Chapter 10 - Risk Assessment PDF

Chapter 10 Risk Assessment Learning Objectives ■■ Define the components of risk assessment. ■■ List what role risk assessments play in Canada. ■■ Describe the types of correct and incorrect risk predictions. ■■ Differentiate among static, stable, and acute dynamic risk factors. ■■ Describe...

Chapter 10 Risk Assessment Learning Objectives ■■ Define the components of risk assessment. ■■ List what role risk assessments play in Canada. ■■ Describe the types of correct and incorrect risk predictions. ■■ Differentiate among static, stable, and acute dynamic risk factors. ■■ Describe unstructured clinical judgment, actuarial prediction, and structured professional judgment. ■■ Address some of the important differences amongst unique sub-populations of offenders. Joanne Marshall has served two years of a three-year sentence for aggravated assault. The assault occurred late one evening after Joanne had returned home from drinking with her friends. She got into a heated argument with her boyfriend, grabbed a knife from the kitchen, and stabbed him in his shoulder. An institutional psychologist completed a risk assessment and is supporting her application for parole. Joanne is going to appear before the three-member parole board to discuss the offence she committed, her plans if released, and what intervention programs she has participated in. The parole board members will need to also consider what level of risk Joanne poses for reoffending, including whether or not she will engage in another violent act. They will also attempt to determine whether she has developed more appropriate ways of dealing with interpersonal conflict. Every day, individuals make judgments about the likelihood of events. Predictions are made about being admitted into law school, recovering from depression, or committing a criminal act after release from prison. Our legal system frequently requires decisions about the likelihood of future criminal acts that can significantly influence the lives of individuals. With the possibility that offenders could spend years or even the remainder of their lives in confinement, decisions by psychologists can have a significant impact. Predicting future violence has been described as “one of the most complex and controversial issues in behavioral science and law” (Borum, 1996, p. 945). M10_POZZ8067_05_SE_C10.indd 289 28/09/16 7:23 PM Although it is clear that significant advances have taken place since the 1990s, risk assessment and prediction remains imperfect. Bonta (2002) concluded that “risk assessment is a double-edged sword. It can be used to justify the application of severe sanctions or to moderate extreme penalties. . . . However, the identification of the violent recidivist is not infallible. We are not at the point where we can achieve a level of prediction free of error” (p. 375). Nonetheless, the systematic assessment of risk provides judicial decision makers, such as judges and the National Parole Board, with much-needed information to help them make challenging decisions. The goal of this chapter is to explore the major issues associated with risk prediction in a forensic context. In particular, the focus will be on understanding the task of assessing risk and predicting violence. What is Risk Assessment? In the past two decades, we have seen a change in the way risk is viewed. Prior to the 1990s, risk was seen as a dichotomy—the individual was either dangerous or not dangerous. Nowadays, risk is regarded as a range—the individual can vary in the degree to which he or she is considered dangerous (Steadman, 2000). In other words, this shift has added a dimension of probability to the assessment of whether a person will commit violence. The focus on probability reflects two considerations. First, it highlights the idea that probabilities may change across time. Second, it recognizes that risk level reflects an interaction among a person’s characteristics, background, and possible future situations that will affect whether the person engages in violent behaviour. The process of risk assessment includes both a “prediction” and “management” component (Hart, 1998). The prediction component describes the probability that an individual will commit future criminal or violent acts. The focus of this component is ScienceCartoonsPlus.com 290 M10_POZZ8067_05_SE_C10.indd 290 Chapter 10 28/09/16 7:23 PM on identifying the risk factors that are related to this likelihood of future violence. The management component describes the development of interventions to manage or reduce the likelihood of future violence. The focus of this component is on identifying what treatment(s) might reduce the individual’s level of risk or what conditions need to be implemented to manage the individual’s risk. As described by Hart (1998), “the critical function of risk assessments is violence prevention, not violence prediction” (p. 123). Risk Assessments: When are They Conducted? Risk assessments are routinely conducted in civil and criminal contexts. Civil contexts refer to the private rights of