Hard Truths About Suicide Prevention

Suicide accounts for nearly 50 000 deaths annually in the US, making it the second leading cause of death among persons 10 to 34 years of age.¹ Although psychiatric illness is associated with elevated rates of death from a range of causes, from cardiovascular disease to cancer,² suicide stands out: it occurs precipitously, disproportionately involves younger individuals, and is generally viewed as more preventable.

Suicide represents a particular challenge in the military because soldiers are placed in extremely stressful situations, often without adequate physical or emotional support. Their risk remains elevated even after they leave active service and attempt to reenter a society ill-equipped to acknowledge their special needs. For this reason, the US Department of Defense and the US Department of Veterans Affairs (VA) have invested billions of dollars to reduce the incidence of suicide. Every VA secretary for the past 15 years has made suicide prevention a top priority and vowed to eliminate suicide. They have launched initiatives that included hiring more than 400 suicide counselors; establishing hotlines that receive more than 600 000 calls per year; screening patients for depression and posttraumatic stress disorder at nearly 60 million primary care, emergency, and mental health visits each year; and ensuring that every person discharged from the military is contacted personally. Despite these heroic efforts, the number of veterans who die by suicide every year has actually increased during the past decade.³

These circumstances highlight how difficult and complex a task it is to prevent suicide more broadly and why limiting interventions to the moment when individuals present with suicidal thoughts does not suffice. Accordingly, suicide has come to be viewed as a public health challenge for which the solution begins with screening and prevention. In this issue of JAMA Network Open, 2 reports^4,5 examine complementary approaches to screening for suicide risk, and both illustrate how daunting the logistics are. Bahraini et al⁴ describe the incorporation of suicide screening questions into the VA’s standard workflow in outpatient and emergency department settings during a 12-month period during 2018 and 2019. In outpatient settings, primary screening results for suicidality were positive in 3.5% and secondary screening results for suicidality were positive for 0.4% of 4.1 million patients. Among 1 million patients in emergency departments, rates of positive screening results were 3.6% and 2.1%. Although the VA is not representative of general clinical practice, these numbers provide a useful reference for estimating the yield of routine screening in these settings. Even so, nearly two-thirds of veterans who die by suicide have not sought any type of health care from the VA.³ Thus, even in a perfect world of optimal screening and intervention by the VA, two-thirds of suicide deaths among veterans would not be prevented.

Kline-Simon et al⁵ describe another approach to screening, based on extraction of data from electronic health records to identify high-risk individuals. Their model is characterized as highly accurate but also embodies key challenges in suicide screening. The positive predictive value of their model is 6%, which means that 17 individuals would need to receive an intervention to prevent a single suicide attempt. High false-positive rates have been a persistent challenge in screening efforts⁶ that have not been overcome by machine learning prediction models.⁷ Even in a system willing to invest those resources, the model’s sensitivity was only approximately 41%, indicating that nearly 60% of suicide attempts would be missed. The authors also recognize that the high number of alerts could readily lead to distraction and alert fatigue, which has been associated with overlooking important findings and results.⁸ Acknowledging the public health need cannot blind us to the profound limitations of existing approaches.

For any screening effort to be clinically useful, there must also be effective and accessible interventions. The US Preventive Services Task Force notes that depression screening should be implemented “with adequate systems in place” (ie, resources to ensure diagnosis and treatment).⁹ That is, there must be resources to respond to positive screening results (ie, not simply notifying a beleaguered practitioner) and to provide patients with effective interventions. Otherwise, the expense and burden on patients, families, clinicians, and staff are to no avail.

Identifying such effective interventions represents another major challenge because suicide prevention studies face substantial feasibility hurdles. For example, although many pharmacologic treatments improve symptoms of depression, few randomized clinical trials have clearly demonstrated a reduction in risk of suicide, even when nonrandomized studies and meta-analyses are suggestive of a reduction.¹⁰ Results of individual psychosocial interventions are likewise disappointing,⁶ although brief interventions have shown modest benefit in reducing future suicide attempts.¹¹ Regardless, the cornerstone of treatment for suicidality remains treating the underlying psychiatric illness, often major depressive episodes.

A third article¹² in this issue addresses this need for development of effective interventions by evaluating a group-based approach to suicide reduction. Wyman et al¹² describe a cluster-randomized trial of a multifaceted program that encouraged formation of supportive networks along with didactic training among US Air Force personnel. They observed modest, but statistically significant, reductions in depressive symptoms, including suicidality, at 6 months. Of importance, they also identify mediating effects for unit cohesion, suggesting a mechanism of action for their intervention. This study adds to a literature that group-based interventions are effective in reducing depressive symptoms and may have advantages in resource-constrained environments. Notably, the study does not actually demonstrate prevention of suicide per se. Whether targeted strategies to reduce suicide are worthwhile, rather than simply developing better treatments for depression, remains to be studied.

In aggregate, these 3 articles highlight the major obstacles in suicide prevention: whether automated or not, screens pose problems both of false-positive and false-negative results, and well-documented, specific, and scalable interventions remain elusive. Like most serious public health problems related to chronic illness, a multipronged approach will be necessary. This strategy entails not only accurately and efficiently identifying individuals at risk but also using a range of targeted options to reduce that risk. It will require concerted initiatives by government and health care delivery organizations to address underlying, pervasive problems that contribute to suicide risk, such as substance use disorders, and to bolster the availability of mental health services. As the VA experience suggests, it will be difficult to reduce the number of suicides without tackling the even more challenging tasks of improving access and developing a new generation of treatments. The studies in this issue underscore the challenges; if the will to invest in mental health is there, these represent hard but ultimately tractable problems.

Accepted for Publication: August 26, 2020.

Published: October 21, 2020. doi:10.1001/jamanetworkopen.2020.22713

Correction: This article was corrected on November, 9, 2020, to fix the spelling of the last name of the author of an article discussed in the editorial.

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Perlis RH et al. JAMA Network Open.

Corresponding Author: Roy H. Perlis, MD, MSc, Department of Psychiatry, Massachusetts General Hospital and Harvard Medical School, 185 Cambridge St, Simches Research Building, 6th Floor, Boston, MA 02114 (rperlis@mgh.harvard.edu).

Conflict of Interest Disclosures: Dr Perlis reported receiving personal fees from RID Ventures, Takeda, Genomind, Outermost Therapeutics, Psy Therapeutics, and Burrage Capital outside the submitted work. No other disclosures were reported.

Disclaimer: Dr Fihn is a deputy editor and Dr Perlis is an associate editor for JAMA Network Open, but neither was involved in any of the decisions regarding review of the manuscript or its acceptance.