Stark figures tell us over 804,000 people die globally by suicide each year, equating to one suicide death every 40 seconds (WHO, 2014). It is not possible to accurately predict which individuals will die by suicide using a single tool. Understanding which factors differentiate between those who will have thoughts of suicide, and those who will act upon those thoughts and attempt suicide, is still elementary (Klonsky & May, 2014; O’Connor & Nock, 2014). Demographic risk factors increase the suicide risk of a whole population across its lifetime, but do not predict suicide in an individual at a single time-point. Absence of risk factors, however, does not mean absence of risk of suicide (Cole-King et al., 2013).
Current suicide risk assessment tools use mainly demographic risk factors (which may be as common in the general population) and have largely been developed without a solid empirical basis. This is the finding of a recent BMJ ‘state of the art’ review of suicide risk assessment and intervention in people with mental illness (Bolton, Gunnell & Turecki, 2015). The review provides a critical overview of the current state of the literature for suicide risk assessment and interventions, with a specific focus on predicting suicide in those individuals with a mental illness.
PubMed was the primary database searched, with a broader search of Google and hand-searching of the reference lists from articles also conducted. All studies selected were in English, and published between January 1990 – February 2015.
BMJ state of the art reviews are ‘generally commissioned by the editor’ (BMJ website, 2015) and differ from systematic reviews in several ways:
- They include a selection of materials, instead of all published literature on a topic
- They do not have to adhere to the same guidelines for conducting and reporting their literature searches, e.g. using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) criteria
- They do not necessarily use a quality assessment tool to evaluate study strengths and weaknesses.
Ten ‘conventional’ risk assessment scales (e.g. the Beck Hopelessness Scale) and 3 empirically based tools (e.g. The Manchester self-harm rule) are reviewed. Emphasis is placed upon the metrics of positive and negative predictive values.
Positive predictive values (PPV) are:
‘“True positives”’: The proportion of people who actually go on to die by suicide, relative to the number of people that a measure identifies as being at risk of suicide (either correctly, or incorrectly). PPV can be affected if there is a low base-rate for the event that is being predicted (as is the case for suicide), and also if the measure generates a lot of ‘false positives’, i.e. identifying people as high risk of attempting suicide when they are actually not.
Negative predictive values (NPV) are:
‘“True negatives’”: The proportion of people who do not die by suicide, compared to the number of people a measure identifies as not being at risk of suicide (either correctly or incorrectly). NPV can be inflated by a low incidence of the event, which is true for suicide, or if a test produces many ‘false negatives’, and fails to identify people as imminently suicidal who actually are.
Even tests that are highly sensitive and highly specific (see here for a brief explanation of these terms), can have a low PPV, due to the fact that suicide is a statistically uncommon event. This is true of the overwhelming majority of the well-established scales included within the review, such as the Beck Depression Inventory, Suicide Intent Scale, Suicide Assessment Scale and also, moving away from ‘conventional’ scales, the ReACT self-harm rule. In short, these tests are unlikely to determine with any accuracy whether someone will make a suicide attempt.
Empirically based methods of risk assessment, including the suicide Implicit Association Test (IAT) and Suicide Stroop, are also discussed. Early studies of the IAT show that it is predictive of suicide at 3-month (Randall et al., 2013) and 6-month follow-up (Nock et al., 2010), and the Suicide Stroop prospectively predicted suicide attempts within 6-months (Cha et al., 2010).
The authors conclude that given limited predictive ability of existing suicide risk assessment tools, we need to develop resource-light interventions and acknowledge the elevated risk in some clinical populations.
Strengths and limitations
The authors do not shy away from difficult truths; risk assessment tools are highly variable and inconsistent. It is also heartening to see Bolton and colleagues busting myths: asking about suicide does not promote a suicide attempt.