individuals and the legal proceedings connected with such rights. Criminal contexts refer to situations in which an individual has been charged with a crime. Common to both contexts is a need for information that would enable legal judgments to be made concerning the probability of individuals committing some kind of act that would disrupt the peace and order of the state or individuals within the state. Civil Setting A number of civil contexts require risk assessment: ■■ Civil commitment requires an individual to be hospitalized involuntarily if he or she has a mental illness and poses a danger to him- or herself or others. A mental health professional, usually a psychiatrist or psychologist, would need to know the probability of violence associated with various mental illness symptoms and disorders and be able to identify whether the circumstances associated with individual patients would affect the likelihood that they would harm others or themselves. In Canada, only a psychiatrist can civilly commit someone to a hospital. ■■ Assessment of risk in child protection contexts involves the laws that are in place to protect children from abuse. The risk of physical abuse, sexual abuse, or neglect is considered when a government protection agency, such as the Children’s Aid Society, decides whether to temporarily remove a child from his or her home or to terminate parental rights. To provide assistance to protection agencies, professionals need to be familiar with the risk factors that predict childhood maltreatment. ■■ Immigration laws prohibit the admission of individuals into Canada if there are reasonable grounds for believing they will engage in acts of violence or if they pose a risk to the social, cultural, or economic functioning of Canadian society. ■■ School and labour regulations also include provisions to prevent any kind of act that would endanger others. ■■ Mental health professionals are expected to consider the likelihood that their patients will act in a violent manner and to intervene to prevent such behaviour. This responsibility is called duty to warn. The Canadian Psychological Association’s Code of Ethics for Psychologists includes a guide to help psychologists decide the most ethical action for any potential dilemma. Risk Assessment M10_POZZ8067_05_SE_C10.indd 291 291 28/09/16 7:23 PM Criminal Settings The assessment of risk occurs at nearly every major decision point in the criminal justice and forensic psychiatric systems, including pretrial, sentencing, and release. A person can be denied bail if there is a substantial likelihood that he or she will commit another criminal offence. In the case of adolescent offenders, the judge can decide to apply adult criminal sanctions depending on the age, type of offence, and risk level posed by the youth. Risk also plays a role in decisions about whether a youth should be sent to secure custody. For example, adolescent offenders should be committed to secure custody only if they are considered high risk; if not, they should be placed in open custody or serve a probation term in the community. An important issue in risk assessment in criminal settings is the disclosure of information about potential risk. This disclosure must be considered in light of the solicitor–client privilege that is fundamental to criminal proceedings. For lawyers to adequately represent their clients, they must be able to freely discuss the case with the clients. This privilege is also extended to experts retained by lawyers. A case in Canada has clarified when solicitor–client privilege and doctor–patient confidentiality must be set aside for the protection of members of the public. Smith v. Jones (1999) involved a psychiatrist who was hired to aid a defence lawyer in preparing a case. The client was a man accused of aggravated sexual assault on a prostitute. The accused told the psychiatrist of his plans to kidnap, sexually assault, and kill prostitutes. The psychiatrist told the defence lawyer about his concerns that the accused was likely to commit future violent offences unless he received treatment. When the psychiatrist found out that the defence lawyer was not going to address his concerns at the sentencing hearing (the accused pled guilty to the charge of aggravated assault), he filed an affidavit providing his opinion about the level of risk posed by the accused. The trial judge ruled that because of concerns about public safety, the psychiatrist was duty-bound to disclose to the police and the Crown counsel the information he obtained. The case was appealed to the Supreme Court that ruled that in cases where there is “clear, serious, and imminent danger,” public safety outweighs solicitor–client privilege (Smith v. Jones, 1999, para. 87). Although risk assessment is a routine component of many sentencing decisions, it is a critical component of certain kinds of sentencing decisions. For example, after 1947, when habitual criminal legislation was introduced, offenders could be sentenced to an indefinite period of incarceration. In 1977, dangerous offender legislation was