This prognostic study investigates if administrative data available at periodic health assessments can be used in machine learning models to identify US Army soldiers at risk of suicide deaths and nonfatal suicide attempts for targeting a primary prevention.