Suicide Prediction – A Shift in Paradigm Is Needed

(cited in Shea, 2011, p. xii).

Predicting Suicide: Science, Art, or Chance?

A landmark study by Pokorny (1983) demonstrated the difficulties associated with predicting suicide. He tried to predict suicide in a sample of 4,800 consecutive Veteran patients, all assessed against 21 known risk factors for suicide risk. More than 800 of these patients were assessed as being at high risk for suicide; however, only 3.75% (n = 30) of these individuals actually died by suicide within a 5-year follow-up period. Unexpectedly, 37 suicides occurred (out of 67) among those that were not previously identified as being at increased suicide risk (more false negatives than true positives). Pokorny's findings revealed an over-identification of persons at risk who never went on to suicide, and a high number of false negatives, constituted by individuals whose high suicide risk was missed or not identified. The overall predictive validity of his assessment was 2.8% (Pokorny, 1983).

The low base rate of suicide makes it particularly difficult to accurately predict suicide risk, and thus to identify suicidal individuals. Twenty-five years ago, Goldstein, Black, Nasrallah, and Winokur (1991) pointed out that there was no single psychometric tool, psychological test, or clinical technique that could accurately predict short-term suicide risk. Improvement in the sensitivity of suicide risk tools (the ability to accurately detect suicidal behavior) has apparently only led to increased rates of false positives. On the other hand, increased specificity of suicide risk tools (the ability to accurately rule out suicidal behavior when it is not present) has led to increased rates of false negatives, whereby individuals who are at risk have not been detected as such (Jacobs, Brewer, & Klein-Benheim, 1999). Clearly, this aspect is much less acceptable than a high number of false positives.

An obvious challenge in gaining reliable data on a person's suicidal status is the fluctuating nature of suicidality. It is almost impossible to conduct research on the more proximal or immediate indicators of a person's suicidal behavior, short of asking directly this information from the client in the aftermath of a suicide attempt, or gaining the reports of events observed just prior to suicide from those who were close to the deceased. The changing state of suicidality cannot be understood in the same way as empirically based risk factors for suicide that inform longer-term vulnerability. Therefore, traditional and categorical risk assessment tools, administered today in clinical practice and hospital admission, do little to inform of actual risk.

So, if we cannot reliably predict suicide in those presenting in vulnerable suicidal states, why should we spend time, resources, and energy in trying to do so? And, if suicide risk assessment results in so much ineffectiveness, why do we continue to make categorical-based judgements about suicidal status (categorizing people at low, medium, and high risk) when a recent meta-analysis on studies of suicide risk assessment shows that about half of those identified as at low risk are likely to go on and kill themselves (Large et al., 2016; Large & Ryan, 2012)?

These and similar questions have become increasingly frequent among mental health professionals and researchers. Large, Sharma, Cannon, Ryan, and Nielssen (2011), Ryan, Nielssen, Paton, & Large (2010), Undrill (2007), and Mulder (2011) all argue about the usefulness of traditional risk assessment tools and procedures to identify suicidal people. Recently, a systematic review and meta-analysis on the ability of assessment scales to predict suicide following self-harm drew the conclusion that none of the tested scales provided sufficient evidence to support their use in predicting suicide (Chan et al., 2016). In addition, "the use of these scales, or an over-reliance on the identification of risk factors in clinical practice, may provide false reassurance and are, therefore, potentially dangerous" (Chan et al., 2016, p. 277).

On the other hand, as pointed out by Draper (2012), it is a bit risky to dismiss suicide assessment tools altogether. Apart from their use in clinical trials, introduction of suicidality questions at hospital admission – where implemented – was generally well received. Anyhow, over the past 10–15 years the conceptualization of suicide risk assessment has graduated significantly in the clinical domain to a more sophisticated and systematic process for learning about a person (including obtaining a psychosocial needs-based understanding of suicidal status) and mobilizing strengths and supports to keep the person safe.

So while the use of suicide prediction scales remains unsatisfactory, there is growing support for a renewed risk assessment approach that is collaborative, client-centered, and needs-based for uncovering the suicidal status of a person. The Aeschi Working Group – founded in 2000 – strongly emphasized the client-oriented approach, putting the client at the center of the stage as expert of his/her story, to tell about their own suicidality (Michel & Jobes, 2010). This approach also underlined the importance of the therapeutic alliance and understanding the client's conceptualization of suicidality. It originated in opposition to the traditional approach on risk assessment and treatment of the suicidal person, considered to be too medicalized and impersonal.

