Lecture 3A: Risk Assessment Flashcards
What are distal determinants of crime?
Historical factors
What are proximal determinants of crime?
Immediate and situational factors
How do we identify the factors that are most strongly associated with criminal behaviour?
use of meta-analysis
What is a correlate of crime and how do we discover them?
- A variable that is positively associated (correlated) with criminal behaviour.
- Discovered using cross-sectional research design.
When measured at the same time (number of criminal peers and involvement in crime), there is a strong association (e.g., r > .20)
What are the problems with correlates?
- Cannot infer causality
- Reverse causation, which came first? (criminal conduct or criminal peers).
- Some other factor may be the true causal factor (e.g., criminal thinking).
- BUT…a cost-effective starting point to test a research hypothesis
What is a risk factor (predictor)?
- A variable that precedes an outcome of interest and increases the probability that the outcome will occur (Kraemer et al., 1997)
- Discovered using longitudinal designs.
What are the problems with risk factors?
- Still cannot infer causality.
- Risk factors are typically correlated with other risk factors that also predict the outcome of interest.
- Maternal smoking during pregnancy is a risk factor for childhood conduct problems, BUT…
- But… not very important in criminal risk assessment.
What is a causal risk factor?
- the holy grail
- A risk factor that precedes criminal conduct and when changed/manipulated results in changes in the outcome of interest.
- Discovered using:
- Randomized controlled treatment designs (RCTs)
- Quasi-experimental designs (match controls and experimental groups)
- Propensity score matching
Why do we strive to identify causal factors?
-Explains crime and thus helpful for theory development and validation.
-Leads to effective intervention and (should) discourage
ineffective intervention.
What are the first and second-order correlates of criminal conduct?
- Andrews et al: Central Eight risk/need factors with embedded Big Four (major causal variables) - > 1000 studies; > 1770 Pearson r’s
- Minor, moderate and major risk factors
- Static and dynamic risk factors
- Assessment & treatment must attend to risk/need factors
What are the eight factors in the central eight?
- history of antisocial behaviour
- antisocial personality pattern
- antisocial cognition
- antisocial associates
- family and/or marital
- school and/or work
- leisure/recreation
- Substance use
What does the risk and need model of criminal conduct assist in?
- Provincial and federal corrections utilize risk and need assessments (versus psychopathology)
- Assist in identification of levels of criminal risk, treatment targets, treatment planning
What are some examples of risk assessment scales?
SIR-R1, SARA, VRAG
What are the 4 generations of risk assessment?
1st Generation: Unstructured professional judgment
2nd Generation: Actuarial, empirical, typically static factors
3rd Generation: Typically actuarial. Static AND dynamic factors
4th Generation: Comprehensive guide from intake to case closure. Assessment, management, treatment progress
What are the 3 approaches to risk assessment?
1) unstructured clinical judgement
2) actuarial tools (2nd to 4th generation)
3) structured professional judgement
What is unstructured clinical judgement?
Subjectively select, analyze, and interpret risk factors
What are the advantages and disadvantages of unstructured clinical judgement?
Advantages: Flexible, Idiographic
Disadvantages: Inconsistent, Low accuracy
What are actuarial tools?
Collect pre-specified risk factors and enter them into a statistical model that combines and weights them
What are the advantages/disadvantages of actuarial toold?
Advantages: Consistent, High accuracy, Often has recidivism estimates
Disadvantages: Nomothetic, Validity across different samples
What are some issues regarding actuarial scales?
- Items are statistically chosen and do not greatly inform treatment (mainly static items)
- Most items are equally relevant for general recidivism
- Few items are unique to violence
Where does SPJ fit into the generations?
Andrews, Bonta, & Wormith (2006)
-Variation of first generation
Quinsey, Rice, Harris, & Cormier (2006)
-“Actuarial methods are too good and clinical judgment too poor to risk contaminating the former with the latter”
-BUT, qualitatively and empirically different from unstructured judgment. SPJ likely fits somewhere within the 2nd and 3rd generation.
What are the more recent approaches to risk and need assessment?