enacted that requires mental health professionals to provide an assessment of risk for violence. Changes to the legislation in 1997 made indefinite incarceration the only option if an offender is found to be a dangerous offender (see Chapter 9 for more information about dangerous offenders). At the same time, a new category of dangerous persons was created, referred to as long-term offenders. To be declared a long-term offender, a person must pose a substantial risk for violently reoffending. Thus, risk assessment is also a core component of this legislation. Risk assessment is also required for decisions concerning release from correctional and forensic psychiatric institutions, such as parole. If a person is sentenced to prison in Canada, he or she can apply to the National Parole Board to get early release. 292 M10_POZZ8067_05_SE_C10.indd 292 Chapter 10 28/09/16 7:23 PM Parole board members use a variety of sources of information (including risk assessments provided by institutional psychologists) to decide the likelihood that the offender will commit another offence if released. Although most offenders get released on statutory release (after serving two-thirds of their sentences), statutory release can be denied if the offender is likely to commit further violent offences. Finally, a patient who has been found not criminally responsible on account of a mental disorder (see Chapter 8) can be released from a secure forensic psychiatric facility only if a risk assessment is completed. Clearly, risk assessment plays an integral role in legal decision making, in both civil and criminal settings, allowing informed decisions that weigh the likelihood that an individual will engage in a dangerous or criminal act in the future. In the sections that follow, we will look at the predictive accuracy of these assessments, as well as the factors that actually predict antisocial outcome. Types of Prediction Outcomes Predicting future events will result in one of four possible outcomes. Two of these outcomes are correct, and two are incorrect. The definitions provided below are stated in terms of predicting violent acts but could be used for any specific outcome (see also Table 10.1): True positive: A correct prediction that occurs when a person who is predicted to engage in some type of behaviour (e.g., a violent act) does so ■■ A true positive represents a correct prediction and occurs when a person who is predicted to be violent engages in violence. ■■ A true negative is also a correct prediction and occurs when a person who is predicted not to be violent does not act violently. True negative: A correct prediction that occurs when a person who is predicted not to engage in some type of behaviour (e.g., a violent act) does not ■■ A false positive represents an incorrect prediction and occurs when a person is predicted to be violent but is not. ■■ A false negative is also an incorrect prediction and occurs when a person is predicted to be nonviolent but acts violently. The last two types of errors are dependent on each other. Minimizing the number of false positive errors results in an increase in the number of false negative errors. The implication of these errors varies depending on the decisions associated with them, and in many cases the stakes are high. A false positive error has implications for False positive: An incorrect prediction that occurs when a person is predicted to engage in some type of behaviour (e.g., a violent act) but does not False negative: An incorrect prediction that occurs when a person is predicted not to engage in some type of behaviour (e.g., a violent act) but does Table 10.1 Predictions: Decisions versus Outcomes Outcome Decision Does not reoffend Reoffends Predicted not to reoffend True negative (correct prediction) False negative (incorrect prediction) Predict to reoffend False positive (incorrect prediction) True positive (correct prediction) Risk Assessment M10_POZZ8067_05_SE_C10.indd 293 293 28/09/16 7:23 PM A gun-free sign at a high school Sascha Burkard/Fotolia the individual being assessed (such as denial of freedom), whereas a false negative error has implications for society and the potential victim (such as another child victimized by a sexual offender). In some cases, it is perhaps tolerable to have a high rate of false positives if the consequences of such an error are not severe. For example, if the consequence of being falsely labelled as potentially violent is being supervised more closely while released on parole, the consequence may be acceptable. However, if the consequence of being falsely labelled as potentially violent contributes to a juror’s decision to decide in favour of the death penalty, then this price is too high to pay. As in many legal settings, the consequences for the individual must be weighed in relation to the consequences for society at large. The Base Rate Problem Base rate: Represents the percentage of people within a given population who commit a criminal or violent act 294 M10_POZZ8067_05_SE_C10.indd 294 A problem with attempting to predict violence is