Association between suicidal ideation and suicide

Most NHS psychiatrists work in GAP, and suicide risk assessment forms a substantial part of the GAP workload.^{Reference Kader, Richardson, Hudson, Langan-Martin, Allam and Ayre2} However, the prevalence of suicidal ideation in people who present to GAP had not been previously reported.

In the present study, the suicide rate during a new episode of care was 66 per 100 000 episodes. In comparison, between 2010 and 2020, the rate of suicide among mental health service patients across the UK was 48 per 100 000,^{Reference Appleby, Kapur, Shaw, Turnbull, Hunt and Ibrahim3} whereas in 2021, the annual suicide mortality rate in the general Scottish population was 13.7 per 100 000 people.⁴ Therefore, our data corroborate reports of a higher rate of suicide in people known to mental health services.

Here, 47.4% of patients referred to GAP reported suicidal ideation when they were assessed (including 76.4% of patients referred as emergencies), yet suicide during the contingent episode of care was very rare. Our study then is in general agreement with others on risk assessment. For example, a case–control study of risk factors for suicide among in-patients and recently discharged out-patients with affective disorders reported 48% of patients having ‘suicidal thoughts at admission’,^{Reference Hoyer, Licht and Mortensen5} and a meta-analysis of the predictive utility of suicidal ideation for suicide among GAP patients during the comparatively high-risk-for-suicide in-patient and immediate post-discharge phases also determined an extremely low PPV, even with a long average follow-up period of 9.1 years (pooled PPV among 29 cohort studies was 1.7%, 95% CI 0.9–3.2).^{Reference McHugh, Corderoy, Ryan, Hickie and Large6}

The negative predictive value (NPV) of suicidal ideation for suicide is also extremely low, which is to say that the absence of suicidal ideation is not an assurance of safety. For example, one retrospective study found that almost 80% of psychiatric in-patients who died by suicide during that admission had denied suicidal thoughts in their last verbal communication.^{Reference Busch, Fawcett and Jacobs7}

Further international evidence of an extremely weak association between suicidal ideation and suicide is provided by results from the US National Survey on Drug Use and Health, when compared with the Centers for Disease Control and Prevention mortality records. In that 2021 survey, only 1.6% of participants aged 65 and over affirmed ‘thinking seriously about killing themselves in the previous 12 months’, despite that group including the highest annual rate of suicide (40 per 100 000 for White males aged 75 and over). This contrasts with the 18- to 25-year age group, who had a far higher prevalence of suicidal ideation at 6%, but a much lower rate of suicide (17.5 per 100 000 for males and 4 per 100 000 for females).^{Reference Harmer, Lee, Duong and Saadabadi8}

Association between other presumed risk indicators and suicide

The objective evidence then is that suicidal ideation is extremely common in people presenting to GAP, and it is very weakly associated with suicide at the group level. This is also true for other presumed risk indicators for suicide. For example, meta-analysis of the predictive utility of various suicide risk assessment scales (Beck Hopelessness Scale, Suicide Intent Scale, and Scale for Suicide Ideation) and clinical risk factors (including alcohol misuse, concomitant physical health ailments, unemployment and previous self-harm) after self-harm showed them not to be clinically useful.^{Reference Chan, Bhatti, Meader, Stockton, Evans and O'Connor9,Reference Runeson, Odeberg, Pettersson, Edbom, Jildevik Adamsson and Waern10} Furthermore, meta-analysis has also concluded that ‘suicidal behaviours’, including ‘suicide attempts’, which might be considered strong indicators of suicide risk, are no more strongly associated with suicide than suicidal ideation.^{Reference Large, Corderoy and McHugh11} Despite this, association statistics continue to be accumulated and, although they can appear compelling (for example, one study found the age-adjusted risk of suicide in the first year following self-harm to be 49 (95% CI 43–57) times greater than in the annual general population^{Reference Hawton, Bergen, Cooper, Turnbull, Waters and Ness12}), they do not help the practitioner determine which of the people they assess will die by suicide, when or what might keep them safe.

