Positive Predictive Values and Potential Success of Suicide Prediction Models

To the Editor We write to disagree with the pessimism of Belsher et al¹ regarding the potential utility of models predicting suicidal behavior. They argue that no existing models have positive predictive value high enough to guide prevention efforts.

Some existing models predicting suicide attempt have positive predictive values equal to or exceeding those of widely accepted risk prediction tools. Among mental health outpatient visits, we can accurately identify those with a 5% risk of suicide attempt over the following 90 days.² For comparison, the US Preventive Services Task Force recommends tamoxifen in women with a predicted risk of breast cancer exceeding 3% over 5 years³ and recommends statins for people with predicted risk of cardiovascular event exceeding 10% over 10 years.⁴