in early detection and prediction capabilities. Large language models demonstrate significant potential for identifying and detecting suicidal behaviors and for saving lives. Nevertheless, ethical problems still need to be examined and cooperation with skilled professionals is essential.

ImportanceSuicide ideation and suicide-related behavior (eg, suicide attempts) are increasingly prevalent among adolescents. Evidence supports safety planning interventions for adults at risk for suicide; the effectiveness for adolescents is unknown.ObjectiveTo evaluate the effectiveness of safety planning as a standalone intervention for adolescents with suicide ideation and/or suicide-related behavior.Data sourcesOvid MEDLINE, OVID PsycINFO, EBSCO CINAHL, and Scopus (Elsevier) from January 1, 2008, to March 26, 2024.Study selectionIncluded were all studies that examined safety planning as a standalone treatment for adolescents and assessed suicide ideation, suicide-related behavior including suicide attempts, and/or re-presentations to health care settings. Included were peer-reviewed studies evaluating intervention effectiveness or efficacy, with or without a control condition. Nonempirical studies, gray literature, and studies not available in English were excluded.Data extraction and synthesisTwo reviewers performed data extraction and quality assessment independently. Meta-analytic random-effects models were used to calculate effect size estimates. Risk of bias was assessed using Joanna Briggs Institute Critical Appraisal tools.Main outcomes and measuresThe outcomes planned for extraction were suicide ideation, suicide-related behavior (eg, attempts; planning with intent to act), and re-presentation to health care settings (ie, emergency department visit; inpatient admission) during the follow-up period.ResultsTen studies including 1002 adolescents (mean [SD] age, 15.0 [0.4] years; 76.0% female) met inclusion criteria; 5 were included (n = 619) in the meta-analysis. There was no significant association between safety planning interventions and suicide ideation (Hedges g = 0.11; 95% CI, 0.01-0.21), behavior (Hedges g = -0.09; 95% CI, -0.20 to 0.02), attempts (risk ratio [RR], 1.03; 95% CI, 0.12-8.88) or suicide-related re-presentation (RR, 0.99; 95% CI, 0.29-3.35) at follow-up. Risk of bias for the majority of studies was moderate to high.Conclusions and relevanceThis systematic review and meta-analysis found limited research evaluating safety planning with adolescents. Although available data do not support safety planning as a standalone treatment for reducing suicide-related outcomes among children and adolescents, well-powered studies are needed to address this common intervention for suicide prevention in adolescents.

ImportanceSuicide is a major public health concern, and as most individuals have contact with health care practitioners before suicide, health systems are essential for suicide prevention. The Zero Suicide (ZS) model is the recommended approach for suicide prevention in health systems, but more evidence is needed to support its widespread adoption.ObjectiveTo examine suicide attempt rates associated with implementation of the ZS model in outpatient mental health care within 6 US health systems.Design, setting, and participantsThis quality improvement study with an interrupted time series design used data collected from January 2012 through December 2019, from patients aged 13 years or older who received mental health care at outpatient mental health specialty settings within 6 US health systems located in 5 states: California, Oregon, Washington, Colorado, and Michigan. Analyses were conducted from January through December 2024.ExposureThe ZS model was implemented in 4 health systems at different points during the observation period (2012-2019) and compared with health systems that implemented the model before the observation period (postimplementation). Implementation included suicide risk screening, assessment, brief intervention (safety plan, means safety protocol), and behavioral health treatment.Main outcomes and measuresThe primary outcome was a measure of standardized monthly suicide attempt rates captured using health system records and government mortality records. Suicide death rates were also measured as a secondary outcome.ResultsThere was a median of 309 107 (range, 55 354-451 837) unique patients per month. In 2017, there were 317 939 eligible individuals (63.2% female). Baseline suicide attempt rates were at least 30 to 40 per 100 000 individuals at each implementation site and decreased to less than 30 per 100 000 individuals at 3 sites by 2019. Decreases in suicide attempt rates were observed at 3 intervention health systems after site-specific implementation: health systems A and B had decreases of 0.7 per 100 000 individuals per month and C, 0.1 per 100 000 individuals per month. System D evidenced a similar suicide attempt rate after implementation (before implementation: median rate: 35.0 [range, 11.0-50.3] per 100 000 patients per month; after implementation: median rate: 34.3 [range, 18.5-42.0] per 100 000 patients per month). The 2 postimplementation health systems maintained low or declining suicide attempt rates throughout the observation period. The rate at system Y decreased by 0.3 per 100 000 individuals per month across the observation period. The rate at system Z began at 11 per 100 000 individuals per month and declined by 0.03 per 100 000 individuals per month during the observation period. Two systems evidenced reductions in the suicide death rate after implementation: system B declined by 0.2 per 100 000 individuals per month and system C by 0.1 per 100 000 individuals per month.Conclusions and relevanceIn this quality improvement study, ZS model implementation was associated with a reduction in suicide attempt rates among patients accessing outpatient mental health care at most study sites, which supports widespread efforts to implement the ZS model in these settings within US health systems.

BackgroundMen have 2-4 times higher suicide mortality than women and are less likely to seek help, with disparities linked to restrictive masculine norms and stigma. This review aimed to synthesize the effectiveness of suicide prevention interventions targeting men and identify gender-tailoring strategies.MethodsStudies evaluating suicide prevention interventions targeting men were included if conducted in male-only samples and reporting outcomes related to suicidality (e.g., suicide deaths, suicide attempts, suicidal ideation), depression, or help-seeking (e.g., behaviors, intentions, attitudes). Six databases (MEDLINE, Embase, PsycINFO, Cochrane Library, Web of Science, and CINAHL) were searched for peer-reviewed English-language studies published up to 31 December 2025. Risk of bias was assessed using RoB 2 and ROBINS-I. Due to heterogeneity, findings were synthesized narratively. The protocol was registered in PROSPERO (CRD420250655554).ResultsSeventeen articles evaluating 14 interventions were included. Only one study assessed suicide deaths, and none occurred. Among RCTs, no significant effects were observed for suicide attempts, suicidal ideation, or depression. Help-seeking behaviors and intentions showed mixed evidence of improvement, with no improvement in help-seeking attitudes. Gender-tailoring strategies were synthesized into seven categories: intervention design and development, risk targeting, content and messaging, outreach and recruitment, delivery modalities, male representation and role modeling, and communication and action orientation.ConclusionsEvidence for the effectiveness of suicide prevention interventions targeting men remains limited, with no clear effects on suicidality or depression and mixed evidence for improvements in help-seeking. Risk of bias was generally moderate to high. A wide range of gender-tailoring strategies was identified. Further development of gender-responsive approaches and more rigorous evaluation are needed to reduce male suicide mortality.