Our data, and these meta-analyses, illustrate the fundamental problems with attempting to predict extremely low-probability events with undiscerning assessments. Commensurately, it has been concluded that ‘the idea of risk assessment as risk prediction is a fallacy […] We are simply unable to say with any certainty who will and will not go on to have poor outcomes’.^{Reference Chan, Bhatti, Meader, Stockton, Evans and O'Connor9} Our study extends this conclusion to patients at the point of assessment by GAP. At that juncture too, suicidal ideation does not predict suicide with a specificity, nor within a time frame, that is clinically useful.

Suicidal ideation does not necessarily imply mental disorder

Additionally, it should not be assumed that suicidal ideation implies the presence of a mental disorder that requires the involvement of GAP services. To illustrate, the prevalence of suicidal ideation varies widely between regions (for example, from 2.09% in Beirut, Lebanon, to 18% in Christchurch, New Zealand), but the rates of psychiatric disorders do not,^{Reference Weissman, Bland, Canino, Greenwald, Hwu and Joyce13} and, further, the prevalence of suicidal ideation does not mirror that of depressive disorders.^{Reference Bebbington, Minot, Cooper, Dennis, Meltzer and Jenkins14}

Unless suicidal ideation is a consequence of a psychiatric disorder, GAP services do not have a unique contribution to make to reducing any associated risk of suicide. Qualitative evidence for this includes that when GAP staff were asked to consider in retrospect whether there was anything distinctive about their patients who died by suicide, what was often reported was a significant clinical improvement, but the persistence of concomitant stress factors.^{Reference Morgan and Stanton15} Social workers and third-sector organisations are better placed than GAP services to address social factors, so they may be more likely to reduce the subjective burden of suicidal ideation and the risk of suicide, including in emergencies. Indeed, ‘informal social support’ was considered a ‘particularly promising broad-based intervention’ after a Scottish Government commissioned systematic review of interventions to prevent suicide and suicidal behaviour.^{Reference Leitner, Barr and Hobb16}

Despite this, the National Institute for Health and Care Excellence (NICE) still recommends that a ‘mental health professional’ performs a ‘risk formulation’ after an occasion of self-harm (but not a ‘risk assessment’, which is, presumably, different because it is constrained by a tool or scale). What a risk formulation should entail is only specified in vague social terms, such as ‘historical experiences, recent problems, and existing strengths and resources’.¹⁷ Social variables may indeed correlate statistically with suicide in large enough samples, but significant group- and population-level associations do not translate into useful predictors at an individual patient level, and they say nothing about how useful a particular therapeutic response might be.

Mistakenly inferring or assuming that studies reporting association statistics have predictive utility contributes to an unrealistic understanding of what suicide risk assessments can achieve^{Reference Mulder, Newton-Howes and Coid18} and this has serious consequences for patients and healthcare resources. Precise terminology is required: when only association statistics have been undertaken, it is important to avoid using terms with no clear definition or meaning, such as ‘predictive value’ and ‘predictive power’ or ‘risk assessment’ and ‘risk formulation’.^{Reference Poldrack, Huckins and Varoquaux19–Reference Varga, Niss, Estampador, Collin and Moseley21}

Uniquely, GAP services have the potential to contain people in hospital, but it would be numerically impossible to admit for further scrutiny all those who are assessed with suicidal ideation (approximately 50% of people referred to GAP throughout the UK would have to be admitted). Further, there is little evidence that GAP ward admissions can prevent suicide in general. For example, a recent study of psychiatry in-patients concluded that the majority of those who died by suicide were considered very unlikely to do so, including 14% having no ‘risk factors’ and 30% having only one ‘risk factor’. False-positive rates like this render such assessments clinically useless.^{Reference Powell, Geddes, Deeks, Goldacre and Hawton22}

Finally, it could also be that hospital-based care, especially compulsory treatment, is in itself dangerous for people who are thinking about suicide because it could compound those feelings of defeat, humiliation and entrapment that are considered motivators for suicide.^{Reference O'Connor and Kirtley23} Evidence for ‘nosocomial suicide’ is that patients who received psychiatric medication had 5.8 times the risk of suicide; patients who received out-patient psychiatric treatment had 8.2 times the risk of suicide; patients who had contact with emergency psychiatric services had 27.9 times the risk of suicide; and patients who were admitted had 44.3 times the risk of suicide; and the magnitude of these risk ratios greatly exceeds both the risk of suicide associated with major psychiatric disorder and the presumed clinical risk factors for suicide among admitted patients. This could suggest that current medical practice in the context of suicidal ideation increases the risk of subsequent suicide.^{Reference Large and Ryan24}