It would be useful to have details regarding how decisions were made about the ‘clinical relevance’ of material reviewed. Given the ‘quality’ of the overwhelming majority of conventional risk assessment tools reviewed in the current paper is low, a greater focus on what does work, and possibilities for advancing the science of assessing and responding to risk would have been beneficial.
Consideration of protective factors is often neglected when evaluating whether or not an individual is at risk of suicide, and the interplay between risk and protective factors is poorly understood (McLean, Maxwell, Platt, Harris & Jepson, 2008). The choice of database for the search (PubMed), even with supplementation from Google, could mean that methods of assessing suicidality that relied upon broader psychological factors (e.g. impulsivity, degree of emotional pain) were overlooked.
Currently the main focus of a suicide risk assessment to characterise and quantify risk through identifying demographic risk factors that may be as common in patients, as in the general public (Cole-King and Lepping, 2010), is clearly unhelpful. Furthermore, suicide risk assessment is itself a complex intervention, which means that it is not totally predictable and the process is influenced by practitioner, patient and organisational factors.
Suicidal thoughts per se are poorly understood, particularly their development and progression. Focusing efforts on people identified by risk assessment tools is unsafe, both due to false negatives and genuine fluctuations in distress. Suicidality is a dynamic process, with suicidal thoughts waxing and waning over time; something that conventional ‘check-box’ risk assessment tools often do not take into account. Irrespective of competence and experience, attempting to predict whether a patient will attempt to end their life is always challenging. Stigma or fear around suicide may prevent clinicians from asking patients about their suicidal thoughts. Stigma can also deter patients from disclosing suicidal thoughts, or lead to apparent inconsistencies in suicidality, whereby patients report suicidal thoughts to one clinician with whom they feel comfortable, but then are reluctant to disclose to another clinician.
The review cites psychological autopsy work showing that greater than 90% of people who die by suicide have a mental illness at the time of their death (Cavanagh et al., 2003) as evidence that mental illness is a highly important and modifiable risk factor for suicide. Depression is the most common mental illness in those who die by suicide (Randall et al., 2014), however, less than 5% of people with depression die by suicide (Bostwick & Pankratz, 2000); presence of mental illness alone is not sufficient explanation for an individual choosing to take their own life. Moreover, mental illness is more strongly associated with suicidal ideation than with suicide attempts (Nock et al., 2009). A previous suicide attempt was also identified by the review as being a key predictor of future suicide risk.
Despite every suicide being a tragedy, it remains an event with a low base-rate, making it a significant challenge to devise measures with adequate positive predictive value to predict suicide. Bolton and colleagues recommend considering “suicidal thoughts and behaviour as an important therapeutic target”. The best ‘low-level intervention’ is a positive and compassionate clinical encounter, whereby practitioners diligently identify and mitigate all risks, promote protective factors, instil hope and co-create a safety plan with explicit reference to removal of means.
This review highlights the lack of evidence-based suicide risk assessment tools available to clinicians, and the ‘hit and miss’ nature of existing instruments. It also draws attention to the critical need for a paradigm shift in the way that we conceptualise suicide risk assessment and intervention development. A map that uses presence or absence of mental illness as its ‘north star’ when attempting to predict suicide or to define opportunities for intervention development, will only take us so far. We must expand our horizons beyond traditional risk factors models of assessment to include other variables that may play a key role in determining who will go on to make a suicide attempt.
Emerging methods of supporting a clinical assessment (e.g. the Implicit Association Test), which have an empirical basis, and are interview independent, may provide an important new avenue for supplementing clinical suicide risk assessment. Whilst the review may paint a somewhat bleak picture of the ‘state of the art’ of suicide risk assessment, this is not a message of hopelessness, but rather a clarion call to action for further research into how we can refine risk assessment and intervention development.
Given the fluctuating nature of suicide risk and the fact that it is a rare event which cannot accurately be predicted, the only safe response is to take all suicidal thoughts seriously and respond appropriately. NICE advocate a needs and assets-based approach after self-harm rather than focusing on risk. We welcome the day when everyone at risk of suicide is responded to with compassion, confidence, and competence, and have a co-created safety plan to ensure their safety.
