Project Overview

Many assessments of suicide risk focus mainly on assessing symptoms of depression, past suicidal behavior, and suicidal ideation because these traits are strongly associated with suicide attempts. Although an effective suicide assessment should be relatively brief, assessments such as these may be neglecting to include other behavioral and demographic variables which are strongly associated with attempted suicide. By referencing psychological studies on variables associated with suicide attempts and implementing these in a machine learning model, we created a model which relatively accurately predicts who is likely to attempt (or who has already attempted) suicide based on responses to the National Comorbidity Survey - Baseline. If a model like this is incorporated into a standard computerized health assessment, it would be possible to automatically alert the practitioner if their client’s assessment responses are similar to people who have attempted suicide in the past.