Risk Assesment

Risk Assessment

Examples

• A police officer escorts a young man acting in a threatening manner to hospital emergency

• A 15 year old girl has allegedly attacked and seriously injured a classmate and a crown prosecutor now petitions the court to raise her to adult court

• A person suffering from schizophrenia murdered his parents was found criminally responsible and is now requesting a transfer to a lower security facility

Assessment

• Assessment = an evaluation of the person’s cognitive, emotional, and behavioural functioning

• Goal = to obtain information that provides a better understanding of the individual

  • This information is used to guide decisions, interventions, and management of offenders

• Risk Assessment

  • Considers the information used

• Risk Prediction

  • Uses this information to assess the risk that people will commit a crime in the future

• Risk Management

  • Develop effective intervention strategies to manage that risk

Assessment Measures

• Clinical interviews

• Self-reports

• Rating scales

• Peer ratings

• Direct observations

• Actuarial instruments

• Physical exams

• Lab tests

• Psychological instruments

Why Do Assessments?

• Civil settings

• Fitness to stand trial

• Sentencing decisions

  • Young offenders, dangerous offenders

• Suicidal ideation

• Psychological disorders

• Classification

• Treatment intervention strategies

• Parole decision-making

• Pre-release

Sources of Information

• Structured interview

• Self reports

• Collateral contacts (family, friends)

• Police reports, prior criminal justice reports

• Court transcripts, Judge’s Reasons for Sentencing, Pre-Sentence Reports, Victim Impact Statements

• Prior mental health reports, psychological tests, actuarial measures

What is Being Assessed?

• Criminogenic Risk Factors

  • Factors that are static

• Criminogenic needs

  • Factors that are dynamic and place an offender at greater risk of reoffending

  • Stable vs. acute dynamic factors

• Treatment intervention strategies

  • Identified needs become treatment targets

  • Noncriminogenic needs may also be treated

Assessment Challenges?

• Informed consent (voluntary, informed, and understood)

• Limits of confidentiality (researcher vs. clinician)

• Rapport

• Earning trust

• Evaluation of honesty

• How are these challenges overcome?

  • Collateral sources of information

  • Use standardized actuarial instruments

  • Compare information from these sources with clinical judgements

Civil Setting: Duty to Warn/Protect

• Liable for failing to protect a potential victim

• Canadian Code of Ethics for psychologists requires psychologists to “do everything possible to prevent serious physical harm or death of others…may include reporting to the appropriate authorities or an intended victim”

Approaches

• Unstructured clinical judgement

  • 1st generation

  • Selection and combination of information are not guided by any explicit rules

  • Decision rules unclear

  • Informal, subjective

  • No specific risk factors, low accuracy

  • Problems:

    • Illusory correlations (believe an association occurs between predictor and outcome)

    • Base rates, reliance on salient cues, overconfidence, gender

Dr. “DEATH”

• Grigson in Texas Capital Sentencing Proceedings

• Doctor...do you have an opinion within reasonable psychiatric certainty whether or not there is a probability that the defendant will commit criminal acts of violence in the future?

• Yes, he most certainly would.

• Would you state whether or not that would be true regardless of where he is?

• It wouldn’t matter whether he is in the penitentiary or whether he was free. Wherever he is he will continue commit violence.

• Would you state whether or not, Doctor, you have an opinion within reasonable psychiatric certainty as to the degree of that probability that you have just expressed to this jury?

• Well, yes sir, I would put it at one hundred percent and absolute.

Predictions of Recidivism

• Predictions of future dangerousness has proven to be a difficult task for professionals.

• Meta analyses have found actuarial instruments to be 10-13% more accurate than clinical judgments for general recidivism.

• Average effect size for the prediction of violence around.30 and .46 for sexual offenders

Approaches

• Actuarial

  • 2nd generation

  • Coollect pre-specified risk factors and enter them into a statistical moodel that combines and weights them

    • Formal, objective

    • Empirically derived factors

    • Same factors used for each case

    • Speecific cutoffs for decisions

    • Focus on static measures

  • Actuarial more accurate than clinical

SIR-R

• Type of current offinse

• Age at admissioon

• Previous incarcerations

• Revocation of conditional release

• Escape from custody

• Security classification

• Age first adult conviction

• Previous conviction for assault

• Marital status at admission

• Interval at risk since last conviction

• Number of dependents

• Current length of sentence

• Previous conviction for sex offences

• Previous conviction for break and enter

• Employment status at arrest

VRAG

• Actuarial instrument designed to predict violent recidivism in serious offenders

• Empirically-derived actuarial risk assessment device

• Developed on a single large sample of forensic psychiatric patients (n = 618)

