Some specific Risk Assessment instruments


Actuarial instruments


a. Risk Matrix 2000 (for sexual and violent offenders: predicts risk of sexual, or violent reoffending, or either)

b. Static-99 (sex offenders)

c. VRAG (violent offenders)

d. SORAG (sex offenders)

Structured anchored clinical judgements

e. HCR-20 (violent offenders)

f. SVR-20 (sex offenders)

g. SARA (domestically violent offenders)

h. SARN (sex offenders)

i. PCL-R (strictly speaking, not a risk assessment instrument, but a means of assessing someone for psychopathic traits; scored similarly to SACJs, and often treated as a risk assessment instrument)



Actuarials: the Risk Matrix 2000

The Risk Matrix 2000 exists in three versions, predicting violence, sexual offending, or both. It uses a small number of items relating to offence history, age on release and type of victims. The matrix was developed on a sample of prisoners released in the late 1970s and followed up over 20 years. It places an individual in one of four risk bands (low, medium, high, very high). The titles of these risk bands may seem a little odd to a lay person, who might (for example) expect a high risk individual to be more likely than not to convict for further sexual offences, whereas the opposite is the case.

Problems with the Risk Matrix 2000

a. Reconviction rates for sex offenders (at least) have reduced since the 1970s; the effect is to exaggerate risk in modern offenders.

b. It doesn’t take age into account past 35, but age-related risk reduction (which is very large) mostly takes place after this; again, the effect is to exaggerate risk.

c. Hart (2005) showed it is more likely than not to place people in the right risk category, but only marginally so; the effect is to increase error.

d. It has not been standardised on lifers, to whom it is mostly applied; the effect is anyone’s guess.

e. It doesn’t distinguish between imminent risk and remote risk (it predicts over a 20-year period); remote prediction cannot in principle be accurate.



Actuarials: the Static-99 (or 2002?)

The Static-99 uses a similar method to the Risk Matrix 2000; the two have one author in common (David Thornton). An updated scoring system was produced in 2002/3, which tries to take additional factors into account, including reduction in risk because of offence-free time (ie, risk is less if there are no recent offences). In practice, many of the recommendations made in connection with this scoring are not supported by empirical research. In other words, there is a large element of the authors' opinion in the scoring scheme, and we still await scientific evidence on how good the authors' opinion actually is. As with the Risk Matrix 2000, the Static-99 uses items relating to age, victim, and offence history.

Problems with the Static-99

a. It doesn’t take age into account after 25; Wollert (2006) showed this seriously exaggerates risk in older offenders.  

b. Revised scoring rules have introduced confusion over how some things should be scored; the effect is to increase error. 

c. Other problems are similar to RM 2000 (which it greatly resembles).

 

Actuarials: the VRAG and SORAG

These are the Violence Risk Assessment Guide and Sex Offender Risk Assessment Guide (Quinsey, Harris, Rice and Cormier, 2006). They are considered together because they use exactly the same rationale, and were devised by the same research group. They follow a similar rationale to the Risk Matrix 2000 and Static-99, but use more items, including some personality and mental health items, such as the PCL-R. Both were originally developed on forensic psychiatric patients in Canada, and this may affect their applicability to prisoners in the UK. Compared with the Risk Matrix and the Static, they give age more weight, though still not enough, failing to allow for any increase in age beyond 35 years. Both allot people to risk categories, resulting in predictions over 7 years and 10 years. The SORAG, although intended for sex offenders, actually predicts the risk of violent reoffending. The authors class all sexual offending as violent in some sense.

Problems with the VRAG and SORAG

a. Development on psychiatric patients leaves doubt as to their applicability to prisoners

b. They may not travel well: several studies suggest over-estimates of risk in British samples by a factor of two! 

c. Studies suggest that the main predictive value comes from the inclusion of the PCL-R, and drops dramatically if this is removed or not available
 
d. Otherwise, there are similar problems to other “actuarials”.



Actuarials generally: verdict

a.    None are really up to the task. Wollert’s (2006) blunt verdict on them all as “useless” except with the youngest groups of adults is supported by the work of other researchers, such as Barbaree, Blanchard and Langton (2003). 

b.    All try to predict over too long a period, and long term prediction simply doesn’t work: it might be better to concentrate on 2-5 yrs

c.    All are widely over-applied: they should only be used on people similar to the original sample

d.    They all mix different offender types too much: incest is very different from serial rape, for example

e.    As for age, not one of them takes this sufficiently into account. They are widely applied to older offenders, despite abundant evidence now that recidivism rates in this group are far lower than previously realised

f.    They are indiscriminately applied to all types of prisoner, including lifers, although none of them has been standardised on this group. Lifers are different from other prisoners in many ways, not just in terms of sentence length.



Comments