(2) Risk Assessment Lectures Flashcards
What is the Importance of Risk Assessment?
The most important aspect of forensic risk
assessment is its impact on public safety
by managing offenders based on their
level of risk to the community and
reoffending.
Predicting violence is a complex and
controversial issues in behavioural
science in law because there is always
room for error and risk assessments play a
huge role in making important decisions
at all stages of the criminal justice system.
For example, forensic risk assessments
inform:
• Sentencing,
• Classification,
• Treatment targets and intensity,
• Parole decisions,
• Level of supervision,
• Notification decisions,
• Release conditions.
What is the Purpose of Risk Assessment?
It has dual purposes; both are needed and rely on one another-
What is a risk estimate based off of? What formats do they come in?
- Prediction Classification:
• An analysis of the likelihood of future
criminal and violent acts (i.e., estimate of
risk that they will commit another offense
in the future).
• The risk estimate is based on the number
of risk factors present and their severity.
• Risk estimates are communicated
numerically (i.e., decimal point or
percentage) or as a risk band/levels of risk
(low, medium or high risk), they can be
offense-specific or general (i.e., sexual
recidivism, violence or general recidivism)
within a specified time frame (i.e., fixed 8-
year period post-release). - Informing Risk Management through
Interventions:
• The development of strategies to manage
risk levels primarily through supervision,
restrictions, and providing access to
intervention.
• It requires knowing the contextual and
specific factors is needed to understand
the relationship between risk factors and
criminal behaviour.
• This is where theory is important because
it can explain “how” these factors lead to
offending rather than just telling us what
factors predict recidivism.
History of Risk Assessment:
(A) 1960s-1970s:
• Unstructured Clinical Judgement was used
during this time period and is associated
with the Baxstorm and Dixon studies which
showed that the base rate of violence is
relatively low in released psychiatric
patients and ranged between 7-15% and
false positives are high 85-86% (i.e.,
classifying someone as high risk of violent
but they’re not violent) which meant that
people commonly had their rights and
freedoms taken away from them when
they were not actually violent.
• This method is entirely subjective and has
accuracy of barely above chance levels
and was term with the phrase “like flipping
coins in the courtroom”.
(B) 1980’s:
• People’s pessimism towards the accuracy
of unstructured clinical judgments “not
always wrong, but most of the time” lead
them to turn to Andrew and Bonta’s “what
works” research to identify the (8) central
risk factors to structure risk assessments
and remove clinicians’ biases from the
decision.
• This is the Actuarial Risk assessment
method which is highly structured and
based on empirically supported risk
factors.
(C) 1990’s:
• Structured Professional Judgements were
introduced in the 1990’s and is
characteristic of a shift in viewing risk
factors as a range (i.e., low, medium and
high risk) rather than dichotomous (i.e.,
dangerous and not dangerous) probability
estimates.
*overtime risk assessments have become more structured and grounded in empirically
supported risk factors which has increased its accuracy in predicting risk of recidivism.
Measuring Risk:
• There are multiple assessment tools
available each designed for their own
specific purpose and populations (i.e., no
one scale can do everything or work well for
everyone).
• Risk assessment tools use statistical
techniques that identify risk factors and
combine them. This can be done by
weighting some factors more than others if
they’re more closely tied to recidivism or
reverse coding them (i.e., factors which are
negatively associated with recidivism, so
they need to be reverse coded).’
• Risk assessment tools require training to be
effectively administered and provide an
accurate risk estimate.
• There are different types of risk factors to
consider such as static/historical and
dynamic/criminogenic needs).
*scales should only be used in the groups they’re designed in. We should not assume that they work for all subgroups of offender, cultures, times or locations.
There are (4) subgroups of risk factors:
o Dispositional:
Demographic variables (i.e., younger/age
and gender/male)
Personality variables (i.e., psychopathy and
impulsivity)
o Historical:
Past Behaviour (i.e., violent and non-
violent)
Age of onset (i.e., younger than 14)
Childhood history of maltreatment (i.e.,
physical abuse and neglect but not sexual
abuse)
o Contextual:
Lack of social support.
Access to victims or weapons.
o Clinical
Substance abuse (i.e., polysubstance use
and high-risk of violence)
Mental illness (i.e., controversial,
relationship remains unclear and most
people who have a mental illness are not
violent).
