Forecasting recidivism in mentally ill offenders released from prison

Area of Science:

• Criminology
• Forensic Psychology
• Corrections

Background:

• Limited research exists on recidivism risk assessment for mentally ill offenders (MIOs) post-release from state prisons.
• Accurate risk assessment is vital for public safety and effective offender management.

Purpose of the Study:

• To identify variables that accurately forecast felony and violent reconviction among mentally ill offenders.
• To evaluate the predictive accuracy of simple correctional variables compared to complex risk assessment instruments.

Main Methods:

• Logistic regression analysis was employed on data from 333 mentally ill offenders released from Washington State prisons.
• Simple, recoded versions of identified variables were used to predict reoffense.
• Comparison of predictive accuracy between simple variable sums and complex logistic regression equations.

Main Results:

• Logistic regression models accurately forecasted felony and violent reconviction.
• Sums of simple recoded variables achieved predictive accuracy comparable to state-of-the-art risk assessment instruments.
• Five of the nine identified variables demonstrated protective effects against reoffense.

Conclusions:

• Standard correctional variables possess significant value in forecasting risk for mentally ill offenders.
• Actuarial risk assessments should be based on local data for improved accuracy.
• Protective factors are important considerations in assessing MIO risk.
• Dynamic, situational, and clinical variables may further enhance predictive accuracy for emergent community risk.