- Improved domains (acute dynamic, mental health)
- Risk status (group) versus state (acute vs. stable factors)
- Longer follow-ups and larger samples
- Use of meta-analytic studies to show trends in the research
- Development of purpose-specific measures (SO, DV, gender-specific)
How do we understand risk ratings?
- (Nominal Categories – L, M, H)
- No agreement regarding what these categories mean.
- Base rates vary by type of outcome (general recidivism, violence, sexual, technical violations).
What is Hanson et al., emerging model for risk communication?
- Very low (equivalent to young males in community)
Circumstantial crime, no overt criminogenic needs. - Low (½ the BR of the moderate group)
Transient criminogenic needs, problem solving issues, not criminal lifestyle. - Moderate (average BR on risk scale used)
Multiple needs, requires structure programming (~200 hours). Months to get to Level 2. - Moderate-high (twice the BR)
Chronic with multiple severe needs, requires multiple programs (~300 hours). Years to get to level 3. - Extremely High (~80-85% failure rate on risk scale)
Persistent, active criminal (in and out of prison), requires motivational change prior to other programming.
Why should we care about risk assessment?
- Recent news articles –> media
- Risk assessment informs: Sentencing (especially D.O. hearings), classification, treatment needs, treatment intensity, parole decisions, level of supervision, notification decisions, release conditions…
- risk is at the foundation of everything that occurs in the CJS after you get arrested, it informs multiple decisions
What is risk assessment vs. risk management?
- Risk assessment refers to process: Content to consider (instruments), Linked to question(s) of interest throughout sentence
- Risk prediction refers to using information to distinguish likelihood of failure. Weighting information for accuracy
- Risk management requires using static & dynamic risk information. Who is at risk, when they are at risk, and how/
What are the goals of risk assessment?
- Improve accuracy of decision-making
- Improve transparency
- Improve consistency
- Meet CCRA requirements
- Inform case management
- Limit liability
How do we measure predictive accuracy?
- Receiver operating characteristic (ROC) analysis
- A technique for measuring the accuracy of risk assessments by examining false positives and true positives across decision thresholds
What are the 4 decision thresholds in roc analysis?
- True positive: when you predict an outcome to be true and your prediction was correct
- False positive: when you predict an outcome to be true and you were incorrect
- False negative: when you predict an outcome to not be true but it was true (eg., predict not violent but they were violent)
- True negative: When you predict that an outcome is not true and you are correct
How does a roc graph work?
- For each possible cutoff value (score), plot false positive rate (x axis) as function of true positive rate (y axis)
- Connect the dots and get a curve
- We can then measure the area under that curve to get an overall measure of predictive accuracy
How do we interpret a ROC analysis?
- AUC ranges from 0.50 (indicating chance accuracy) to 1.00 (indicating perfect accuracy)
- In other words, the probability that a randomly selected recidivist will have a higher risk score than a randomly selected non-recidivist
- Also, can be directly converted to a Cohen’s d effect size (no math required! Just look it up in a table!!) r =.33 / AUC=.80
What are the strengths and limitations of the ROC?
-A procedure that allows researchers to establish accuracy scores that are not biased by decision thresholds (i.e., scores on an assessment tool) or impacted by base rate (like the correlation)
what are the levels of predictive accuracy for the different risk assessment methods?
- Clinical judgments: AUC = 0.55
- Actuarial tools: AUC = .68 to .80
- Structured Professional Judgment: AUC = .62 to .75
What is the bottom line about risk assessment?
- The risk scales used by CSC and Provinces have acceptable predictive validity.
- Ignoring them will increase decision errors and liability (without compelling rationale).
- Scales (VRAG, PCL-R, LSI-R, LS/CMI) used by psychologists may be slightly more accurate BUT their application in terms of gender and race is hotly debated.
What is the Ewert Decision?
Supreme court decision, ruled that risk assessment cant be used on indigneous offenders (Sept. 18, 2015)
What do we find when we compare risk assessment generations for violent recividism?
-third generation is slightly better
What do we find when we compare administrative methods for prediction of violent recidivism?
- File (.26)
- interview (.11)
- Self Report (.12)
- File & Interview (.30)
- May not want to rely on interviews