The (4) Steps of Developing a Risk Assessment Measure:
- Identifying relevant risk factors from
existing theory to meet the needs of the
scales intended purpose (i.e., general or
specific etc.) - Determine the scoring system of the
scale (i.e., numerical score, yes or no
format, or rating scales). - Determine who risk factors will be
combined (i.e., each risk factor is an item,
and we need to decide how the items will
be collectively interpreted). Do some of
them be weighted or reverse coded? - Provide an overall assessment of risk (i.e.,
low medium, high or numerical risk
estimate total score).
Static Risk Factors:
What are they? Includes \_\_ and \_\_\_ They're easy to.... More --- than dynamic Risk Factors Good for.... Example scales (3)
Are fixed and not amenable to change
even with intervention (i.e., demographic
and historical risk factors).
Since they’re historical variables or
demographic variables they’re relatively
easy to score because the information is
kept on official records and can be
assessed without interviewing the
offender.
The most frequently used type of risk
factors for triaging offenders into
treatment based on risk.
They can be reliably measured and are
very predictive because the best predictor
of future behaviour is past behaviour (i.e.,
violent and non-violent).
Example Static Risk Factors are:
Young (age), Male (gender), age at first
offence (criminal history), victim type.
Example risk assessment tools which use
static risk factors:
Risk of Conviction and Risk of
Incarceration (ROCROI) is a statistical
software which automatically takes file
information to generate a risk estimate to
accurately and effectively triage offenders
into the appropriate risk level treatment
(i.e., general recidivism a year after
release).
ASRS Algorithms is the “automated sexual
recidivism scale” is the sexual recidivism
specific scale of the ROCROI.
The static-99 is another sexual offender
specific risk scale (i.e., only uses static risk
factors that have been empirically
supported and placed offender into a risk
band).
Dynamic Risk Factors:
What are they…
They add ____ over static Risk Factors
Useful for…..
May include (4)
They change over time through
maturation, intervention, changes in
environment and dispositional factors (i.e.,
treated or untreated).
For example, impulsivity, attitudes,
antisocial peers and substance use.
Less convenient to assess because it
requires a clinical assessment with the
offender and are thereby less reliable as
well (i.e., observation, self-report and
clinical judgement) because it is
subjective- clinicians will vary on their
judgments of whether dynamic risk factors
are present.
They are more useful because they’re
sensitive to change and can be used as
target interventions (i.e., unlike static risk
factors which are not useful for this
purpose).
There is evidence that adding DRF’s onto
Static RF’s increases its incremental
predictive validity.
(2) Types of Dynamic Risk Factors:
what is a propensity model?
- Stable:
• Stable dynamic risk factors are persistent
and change slowly if at all (i.e., traits and
lifestyle choices; substance abuse and
employment).
• Over months or years and are resistant to
change. - Acute:
• Acute dynamic risk factors fluctuate rapidly
and pose an imminent risk of violence (i.e.,
emotional states or contextual triggers; job
loss, intoxication).
• Occurs over minutes, hours or days.
• Most useful for probation officers to
identify changes in risk which poses an
immediate threat and requires prompt
intervention.
*A Propensities Model of risk which states that individuals have stable dynamic risk factors which are enduring and are triggered by environmental triggers or contextual factors which lead to an increase of risk and reoffending behaviour.
The Central (8) Risk Factors for General Offending
Andrew & Bonta (2010)
- Criminal History (static)
- Pro-criminal Personality (impulsivity,
aggressive, sensation seeking) - Pro-criminal Attitudes
- Pro-criminal Associates
- Family/Marital Problems
- School/Work Problems
- Use of Leisure Time
- Substance Abuse
*each of the (8) factors have their own sub-scale of items. These target risk factors for
general offending but other offense-specific risk factors can be added onto the scale.
(4) NOT EMPIRICALLY SUPPORTED RISK FACTORS:
• Low SES
• Personal distress or mental illness (i.e., low
self-esteem, depression, psychosis etc.)
• Lack of remorse or empathy
• Low verbal intelligence
*these are not risk factors that predict offending behaviour at the population level but may be important at an individual level.
Offense Specific Risk Factors:
(5) Sexual Offending:
(3) Domestic Violence:
(5) Terrorism:
(A) Sexual Offending:
a. Sexual preoccupation
b. Deviant sexual interests (i.e., paraphilia)
c. Emotional identification with children
d. Stranger victims
e. Male victims
(B) Domestic Violence:
a. Jealousy
b. Number of stepchildren
c. Hostility towards women
(C) Terrorism:
a. Social wellbeing
b. Economic indicators
c. Governance
d. Law enforcement
e. Armed conflict
Protective Factors:
Whatever are they…
Are they integrated into risk assessment scales…
(4) examples
Factors which predict reduced risk of
recidivism.
