Lecture 6 Flashcards
1
Q
What is the point of risk assessment?
A
- Can remove someone’s freedom on basis of outcome OR discharge people who do bad things = very important
- Must be conducted in a professional, accountable, and evidenced-way
- Defend decision and show transparency in how you arrived at those decisions
2
Q
How do you express risk?
A
- ROC analysis - receiver operator characteristics
- If predicted to happen, and happens = hit
- If predicted to happen, and didn’t = false positive
- Not predicted, happens = miss
- Not predicted, didn’t = correct rejection
- Good prediction has lots of hits and few false positives
- Can be determined by graphing the probability of a hit vs false positives
- In order to get lots of hits, you could say they could be dangerous, you would also have many false positives
- As well as ability to tell one group from another, there is a bias for hits/false positive
- To quantify this, you calculate the area under the curve (AUC)
- AUC = 0.5 is chance performance, 1.0 is perfect performance, 0.56 is weak, 0.71 regarded as strong
3
Q
What are the pros/cons of ROC analysis?
A
- Immune to baseline changes = doesn’t matter if it is rare or common event = keeps continuous nature of assessment scale
- Lose the quality of event e.g how long to reoffend
4
Q
What major factors predict violence?
A
- Mental illness: schiz, mood disorders, PTSD
- Mental disorders: personality disorders, substance misuse, learning disabilities
5
Q
What specific factors predict violence?
A
- Previous violence
- Negative attitudes
- Poor temper control
- Poor relationships
- Unemployment
- Victim of abuse
6
Q
What are the methods of risk assessment?
A
- Clinical judgement
- Actuarial measures
- Structured professional judgement
- Formulation based approach
7
Q
What is unstructured clinical judgement?
A
- Professional makes a decision based on impression, intuition and experience
- Most widely used method as it allows idiographic analysis of offenders behaviour and specific formulation informing treatment, prevention, and management of individual
- Allows biases, overlooking of important factors, not based on empirical evidence of risk
- It does not work
8
Q
What was the evidence that clinical judgement does not work initially
A
- Baxtrom was someone held in a NY state prison beyond length sentence due to him being dangerous
- Appealed and won case = NY state decided that 966 other dangerous patients should also be released
- Levels of violence in these prisoners was slightly more than the average person in the general population
- Very few committed any offence and only 20 were later arrested for any violent crime
9
Q
What was a study looking at why clinical judgements were so bad?
A
- Gave professional information about a patient and asked to make various risk judgements
- Events had already happened
- Asked to say why they made their decisions and cues were unrelated to violence of patient
- Interrater reliability was very poor
- AUC was 0.5
10
Q
Why are they/we so bad at clinical judgement (no study)
A
1) They are blind to their outcomes for the vast majority of cases - small number of those released are violent and they may not hear about it - impression that they were right
2) Tendency to weigh bizarre factors heavily - hallucinations seem important but they might not be = tend to neglect criminogenic factors
3) Too many variables - research in domain of cognitive psychology reveals that we can only keep track of a small number of variables when making decisions
4) We tend to make judgements quickly and then seek support for these
11
Q
What is actuarial assessment?
A
- Factors thought to be predictive of risk are put together using a pre-ordained method = normally based on a construction sample or on the scientific literature
- Avoids individual bias
- Does not need clinical skills to formulate: computer does it
- Fast
- Lacks ideographic information
- Does not easily suggest risk management = does work
12
Q
What is the VRAG? And how did it arise? (form of actuarial assessment)
A
- Violence risk appraisal guide: based on 618 men from Canadian max security psych unit = all committed one serious antisocial act
- Released to community/min security/ half-way house
- Target behaviour was violent incident or return to secure unit
- Violence is broadly defined - common assault, SA, armed robbery or pointing a firearm
- Not robbery or possession of weapon or arson
13
Q
How does the VRAG measure?
A
- Range of potential predictors
- Dropped any items that did not predict the violence
- If two items were very correlated, higher correlation was used
- Regression to see which variables added independently to risk prediction model
- These 12 items are VRAG items = allows for prediction, as it predicted the Canadian Prisoners
14
Q
What are the categories and risk probabilities of VRAG?
A
- Scores range from -26 to +38
- Scores used to put people into 9 categories = and works (looked at %committed crime 7yo and 10yo after)
- Reliability is very high = 0.9
- AUC = 0.76 = predicts violent incidents
15
Q
What is the evidence for VRAG? (Samples)
A
- In similar sample = can have slippage = original analysis can take advantage of random correlations
- Different samples might have different risk factors
16
Q
What was a study looking at the validity of the VRAG?
A
- Followed 421 male patients being discharged from medium security units
- Found records of people who were discharged on dates in the future, and scored VRAG by case-note reviews at time of discharge and blind to outcome, can see in present the outcome
- After first year AUC = 0.86, and as time goes on, performance went down to 0.76
- VRAG has good validity for predicting crime, especially violent crime
- Absolute rates of violence in this group are lower than Canadian Sample
17
Q
Why were the rates in the UK lower than in Canada?
