Research Areas

  • ML/NLP for Suicide Risk Prediction & Detection

  • Developing Interventions That Work

  • Mental Disorders and Suicide Risk

  • Testing Theories of Suicide

ML/NLP for Suicide Risk Prediction & Detection

Suicide is very hard to predict. In fact, despite decades of research, the ability to predict suicide remains only slightly better than chance. One reason for the lack of progress in this area is that many studies have examined isolated risk factors (e.g., depression) as predictors of suicide. My work leverages machine learning (ML) and natural language processing (NLP) to develop more precise, complex models for predicting and detecting suicide risk.

Given the meteoric rise in using ML/NLP to predict suicide, many questions have emerged about the clinical utility (or lack thereof) of these methods. To that end, my research has also considered how to best harness ML-based risk models in clinical settings.

Currently, my work at Crisis Text Line focuses on using NLP to better understand and help people in suicidal crisis using a dataset of more than 11 million text message conversations. You can read more about our work here.

    • Zuromski KL, Low DM, Jones NC, Kuzma R, Kessler D, Zhou L, Kastman EK, Epstein J, Madden C, Ghosh SS, Gowel D, Nock MK. Detecting suicide risk among US servicemembers and veterans: a deep learning approach using social media data. Psychol Med, 2024 Sep;54(12):3379- 3388. PMID: 39245902. PDF

    • Zuromski KL, Bernecker SL, Gutierrez PM, Joiner TE, King AJ, Liu H, Naifeh JA, Nock MK, Sampson NA, Zaslavsky AM, Stein MB, Ursano RJ, Kessler RC. Assessment of a risk index for suicide attempts among US Army soldiers with suicidal ideation: Analysis of data from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). JAMA Netw Open. 2019 Mar;2(3):e190766. PMID: 30874786. View Online

    • Ross EL, Zuromski KL, Reis BY, Nock MK, Kessler RC, & Smoller JW. Accuracy requirements for cost-effective suicide risk prediction in the United States. JAMA Psychiatry. 2021 Jun 1;78(6):642-650. PMID: 33729432. View Online

    • Low DM, Zuromski KL, Kessler D, Ghosh SS, Nock MK, Dempsey W. It's quality and quantity: The effect of the amount of comments on online suicidal posts. Proc Conf Empir Methods Nat Lang Process. 2021 Nov:95-103. PMID: 35224567. View Online

    • Bentley, KH, Zuromski KL, Fortgang RG, Madsen EM, Kessler D, Lee H, Nock MK, Reis B, Castro VM, & Smoller JW. Focus group study on implementing machine learning models for suicide risk prediction in clinical practice. JMIR Form Res. 2022 Mar 11;6(3):e30946. PMID: 35275075. View Online

Developing Interventions That Work

Most people who die by suicide do not have contact with a mental health provider in the year leading up to their death. How can we ensure people who need help are receiving it? My work in this area has focused on better understanding the reasons why people do not seek out mental health treatment, including stigma and other barriers (e.g., cost of treatment), identifying cost-effective treatments, and exploring novel intervention approaches.

I have focused on building user-friendly, technology-delivered interventions that people can access outside of traditional healthcare settings. In an ongoing grant from the National Institute of Mental Health (R34MH124973), my team is conducting a randomized controlled trial to test the effectiveness of a novel, game-like smartphone app to reduce suicide risk in adolescents after they leave psychiatric hospitalization.

The development of scalable, technology-based interventions like this will enable more people to receive potentially life-saving care.

    • Zuromski KL, Dempsey CL, Ng THH, Riggs-Donovan CA, Brent DA, Heeringa SG, Kessler RC, Stein MB, Ursano RJ, Benedek D, Nock MK. Utilization of and barriers to treatment among suicide decedents: Results from the Army Study to Assess Risk and Resilience Among Servicemembers (Army STARRS). J Consult Clin Psychol. 2019 Aug;87(8):671-683. PMID: 31008631. View Online.

    • Bernecker SL, Zuromski KL, Curry JC, Kim JJ, Gutierrez PM, Joiner TE, Kessler RC, Nock MK, Rudd MD, & Bryan CJ. Economic evaluation of brief cognitive-behavioral therapy vs treatment as usual for suicidal U.S. Army soldiers. JAMA Psychiatry. 2020 Mar;77(3):256-264. PMID: 31774485. View Online.