Not in most risk assessment scales but
can be added as an additional sub-scale
(e.g., SAProF).
For example, self-control, social support,
medication intelligence.
*protective factors should always be considered. Strength-based approach is better than a risk avoidance deficit approach.
(4) Possible Outcomes of Risk Assessment:
Link between false positive and flase negative?
- True Positive:
• Correct predictions that offender will be
violent, and they are. - True Negative:
• Correct prediction that an offender will not
be violent, and they aren’t. - False Negative:
• Inaccurate prediction that someone will
not be violent, and they are.
• Harm to victim. - False Positive:
• Inaccurate prediction that someone will be
violent, and they aren’t.
• Loss of rights and liberty to the offender.
*False positive and false negatives are linked; by aiming to reduce making false positives we increase the risk of making false negatives. The aim of the game is to manage risk to offender and risk to the community.
What is predictive accuracy:
testing predictive accuracy of a scale:
Is a probability estimate in numerical form
as a decimal, percentage or level of risk
(i.e., 0.50, 70% or levels of risk to
communicate risk to CJS).
Risk assessment is imperfect; it is a
prediction so there is always room for
error (i.e., false positives and false
negatives) are inevitable thus we must do
our best to balance these risks.
Statistical methods can be used to test us
how well a specific risk assessment tool
can predict recidivism by comparing it to
actual recidivism outcomes.
The predictive accuracy of a scale is
communicated with an effect size;
AUC/ROC curve or the Cohen’s D.
One way is to compare a group assessed
on a tool where higher scores indicate
higher likelihood of recidivism.
See who reoffends and who does not,
compare these groups on their scores
within a follow-up period.
AUC and Cohen’s d tell us whether the
tool has predicted correctly based on the
proportion of true positives or mean
scores across groups.
*AUC and Cohen’s d values are not associated with the scores on the instrument! It only tells us the relationship between higher scores (risk) and recidivism (outcomes).
AUC/ROC:
Area Under the Curve (AUC) or Receiver Operating Characteristic (ROC)
It assesses the proportion of true positives
Accuracy values range between 0-1
o 0.5 is chance levels (i.e., scale predicts
recidivism at chance levels)
o 1 is perfect accuracy (i.e., 100% true
positives)
o Less than 0.5 means that the scale is
predicting recidivism in the wrong
direction (i.e., predicts recidivism at less
than chance levels, lower scores predict
recidivism better than higher scores on
the scale).
Interpretation of Numerical Value:
o The probability that a randomly selected
recidivist will have a higher risk score than
a randomly selected non-recidivist when
using a risk-assessment tool.
o Effect Sizes:
0.56 (small effect size)
0.64 (moderate effect size)
0.71 (large effect size)
o The scale’s ability to accurately to predict
recidivism is small, moderate or large.
Cohen’s d:
Compares recidivists to non-recidivists on
their mean score (i.e., on the risk
assessment tool), relative to the standard
deviation.
0 = no difference between recidivist and
non-recidivists mean scores of risk
more than 0; is the desired result. Where
recidivists have a higher mean score than
non-recidivist groups on risk.
less than 0; is an undesirable result.
Where non-recidivist score higher on risk
than recidivist groups. Scale predicts in
wrong direction!
Interpretation:
o 0.20 = small effect size
o 0.50 = moderate effect size
o 0.80 = large effect size
The minimum cohen’s d effect size is 0.15
for risk factors to be psychologically
meaningful. This is equivalent to an AUC of
0.54 (small effect size).
If a risk assessment tool predicts better than chance levels and in the correct direction:
AUC score will be ___
ROC score will be ___
If it predicts worse:
AUC score will be ___
ROC score will be ___
If a risk assessment tool predicts better than chance levels and in the correct direction: o AUC will be above 0.5. o Cohen’s d will be above 0.
If it predicts worse:
o AUC will be below 0.5, more false
positives than true positives than true
positives if you randomly select a
recidivist and non-recidivist the recidivist
is more likely to have a lower score.
o Cohen’s d will be below 0 (the average
score for recidivists is lower than non-
recidivists).
