Can we identify those who should receive suicide-prevention intervention?

We know that, for example, in the United Kingdom, about 75% of those who die by suicide were not under the care of mental health services at the time of their death [Reference Windfuhr and Kapur18]. A reasonable inference from this may be that society, and particularly those closest to the person who died by suicide, were unaware of increased concerns sufficient to prompt referral to mental health services. This is consistent with the hypothesis that it is very difficult to accurately identify whether somebody is going to die by suicide. Much research has been done regarding people in contact with mental health services to identify those most at risk of suicide.

This has identified many factors that are associated with an increased risk of suicide [Reference Windfuhr and Kapur18–21].

Risk factors for future suicide are generally formulated as either being associated with a lifetime risk of suicide [Reference Favril, Yu, Uyar, Sharpe and Fazel19] or with a risk of suicide within the following year [Reference Hjorthøj, Madsen, Agerbo and Nordentoft20].

Windfuhr and Kapur [Reference Windfuhr and Kapur18], in a review of review of 15 years of UK National Confidential Inquiry into Suicide findings, found that an increased risk of suicide is related strongly to the presence of mental disorder; advancing age; and to gender, with males significantly more likely to die by suicide. This is consistent with World Health Organization data [21].

A systematic review and meta-analysis of psychological autopsy studies [Reference Favril, Yu, Uyar, Sharpe and Fazel19] found that the highest OR of >10 was for those receiving psychiatric treatment, those with a previous history of self-harm, and those with an adverse life-event within the previous month. They found that the following are all associated with an OR for suicide of >5: family history of mental disorder, depression, schizophrenia, borderline/paranoid/dependent personality disorders, receipt of psychiatric treatment, previous self-harm, previous suicide attempt, and relationship conflict.

A large meta-review of suicide in people with mental disorders [Reference Chesney, Goodwin and Fazel22] found that “for suicide mortality, borderline personality disorder, depression, bipolar disorder, opioid use, and schizophrenia, as well as anorexia nervosa and alcohol use disorder in women, had substantially increased rates (greater than 10 times) compared with the general population.”

Ryan et al. [Reference Ryan, Large and Callaghan23] stated in an editorial that “Patients who present in psychological crisis or after a suicide attempt are more than 50 times more likely than the general population to die by suicide in the following year.” This has been confirmed in multiple studies, such as a retrospective cohort study of 3987 Korean patients by Choi et al. [Reference Choi, Park, Yi and Hong24] and a prospective cohort study of 29,571 patients by Steeg et al. [Reference Steeg, Kapur, Webb, Applegate, Stewart and Hawton25].

It appears intuitively that the best opportunity for intervention would be with those at the highest risk of suicide.

Many forms of suicide risk assessment have been developed to identify those at highest risk. Examples include the Patient Safety Screener-3 [Reference Boudreaux, Jaques, Brady, Matson and Allen26], the Pediatric Emergency Care Applied Research Network Computerized Adaptive Screen for Suicidal Youth [Reference King, Brent, Grupp-Phelan, Casper, Dean and Chernick27], Columbia-Suicide Severity Rating Scale (C-SSRS) [Reference Posner, Brown, Stanley, Brent, Yershova and Oquendo28], Ask Suicide-Screening Questions [Reference Horowitz, Bridge, Teach, Ballard, Klima and Rosenstein29], and the Beck Scale for Suicide Ideation [Reference Beck, Kovacs and Weissman30].

Unfortunately, there is no known suicide risk assessment that has been shown to effectively identify those individuals who will die by suicide and thereby target interventions and prevent suicides.

There are several reasons for this. At the time of any risk assessment, most people who go on to die by the intentional act of suicide will not yet have made the suicide decision [Reference Williams, Davidson and Montgomery31–Reference Paashaus, Forkmann, Glaesmer, Juckel, Rath and Schönfelder35]. If they themselves do not yet know, then it is presumably difficult for a clinician to discover what that person’s future decision will be. Therefore, clinicians are guided by the suicide risk factors identified and by their clinical assessment of the individual.

Suicide rates are generally low, with age-standardized annual rates ranging worldwide between 0.31 and 87.48 according to 2019 World Health Organization [36]. The rate in most of Europe, Africa, Asia, the Americas, and Australia is less than 15/100,000/year.