A
- Use of formal reconviction may miss many violent incidents
- More supervision?
- Not able to correct for any further ti,e spent incarcerated
- We are less violent than Canadians
- Less competent detection/prosecution of incidents
18
Q
Why do Clinicians argue that the comparisons are not fair?
A
- Clinicians focus on STM outcomes
- Clinicians have different information available
- Clinicians used to be bad, but now they know so can do better
- Clinicians concentrate on really dangerous - not single act that might occur years ahead
- Some Clinicians are bad
19
Q
What was a study looking at if this was fair?
A
- Clinicians who knew patient well were convened = had access to all information
- Gave patients rating of patients security needs and compared this rating to VRAG
- AUC for Clinicians = 0.59, VRAG = 0.8
- Same when compared for short follow-ups
- Used composite scores - not letting one Clinicians letting the side down
- Only VRAG significantly predicted homicides - clinicians were at chance levels
20
Q
What are other Actuarial?
A
- ORGS - predicts general offending
- RM2000 - predicts sexual offences
- COVR - uses classification tree and easy to get variables
21
Q
Why do people not use actuarial?
A
- Many clinicians feel that such a method ‘de-individuates’ the patient
- Many clinician feel that there status is being eroded – if a formula / computer can make the decision then what is the clinician for?
- Many clinicians feel that risk assessment is NOT their job – they are there to treat the person’s illness.
- Person has to fit with sample – broken leg problem
- Tend to use static measures – not good at showing any change in dangerousness – but does not have to be so!
- Tend to be based on most common violence – though this does not have to be so – however, empirically they seem to do very well with rare violence (murder) as well
- Getting a number is not enough – we really also need to know why risk, when a risk, what can we do about it
22
Q
What is a structured clinical assessment?
A
- Combines clinical and actuarial measures
- Clinicians scores set of preset items on whether present or not, but makes decision or risk in a clinical manner
- Avoids individual bias only to some extent
- Contains ideographic info and good and making evidence-based formulation
- Time consuming and needs clinical skills to interpret
23
Q
What is the HCR20?
A
Using historical, clinical and risk management = looking at 20 items
24
Q
What was a study looking at prediction with the HCR20?
A
- Most violence missed when in community
- Most studies are retrospective - biased by event they are trying to predict
- Often trying to predict behaviour in a different setting to assessment
- Prospective study of violence in local Forensic Mental Health Setting
- Assessed at admission, followed for 3 months and all incidences of violence noted - classified according to type
- In physical aggression = HCR20 had AUC of 0.81
25
Q
What was the recidivism following discharge with HCR20?
A
- UK male mentally disordered offenders discharged from MSUs for at least 2 years
- Traced via home office
- Assessments made at time of release so study is prospective
- HCR20 is a good predictor of future violence: 6mo after = AUC of 0.76, after 2y = 0.71
- Also predicted any offences
- But clinical scale was not very good when people when leaving the unit, but very good when they arrives in the other study
26
Q
What was the HCR20 like for female patients?
A
- HCR20 has some predictive validity in women
- Men AUCs ranged from 0.75-0.88, but women: 0.52-0.63 but the clinical judgement for women was much better
27
Q
What was the HCR20 like for other populations?
A
- Learning Disability: PiC (Partnerships in Care) sample.
- HCR-20 very good (AUC > 0.80) – better than other offenders. (UK sample). C-scale good predictor as their condition does not change
- Ethnic Minorities: Male PiC sample.
- HCR-20 has very similar properties in UK ‘Black’ population as in the ‘White’ sample
28
Q
What was a study looking at the diagnosis using HCR20 and the effect on mental illnesses?
A
- Diagnosis is an indicator of future violence - those with PD are most dangerous
- HCR20 us a good predictor in those with Schiz but worse in those with PDs
- Maybe because they are more impulsive or because the files have been lost due to suicide/fights or different prison
29
Q
What are the limitations of the HCR20?
A
- Need more evidence of its utility in specific populations and violence e.g violence to children or DV
- Need to see if treatment produces changes in HCR20 and if this is indicative of less dangerousness
- Homicide is rare and therefore few studies of its prediction
30
Q
What is the HCR20v3?
A
- Update in 2014
- Small changes to items
- Seven steps:
- Gather info
- Presence of item
- Relevance of item
- Formulation of Risk
- Scenario planning
- Case management plan
- Develop final options
31
Q
What is the other side of the HCR20?
A
- Not all clinicians are happy about the use of any instruments
- Appears to argue that all attempts at risk assessment are useless
- Argues to stop risk-assessment but to focus on treatment like they do in self-harm and suicide
