LECTURE 1 – INTRODUCTION
Risk assessment = probability calculation that a harmful behaviour or event will occur, and involves
an assessment about the frequency of a behaviour or event, its likely impact and who it will affect =
attempt to predict the likelihood of future offending in order to identify individuals in need of
intervention (risk management)
Reasons
o Safety (self / fellow residents / society)
o Public interest (prevention of (severe) recidivism)
o Ethical importance (same method for every patient)
o Therapeutic importance (RNR model) (adequate risk assessment leads to
Insight into risk factors
Risk management plan
o Communication (more transparency & uniformity in decision-making, improving
communication about risk & risk management)
History
1876, Lombroso: criminals have certain facial characteristics (hawklike nose, large protruding jaws, ..)
1928, Burgess: first RA tool
Developed to determine risk of re-offence of offenders being released from Illinois prison
Actuarial; marital status, criminal & employment history, institutional misconduct
76% of high risk status recidivated within 5 years
1966, Baxtrom v Herold: 967 ‘dangerous’ psych. Patients transferred to regular psych. hospitals, 121
subsequently released into society.
Followed up by Stadman & Cocozza: after 4 years, 2.7% were sent back to forensic psych.
Hospitals. Of the patients in society 17% got arrested and 7% was convicted
1970s-1980s: unsatisfactory state of art regarding predictions of violence. Ability of professionals to
reliably predict future violence = unproven
Teachers & psychiatrists got asked questions on recidivism based on given information
o IRR of teachers was low, but better than IRR of psychiatrists
o On average large similarity between both groups
o More information more conservative judgement
1990s, Phillipse: Dutch TBS system enjoyed great respect; praised for innovative approach to forensic
patients. Researched how Dutch behavioral experts from forensic fields made RA in practice
None of the predictors were predictive (denial factor, empathy factor, …)
No link between clinical assessment of relapse risk & actual recidivism
Theoretical & empirical framework
Risk, Need & Responsivity (RNR) model by Andrews & Bonta (1990): evidence based practice,
interventions based on results from research into their effectiveness
Risk principle: who to treat?
o Risk of recidivism is central, not the disorder
o Misalignment of treatment intensity may increase recidivism risk
High risk offenders benefit most from more intensive treatment
When same treatment applied to high & low risk groups, treatment effect is
greater in the high risk group
, Need principle: what to treat?
o Each indiv has own combination of factors that led to the crime (criminogenic needs)
Only factors associated with recidivism reduction should be addressed
o Dynamic risk factors are modifiable by treatment interventions
Focus on high dynamic risks, adequate treatment of criminogenic needs is
associated with higher recidivism reduction
Responsivity principle: interventions tailored to offenders’ characteristics (e.g. motivation
level, learning style, personal circumstances
Category Central risk Risk Dynamic need
factor
Big 4 History of early & ongoing involvement work on non-criminal alternative
antisocial in a number and variety of behaviours in high-risk situations
behaviour antisocial acts in a variety of
settings
Antisocial thrill seeking, low self-control, work on problem-solving skills, self-
personality restlessness, aggression management skills, anger
pattern management, coping skills.
Antisocial attitudes, values, beliefs and reduce antisocial cognitions,
cognitions rationalizations that support recognize risky ways of thinking
criminal behaviour; cognitive- and feeling, develop alternative &
emotional conditions of anger, less risky ways of thinking and
resentment and pride: feeling; adopt a renewed and/or
criminal vs anti-criminal anti-criminal identity
identity
Antisocial close contact with criminal reduce contacts with criminal
peers others, relative social isolation others, build contacts with non-
with non-criminal others; criminal others
direct social support for crime
Mod 4 Family/marital education and/or care, control reduce conflict, build positive
relationships and/or supervision relationships, improve control &
oversight
School and/or low levels of achievement and improve performance, rewards,
job satisfaction and satisfaction
Prosocial low levels of engagement and improve engagement, rewards and
recreational satisfaction in non-criminal satisfaction
activities leisure activities
Substance Abuse of alcohol and/or drugs Reduce drug abuse, reduce
abuse personal and interpersonal support
for addictive behaviour, develop
alternatives to drug abuse.