(3) Methods of Risk Assessment:
- Unstructured Clinical Judgement:
a. Practitioners subjectively select, analyze,
and interpret factors deemed by them to
be relevant to risk on any given case.
b. No formal guidance or rules to structure
judgements.
c. Completely at the discretion of the
practitioner.
d. Often guided by past experience,
heuristics, and traditional models of
psychopathology (i.e., personality, mental
disorders and personal distress as
perceived risk factors even though they’re
not empirically supported at group level).
Example: Dr Death:
o Gave unstructured clinical assessments to
advise sentencing for death penalty.
o He sometimes used interviews, otherwise
looked at official data on their family
background, remorse, diagnosis, IQ, and
responses to hypothetical situations.
o He was very over-confident and claimed
his judgments were 100% accurate.
- Actuarial Risk Scales:
a. Empirically supported risk factors, mainly
static, are used to develop risk
assessment scales to provide structure to
risk assessment and reduce the
subjectivity of clinical judgments and
decision-making errors.
b. They use statistical methods to make an
equation, formula, graph, or actuarial table
to arrive at a probability of risk of
recidivism.
c. Usually place offenders into risk bands
(low, medium, high).
d. E.g., static-99, Violence Risk Appraisal
Guide (VRAG) and ROC*ROI. - Structured Professional Judgments:
a. Developed in mid 1990s in response to
problems with unstructured and actuarial
methods of risk assessment.
b. Typically uses 20-30 predetermine
dynamic risk factors that are guided by
theory, research and clinical experience
which increases the scope of the scale to
be applicable to more groups.
c. The items are selected in a logical and
rational manner.
d. Similar narrative approach to
communicating risk and uses case
specific details to express the presence
and severity (i.e., may or may not be rated
numerically with no predetermined cut
offs of risk).
e. Narrative means to communicate risk with
(i.e., high, medium and low) based on a
combination, relevance, severity, and
number of risk factors present.
f. Geared towards informing risk
management plans (i.e., treatment and
formulation and scenario planning).
(3) Advantages and (3) Disadvantages of Unstructured Clinical Judgement:
Advantages
Is a convenient and cheap risk
assessment tool because it requires no
scale, training for a judgement to be
made.
It is flexible and done at the practitioner’s
discretion.
Idiographic; the risk assessment is very
individualized rather than based on group
level supported risk factors.
Disadvantages
Highly influenced by judgement biases
(i.e., illusionary correlations between
mental health and violence etc., salient
aspects such as bizarre delusions are
focused on).
Low Inter-rater reliability (i.e., clinicians’
judgements are subjective and
unstructured so the agreement between
independent raters is poor).
Lower accuracy than other methods (i.e.,
barely above chance accuracy).
(4) Advantages and (7) Disadvantages of Actuarial Risk Assessment:
Advantages
Its objective (i.e., removes subjectivity in
clinical risk assessment by structuring the
RF’s assessed).
It produces a probabilistic estimate of risk
(i.e., numerical value rather than narrative
to communicate risk).
More consistent and valid at population
level (i.e., looking at empirically validated
RFs at the population level rather than
individual level is more accurate).
Speed: its quick to assess risk because it’s
done automatically in some cases, with
official records and doesn’t require an
interview with the offender.
Disadvantages
Nomothetic: “broken leg” dilemma- the
scale identifies the risk factors present but
doesn’t take into account their protective
factors or barriers they face that inhibit
their likelihood to commit a crime (i.e., can
be labeled high risk when they have a
broken leg that prohibits their ability to
offend).
Overlooks contextual factors that
precipitate crime; rigid structure and
doesn’t allow for clinical judgment to focus
on factors most relevant to the specific
individual being assessed.
Room for bias/judgement in applying
coding rules (i.e., there is subjectivity in
the clinician’s judgement of how that code
the risk severity on each factor based on
the quality of the official records available
to them).
Numerical estimates vary across samples:
the scales predictive accuracy is not the
same in all populations.
Weighting presumes that the relationship
between RFs and recidivism is the same
across all samples and individuals.
Often do not include dynamic risk factors
so there is limited utility for guiding
intervention.
Important risk factors omitted if not
predictive in calibration sample (i.e.,
sample the scale is developed in).
Example: Static-99R Scale
(4) Advantages and (3) Disadvantages of Structured Professional Judgments:
Advantages
Can adapt faster to new research.
Flexible; allows for more professional
discretion.
No “broken leg” dilemma; responsive to
individual differences in the relevance or
function of risk factors.
Reduced sample dependency.
Disadvantages
Re-introduces judgement biases.
