Harnessing Computerized Adaptive Testing, Transdiagnositic Theories of Suicidal Behavior, and Machine Learning to Advance the Emergent Assessment of Suicidal Youth (EASY) Study
Back to Dr. Brent's Research
Dates: 4/1/19-3/31/2023
Assessing acute suicidal risk in adolescents during emergencies is a significant clinical challenge because current methods for predicting suicide attempts are limited, and the risk among suicidal adolescents is high. No studies have systematically examined the most effective approaches for predicting suicidal behavior in youth presenting to psychiatric emergency departments (PEDs). To address this gap, we proposed a study involving 1,800 youth seen at a regional PED, including 1,350 evaluated for suicidal risk, with follow-up at 1, 3, and 6 months to identify subsequent suicide attempts.
The study tested three complementary approaches: (1) computerized adaptive tests (CATs) for suicidal risk and six diagnostic groups, completed by youth and parents in 10–15 minutes; (2) theory-based measures of suicide risk, including Shneidman’s concept of psychache and Joiner’s Interpersonal Theory of Suicide; and (3) machine learning and natural language processing applied to electronic health records. We hypothesized that each approach will outperform standard clinical assessment and that combining all three will provide the most accurate predictions. This research is innovative because it integrates CATs, theory-driven measures, and advanced data analytics in a large, high-risk adolescent sample, with the potential to develop brief, accurate tools for identifying youth at imminent risk and improving resource allocation in emergency mental health care.
Selected Primary Outcome Papers:
- PubMed - Predictive Validity of the K-CAT-SS in High-Risk Adolescents and Young Adults
- ScienceDirect - A Prospective Examination of the Predictive Validity of the Three Transdiagnostic Assessments of Risk for Suicidal Behavior: Psychache, the Interpersonal Theory of Suicide, and Reasons for Living