, LECTURE 2 – RISK & PROTECTIVE FACTORS
RA tools & predictive validity
Sensitivity: prop. Of recidivists previously assigned as high risk (correct pos)
Specificity: prop. Of non-recidivists previously assigned as low risk (correct neg)
Area Under the Curve (AUC): trade-off between sensitivity & specificity
o Ranges from 0-1, 1 being perfect discrimination (perfect prediction)
o SPSS: risk score & dichotomous variable (reoffence yes/no)
o Cutoff scores may differ, we use:
0.56-0.64 small pv
0.64-0.71 moderate pv
0.71-1.00 strong pv
Negative Predictive Value (NPV): prop. Of offenders assessed low risk, who later did not
reoffend
Positive Predictive Value (PPV): prop. Of offenders assessed high risk, who later reoffended
Number Safely Discharged (NSD): number of low-risk patients who can be discharged before
a recidivism occurs
Number Needed to Detain (NND): number of high-risk patients who need to be locked up in
order to avoid recidivism
Global accuracy High risk accuracy Low risk accuracy
Retrospective AUC Sensitivity Specificity
(insensitive to base
rate)
Prospective PPV NPV
(sensitive to base rate) NND NSD
RA tool with sensitivity = 67% and specificity = 53%
o Within the group of recidivists, 67% were correctly classified as high risk
o Within the group of non-recidivists, 53% were correctly classified as low risk
RA tool with PPV = 20% and NPV = 90%
o Among those classified as high risk, 20% will reoffend
o Among those classified as low risk, 90% will not reoffend
The first tool is better at classifying high risk patients, the second is better at classifying low
risk patients. Altering the threshold will positively affect one of the abilities, and negatively
affect the other. The AUC is used to determine this tradeoff.
First generation RA tools
First gen used unstructured professional judgement: no use of tools
Advantage: no costs, flexible & easy, no training required
Disadvantage: lack of consistency & transparency, sensitive to prejudice / bias /
countertransference, lack of scoring integrity, poor reliability & validity, not accurate, no
training required
Second generation RA tools
Second gen used actuarial instruments (“statistical”): evidence based, mostly static items
Advantage: evidence based, less open to interpretation, transparent, objective, replicable,
reliable, good predictive validity, quick & easy to execute
, Disadvantage: atheoretical, invariable, limited identification of treatment targets & limited
integration of intervention, only risk factors (no protective), no clinical override, statements
at group level
Instrument Intended for
General VRAG(R) Violence
StatRec General recidivism
Specific group J-SOAP D 12-18y. sex offenders
Static(R) Adult sex offenders
STABLE Adult sex offenders
Honorable mention: PCL-R (Psychopathy CheckList Revised): not a RA tool, but has good predictive
validity for reoffending
Third generation RA tools
Third gen use structured clinical judgement (SCJ/SPJ): combination of actuarial & clinical method.
Advantage: more transparent than unstructured CJ, sensitive to changes over time, good
reliability & predictive validity, identification of treatment goals, professional is always in
charge, context dependent, multi-dimensional, risk formulation
Disadvantage: more open to interpretation, repeated measure to examine change, limited
integration of intervention.
Instrument Intended for
General HCR-20 (V2) Violence
HKT-30 General/violence
SAPROF Violence, protective
Youth EARL-20B Boys 6-12y, violence
EARL-21G Girls 6-12y, violence
SAVRY 12-18y, violence
SAPROF-YV 12-23y, protective
Women FAM Women, violence
OID (intellectual disability) DROS Diverse
Outpatients FARE +18y, outpatient
RAF GGZ 12-18y, outpatient
Specific type of offense SVR-20 Sexual violence
SARA Partner violence
B-SAFER Partner violence
CARE-NL Child abuse
Fourth generation RA tools
Fourth gen are second or third gen tools + case management: includes case planning & intervention
Advantage: same as 2nd/3rd gen
Disadvantage: more time-consuming, repeated measurement to examine change, training
Actuarial 4th gen tools:
Instrument Full name
LS/CMI Level of Service / Case Management Inventory
MnSTARR2.0 Minnesota Screening Tool Assessing Recidivism Risk
RISc3 Recidive Inschattings Schalen
VRS Violence Risk Scale
VRS-YV Violence Risk Scale Youth Version
Risk assessment = probability calculation that a harmful behaviour or event will occur, and involves
an assessment about the frequency of a behaviour or event, its likely impact and who it will affect =
attempt to predict the likelihood of future offending in order to identify individuals in need of
intervention (risk management)
Reasons
o Safety (self / fellow residents / society)
o Public interest (prevention of (severe) recidivism)
o Ethical importance (same method for every patient)
o Therapeutic importance (RNR model) (adequate risk assessment leads to
Insight into risk factors
Risk management plan
o Communication (more transparency & uniformity in decision-making, improving
communication about risk & risk management)
History
1876, Lombroso: criminals have certain facial characteristics (hawklike nose, large protruding jaws, ..)