No probabilistic estimate of risk.
Requires an understanding of the function
of risk factors (i.e., underlying theory).
Example: HCR-20 o Historical and Clinical Risk-20 Developed to assess risk of violence. Consists of 20-items to guide formulation of risk assessment. Rated 0-2 (with a max score of 40) Can include any other case-specific information that the professional feels is important.
____ risk assessment methods outperform both ___ and ___ judgments.
Actuarial risk assessment methods outperform both unstructured and structured professional judgments.
What is the base rate problem?
Predicting recidivism of an offense type with a low base rate increases the risk of making false-positive risk judgments and leads to offenders being kept being detained because they’re perceived to be high risk but actually don’t re-offend.
With very low base rates it would be more accurate to predict when someone won’t offender than does.
The more frequent the behavior the easier it is to predict its future frequency.
How do we Report the Results?
There are two main methods used to communicate the results of a risk assessment, which by itself, is not useful for decision making in the CJS:
What are they?
and provide an example statement with joe who scored (7) on the Static-99R-
What are two considerations to keep in mind when communicating a risk estimate to the parole board?
1. Relative Risk: o Where the offender’s level of risk relative to other offenders. o E.g., Joes’ score of 7 on the Static-99R places him in the 97th percentile which is within the high-risk band for sexual recidivism.
2. Absolute Risk: o Where the offender’s level of risk is compared to other offenders who have the same risk score as them. o E.g., 27% of offenders who also score a 7 on the Static- 99R and share similar characteristics to Joe tend to sexually re-offend in 5 years.
Things to keep in mind:
o We need to be careful that
we use simple terminology
clear of jargon to ensure
that we are effectively
communicating the
offender’s risk to individuals
in the CJS who are making
parole and sentencing
decisions.
o Research is currently being
conducted on developing
“common risk language”
this is to address the fact
that the score cut offs for
bands of risk are different
for each scale. Thus, it
would be beneficial to have
a universal level of risk
classification system.
What are the (4) Generations of risk assessment?
Which of the (3) risk assessment methods is NOT included?
1. The First Generation: o 1960s-1970s: o Unstructured Professional Judgement method of risk assessment, which is very biased, with no structure to the factors associated with risk used, low interrater reliability and poor predictive accuracy!
2. The Second Generation: o 1970s-1980s: o Actuarial risk assessment method which uses structured risk assessment based on empirically supported static factors to reduce the subjectivity of clinical judgement and increased its predictive accuracy.
3. The Third Generation: o 1980s-1990s: Andrew and Bonta’s Central (8) Risk Factors. o Actuarial risk assessment method which used both static and dynamic risk factors.
4. The Fourth Generation: o 1990s-till now: o Comprehensive risk assessment tools which guide case management: assessment, management, and treatment progress.
Where does SPJ fit? o It is not included in the (4) generations of risk assessment: However, Andrews et al. (2009) argue that SPJ is a variation of the first- generation tools, but evidence supports that SPJ has way more predictive accuracy than UCJ’s.
Do actuarial risk assessments only use static risk factors and SPJ only use dynamic and/or static risk factors?
No, as we can see from the (4) generations of risk assessment. Only the second-generation actuarial scale used only static risk factors and the third generation uses both. In terms of SPJ, they can also use both.
Does it Matter which Risk Assessment Scale you Use?
Do they all do the same task?
Can they all inform treatment design?
Are they appropriate for all populations?
Is there consistency or interrater reliability in scores across scales?
The two primary methods (actuarial and SPJ) which have their own respective strengths, weaknesses and purpose. o E.g., the actuarial scale is more precise gives a risk estimate and has issues with “the broken leg dilemma”. o E.g., if advising the parole board, you may use both actuarial and SPJ. There are scales for general or offense specific risk assessment scales (i.e., terrorism, IPV, Sexual Offending, Violence etc.). There are different scales for different populations (i.e., gender, cultures, age etc.). Not all scales can inform interventions (i.e., actuarial with only predominantly static factors).
*even scales designed for the same purpose provide different answers (i.e., there is a lack of consistency between them; predominantly those with DRF’s in them that increases the subjectivity in ratings by various clinicians).
(4) Criteria for Evaluating Risk Assessment Instruments:
Which two are most important?
1. Predictive Accuracy: does it make more true positives/negatives than false positives/negatives (i.e., error; are they balanced?). 2. Validation: has the scale been validated in different contexts than western settings across time?