• 12 static items

• scores range between -27 and +35

• scores categorized into 9 risk bins

• higher scores/bins = higher risk

1. PCL-R score (+)

2. Elementary school problems (+)

3. Personality disorder (+)

4. Separated from parents (+)

5. Failure on prior release (+)

6. Alcohol abuse (+)

7. Non violent offense history (+)

8. Never married (+)

9. Schizophrenia (-)

10. Victim injury (-)

11. Female victim (-)

12. Age (-)

Actuarial Disadvantages

• Focus on static factors

• Risk level cannot change

• Provides little information about treatment needs

• Must cross-validate risk factors on different samples

Approaches

• Structured Professional Judgement (SPJ)

  • 3rd generation

  • Specific risk factors

  • Derived from literature

  • Includes static and dynamic

  • Includes case critical factors

  • Selection of items is guided by explicit rules but combination of these items is not

  • Additional items may be considered

  • Rater makes final decision about risk level

• Fewer predictive studies

LSI-R

• Designed to predict general recidivism

• 54 risk and criminogenic needs items (i.e., both static and dynamic)

• 10 subcomponents

• scores categorized into 5 risk/need levels

• Higher scores = higher risk

1. Criminal history

2. Education/Employment

3. Financial

4. Family/Marital

5. Accommodation

6. Leisure/Recreation

7. Companions

8. Alcohol/Drug problem

9. Emotional/Personal

10. Attitudes/Orientation

HCR-20

• Structured professional judgment instrument

• Designed for violence risk assessment in criminal and psychiatric populations

• Items selected on basis of lit review and clinical experience

• Historical items

  • Past violent behavior

  • Young age at first violence

  • Relationship instability

  • Employment problems

  • Substance use problems

  • Major mental illness

  • Psychopathy

  • Early maladjustment

  • Personality disorder

  • Prior supervision failure

• Clinical items

  • Lack of insight › Little insight into mental disorder, treatment needs, triggers

  • Negative attitudes › Procriminal, supportive of violence

  • Active symptoms of major mental illness › Specific threat delusions, sadistic fantasies

  • Impulsivity › Affective instability, behavioral acting out

  • Unresponsiveness to Treatment › Respond poorly to treatment, non-compliant, refuse treatment

• Risk Management Items

  • Plans lack feasibility › No plans or unsuited to individual’s needs

  • Exposure to destablizers › Antisocial peers, victims, substance use

  • Lack of personal support

  • Noncompliance › Refuse to take medication, fail to comply with discharge plans

  • Stress › Ability to cope with stress, association between stress and violence

HCR-20: Risk Ratings

• Low risk - monitor and intervene with low priority and intensity

• Mod risk - monitor and intervene with some priority and intensity

• High risk - monitor and intervene with high priority and intensity

SPJ: Strengths

• Predicts likelihood, monitors change, and suggests intervention and management strategies

• Simple, reliable

• Greater flexibility because case-specific info and interactions can be considered

• Predictive and dynamic validity with variety of samples

SPJ Weaknesses

• “Human” judgment may reduce accuracy

• Requires clinical training

Which Approach Better?

• Actuarial is more accurate than unstructured clinical judgment

• Structured professional judgment appears to be similar to actuarial in accuracy

Predicting Recidivism: Dynamic

• Many risk scales have been created to predict various types of outcome and in general are reliable measures of risk.

• One of the major limiting factors is that many of them do not contain dynamic factors and as such are not able to inform risk management.

• Gendreau et al. (1996), in a meta analysis of prospective studies with a minimum follow up of 6 months, found that dynamic risk factors were equally, if not better, at predicting general recidivism (.12 for static and .15 for dynamic).

• More recent meta-analysis with similar criteria predicting violence found a similar pattern of results (.22 for static and .25 for dynamic; Campbell et al., 2007)

• For the most part, the risk management process is subjective, may vary from one clinician/officer to the next, and may be vulnerable to the same limitations as risk assessments based on unstructured clinical judgments.

• Obstacles for the Investigation of Dynamic Risk

  • Lack of confidence in predictive ability

  • Concerns regarding measurement

  • Challenges in analyzing the data

General Recidivism

• Research on dynamic predictors of violence in its infancy.

• Most research is disjointed.

• Summary papers have been useful at consolidating the research. › (e.g., Loza & Dhaliwal, 2005; Douglas & Skeem, 2005)

Violent Recidivism

Sexual Recidivism

Dynamic Risk: Limitations

• Single point measures

• Pre/post measures

• Large domains/scales

• Findings disjointed

• Long term predictions

• Frequency of reassessment

• Ecological validity

• Weak statistical procedures

• Lack of consideration for protective factors

Protective Factors

• Those characteristics or assets of an individual that buffer risk.

• Literature abundant in youth mental health.

• Very few efforts have been made to extend that literature into understanding adult criminal behavior.