1928, Burgess: first RA tool
Developed to determine risk of re-offence of offenders being released from Illinois prison
Actuarial; marital status, criminal & employment history, institutional misconduct
76% of high risk status recidivated within 5 years
1966, Baxtrom v Herold: 967 ‘dangerous’ psych. Patients transferred to regular psych. hospitals, 121
subsequently released into society.
Followed up by Stadman & Cocozza: after 4 years, 2.7% were sent back to forensic psych.
Hospitals. Of the patients in society 17% got arrested and 7% was convicted
1970s-1980s: unsatisfactory state of art regarding predictions of violence. Ability of professionals to
reliably predict future violence = unproven
Teachers & psychiatrists got asked questions on recidivism based on given information
o IRR of teachers was low, but better than IRR of psychiatrists
o On average large similarity between both groups
o More information more conservative judgement
1990s, Phillipse: Dutch TBS system enjoyed great respect; praised for innovative approach to forensic
patients. Researched how Dutch behavioral experts from forensic fields made RA in practice
None of the predictors were predictive (denial factor, empathy factor, …)
No link between clinical assessment of relapse risk & actual recidivism
Theoretical & empirical framework
Risk, Need & Responsivity (RNR) model by Andrews & Bonta (1990): evidence based practice,
interventions based on results from research into their effectiveness
Risk principle: who to treat?
o Risk of recidivism is central, not the disorder
o Misalignment of treatment intensity may increase recidivism risk
High risk offenders benefit most from more intensive treatment
When same treatment applied to high & low risk groups, treatment effect is
greater in the high risk group
, Need principle: what to treat?
o Each indiv has own combination of factors that led to the crime (criminogenic needs)
Only factors associated with recidivism reduction should be addressed
o Dynamic risk factors are modifiable by treatment interventions
Focus on high dynamic risks, adequate treatment of criminogenic needs is
associated with higher recidivism reduction
Responsivity principle: interventions tailored to offenders’ characteristics (e.g. motivation
level, learning style, personal circumstances
Category Central risk Risk Dynamic need
factor
Big 4 History of early & ongoing involvement work on non-criminal alternative
antisocial in a number and variety of behaviours in high-risk situations
behaviour antisocial acts in a variety of
settings
Antisocial thrill seeking, low self-control, work on problem-solving skills, self-
personality restlessness, aggression management skills, anger
pattern management, coping skills.
Antisocial attitudes, values, beliefs and reduce antisocial cognitions,
cognitions rationalizations that support recognize risky ways of thinking
criminal behaviour; cognitive- and feeling, develop alternative &
emotional conditions of anger, less risky ways of thinking and
resentment and pride: feeling; adopt a renewed and/or
criminal vs anti-criminal anti-criminal identity
identity
Antisocial close contact with criminal reduce contacts with criminal
peers others, relative social isolation others, build contacts with non-
with non-criminal others; criminal others
direct social support for crime
Mod 4 Family/marital education and/or care, control reduce conflict, build positive
relationships and/or supervision relationships, improve control &
oversight
School and/or low levels of achievement and improve performance, rewards,
job satisfaction and satisfaction
Prosocial low levels of engagement and improve engagement, rewards and
recreational satisfaction in non-criminal satisfaction
activities leisure activities
Substance Abuse of alcohol and/or drugs Reduce drug abuse, reduce
abuse personal and interpersonal support
for addictive behaviour, develop
alternatives to drug abuse.
