Risk assessment

Question 1

Why are risk assessments important?

Answer 1

Can change the outcome of someone’s life.

Question 2

What does ROC stand for?

Answer 2

Receiver operating characteristic.

Question 3

What is ROC?

Answer 3

Used to look at the usefulness of a risk assessment tool through a graph and figure.

Question 4

What does the curve of an ROC graph depend on?

Answer 4

Different biases/cut-off scores of an instrument.

Question 5

What is an ROC result expressed as?

Answer 5

Area Under the Curve (AUC).

Question 6

What is regarded as a weak, moderate and strong AUC?

Answer 6

Weak - 0.56
Moderate - 0.65
Strong - 0.71

Question 7

What are the pros of ROC?

Answer 7

Immune to baseline changes.

Keeps continuous nature of assessment scale.

Question 8

What are the cons of ROC?

Answer 8

Lose ‘quality’ of event, e.g., how long to re-offend, how violent, etc.

Question 9

What are the types of risk assessment?

Answer 9

Clinical judgement, actuarial measures, structured professional judgement.

Question 10

What is unstructured clinical judgement?

Answer 10

Professional makes a decision based on impression, intuition and experience.

Question 11

What is the biggest limitation of clinical judgements?

Answer 11

Allows biases - not based on empirical evidence of risk.

Question 12

What did the Baxstrom study (Steadman et al., 1970) find?

Answer 12

When 966 ‘dangerous’ patients were released, only very few committed any ohter offences (20 were later arrested for any violent crime).

Question 13

What did Odeh et al. (2006) find when looking at the reliability of clinical judgements?

Answer 13

Different people made widely different predictions form the same vignette - across all professions. People also had different reasons for why they came to their conclusions. Interrater relaibility was very poor.

Question 14

What are 4 reasons why clinical judgements are so bad?

Answer 14

1. Blind to the outcomes of the cases.
2. Tendency to weigh bizarre or unusual facts rather heavily.
3. Too many variables (we can only keep track of a small number of variables when making a decision).
4. Tend to make judgements quickly and then seek support for these.

Question 15

What are actuarial assessments?

Answer 15

Make a decision off a preordained plan.

Factors thought to be perdictive of risk are put together using a preordained method.

Question 16

What are actuarial assessments based on?

Answer 16

‘Construction sample’ or the scientific literature.

Question 17

What are the pros of actuarial assessment?

Answer 17

Avoids individual bias, doesn’t need clinical risk to formulate, fast.

Question 18

What are the cons of actuarial assessment?

Answer 18

Often lacks ideographic information.
Does not easily suggest risk management.
Person has to fit with the sample.
Uses static measures.
Based on most common violence.
Getting a number is not enough.

Question 19

What are some examples of an actuarial assessment?

Answer 19

Violence Risk Assessment Guide (VRAG), OGRS, RM2000, COVR.

Question 20

What is the VRAG based on?

Answer 20

Based on men in a Canadian maximum secure psychiatric unit who were released to the community, minimum security or a half way house.

Question 21

What was the target behaviour looked for when creating the VRAG?

Answer 21

‘Violent incident’ - charge or conviction or violence or return to secure unit for reasons that would otherwise have led to a charge.

Question 22

How was the VRAG created?

Answer 22

Measured a range of potential predictors and dropped any that didn’t predict the violence. If two items were correlated, the one with the highest correlation was used. Then used a ‘least-squares regression’ to see which of the remaining variables added independently to risk prediction model.

Question 23

How many items are on the VRAG?

Answer 23

12

Question 24

Explain the scores on the VRAG.

Answer 24

Range from -26 to +38 which are split into 9 categories.

0 is the average score for people in maximum security so average population would score lower than 0.

Question 25

What is the reliability, standard error of measurement (SEM) and AUC of the VRAG?

Answer 25

Reliability is 0.9, small SEM, and AUC of 0.76.

Question 26

What has been suggested when looking at the AUC of the VRAG in other populations?

Answer 26

Different samples might have different risk factors, even similar samples might show some slippage.

Question 27

What methodology did Snowden et al. (2007) use when exploring the VRAG?

Answer 27

Followed up 421 male patients being dischraged 2 years later from MSUs in the UK - pseudo prespective study. Scored on case note review at time of discharge and blind to outcome.

Question 28

What did the Snowden et al. (2007) find when exploring the VRAG?

Answer 28

After 6 months, VRAG scored .86, decreased over time but only went down to .76.

Question 29

What was different about the VRAG sample and the Snowden et al. (2007) sample and why?

Answer 29

Snowden sample had lower base rates of violence. Less violent than Canadians, use of formal reconnections may miss all violent incidences, may be more supervision/management, didn't correct for any further time spent incarcerated, less competent detection/prosecution of incidences.

Question 30

Why don't people use actuarial assessments?

Answer 30

Feel it 'de-individualises' the patient, clinicans feel their status is being eroded, clinicans feel risk assessment is not their job.

Question 31

Why have people argued that it isn't fair to compare clinical judgements and actuarial assessments?

Answer 31

Clinicians tend to concentrate of 'short-term' outcome, they also have different information available, they concentrate on the really dangerous (not a single act of violence in years ahead), some clinians are bad.

Question 32

What did Harris et al. (2002) find when comparing VRAG scores and clinical judgements?

Answer 32

AUC for clinicans - 0.59, AUC for VRAG - 0.80. Same scores whe compared to short-term follow ups, usec composite scores - not one clinican letting everyone down, only VRAG significantly predicted homicides.

Question 33

What are structured clinical assessments?

Answer 33

Combined clinical judgements and actuarial assessments.

Question 34

What are some examples of structured clinical assessments?

Answer 34

HCR20, SARA, SAVRY, SVR20

Question 35

What are the limitations of the HCR20?

Answer 35

Need more evidence of its utility in specific populations and specific forms of violence. 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.

Question 36

What are some problems when looking at the effectiveness of the HCR20?

Answer 36

Most violence is missed when committed in the community, most studies are retrospective, and often trying to predict behaviour in different setting to assessment.

Question 37

What did Gray et al. (2003) find when looking at HCR20 scores and future violence in a forensic MH setting?

Answer 37

HCR20 significantly predicted physical aggression as a whole but the individual rating scales also did.

Question 38

What did Gray et al. (2008) find when looking at effectiveness of the HCR20 in predicting recidivism following discharge in UK male MI offenders being discharged from MSUs?

Answer 38

HCR20 was a good predictor of future violence according to the AUC scores. But the clinical scale was poor.

Question 39

Why might the clinical scale of the HCR20 be poor when looking at future violence?

Answer 39

It is about current presentation so it wouldn't stay static.

Question 40

What has been found using the HCR20 with females?

Answer 40

It has some predictive value with women but the AUC was smaller. AUC was better when clinical judgement using the HCR20 was much better.

Question 41

What was found using the HCR20 with people with learning disabilities?

Answer 41

It was a very good predictor of violence (in UK sample) - better than other offenders. Clinical scale was also a good predictor, suggested because their clinical presentation is pretty much static.

Question 42

What has been found in ethnic minorities using the HCR20?

Answer 42

HCR20 has very similar properties in UK black population as in the white sample.

Question 43

What has been found when using the HCR20 with different diagnoses?

Answer 43

HCR20 was a good predictor in those with schizophrenia but worse in those with PDs. PDs were the hardest to predict although they are the most violent.

Question 44

What other measure in the Gray et al. (2003) study predicted violence as well as the HCR20?

Answer 44

BPRS predicted physical and verbal aggression.