The low population base rate of suicide leads to risk categorization methods identifying many more people as being at the very highest risk than will die by suicide in the following year. Steeg et al. [Reference Steeg, Kapur, Webb, Applegate, Stewart and Hawton25], in a multicenter prospective cohort study of 29,571 self-harm presentations, found: “Although patients presenting after suicide attempts or in crisis have a greatly increased probability of eventual suicide compared with the general community, fewer than one in 200 will actually die by suicide in the next 6 months.”

Madsen et al. [Reference Madsen, Agerbo, Mortensen and Nordentoft37], in a Danish national, prospective, register-based study, found “that only 12% of suicides occurred among high-risk patients, and that fewer than one in 400 high-risk patients died by suicide.”

The clinical implication is that the best possible system of risk assessment, looking at the highest risk group of psychiatric inpatients, would identify as the very “highest risk” group about 98 people who will, in fact, not die by suicide in the next year for every two who do; and, even more frustratingly, would miss the majority of those who do go on to die by suicide because they would be assessed as being at “low” risk [Reference Large, Smith and Sharma38, Reference Kessler, Bossarte, Luedtke, Zaslavsky and Zubizarreta39].

This was looked at in detail in a systematic review and meta-analysis by Large: “Even making assumptions that optimize the power of suicide risk assessment, we estimated that while some (in)patients could be categorized as “ultra-high-risk,” as few as 3% of these would go on to suicide in the year after discharge. Meanwhile, about 60% of people who did eventually suicide during the same period would have been categorized as being at lower risk.” “Despite the apparently strong association between high-risk categorization and subsequent suicide, the low base rate of inpatient suicide means that predictive value of a high-risk categorization is below 2%” [Reference Large, Smith and Sharma38].

In the United Kingdom, the National Institute of Health and Care Excellence has acknowledged the reality that suicide risk assessment tools are not clinically useful by recommending:

The challenge for suicide prevention is that, despite all efforts, no tool or method has been found to identify those who will go on to die by suicide with clinically useful accuracy. The presence of suicide risk factors cannot tell us whether or when an individual may attempt suicide [Reference Large, Smith and Sharma38–Reference Gillies, Chicop and O’Halloran42].

Studies, using blind ratings of suicide risk by case note review, appear to confirm this inability to usefully predict suicide and the danger in investigations of hindsight bias [Reference Fahy, Mannion, Leonard and Prescott43, Reference Conlon, Garland, Prescott, Mannion, Leonard and Fahy44].

Can we identify the best time for delivery of a suicide-prevention intervention for potentially suicidal individuals?

Common sense suggests that suicide-prevention interventions, that are delivered after the decision to die by suicide has been taken, may most likely succeed in preventing the death. Examples include talking down a person about to jump from a high building or treating for an overdose.

Unfortunately, research (Table 2) suggests that for very many people who attempt suicide, there is often a period of only minutes between the decision and the act [Reference Williams, Davidson and Montgomery31–Reference Paashaus, Forkmann, Glaesmer, Juckel, Rath and Schönfelder35]. This means there would typically be a very brief window of opportunity for intervention. A clinician would have to discover the suicide decision during this brief window to act. Realistically, this makes intervention at the right time very challenging and often impossible to achieve.

Table 2. Suicide: Time from the decision taken, to the suicidal act made

There currently appears to be no practical way of identifying when those most at risk should be targeted for intervention. The typically very short window of opportunity to intervene with suicide prevention between the suicide decision and the suicide act adds considerably to this difficulty. Clinical staff would need to be in the right place at the right time to successfully intervene, which is unfeasible given the very small proportion of those deemed at highest risk who go on to die by suicide. Put differently, a flawless suicide-prevention service would have to treat 100 of the highest risk patients for a whole year with perfect treatment to prevent two or three suicides [Reference Steeg, Kapur, Webb, Applegate, Stewart and Hawton25, Reference Madsen, Agerbo, Mortensen and Nordentoft37, Reference Large, Smith and Sharma38].

Do we have effective clinical interventions, that are practical and achievable, that would prevent suicides or at least significantly reduce the risk?