However, not all patients are the same and medical interventions invariably incur both potential benefits and harms: there has long been a debate in GAP about the benefit/harm ratio of NHS interventions, including compulsory detentions, for people with personality disorders unresponsive to medications and acute psychotherapies and who have repeatedly hurt themselves, usually in response to interpersonal difficulties. In contrast, for patients with severe and enduring psychiatric illnesses, such as schizophrenia or psychotic depression, driven to kill themselves by psychotic symptoms but who respond to medications, the benefit/harm ratio of GAP interventions, including compulsory detentions, is far more likely to be understood by society, the patient's family and, most crucially, the patients themselves when recovered.

Limitations

The broad provisional clinical categories used in this study reflect the limited information available at the time of referral and preliminary assessment, but this is usual in NHS medical practice. The 6-month duration of the study is a limitation and annual mortality rates (the usual measure) cannot be reported. However, the predictive utility of suicidal ideation for subsequent suicide during the contingent episode of care can be reported and this is a more clinically useful statistic (in this context, a meta-analysis of similar patient populations concluded that the duration of follow-up does not significantly moderate the predictive utility of suicidal ideation for suicide^{Reference McHugh, Corderoy, Ryan, Hickie and Large6}). Suicidal ideation was significantly associated with assessments performed by the emergency assessment service, but less so with referrals considered at the weekly non-emergency written referral meeting. This could be due to the differences in how suicidal ideation was determined by each service. Optimally, this study would have been done as a large multi-site study to capture any regional variations, and we suggest this with the national audit recommendation that follows.

Clinical implications and recommendation

Every few years there is debate in the media about the suicide of a person who had been in contact with GAP and was expressing suicidal thoughts, and a recent NHS Resolution review highlighted ‘risk assessment’ as a prominent theme in suicide-related legal claims.^{Reference Oates25} The inevitable question is ‘why did services not admit the person to hospital when they said they were thinking about suicide?’

Our study goes some way to answering this: (a) it is impossible, given the prevalence of suicidal ideation in people presenting to GAP, and potentially harmful to admit every patient who presents with suicidal ideation; (b) suicide cannot be predicted with a specificity and sensitivity, nor within a time frame, that is clinically useful (the group-level associations between suicide and psychosocial variables that have been reported provide only probability estimates which translate poorly, if at all, to the individual, especially in relation to such rare events).

The fact that suicide risk prediction is ineffective has been asserted before, further to the highest standard of evidence.^{Reference Large, Soper and Ryan26} Additional examples to those already discussed include a meta-analysis of more than 3000 risk factors for suicidal thoughts and behaviours which concluded that ‘prediction was only slightly better than chance’,^{Reference Franklin, Ribeiro, Fox, Bentley, Kleiman and Huang27} including, for example, prior psychiatric hospital admissions and suicide attempts. Moreover, an examination of 64 models of suicide prediction found that the ‘accuracy of predicting a future [suicide] event is near 0’.^{Reference Belsher, Smolenski, Pruitt, Bush, Beech and Workman28} What needs to be examined is not the utility of risk predictors, but the efficacy of risk management plans to reduce distress and provide relief.

In practice the NHS does not to admit to hospital the vast majority of the large number of people presenting to GAP with suicidal ideation. The main reason is that admission is unlikely to be beneficial for most people with suicidal ideation. However, every few years every clinician working in a GAP service with a usual clinical workload will likely have a patient die by suicide because they interact with such a large number of people with suicidal thoughts. This may not be well-known outside of GAP, and the mismatch could contribute to the trauma GAP staff experience in the aftermath of a suicide.^{Reference Gibbons, Brand, Carbonnier, Croft, Lascelles and Wolfart29}

The National Confidential Inquiry into Suicide and Safety in Mental Health (NCISH)^{Reference Appleby, Kapur, Shaw, Turnbull, Hunt and Ibrahim3} does not record information on suicidal ideation for people presenting to GAP, despite NHS staff spending a great deal of time attempting to assess its significance for individual patients. An audit of the management of suicidal ideation in GAP could provide a baseline of current custom and evidence to support quality improvement and transparent evidence-based practice.