As Draper (2012) suggests, where the risk assessment is a systematic process focusing on the individual's circumstances (including a range of social, environmental, situational, family, and other domains), and involving a series of convergent data-gathering procedures (e.g., interviewing family, significant others, workplace etc.), this becomes more effective. Of course, technical competence and an empathetic attitude make the client–clinician relation a trustworthy one. These characteristics can be extremely helpful in responding to and managing current and short-term suicidality. Perceived this way, the aim of risk assessment should be focused on mitigating risk and informing effective and personalized care of the individual, with an ongoing view to support and facilitate growth and recovery; ultimately, to improve and enhance the person's self-stated reasons and desires for living.

The Building Up of STARS

Sharing the same approach, some suicidologists of the Australian Institute for Suicide Research and Prevention (AISRAP) at Griffith University, Queensland, have compiled a protocol for screening potential suicidal persons, named the "Screening Tool for Assessing Risk of Suicide" (STARS; Hawgood & De Leo, 2015). STARS was presented after several years in development; three previous, unpublished versions underwent consecutive edit iterations prior to the current form. Rather than a quantitative instrument, STARS privileges a qualitative dimension aimed at gaining data based on the client's narrative around current suicidal state, to inform needs-based commensurate care responses. STARS therefore represents a roadmap to a systematic suicide risk assessment to be undertaken by specialized mental health professionals able to incorporate mental status examination and psychiatric evaluation. However, the administration of STARS does not require a formal training in psychiatry. On the contrary, its administration desirably implies critical knowledge and understanding of the client-oriented approach to risk assessment, which relies on the human-to-human connection and seeks collaborative understanding of the troubles that bring about the intense suffering.

Often the clinician loses sight of the true meaning and objective of understanding a person's suicidal state, owing to anxieties and fears around potential litigation as well as the shear challenges that come with immersing oneself in the client's world of psychache (Shneidman, 1993). The loss of a client due to suicide (or fear of such a loss) is also well known as one of the most stressful of all events in the life of a clinician (Ratkowska, Grad, De Leo, & Cimitan, 2013; Seguin Bordeleau, Drouin, Castelli-Dransart, & Giasson, 2014). Jobes (2006), Linehan, Comtois, and Ward-Ciesielski (2012), Shea (2011), and numerous other researchers in the field have discussed clinicians' unwillingness to treat suicidal individuals because of these issues, which often lead to defensive practices, suboptimal responses with suicidal patients, and sometimes termination of treatment once suicidality is disclosed. Some suicide risk assessment measures have recently been developed that focus on benefits to both the patient and the clinician (Simon & Shuman, 2009). At least in part, these instruments aim to reduce clinician anxiety by providing assistance in data gathering and informing client safety and appropriate treatment (Jobes, 2006; Linehan et al., 2012). Some examples of today's leading suicide risk assessment and management approaches include David Jobes's Collaborative Assessment and Management of Suicidality (CAMS; Jobes, 2006, 2016), Chronological Assessment of Suicide Events (CASE) by Shaun Shea (2011), Attempted Suicide Short Intervention Program (ASSIP) by Michel and Gysin-Maillart (2015), and Marsha Linehan's University of Washington Risk Assessment Protocol (UWRAP for assessors; Linehan, Comtois, & Murray, 2000). Common to all of these approaches are the importance of the therapeutic alliance and an emphasis on collaboration as central to obtaining critical information on suicidality. The belief here is that commonly experienced shamefulness about the experience of suicidality, which often prevents such disclosures, may be alleviated by a trusting and empathetic therapeutic alliance.

Likewise, STARS essentially focuses on collaboratively obtaining important information from the client's perspective, using clinician probes and client-weighted needs or priorities for informing either further systematic assessment or immediate safety responses. The STARS protocol is not a psychometric tool administered to categorize or stratify level of suicide risk in an individual; or to predict short- or longer-term suicidality. On the other hand, no single suicide risk assessment tool or procedure has been demonstrated to correctly identify future suicide risk (Chan et al., 2016; Simon, 2011), and our capabilities in predicting suicide (SBU, 2015), and subsequent self-harm episodes (Bolton, Gunnell, & Turecki, 2015; Kapur et al, 2005), remain very limited.