There are a multitude of care initiatives and clinical interventions that are aimed at preventing suicide, with many publications showing effectiveness at reducing suicide rates [Reference Conlon, Garland, Prescott, Mannion, Leonard and Fahy44–Reference Hawton and Sinyor46].

Hawton and Sinyor commented in an editorial on the challenge of showing whether they work:

Various environmental interventions have played a role in different parts of the world. Multiple methods of restricting access to means of suicide have shown evidence of effectiveness. In the United States, increased access to guns has been associated with higher suicide rates, and decreasing gun access has been associated with an overall reduction in suicides [Reference Anestis and Anestis47, Reference Anestis, Khazem, Law, Houtsma, LeTard and Moberg48].

In the United Kingdom, coal gas was previously used as fuel for domestic cooking and heating. This was highly toxic because of the high levels of carbon monoxide produced during combustion, leading to a high number of suicides by inhalation. When coal gas was replaced with less toxic natural gas between the 1950s and 1970s, this change was associated with an overall reduction in the suicide rate [Reference Kreitman49].

A large international systematic review of multiple different means-restriction interventions found short-term benefit in some studies, but did not examine whether this effect was sustained in the long term [Reference Lim, Buckley, Chitty, Moles and Cairns50].

A Finnish study using 43 years of suicide data found that multiple means restrictions were gradually followed by an increased use of other methods [Reference Ohberg, Lonnqvist, Sarna, Vuori and Penttila51].

Admission to a psychiatric hospital is often used to try to prevent suicide. Some research evidence suggests that inpatient admission itself may not prevent suicide. [Reference Large, Chung, Davidson, Weiser and Ryan52] Evidence suggesting a lack of efficacy of admission in preventing suicide includes a natural experiment in Fulton County, Georgia, USA, when admission to a psychiatric hospital was cut by 56% due to reduced funding. During this period, the suicide rate fell from 12 to 10 per 100,000 population [Reference Garlow, D’Orio and Purselle53]. There is debate about whether admission is protective or may increase suicide risk. [Reference Large and Kapur54]

Other individual-level approaches include treating associated conditions, such as depression, with medications, cognitive behavioral therapy (CBT), dialectical behavioral therapy (DBT), and collaboratively developing a safety plan [Reference D’Anci, Uhl, Giradi and Martin55]. A systematic review of evidence-based suicide prevention [Reference Mann, Michel and Auerbach56] found evidence for training of primary-care physicians in depression recognition and treatment, and for means restriction such as firearms controls in the prevention of suicide. Educating youth regarding depression and suicidal behavior, active mental health service outreach postdischarge and post-self-harm, antidepressants, ketamine, and CBT and DBT were all found to reduce suicidal behaviors [Reference D’Anci, Uhl, Giradi and Martin55, Reference Mann, Michel and Auerbach56]. There is some evidence suggesting that clozapine may have a protective effect [Reference Lee, Cotes, Mojtabai, Margolis, Nucifora and Nestadt57].

Multimethod interventions have been proposed as a more effective way of preventing suicide, for example, the “Zero Suicide Model” [Reference Brodsky, Spruch-Feiner and Stanley58] and “Alliances against Depression” [Reference Hegerl, Rummel-Kluge, Värnik, Arensman and Koburger59], but there is an absence of convincing evidence of clinical efficacy [Reference Ouliaris, Gunaratne, Ryan, Rosenberg, Salvador-Carulla and Gill60], and criticism of the ethical [Reference Sjöstrand and Eyal61] and research [Reference Sisti and Joffe62] consequences.

The published evidence, usually based on associations, suggests that many of these interventions prevent suicides [Reference Zalsman, Hawton, Wasserman, van Heeringen, Arensman and Sarchiapone45, Reference Hawton and Sinyor46, Reference Mann, Michel and Auerbach56]. Methodological issues include defining and separating outcome measures of suicidal ideation, suicidal behaviors with or without intent, and suicide itself [Reference Ohberg, Lonnqvist, Sarna, Vuori and Penttila51]. The challenges in conducting gold-standard double-blinded, randomized controlled trials of these interventions means proving the cause–effect relationship is difficult, with a lack of robust evidence of sustained efficacy [Reference Huang, Harris, Funsch, Fox and Ribeiro63, Reference Fox, Huang, Guzmán, Funsch, Cha and Ribeiro64].

