We developed a dynamic prediction model to predict criminal recidivism in individuals under community supervision. The model accounts for adverse events, changes in circumstances, and offense-free time, resulting in improved discrimination performance. Our comparison of the dynamic model with its static equivalent highlights the benefits of incorporating offense-free time in actuarial risk assessment tools. Our findings reveal that traditional actuarial tools, when used as monitoring instruments, systematically overestimate recidivism risk over time.

Educational Objectives

• Describe the importance of incorporating adverse life events and desistance effects in actuarial risk assessment tools
• Identify the limitations of actuarial tools not specifically developed for risk monitoring, including the potential for bias towards overestimation
• Discuss the process of developing a dynamic risk assessment tool