Main Features of STARS

As mentioned, STARS is designed for collaborative administration, beginning with the clinician requesting the client to share his/her own narrative of current problems. It is an assumption that clinicians or mental health workers will be well versed in micro-counseling skills as well as basic interviewing abilities. Each section of STARS includes questioning probes against specific items pertaining to several domains of enquiry that fall under three separate sections: Part A – Suicidal Behavior Enquiry (critical questions asked first in a crisis situation); Part B – Risk Factor Enquiry (in cases where suicidality is not imminent, enquiry may start with nonintrusive questioning); Part C – Protective Factors Enquiry (again, where risk is not imminent, enquiry may start with nonintrusive questioning). These three sections are neither prescriptive nor linear, and may be covered in any particular order. Probes were designed simply as examples to facilitate access to client experiences, which are often difficult to label and disclose (such as suicidal status, feelings about self, burdensomeness, shame). In general, the tool provides a pathway to enquiry into specific and broader areas of an individual's life known to impact on the inner psychological state. The final Clinical Notes section of STARS (discussed later) provides direction on documentation of the obtained data in a way that might best reflect justified response and safety planning.

The background of STARS is based on the combined theoretical works of Edwin Shneidman (1988), Aaron Beck (Beck, Rush, Shaw, & Emery, 1979), Roy Baumeister (1990), Marsha Linehan (1993), David Jobes (1995), Konrad Michel (Michel & Jobes, 2010), Shawn Shea (2011), and David Rudd, Thomas Joiner, and Hasan Rajab (2001). It is beyond the scope of this editorial to expand further into specific theories and principles underpinning STARS domains of enquiry and items; suffice to say that accumulating empirical evidence from the works of many researchers guided our selection of STARS items.

The meaning of the items assessed by STARS will be unique to each individual and as such an opportunity exists for either the client or the clinician to document (under specify columns) responses to each item/factor separately. There is no scoring of these responses, just a determination of current experience and priority of items for informing care.

Confidence ratings are requested so that the clinician is: (a) encouraged to reflect upon his/her decisions rather than merely accepting first-impression inferences; (b) considers acuity and changeability of the suicidal state (including influences on client capacity to engage); and (c) able to indicate to colleagues and others the level of certainty associated with item responses. Higher clinician confidence ratings gained from more comprehensive information can result in more accurate documentation of client needs, including a clear rationale and purposeful management and caring of the client.

Data obtained from the use of STARS could not be sufficiently reliable when derived from the client alone, despite recognition that the client is the best author of his or her story. Suicidal persons do not always share their true intentions; even in the context of the deepest clinical engagement, the actual intent to die may not be revealed. Several reasons for this are well documented (Jobes, 2006; Shea, 2011). Convergent information from other family members, including next of kin, partner, peer, and sibling, is therefore recommended for more confident documentation and more systematic suicide risk formulation. The STARS data recording form (Clinical Notes) requires clinicians to source this information, where possible. While seemingly time-consuming, such type of commitment may not only prevent fallible or suboptimal management responses but may save a client's life.

The clinical notes section of STARS is primarily for clinician reporting on a synthesis of the observed suicidal status of the client and the negotiated actions determined between the client, clinician, and other health professionals and/or carer for best meeting the client's needs. Specifically, clinician documentation is requested on: (a) self-reported warning signs indicating immediate perceived indicators of suicidality; (b) self-reported life circumstances and situational factors perceived as most relevant to current experience of distress and suicidality; and (c) presence and absence of risk factors and protective factors reflected by the client and via available collateral data.

A safety planning and response section follows, which offers the opportunity for clinical documentation of outcomes from clinician–client negotiation around ways to keep the client alive, and issues concerning client willingness to engage in help-seeking as well as help-accepting behaviors during and leading up to crises or distressing suicidal states. The main emphasis in this section is ensuring that safety planning responses are well aligned with client-reported priority needs associated with immediate and short-term potential suicidality. An additional focus is on documentation of actions taken by the clinician and negotiated with the client on both the existing and organized crisis supports and activities planned for mitigating future suicidality. Documentation is also sought for evidence of clinician consultation with other health professional(s) and/or caretaker or next of kin (where possible and appropriate).

The main part of the STARS clinical notes section is then concluded with a request for clinician summation of observed and current client suicidality and clinician confidence in the judgment of the suicidal status and consequent associated management plan. The clinician's rationale for actions taken to meet client needs and increase safety is brought to awareness in this section. Follow-up assessment and review periods as well as documentation of evidence of peer consultation are also requested in this final section.

Conclusion

STARS provides an opportunity for collaboratively and compassionately engaging a client in discussion about some of the most hidden or least communicated experiences of his/her life. It offers the clinician a timely opportunity to access information on a range of risk, protective, and circumstantial elements known to potentially contribute to suicidality, in order to delicately explore important personal areas that may require further support and intervention.

With the use of STARS, we hope that clinician anxiety may be reduced, while more reliable and meaningful documentation and effective client care may increase.