Unfortunately, there is little evidence of a significant reduction in population-level mortality that might have been expected to flow from these promising interventions.

In the United Kingdom, suicide rates have not dropped since the National Confidential Inquiry into Suicide and Homicide (NCISH) was established in 1996. The suicide mortality rate was 10.9/100,000 in 1996 and slightly higher 11.4/100,000 in 2023 [65].

In contrast, deaths in surgery dropped initially, following into Peri-Operative Death from 121/10,000 in 1998 to 76/10,000 in 2014 [Reference Nepogodiev, Omar and Bhangu66]. Following the Confidential Enquiry into Maternity Services Deaths mortality rates dropped from 90/100,000 in 1952 to 8.79/100,000 in 2017–2019) [Reference Knight, Bunch, Felker, Patel, Kotnis and Kenyon67].

Encouraging results from suicide prevention research and interventions have been described [Reference Zalsman, Hawton, Wasserman, van Heeringen, Arensman and Sarchiapone45, Reference Hawton and Sinyor46, Reference Mann, Michel and Auerbach56]. However, these have had comparatively little impact on the overall suicide mortality rate when compared to the reductions in those for physical disorders.

This is revealed by an examination of national population mortality rate trends, for example, in England [68] and the United States [Reference Insel2], where there is high-quality mortality data available.

In England [68] and the United States [Reference Insel2], there has been an often dramatic improvement in mortality rates for most medical conditions over the past couple of decades. Major decreases have been seen, for example, in the SMR for heart disease and cancers.

Yet this is not true of suicide. The suicide mortality rate has not reduced in the way it has for other conditions, as illustrated in Figures 1 (England) and 2 (United States). The many different interventions, past and present, including initiatives to improve medications, psychological therapies, and in-/out-patient care have not led to a similar meaningful reduction in overall suicide rates.

Figure 1. Changes in mortality rates – England [68].

Figure 2. Changes in mortality rates – United States (Insel/NIMH) [Reference Insel2].

Figure 1 shows data for England illustrating the comparative trend from 2001 to 2023 in SMR for suicide compared to colorectal cancer, lung cancer, chronic obstructive pulmonary disease, stroke, and ischemic heart disease [68].

Tom Insel, then Director of the United States NIMH, pointed out what he called the “inconvenient truth” that all the efforts of mental health clinicians and researchers have essentially had little impact on suicide mortality compared to some other conditions. He infers from this that psychiatry has not yet discovered how to prevent suicide [Reference Insel2].

Mortality rate trends in England for several physical disorders show major reductions over recent decades. However, these data (Figure 2) reveal that despite all our efforts, all the research, and all the clinical and environmental interventions, there is little evidence of therapeutic interventions that have similarly reduced suicide mortality rates.

This illustrates how hard it is to prevent suicide. The assumption that suicide can be systematically prevented if only we identify and treat people who may suicide is therefore not supported by the available research evidence.

Do health services correctly target those most at risk of suicide?

We should consider whether mental health services correctly target those who are at highest suicide risk.

More intensive treatment is associated with higher rates of suicide, suggesting that healthcare is to some extent correctly targeted at those most at risk. Hjorthøj et al. [Reference Hjorthøj, Madsen, Agerbo and Nordentoft20], in a nested case–control study of 2429 cases and 50,323 controls in Denmark, found that the more intensive the treatment, the higher the suicide risk in the following year. This association of higher levels of care with higher suicide risk appears to suggest that health services do, to some extent, target those at higher suicide risk.

The UK NCISH found that the risk of suicide after hospital discharge is at its highest in the first days after discharge and then decreases day-by-day thereafter [Reference Appleby, Shaw, Kapur, Windfuhr, Ashton and Swinson69]. While these data cannot show causation, it appears to be consistent with the hypothesis that patients are generally discharged at a time when their suicide risk is reducing.

Despite this, there has been no reduction in the overall suicide mortality rate.

Counterintuitively, and without evidence of causality, some even infer that the treatments themselves cause suicide, for example, antidepressant medication [Reference Healy and Aldred70, Reference Plöderl, Hengartner, Bschor and Kaminski71].