Risk Variables: Risk variables are personal characteristics, situations, or environmental conditions that predict the onset, continuity, or escalation of sex offending behavior.

What are Static Risk Variables? Static risk variables are historical and reflect prior life experiences and previous behaviors that are associated with a statistically increased likelihood or probability of sex offending behavior. Thus, static risk variables are fixed or inert. Whatever degree of risk is implied by these static variables can only change with the introduction of dynamic risk variables.

What are Dynamic Risk Variables? A true dynamic risk variable must satisfy a number of conditions: 1st It must precede and be associated with sex offending behavior, 2nd It must be capable of changing, and 3rd Changing the variable must change or influence or effect sex offending behavior. Thus, dynamic risk variables must be capable of mitigating (reducing) or exacerbating the risk suggested by static variables.

Are Risk Variables Causes? Is it proper to say that risk variables cause sex offending behavior? ONLY when it has been demonstrated empirically that changing the risk variable results in changes in the onset or continuity of sex offending. Otherwise, we must assume that risk variables are simply correlated with sex offending behavior.

What are Actuarial risk assessment scales? Actuarial refers to the work done by actuaries. Actuaries are individuals who are trained to calculate risks using statistics, usually for insurance companies. Actuarial scales are developed using statistical analyses of groups of individuals with known outcomes (such as men who have been convicted of a new sex offense and men who apparently have not re-offended sexually). These analyses tell us which items ("predictor variables" or "risk variables") do the best job of differentiating between those who re-offended and those who apparently did not reoffend. Since some variables inevitably do a better job than others, these analyses can also tell us how much each variable should be weighted. These variables are combined to form a scale. The scale is then tested to see how well it works at predicting those who reoffend sexually. These studies provide support for the predictive validity of the scale. The scales are then used on completely different samples to see how well they work. This is referred to as cross-validation. When a scale has been used on a sufficiently large number of offenders, the scale score may be expressed as an estimate of the probability that the individual may reoffend sexually within a specified time frame (e.g., someone with this scaore has a 30% probability of reoffending sexualy within three years).

What about item weighting? Actuarial scales often seem to work better when variables are properly weighted. Item weighting takes into consideration that some variables simply are more important than others when it comes to predicting outcome. Proper item weighting is done with a statistical procedure called multiple linear regression. The result is a "weighted linear prediction." Item weighting, however, is not required. Some argue that simple unit item weighting (each variable is weighted equally) is just as effective. The PCL-R, for instance, is a simple unit item weighting. All items are scored 0, 1, or 2.
   Item weighting is, of course, an empirical question. In order to do proper item weighting, large samples of offenders are needed to determine the item weights, and the judgments must be made empirically. A number of risk assessment scales use "clinically derived" item weights (i.e., clinicians decided how important each item was and weighted each item according to their judgments). This is absolutely not acceptable. Without sufficient data to suggest otherwise, items should have equal weights.

Can I "adjust" the scores of risk scales? The question of whether or not a clinician should "adjust" the numerical score from an actuarial scale based upon knowledge of important information that the scale does not take into consideration or does not emphasize has been a matter of some controversy. Those who recommend adjustment do so, because the clinician/examiner may be aware of potentially critical information that could affect risk but which is not addressed (or not adequately addressed) by the scale. Examples might include being very depressed or very angry after being kicked out of the house, having a girlfriend suddenly end a relationship, or having a very important person die (or be killed). When the risk assessment is close in time to these acute insults, it is obviously important to take note of them and factor them in to the overall risk picture.
   We do not recommend, however, that the score of any empirically-validated risk assessment scale ever be changed. The most important reason is that it is very difficult, if not impossible, to provide adequate ground rules or guidelines for insuring uniform adjustment. Without such guidelines, how much a score is adjusted and under what circumstances the adjustment occurs is left up to the individual, introducing error and unreliability. Changing the score of a risk assessment scale based on "new" information is equivalent to changing the T-score on an MMPI scale after realizing that the client's score on that scale does not reflect this "new" information. It would obviously be highly unethical to change T-scores on standardized tests. It is, in our estimation, equally unethical to change the scores on risk assessment scales.
   So, how does one take into account information that clearly appears to be of importance in assessing risk? We recommend that critical, risk-relevant information be incorporated into a comprehensive assessment of risk. Rather than adjusting the numerical score of the scale, we recommend that such information be used to "adjust" the conclusions. All information obtained is presented, along the score that most accurately reflects the scale. The resulting conclusions might, if deemed appropriate, include a sentence such as, "Although the risk assessment scale score is relatively low, there are clear aggravating factors in the individual's life that may increase his risk..."

What can I do to improve my reliability? Assuming that the items on the scale are clearly stated and the criteria for scoring the items are clearly stated, the single most important factor contributing to unreliability is the ambiguity of the information that is being used to score the items. How clear or how ambiguous the information is may vary enormously from one case to another. There are no foolproof methods for dealing with ambiguous information.
   To enhance reliability, we strongly recommend that examiners use as many sources of information as possible when scoring the scale variables. In addition, although it is often not feasible, we also recommend that the scale be scored by two independent clinicians who then compare and discuss their scores. The agreed upon scores should be used. When the available information is very limited, unclear, or incomplete, items should be scored "conservatively" (that is, in the direction of lower risk), and it should be noted that the resulting score may underestimate the risk.
   Clinicians should, of course, study the Manual prepared for the scale. Lastly, it is always helpful to complete training cases before using the scale on a "real" case. The importance of adequate training on practice cases cannot be overstated.