The authors would like to thank Professor Rory O’Connor and Eva Dumon for their invaluable feedback on drafts of this blog post.
If you need help
If you need help and support now and you live in the UK or the Republic of Ireland, please call the Samaritans on 116 123.
If you live elsewhere, we recommend finding a local Crisis Centre on the IASP website.
We also highly recommend that you visit the Connecting with People: Staying Safe resource.
Bolton, J. M., Gunnell, D., & Turecki, G. (2015). Suicide risk assessment and intervention in people with mental illness. BMJ, 351(nov09 1), h4978–h4978. http://doi.org/10.1136/bmj.h4978
BMJ website (2015) Article types. Last accessed 2 Dec 2015.
Bostwick, J. M., & Pankratz, V. S. (2000). Affective disorders and suicide risk: a reexamination. American Journal of Psychiatry, 157, 1925–1932. [PubMed abstract]
Cavanagh, J. T. O., Carson, A. J., Sharpe, M., & Lawrie, S. M. (2003). Psychological autopsy studies of suicide: a systematic review. Psychological Medicine, 33(3), 395–405.
Cha, C. B., Najmi, S., Park, J. M., Finn, C. T., & Nock, M. K. (2010). Attentional bias toward suicide-related stimuli predicts suicidal behavior. Journal of Abnormal Psychology, 119(3), 616–622. http://doi.org/10.1037/a0019710
Cole-King A, Garett V, Williams H, Hines K, Platt S. (2013). Suicide mitigation embedding compassion in clinical care. Advances in Psychiatric Treatment, 19, 276-283.
Cole-King A, Lepping, P. (2010). Suicide mitigation: time for a more realistic approach. British Journal of General Practice, 60(570), 3-4
Klonsky, E. D., & May, A. M. (2014). Differentiating Suicide Attempters from Suicide Ideators: A Critical Frontier for Suicidology Research. Suicide and Life-Threatening Behavior, 44(1), 1–5. http://doi.org/10.1111/sltb.12068
McLean, J., Maxwell, M., Platt, S., Harris, F. M., & Jepson, R. (2008). Risk and protective factors for suicide and suicidal behaviour: A literature review. Retrieved from http://www.storre.stir.ac.uk/handle/1893/2206
Nock, M. K., Hwang, I., Sampson, N., Kessler, R. C., Angermeyer, M., Beautrais, A., … Williams, D. R. (2009). Cross-national analysis of the associations among mental disorders and suicidal behavior: Findings from the WHO World Mental Health Surveys. PLoS Medicine, 6(8), e1000123. http://doi.org/10.1371/journal.pmed.1000123
Nock, M. K., Park, J. M., Finn, C. T., Deliberto, T. L., Dour, H. J., & Banaji, M. R. (2010). Measuring the Suicidal Mind: Implicit Cognition Predicts Suicidal Behavior. Psychological Science, 21(4), 511–517. http://doi.org/10.1177/0956797610364762
O’Connor, R. C., & Nock, M. K. (2014). The psychology of suicidal behaviour. The Lancet Psychiatry, 1(1), 73–85. http://doi.org/10.1016/S2215-0366(14)70222-6
Randall, J. R., Rowe, B. H., Dong, K. A., Nock, M. K., & Colman, I. (2013). Assessment of self-harm risk using implicit thoughts. Psychological Assessment, 25(3), 714–721. http://doi.org/10.1037/a0032391
Randall, J. R., Walld, R., Finlayson, G., Sareen, J., Martens, P. J., & Bolton, J. M. (2014). Acute Risk of Suicide and Suicide Attempts Associated With Recent Diagnosis of Mental Disorders: A Population-Based, Propensity Score–Matched Analysis. Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie, 59(10), 531–538.
World Health Organization. (2014). Preventing suicide: a global imperative.