Prognostic models to detect and monitor the near-term risk of suicide: state of the science

Abstract

Aspirational Goal 3 of the National Action Alliance for Suicide Prevention's Research Prioritization Task Force research agenda is to "find ways to assess who is at risk for attempting suicide in the immediate future." Suicide risk assessment is the practice of detecting patient-level conditions that may rapidly progress toward suicidal acts. With hundreds of thousands of risk assessments occurring every year, this single activity arguably represents the most broadly implemented, sustained suicide prevention activity practiced in the U.S. Given this scope of practice, accurate and reliable risk assessment capabilities hold a central and irreplaceable position among interventions mounted as part of any public health approach to suicide prevention. Development of more reliable methods to detect and measure the likelihood of impending suicidal behaviors, therefore, represents one of the more substantial advancements possible in suicide prevention science today. Although past "second-generation" risk models using largely static risk factors failed to show predictive capabilities, the current "third-generation" dynamic risk prognostic models have shown initial promise. Methodologic improvements to these models include the advent of real-time, in vivo data collection processes, common data elements across studies and data sharing to build knowledge around key factors, and analytic methods designed to address rare event outcomes. Given the critical need for improved risk detection, these promising recent developments may well foreshadow advancement toward eventual achievement of this Aspirational Goal.