Wednesday, December 8, 2010

Blogging about someone else's blog...

Let me say first that it is not my intention to initiate some kind of a blog war. But, I read something recently in another blog that caught my attention and that probably requires a comment or two (or six).

Actuarial risk assessment scales for sexual and other offenders started to appear in the early to mid 1990s. The Violence Prediction Scheme (an early precursor to the VRAG and SORAG) arguably got the ball rolling, although the LSI was already in existence at the time. It has been claimed that the Static-99 is the most widely used sexual offense risk assessment scale on the planet, which may indeed be true, but it is hardly the only instrument of its kind. Most sexual offender evaluators use at least one of the more popular scales in their assessments of offenders.

Use of these scales is based on the premise that certain factors that, when clustered together, are predictive of who is at higher risk to reoffend. The more factors you have, the higher your risk or, at least, that's the thought. Items are typically chosen through a process of meta-analysis of factors reported in the primary research literature. As I already noted, various scales exist, but the majority of the factors included in them are roughly the same, largely because they are constructed using the same primary research literature. The two biggie factors appear to be sexual deviance and antisociality. Take your pick: SORAG, MnSOST-R, Static-99, etc. -- they all look pretty similar once you get past the idiosyncratic wording of items.

So, back to the other blog...

Dr. Franklin has something of a penchant for the flamboyant. Actually, I often appreciate a bit of bluster, but I think we also need to avoid bias and unnecessary sideswipes at our colleagues. In the spirit of honesty, let me say that I work in a civil commitment center and that I am a certified trainer for the Static-99 group of measures. So, my potential bias is on the table. Let me also say, however, that my work in the civil commitment center is aimed fully at making sure that my residents have the best treatment opportunities available so that they may have a reasonable chance at returning safely to the community to live a balanced, self-determined lifestyle free of future harm to others or themselves. As far as the Static-99 is concerned, as soon as someone shows me a tool that does the job markedly better, I'll jump ship and start using that one.

In her recent blog, Dr. Franklin suggests that her colleagues presenting on actuarial risk assessment at this fall's ATSA conference are "self-appointed gurus". Really, Dr. Franklin, is that actually a fair appraisal? The main personalities in the Static-99 world have contributed immeasurably to our (better) understanding of sexual deviance and risk to the community. I see no good reason to view them as anything other than well-intentioned, honest, and diligent scientists. That others may see them as "gurus" is not of their conscious efforts but, rather, a byproduct of the good work they have done. As far as I can tell, almost every person associated with the Static-99 group has worked hard to demonstrate both the strengths and weaknesses of the methodology and the scale itself. Not a one of them has ever overstated the value of the method or the tools they tout. Dr. Franklin rightly notes that these instruments are freely available on the net. We should also note that research as to their utility is vigorously encouraged.

For the record, all of the research to date suggests that tools of this nature have, at best, moderate predictive accuracy in helping us assess who is or is not likely to reoffend. I'm confused by Dr. Franklin's suggestion that we should use random sampling in the construction or validation of actuarial risk measures. What is it that we should be sampling randomly? The items? The subjects we score on the resultant scale? I may have missed something, but it seems to me that we should rather intentionally seek out those factors best predictive of future offending and that we should use these scales on persons who demonstrate at least some propensity for future offending. Neither of these seems well suited to a random approach.

As to the psychometric properties of the Static-99, I find it curious that Dr. Franklin only takes shots at this particular scale. As I noted above, there are many others available and given that they all sample from the same item pool and are subject to the same foibles and limitations, why not pick on the whole mess of them? What applies to Static-99 applies equally to them all. No risk assessor in his or her right mind should ever view a score (and risk rating) on an actuarial scale as being the only thing you need to know in evaluating a sexual offender's propensity for recidivism. They are tools found in a much larger toolbox of useful protocols/data/etc. you need to consider in providing comprehensive assessments of risk.

I also find it curious that Dr. Franklin labels the Static-99 as unreliable. The research I've seen suggests that it is particularly reliable -- guys who score higher reoffend more often than guys who score lower. This appears to be so no matter where the scale is used. Further, inter-rater reliabilities reported by various researchers have strongly suggested that, when using the coding rules, different assessors come up with the same score on the same guy most of the time. Overestimating risk is not a reliability issue, it's a validity issue -- one that we've already acknowledged in saying that the scale gives us only moderate predictive accuracy.

Andrew Harris (perhaps, somewhat cynically) opined the following:

Since we know from the literature that without the use of actuarials people tend to greatly over-estimate risk – and since all good ATSA members would support the use of actuarials, are we not cheap shills for the defence? The fact remains that a well done sex offender risk assessment is almost without exception a great aid to the offender.

Here we have a member of the Static-99 group encouraging us to use actuarials to combat the sort of generalized tendency towards overestimation of risk that Dr. Franklin lamented in her blog. The last sentence of Dr. Harris' comment is simply brilliant. Maybe, what Dr. Franklin should be highlighting more is the fact that some evaluators are falling prey to partisan concerns. A "well done sex offender risk assessment" should say the same thing, regardless of who you are working for.

As for the effects of aging on reoffense risk, Dr. Franklin's comments are very much in line with my thinking and that of others in the Static-99 group. Indeed, the Static-99R exists solely because of a revised age-weighting that attempts to account for this effect. Too bad Dr. Franklin didn't note that a good bit of the research and discussion regarding this aging-out phenomenon, as it relates to sexual offenders, was instigated, led, or encouraged by members of the Static-99 group.


1 comment:

  1. Though a fan of Dr. Franklin's blog, I found your analysis insightful and refreshing! Thanks for the information. MB Whitehill, PhD.