, LECTURE 2 – RISK & PROTECTIVE FACTORS
RA tools & predictive validity
Sensitivity: prop. Of recidivists previously assigned as high risk (correct pos)
Specificity: prop. Of non-recidivists previously assigned as low risk (correct neg)
Area Under the Curve (AUC): trade-off between sensitivity & specificity
o Ranges from 0-1, 1 being perfect discrimination (perfect prediction)
o SPSS: risk score & dichotomous variable (reoffence yes/no)
o Cutoff scores may differ, we use:
0.56-0.64 small pv
0.64-0.71 moderate pv
0.71-1.00 strong pv
Negative Predictive Value (NPV): prop. Of offenders assessed low risk, who later did not
reoffend
Positive Predictive Value (PPV): prop. Of offenders assessed high risk, who later reoffended
Number Safely Discharged (NSD): number of low-risk patients who can be discharged before
a recidivism occurs
Number Needed to Detain (NND): number of high-risk patients who need to be locked up in
order to avoid recidivism
Global accuracy High risk accuracy Low risk accuracy
Retrospective AUC Sensitivity Specificity
(insensitive to base
rate)
Prospective PPV NPV
(sensitive to base rate) NND NSD
RA tool with sensitivity = 67% and specificity = 53%
o Within the group of recidivists, 67% were correctly classified as high risk
o Within the group of non-recidivists, 53% were correctly classified as low risk
RA tool with PPV = 20% and NPV = 90%
o Among those classified as high risk, 20% will reoffend
o Among those classified as low risk, 90% will not reoffend
The first tool is better at classifying high risk patients, the second is better at classifying low
risk patients. Altering the threshold will positively affect one of the abilities, and negatively
affect the other. The AUC is used to determine this tradeoff.
First generation RA tools
First gen used unstructured professional judgement: no use of tools
Advantage: no costs, flexible & easy, no training required
Disadvantage: lack of consistency & transparency, sensitive to prejudice / bias /
countertransference, lack of scoring integrity, poor reliability & validity, not accurate, no
training required
Second generation RA tools
Second gen used actuarial instruments (“statistical”): evidence based, mostly static items
Advantage: evidence based, less open to interpretation, transparent, objective, replicable,
reliable, good predictive validity, quick & easy to execute
, Disadvantage: atheoretical, invariable, limited identification of treatment targets & limited
integration of intervention, only risk factors (no protective), no clinical override, statements
at group level
Instrument Intended for
General VRAG(R) Violence
StatRec General recidivism
Specific group J-SOAP D 12-18y. sex offenders
Static(R) Adult sex offenders
STABLE Adult sex offenders
Honorable mention: PCL-R (Psychopathy CheckList Revised): not a RA tool, but has good predictive
validity for reoffending
Third generation RA tools
Third gen use structured clinical judgement (SCJ/SPJ): combination of actuarial & clinical method.
Advantage: more transparent than unstructured CJ, sensitive to changes over time, good
reliability & predictive validity, identification of treatment goals, professional is always in
charge, context dependent, multi-dimensional, risk formulation
Disadvantage: more open to interpretation, repeated measure to examine change, limited
integration of intervention.
Instrument Intended for
General HCR-20 (V2) Violence
HKT-30 General/violence
SAPROF Violence, protective
Youth EARL-20B Boys 6-12y, violence
EARL-21G Girls 6-12y, violence
SAVRY 12-18y, violence
SAPROF-YV 12-23y, protective
Women FAM Women, violence
OID (intellectual disability) DROS Diverse
Outpatients FARE +18y, outpatient
RAF GGZ 12-18y, outpatient
Specific type of offense SVR-20 Sexual violence
SARA Partner violence
B-SAFER Partner violence
CARE-NL Child abuse
Fourth generation RA tools
Fourth gen are second or third gen tools + case management: includes case planning & intervention
Advantage: same as 2nd/3rd gen
Disadvantage: more time-consuming, repeated measurement to examine change, training
Actuarial 4th gen tools:
Instrument Full name
LS/CMI Level of Service / Case Management Inventory
MnSTARR2.0 Minnesota Screening Tool Assessing Recidivism Risk
RISc3 Recidive Inschattings Schalen
VRS Violence Risk Scale
VRS-YV Violence Risk Scale Youth Version