- Others:
a. Is it easy to use?
b. Is it easy to learn and train
people on? Is it costly to train
people?
c. Is it quick to rate?
d. Consistent ratings across
practitioners? i.e., is their
strong inter-rater reliability. - Does the scale have
construct validity? Does the
scale measure the risk factors
it claims to?
The most important two are predictive accuracy and validation.
Which risk assessment method is most subject to cultural bias?
o Actuarial tools are subject to cultural biases, more so SPJ (i.e., their more strict/rigorous with less room for adjustments). o It’s an enduring concern about the cultural appropriateness of risk assessment scales for Aboriginals, can have implications for other jurisdictions (including NZ).
Cultural and Ethical Issues is Risk Assessment?
Why is cross-cultural validation important?
Who scores higher? Why is this important?
o Cross-cultural validation: most scales are supported in a white western context and have not been validated in other cultural contexts. o There is inconclusive evidence on the cross- cultural generalizability of scales (i.e., mixed findings; equal predictive accuracy or less accurate for indigenous communities). o Indigenous individuals tend to score higher: which impacts on sentencing and parole conditions (i.e., over- supervision, denied parole, harsher sentences etc.).
- This is important because biased scales that present aboriginals as higher risk in cultural contexts where minorities are overrepresented in the CJS is going to lead to people accepting the adequacy of the scale without rigorously testing its cross-cultural appropriateness.
- Biases in the CJS and scale interact and negatively impact cultural minorities’ quality of life.
(5) Types of Biases in Risk Assessment scales?
1. Item Bias: o Items don’t have the same meaning across cultures, or they do not have the same relationship with offending in all groups (i.e., the items, which measure RFs, are not understood the same across cultures).
2. Construct Bias: o Constructs (i.e., RFs) being measured are not the same across cultures (i.e., the influence of cultural norms means that what is considered to be risk factors differs across cultures and makes some items more or less appropriate.
3. Method Bias: o The administration and assessment conditions the scale is used in can vary across cultures (i.e., language barriers between practitioner and offender, the response styles will vary).
4. Judgement Error: o Biases due to subjective components of assessing risk: heuristics, illusionary correlation, overconfidence and the quality of the information provided (i.e., official records, self-report or observation).
5. Over-Confidence: o Confidence is not associated with accuracy! (i.e., being confidence doesn’t mean the risk judgment is accurate).
How does Culture Impact Risk Assessment?
Three possible problems
Risk assessment is normatively developed on white western people (i.e., Canadian or US samples which is a small percentage of the world population).
It ignores that cultural differences in practice, expectations, beliefs about health and mental illness, what deviancy is, worldviews, social and environmental experiences that influence the predictive accuracy of the scale.
Three possible problems: o Overlooking relevant factors to that population not included on the scale. o Including factors that are not relevant or less relevant to the population. o Variation in the expression and manifestation of risk across populations.
SPJ maybe better able to accommodate to variation (i.e., cultural responsivity) but is also subject to biases (i.e., +/-‘s to clinical discretion).
Actuarial tools are ridged
(i. e., do not allow for cultural
responsivity) .
(4) Solutions to Cross-Cultural Biases in Risk Assessment?
Cross-Cultural Validation (i.e., testing the predictive accuracy of the scale across groups, time and contexts). Ensure that practitioners are trained in cross-cultural differences in understanding risk, its links to recidivism and how it is manifested (i.e., adjust the scale accordingly to be responsive to cultural variation). Reducing Method Bias (i.e., language barriers, understandings of risk, deviancy, well-being, and factors which influence their responsivity to treatment and how the items are explained). Identification of risk factors which are more/less relevant to the population being studied (i.e., use the Maori culture related needs: MaCRNs; cultural identity, cultural tension, whanau, whakapapa and weight the risk factors accordingly or adding/removing them).
(6) Methodological Issues in Validating Instruments Cross-Culturally:
1. Safety Concerns regarding the release of high-risk offenders into the population to test actual recidivism rates which makes it unethical and restricts our samples to low-risk offenders.
2. False Positives are
common but maybe the
result of effective
treatment which reduces
the risk of recidivism.
3. Recidivism is under-
reported (x6) in official
records with self-report
data also under reporting
when recidivists who
haven’t been caught will
not want to admit to a
crime. The use of
dichotomous recidivism
scoring (yes/no) leaves out
important information
about the offense type and
the severity of the offense.
