Predicting general criminal recidivism in mentally disordered offenders using a random forest approach

Area of Science:

• Forensic Psychiatry
• Criminal Psychology
• Mental Health Law

Background:

• Psychiatric expert opinions require scientific validation for assessing offender recidivism risk.
• General recidivism risk assessment in mentally ill offenders is underdeveloped in Continental Europe.

Purpose of the Study:

• To develop a valid instrument for assessing the risk of general criminal recidivism in mentally ill offenders.
• To improve forensic-psychiatric risk assessment for legal and therapeutic applications.

Main Methods:

• Analysis of 259 mentally ill offenders' data and criminal records.
• Identification of risk factors for general criminal recidivism.
• Classification of re-offences using a random forest approach.

Main Results:

• 51% of mentally ill offenders were reconvicted.
• Key predictors identified: number of prior convictions, age, index offense type, criminal history diversity, and substance abuse.
• The model achieved high accuracy in identifying re-offenders (58-95%) and offenses (65-97%), with an AUC of .90.

Conclusions:

• A novel statistical approach for forensic-psychiatric risk assessment was developed.
• This tool aids in evaluating general reoffending risk in mentally disordered individuals, particularly high-risk groups.
• The approach supports expert court opinions, risk management, and therapeutic interventions.