(CANCELLED) Desistance From Crime: A Highly Overlooked Protective Factor in Actuarial Risk Assessment (PID070)

11am – 11.30am EDT, 5 September 2024 ‐ 30 mins

Workshop Session

Individuals with psychiatric disorders are disproportionately represented in the correctional system, often lacking adequate resources. Current risk assessment tools used for monitoring assume static risk levels, overlooking the potential effects of crime desistance and sudden but transient changes in an individual's circumstances. Addressing these issues, we created a dynamic prediction model for criminal recidivism among those under community supervision.

Using data from 59,676 individuals given community sentences in Sweden, we developed the model, considering criminal, sociodemographic, and clinical factors. Employing a dynamic prediction method, we accounted for adverse events, changes in circumstances, and crime desistance.

Our model tool demonstrated robust calibration and discrimination (c-index = 0.74 for violent reoffending, 0.69 for general reoffending). We also demonstrated that if used for continuous monitoring, conventional risk assessment tools tend to overestimate recidivism risk over time, highlighting the need for dynamic monitoring.

Our model offers sensitive risk estimates adaptable to individual changes and distance from crime. Implementing such dynamic models could enhance risk stratification for those with psychiatric disorders under supervision, improving service provision and outcomes.