• Among serious group of youth, positive peer relations, good school performance, participation in organized leisure activity, positive response to authority correlated with lower recidivism and better compliance (Hoge et al., 1996).

• Some preliminary research found that protective factors added incrementally to the prediction of general recidivism

• Structured activities and strong family relations potentially important factors in understanding protection from criminal behavior (DeMatteo et al., 2005).

Outcome Statistics

• Correlations

  • Between risk measure and outcome

  • Range from + 1.00 to - 1.00

  • r = .30

• Odds ratio

  • Take scores above and below median

  • OR = 2.50

  • One group is 2.5 times more likely than other group to possess some criterion

• Analysis of variance

  • Time to reoffend, number of offenses

  • Compare 2 or more groups

• Regression

  • Determine the proportion of variance accounted for by the risk measure

  • Input one set of variables and determine if risk measure accounts for any additional variance

• Survival curve analysis

  • Takes into account length of follow-up

  • Evaluate how quickly participants recidivated

Decision Outcomes

• TP = person correctly predicted to be violent

• TN = person correctly predicted not violent

• FP = person predicted to be violent but is not

• FN = person predicted to be nonviolent but is

• Errors have different consequences › FP – individual

Example 1

Baxstrom study (Steadman & Cocozza, 1974) › 1966 – US supreme court › “dangerous” mentally ill patients released into the community › Follow-up 4.5 years › 98 patients followed › Used age and previous criminal history to classify into low and high risk

Decision Outcomes

• Clinical judgments › AUC = 0.55

• Actuarial tools › AUC = 0.80

• Structured clinical guidelines › AUC = 0.75

Methodological Issues

• Definition of violence › Violence is the actual, attempted, or threatened harm to a person or persons › Type, severity, target of violence

• Length of follow-up period › Longer the follow-up = higher rate of violence

• Most studies use a limited number of predictors › Need to use multiple predictors across domains

  • Historical, neurological, situational, psychological

• Most use Static vs. dynamic › static = historical, factors that do not change › dynamic = factors that fluctuate or can change

Written Reports

• Questions to Consider:

  • What is the likelihood that the individual will engage in future violence?

    • risk and protective factors

    • probabilistic statement

    • time period

    • relative to some specific comparison group

  • What is the probable context, victim, severity and frequency of any future violence?

  • What steps need to be taken to manage the individual’s risk?

  • What circumstances might exacerbate the individual’s risk?

Framing the Prediction

• this person is dangerous

• If [the following risk factors are present] then there is a [high, moderate, low] probability that the person will engage in [some specific] behavior within [specific period of time] that may place [specify victims] at risk for [specify type and severity of harm]

Communicating Risk

• Low violence risk

  • Few risk factors present

  • No further assessment/preventive actions

    • E.g., 60-year-old depressed man with no violent history and no threats of violence

• Moderate violence risk

  • Several risk factors present

  • Gather more information/monitor person

    • E.g. 25-year-old woman who is abusing alcohol, with a history of assaults, but without a recent violent act

• High violence risk

  • Numerous risk factors present

  • Priority given to gathering additional information and close monitoring

  • Make preparations for preventive actions should condition deteriorate

    • E.g. 30-year-old woman who is using street drugs, with a history of assaults, making recent vague threats

• Extreme High Violence Risk

  • Numerous risk factors present

  • Enough information to make a decision

  • Take preventative action (e.g. intensive case management, involuntary hospitalization, warn potential victim)

    • E.g. 35-year-old man who is using street drugs, has a history of recent violence, is threatening his spouse, and has recently purchased a gun

Treatment

• EFFECTIVE CORRECTIONAL TREATMENT SHOULD BE BASED ON THE PRINCIPLES OF RISK, NEED, RESPONSIVITY (Andrews & Bonta, 2010)

Risk Principle (”Who”)

• Assess risk

  • Offenders deemed higher risk of reoffending should be the focus of institutional intervention programs

• Match risk level to treatment services level

  • Low risk offenders are unlikely to reoffend and may actually increase in risk when exposed to treatment

Need Principle (”What”)

• Target criminogenic needs primarily rather than non-criminogenic needs to decrease recidivismThese are criminogenic needs known to contribute to reoffending

  • Antisocial attitudes

  • Substance abuse

  • Antisocial peers

• Non criminogenic needs: Self-esteem, anxiety

Responsivity Principle (”How”)

• Correctional intervention should match the learning styles of offenders

• General: Use structured cognitive behavioural interventions

• Specific: Match treatment delivery to offender’s ability and learning style

Treatment

• Treatment programs that adhere to the RNR model (particularly cognitive-behavior based programs) have been demonstrated to effectively reduce recidivism in various settings and with various types of offending