4. Base Rates: there is more
predictive accuracy when
the offense type has a
higher base rate of
recidivism (i.e., more
frequent behaviours are
easier to predict).
5. Follow-up periods needed
to capture information
about recidivism makes
cross-cultural validation
very time consuming.
6. Subjectivity of SPJ: means
there is variation across
practitioners which makes
cross-cultural comparisons
hard to make because the
groups are not equivalent.
(5) Ethical Issues in Risk Assessment:
Not having scales to account for cultural variation has implications on indigenous people (i.e., risk estimates, CJS decisions, responsivity to treatment and rehabilitation etc.).
Gap between research and practice (i.e., practice is not up to date with research; it takes time to update treatment to suit theory, it takes time and money to train and update practice, frequently changing treatment and create instability and reduce effectiveness of treatment).
Balancing false positives with false negatives. For example, the public would be in favour of false positives in keeping people looked up to reduce risk of harm to the community, but this impairs the rights and freedoms of the offender and impacts their QOL.
Risk bands are arbitrary, and
variable across clinicians
and scales.
Focusing on Risk Avoidance is one side of the puzzle, what about protective factors which reduce the risk of recidivism?
Protective Factors:
Theoretical confusion?
Can they be measured?
(5) examples?
Factors which mitigate or predict the
reduction in risk of recidivism.
There is a lot of conceptual confusion
around what protective factors actually
are: are they the opposite of RFs, the
absence of RFs or something else? The
evidence says they’re their own thing.
Difficulty with extended temporal
relationship between variables being
studied (i.e., youths vs. adults, onset vs.
maintenance).
Protective factors can be measured and
are shown to predict the reduced
likelihood of recidivism.
For example, o Prosocial involvement. o Social Support. o Intelligence. o Attachment. o Employment.
Example Protective Factor Scale:
SAPROF
Adults or children?
Static or dynamic Risk Factors?
Used in conjunction with —- scale
3 substances are…
First protective factor tools designed for
adults.
Used for in-patients alongside the HCR-20
(SPJ).
It has 17 items: static and dynamic which
are broken into (3) sub-scales:
o Internal: Intelligence. Secure attachment style. Empathy. Coping. Self-control.
o Motivational: Work. Leisure activities. Financials management. Motivation for treatment. Attitudes towards authority. Lifegoals. Medication.
o External:
Social network.
Intimate relationship.
Professional care.
Living circumstances.
External control.
There have been additional scales
developed to fit other populations: SO,
Gender, Age or Disability.
Includes goals: space for SPJ where they
develop goals with the offender, and
includes actuarial risk scoring for each
item to provide a risk estimate.
Risk Assessment in NZ:
___ and ___mainly do Risk Assessments
Are there many validated scales in nz?
Examples (5) two already talked about and three new scales
Mainly probation officers and
psychologists make risk assessments
(albeit for different reasoning, monitor risk
vs. treatment efficacy).
Risk assessments are used to advise the
day-to-day management of offenders and
advising the parole board and courts.
ROC*RIC is a static RF Actuarial tool that
makes automatic risk estimates of
offender’s official records and is widely
used in NZ.
ASRS is another static actuarial tool to
automatically estimates offender risk but is
specific to sexual offending (i.e., adjusts
for lower base rates).
Dynamic risk assessment for offender re-
entry (DRAOR) is only used by probation
officers to monitor for changes in risk.
Stable and Acute (2002) is a dynamic
actuarial tool for sexual offending (static
and dynamic risk factors).
Violence risk scale (VRS) includes STURP
treatment planning and monitoring
change in risk and provides estimates of
risk for the parole board
Dynamic Risk Assessment for Offender Re-Entry (DRAOR)
Developed in Canada and NZ
Three-point scale ranging from 0-2. Comes with a manual that provides examples on how to score each factor. The scores from the (3) sub-scales are aggregated to find a total score. Typically, monitors offenders quite regularly (i.e., weekly for high risk or once a month for lower risk offenders).
*Acute sub-scale is for monitoring changes in risk that may require immediate intervention regularly by the probation officer.
Stable DRFs: Peer associates Attitudes towards authority Impulse control Problem-solving Sense of entitlement Attachment with others
Acute DRFs: Substance abuse Anger/hostility Opportunity/Access to victims Negative Mood Employment Interpersonal Relationships Living Situation
Protective Factors: Response to advice Prosocial identity High expectations Cost/Benefits Social support Social control
Conclusions:
Risk assessment is a crucial part of the CJS
which has many cultural and ethical
implications that need to be addressed.
