Predictive Validity of the e-Connect Suicide Risk Classification Algorithm in Youth on Probation

Youth on juvenile probation have alarmingly elevated rates of suicidal thoughts and behaviors (STBs). We recently implemented e-Connect, a clinical decision support system for use by probation officers that includes automated screening, tiered suicide risk classification, and treatment linkage for youth on probation. e-Connect is built on an evidence-informed algorithm that synthesizes information on severity and recency of suicide risk factors to place youth in 1 of 4 suicide risk classes. This paper evaluates the predictive validity of a close analogue of this algorithm for predicting STBs over the 12 months following assessment.

The e-Connect screening instrument is based on 20 items adapted from the Global Appraisal of Individual Needs-Initial (GAIN-I). In the present study, we constructed a close proxy of the e-Connect risk classification using the analogous 20 items from an archival longitudinal dataset of GAIN-I administrations. Selected participants were 4,344 juvenile probationers with baseline and1 to 4 follow-ups available. Logistic regressions assessed relationships of baseline risk class with 4 STBs (ideation, planning, gathering lethal means, and attempts) over the following 12 months.

Baseline risk class predicted likelihood of all 4 STBs over 12 months. Linear contrasts showed that more severe classes were associated with progressively higher risk. Baseline risk class incrementally predicted follow-up STB outcomes even after adjusting for baseline history of STBs.

Results showed strong predictive validity of the tiered risk classification algorithm, and support its use in probation settings for effective triage and treatment linkage of youth at risk for suicide.