    • Wilks CR, Yin Q, Zuromski KL. User experience affects dropout from Internet-delivered dialectical behavior therapy. Telemed J E Health.2020 Jun;26(6):794-7. PMID: 31502945. View Online.

    • Follet LE, Kelly F, Millner AJ, Zuromski KL. Examining the acceptability and feasibility of a Just-In-Time Adaptive Intervention for suicidal adolescents. Poster presented at: 58th Annual Convention of the Association of Behavioral and Cognitive Therapies; 2024 Nov 14-17; Philadelphia, PA.

Mental Disorders and Suicide Risk

Suicidal thoughts and behaviors frequently occur alongside other mental health concerns. The association between suicide risk and mental disorders is often bidirectional, with each influencing and exacerbating the other. Further, mental disorders share underlying mechanisms with suicidal behavior, such as emotion regulation difficulties and impaired cognitive control.

In my research, I have sought to better understand the overlap between suicide risk and mental disorders like PTSD and eating disorders. This has included work focused on how to best measure the presence of these disorders. I have also examined how transdiagnostic factors like sleep problems impact risk for suicide, harnessing technologies such as smartphones and wearable biosensors to study these phenomena in real-time. This work, funded by my recently completed Career Development award from the National Institute of Mental Health (K23MH120439), highlights sleep as a proximal, modifiable risk factor for suicide.

    • Zuromski KL, Cero I, Witte TK, Zeng P. The quadratic relationship between body mass index and suicide ideation: A non-linear analysis of indirect effects. Suicide Life Threat Behav. 2017 Apr;47(2):155-167. PMID: 27291861. PDF

    • Zuromski KL, Cero I, Witte TK. Insomnia symptoms drive changes in suicide ideation: A latent difference score model of community adults over a brief interval. J Abnorm Psychol. 2017 Aug;126(6):739-749. PMID: 28557509. PDF

    • Smith AR, Zuromski KL, Dodd DR. Eating disorders and suicidality: What we know, we don’t know, and suggestions for future research. Curr Opin Psychol. 2018 Aug;22:63-67. PMID: 28846874. PDF

    • Zuromski KL, Ustun B, Hwang I, Keane TM, Marx BP, Stein MB, Ursano RJ, Kessler RC. Developing an optimal short-form of the PTSD Checklist for DSM-5 (PCL-5). Depress Anxiety. 2019 Sep;36(9):790-800. PMID: 31356709. View Online.

Testing Theories of Suicide

The rate of suicide in the U.S. been increasing over the past 20 years. To better understand how and why suicide happens, as well as how to prevent it, a number of theoretical models have been proposed. The Interpersonal Theory is one of the major theories of suicide (Joiner, 2005; Van Orden, Witte et al., 2010). This theory hypothesizes that suicide happens when someone possesses both suicidal desire and the capability to enact self-harm.

I conducted stringent tests of the theory in diverse samples (e.g., psychiatric inpatients, detained adolescent offenders, Army soldiers) during graduate school (working closely with my advisor Tracy Witte, who helped develop this theory) and in my postdoctoral fellowship. My work contributed to the knowledge base for what has arguably been the most influential modern theory of suicide.

    • Zuromski KL, Witte TK. Fasting and acquired capability for suicide: A test of the interpersonal-psychological theory of suicide in an undergraduate sample. Psychiatry Res. 2015 Mar;226(1):61-67. PMID: 25530417. PDF

    • Cero I, Zuromski KL, Witte TK, Ribeiro JD, Joiner TE. Perceived burdensomeness, thwarted belongingness, and suicide ideation: Re-examination of the Interpersonal-Psychological Theory in two samples. Psychiatry Res. 2015 Aug 30;228(3):544-50. PMID: 26099656. PDF

    • Zuromski KL, Cero I, Witte TK. Non-monotonic temporal variation in fearlessness about death: A latent class growth analysis. Psychiatry Res. 2018 Oct;268:46-52. PMID: 29986178. PDF

    • Chu C, Zuromski KL, Bernecker SL, Gutierrez PM, Joiner TE, Liu H, Naifeh JA, Stein MB, Ursano RJ, & Nock MK. A test of the interpersonal theory of suicide in a large, representative, prospective study: Results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Behav Res Ther. 2020 Sep;132:103688. PMID: 32731055. View Online