Validating the predictive accuracy of scales
is important to minimize error (i.e., false
positives/negatives).
It’s important that when picking a scale,
you choose one that is appropriate for the
population being studied and the task at
hand.
The most common and widely used risk
assessment methods that have predictive
accuracy at above chance levels.
It’s important to have risk assessment tools
which have the capacity to monitor change
in risk and inform case management and
intervention (i.e., the (4) generation of risk
assessment).
We need more knowledge on risk factors
and desistance from crime to provide
balanced scales with more predictive
ability, and better treatments for offenders.
Quiz:
True or False-
- Risk factors are used in both prediction
and intervention: - A false negative occurs when people
with a high risk score reoffend: - Static risk factors can be targeted in
treatment: - Dynamic risk factors can change without
treatment: - A “propensities model” of risk states that
all individuals with acute dynamic risk
factors will reoffend: - An AUC of 0.2 indicates a small effect
size: - A cohen’s d of 0.8 indicates a large effect
size: - Structured Professional Judgment is
more accurate than actuarial assessment: - Actuarial assessment uses only static risk
factors: - Protective factors can also be measured:
Yes, the dual aims of risk assessment is
predicting risk of recidivism which can
help triage people into interventions of
the right intensity, and provide targets for
interventions to manage risk.
Yes, a false negative is an incorrect risk
prediction where an offender is expected
to not reoffend but they do.
No, Static RFs are not amenable to
change even with treatment.
Yes, they can change due to changes in
situational factors or maturation over time.
No, the model shows that their risk of
recidividm has increase and intervention
is needed but it doesn’t garuntee
offending will occur!
No, 0.5 score indicates the scale performs
at chance levels. Thus, 0.2 is below
chance accuracy which sugests it is
prediciting in the wrong direction (i.e., low
risk offenders are more likely to reoffend
than high-risk).
Yes, A score of one means perfect
predictive accuracy so a socre of .8
indcates a large effect size.
No, the actuarial scale has more predictive
accuracy because it largely focuses on
static risk factors, more rigorous and less
room for clinician discretion and therefore
biases relative to SPJ.
No, it can use static and dynamic risk
factors!
Yes, they’re measurable.
Quiz:
Short Answer-
1. What are the four types of factor
discussed in this lecture? Not just risk
factors!
2. An example of each?
3. What does the broken leg dilemma
mean?
4. What is the main benefit of assessing
DRF?
5. Why should we consider PF?
6. Why is predictive accuracy important?
7. What else is important in selecting a
tool?
8. What are the 4 generations of risk
assessment?
Static RFs, Stable DRFs, Acute DRFs and
Protective Factors.
Age of first offence, alcoholism,
intoxication, prosocial involvement.
Rigidness of scales to set list of risk factors
(population level) without room for
adjustment to individual factors that may
mitigate risk of offending (i.e., having a
broken leg).
They can be used as treatment targets,
monitoring changes in risks and compare
before/after treatment risk to test
treatment efficacy.
To provide a balanced picture of risk of
offendng. Protective factors mitigate risk
of offending and we need to assess them
to 1) provide strength-based treatment
thats more effective, increases offenders
responsisivity to treatment and can
increase QOL; 2) to reduce risk of broken
leg dilemma which increases risk of
incarcerating people who are lower risk
than the scale predicts and infringing on
their personal rights and freedoms.
It mesaures the true positive/negatives with
false negative/positives. This is important
becasue the conclusion of the risk
assessment impacts decisions at all stages
of the CJS which has a large reppurcisons
on the offender and society (i.e.,
financial/social costs, harm, QOL).
Important issue for cultural concerns; can
the scale be validated across cultural
samples?
Does it fufil the desired task, does it suit
the sample being studied (i.e., gender,
race, age, disability etc.), is it cost-effective,
is it easy to learn or train, is it quick to rate
and does it have good inter-rater
reliability?
Unstructured Clinical Judgement. Actuarial (static) Actuarial (static/dynamic) Case Management and Treatment Planning
The predictive accuracy of a scale is communicated through ____ or ____
Auc/Roc Curve or Cohen’s D
Does Auc/Roc Curve and the Cohen’s D methods of communicating the predictive accuracy of a scale refer to specific items on the scale?
No. They refer to higher scores on the scale being associated with more recidivism or mean group scores on the scale are higher for reoffender than non-reoffender groups.
