Assessment and Management Actuarial and Clinical

This distinction, between the elements and the dimensions of risk, also relates to the debate between the actuarial and clinical prediction of risk. (Here 'clinical' does not necessarily refer to a medically qualified individual but rather to any person professing or practising the skill of making risk predictions about other people, for example probation officers, through studying the individual.) The actuarial approach stresses the importance of background variables, such as age (for a recent exposition of this position see Hare, 2002). The clinical approach stresses knowledge and experience of the individual (for a recent exposition of this position see Maden, 2002). Actuarial approaches claim greater predictive accuracy and power (Quinsey et al., 1998). Clinical approaches claim that actuarial information is impersonal, that at best it informs risk decision-makers about a class of people and not the specific individual in question. Clinicians have a poor record, or reputation, in risk prediction (see Limandri and Sheridan, 1995 and sources cited therein). But that, at least substantially, is a consequence of their predicting risk over long periods. Predicting another person's behaviour over several years, for example, is inviting error, unless you can be sure you will have that person under your close custody during that period! Those who predict risk will seldom have control over the individual in question during the period between prediction and the outcome contemplated occurring, or not occurring. Lawyers, and others, may and should be criticised for seeking inappropriate predictions of risk. (Unfortunately adversarial trial systems allow them the excuse of doing the best for their client.) But clinicians should accept responsibility for inappropriately answering. Questions about risk assessment should never be answered without reference to risk management. There is an urgent need for professional organisations to declare and to explain in public documents, which the courts can recognise are authoritative, the questions it would be professionally inappropriate for their members to answer in court.

The MacArthur Research Network on Law and Mental Health conducted a highly regarded empirical study into the prediction of violence by people with a mental disorder. Their iterative classification tree methodology permits the identification of groups of patients with an increased likelihood of violence (Monahan et al., 2001). This actuarial methodology permits a level of accuracy that exceeds best clinical predictions. But the authors do not, in marked contrast with Quinsey and colleagues

(1998), recommend that their approach, or actuarial approaches in general, should replace or exclude clinical assessments of risk. In particular they note (pp. 130-135) that their research population involved groups of civilly detained white, African American or Hispanic adults between 18 and 40 years old. So a patient whose risk has to be predicted may not fall within, or be represented by, those groups. Or there may be some special factor, for example the patient has broken both arms or has declared a desire to be violent, which makes him or her distinctively different and therefore less or more likely to be violent. They conclude that actuarial tools, such as their own, 'are best viewed as 'tools' for clinical assessment... —tools that support, rather than replace, the exercise of clinical judgment' (Monahan et al., 2001, p. 134, reference deleted). They believe that clinical judgement, aided by actuarial assessment, reflects current good professional practice. But they seek research, which might cause them to review their judgement, on how often, by how much and why, clinicians revise actuarial assessments. Litwack (2001), in a review of the debate, concludes that actuarial methods have not been conclusively proved superior to clinical. One of his points is that clinical decisions, for example about dangerousness, are not always predictions. So comparison is not always proper. In the terms of the model proposed here the clinician's decision may concern management rather than assessment.

But the problem with clinical assessments of risk is not just their comparative lack of predictive power. There are, also, several serious process problems. These relate to how clinical risk decisions are made and communicated. For example a key problem is the misuse, in practical decision-making, of base rates. They are undervalued (Tversky and Kahneman, 1973,1980). Information may be available about how likely a particular outcome is, after an actuarial risk assessment has been conducted, but that information is liable to be ignored or under-used.

There is no question that subjects in psychology experiments tend to ignore base rates

(just as physicians do) even when the base rates are clearly indicated. (Rachlin, 1989, p. 54)

A high-quality actuarial risk assessment may have been undertaken. A clinician may appreciate the value of such information but decide that other information about the particular patient, obtained in a clinical examination, should be taken into account and should be allowed to vary the importance attached to the actuarial assessment. For example the actuarial research, based on a large population, might indicate that men are six times more likely to, whatever, than women. But the decision-maker might fail to give that information its due, and/or he or she might rate clinical information, for example the patient had been taking his drugs without problems for a month, as more important.

The clinician does not, with the current state of research, know how much emphasis to give, or not to give, to that clinical information. That must wait, at the very least, upon the research recommended by the MacArthur team. As they note:

... the principal use of violence risk assessments ... is as an indicator of the need for violence risk management. (Monahan et al., 2001, p. 135; italics in original)

Perhaps we could and should go further, using the model of risk being developed here. Actuarial methods could/should be required, where they are available, for risk assessment. But clinical concerns about that assessment, plus additional case-specific information, should (only) be utilised in the risk management plan adopted.

Risk assessment relates to the elements of risk, which is what the empirical research has concentrated upon. Risk management relates to the dimensions of the risk in the particular case context. Special knowledge, about a particular patient, would justify different controls, more or less, on how the decision is implemented. It should not, it is suggested, be allowed to alter a competent actuarial risk assessment. ('Competent', here, is intended to refer to minimum professionally agreed standards rather than to create a definitional argument.) Lawyers should be encouraged to appreciate the distinction between risk assessment and risk management and only to ask questions appropriate to the particular category. It is appropriate for them to challenge the quality of particular risk assessments, for example to ensure that appropriate actuarial approaches were properly used and communicated. Then it is appropriate for them to enquire into how that risk assessment informed the clinicians' (or other risk-takers') plans for action. Such an approach, emphasising the iterative and integrative roles of risk assessment and risk management, has considerable potential for practitioners, both lawyers and risk-takers. It will help to clarify, improve and justify decision-making.

0 0

Post a comment