Using Another Method of Risk Assessment: Static-99

[Page 388]

As demonstrated above, Parole Board members, in assessing the dangerousness of sex offenders, judged many more of them to be 'high risk' than were in fact reconvicted of a sexual offence over quite a long follow-up period. To what extent might their judgements have been improved if they had had available to them a validated actuarial prediction instrument for sex offender risk assessment?

We calculated risk of reconviction according to the recently devised prediction instrument for male sex offenders, known as 'Static-99', which has been validated in a number of countries including the United Kingdom (Hanson and Thornton 1999). Static-99 classifies offenders (on the basis of data relating to their previous offending history, characteristics of the offences they have committed, age and length of domestic relationship) into four categories: 'low', 'medium-low', 'medium-high', and 'high'. It has been shown to be as accurate, or more accurate, than other sex-offender prediction tools in identifying offenders who are subsequently reconvicted of sexual or violent offences.

We compared the prisoners' risks scores on Static-99 with the judgements made by Board members on whether prisoners were 'high risks' or not. As Table 6 shows, the level of risk on Static-99 was a good predictor of increased probability of being reconvicted and imprisoned for a sexual offence, and especially of being reconvicted and imprisoned for either a sexual or a violent offence. In the four-year follow-up sample, none of those classified by Static-99 as 'low' risks was reconvicted, and only 2 per cent of those classified as 'medium-low' risks were reconvicted. Of those in the 'high' risk band, 14 per cent were reconvicted of a sexual offence and as many as 27 per cent were reconvicted of a sexual or a serious violent offence.  

Table 6
Static-99 assessment of risk and rates of reconviction of a sexual or serious violent offence: four-year follow-up 

Type of conviction

Static-99 'Low'

Static-99 'Medium-Low'

Static-99 'Medium-High'

Static-99 'High'

Total

Table 7 
Comparison of Parole Board indication of 'high risk' with Static- 99 assessment of risk: four-year follow-up  

Parole Board indication of serious risk

Static-99 'Low'

Static-99 'Medium-Low'

Static-99 'Medium-High'

Static-99 'High'

Total

So, would the number of 'false positives' have been reduced had members of the Board been able to use Static-99 to identify the 'high-risk' offenders? And would it have been at the expense of failing to identify some of those who were reconvicted, the 'false negatives'?

[Page 389]

The answer depends on whether one chooses to concentrate only on those with the highest probability of reconviction -- the Static-99 'high' risks, or the larger and broader category of 'high/ medium-high ' risks. The results of the analysis are set out in Table 8. 

Table 8
Assessment of risk by Static-99 and reconviction by type of offence: 
four-year and six-year follow-up  
 

Follow-up period

Type of reconviction

Static-99 'High risk'

Static-99 'Not high risk'

Total

4 years

6 years

They show that roughly 45 per cent of those reconvicted of a sexual or serious violent offence after four or six years (namely six out of 13 and five out of 11 respectively) were classified by Static-99 as 'high' risks. [*6] 

[*6] Those identified as 'high risk' by Static-99 have, according to Hanson and Thornton, a 39 percent probability of a reconviction for  a sexual offence, and a 44 percent probability of reconviction for a violent offence, after a five-year follow-up period. Comparable probabilities for 'medium-high' risk are 30 and 39 percent  respectively.  

But this was lower than the proportion correctly identified by Parole Board members, who identified 11 of the 13 sexual or serious violent recidivists at the four-year follow-up stage and 9 of the 11 at the six-year stage (see Table 5). In other words, there would have been more 'false negatives' if Static-99 'high' risk had been used as the risk assessment criterion by the Board.

On the other hand, using Static-99 'high' risk as the criterion would have produced a somewhat lower proportion of 'false positives' after both four- and six-year follow-up periods (compare Table 5 with Table 8): the reason being that many fewer prisoners were rated 'high risk' by this instrument than by the Parole Board. 

Indeed, Static-99 did not identify a single one of the intra-familial sex offenders against children as a 'high', risk, whereas Parole Board members classified about half of them as 'high risk'. Yet, as already mentioned, none of them was reconvicted of a sexual or serious violent crime resulting in imprisonment. 

Another reason was the substantial proportion of 'deniers' who were considered 'high risk' by Board members, only one of whom was reconvicted of a sexual offence. Two-thirds of these 'high-risk' deniers were classified as 'low' or 'medium-low' risks by Static-99. In fact, only one of the Parole Board's 'high-risk' deniers in both the four-year and six-year follow-up samples was classified in the highest risk category by Static-99. But, of course, the trade-off for the lower 'false positive' rate obtained by using Static-99 'high' risk was that fewer of the sexual and serious violent recidivists were identified.

However, we found no advantage, either way, when the two categories, 'high' risk and 'medium-high' risk, of Static-99 were combined. Indeed the proportions of 'false positives' and 'false negatives' were similar (bearing in mind the small numbers involved) to those resulting from Parole Board members' judgements of 'high risk' in the early 1990s.

[Page 390]

In an attempt to identify whether there were any characteristics of the offences committed by the prisoners in this sample which, if taken into account, would produce a lower rate of 'false positives' , we compared the proportions reconvicted of a sexual or of a sexual or serious violent offence amongst various categories of prisoner: 

This analysis showed that it was possible, amongst those whom members of the Parole Board judged 'high risk' and Static-99 assessed as 'high' risk, to isolate some categories of offender who have about a 40 to 45 per cent rate of reconviction for a serious sexual or violent offence over a six-year follow-up period. In other words, if these characteristics were taken into account, the 'false positive rate' might drop to around 55 to 60 per cent.

Thus, Static-99 would not have improved the ability of the Parole Board to identify the 'true positives': those who became sexual or serious violent recidivists. However, if it had been available as a guide to the Board, it may have been able to cut the proportion of those who were 'false positives' if account had been taken of the lower levels of risk that Static-99 assigned to intra-familial sexual offenders against children and 'deniers'.