determining base rates. The base rate represents the percentage of people within a given population who commit a criminal or violent act. It is difficult to make accurate predictions when the base rates are too high or too low. A problem that emerges when attempting to predict events that have a low base rate is that many false positives will occur. For example, the past decade has seen several high-profile school shootings. However, although these events generate much media coverage, they occur infrequently. Any attempt to predict which individual youths might engage in a school shooting would result in many youths being wrongly classified as potential shooters. The base rate can vary dramatically depending on the group being studied, what is being predicted, and the length of the follow-up period over which the individual is monitored. For example, the base rate of sexual violence tends to be relatively low, even over extended follow-up periods, whereas the base rate for violating the conditions of a conditional release is very high. The base rate problem is not such a concern if predictions of violence are limited to groups with a high base rate of violence, such as incarcerated offenders. The general rule is that it is easier to predict frequent events than infrequent events. Chapter 10 28/09/16 7:23 PM A History of Risk Assessment Before 1966, relatively little attention was paid to how well professionals could assess risk of violence. In the 1960s, civil rights concerns provided the rare opportunity to study the accuracy of mental health professionals in predicting violence. In the case of Baxstrom v. Herald (1966), the U.S. Supreme Court ruled that the plaintiff Johnnie Baxstrom had been detained beyond his sentence expiry and ordered him released into the community. As a result of this case, more than 300 mentally ill offenders from the Dannemora State Hospital for the Criminally Insane and another state hospital were released into the community or transferred to less secure institutions. Steadman and Cocozza (1974) followed 98 of these patients who were released into the community but had been considered by mental health professionals as too dangerous to be released. Only 20 of these patients were arrested over a four-year period, and of these, only 7 committed a violent offence. In a larger study, Thornberry and Jacoby (1979) followed 400 forensic patients released into the community because of a similar civil rights case in Pennsylvania (Dixon v. Attorney General of the Commonwealth of Pennsylvania, 1971). During an average three-year follow-up period, 60 patients were either arrested or rehospitalized for a violent incident. The two studies we have just described are known as the Baxstrom and Dixon studies. These cases and similar ones call into question the ability of mental health professionals to make accurate predictions of violence. Two key findings emerged from the research. First, the base rate for violence was relatively low. For example, in the Baxstrom study, 7 out of 98 (roughly 7%) violently reoffended, as did 60 out of 400 (15%) in the Dixon study. Second, the false positive rate was very high. In the Baxstrom and Dixon studies, the false positive rates were 86% and 85%, respectively. These findings indicate that in the past many mentally disordered forensic patients were needlessly kept in restrictive institutions based on erroneous judgments of violence risk. Ennis and Litwack (1974) characterized clinical expertise in violence risk assessment as similar to “flipping coins in the courtroom” and argued that clinical testimony should be barred from the courtroom. Other researchers have gone even further, concluding that “no expertise to predict dangerous behavior exists and . . . the attempt to apply this supposed knowledge to predict who will be dangerous results in a complete failure” (Cocozza & Steadman, 1978, p. 274). This pessimism continued into the 1980s. John Monahan, a leading U.S. researcher, summarized the literature in 1981 and concluded that “psychiatrists and psychologists are accurate in no more than one out of three predictions of violent behavior over a several-year period among institutionalized populations that had both committed violence in the past (and thus had a high base rate for it) and who were diagnosed as mentally ill” (Monahan, 1981, p. 47). Notwithstanding the above conclusion, both Canadian and U.S. courts have ruled that predictions of violence risk do not violate the basic tenets of fundamental justice, nor are they unconstitutional. In Barefoot v. Estelle (1983), the U.S. Supreme Court determined the constitutionality of a Texas death-penalty appeal decision. Thomas Barefoot burned down a bar and shot and killed a police officer. Barefoot was Risk Assessment M10_POZZ8067_05_SE_C10.indd 295 295 28/09/16 7:23 PM convicted of capital murder and, at the sentencing phase of the trial, testimony was presented from two psychiatrists (one being Dr. James Grigson, whom we will discuss later in the chapter) about the threat of future dangerousness posed by Thomas Barefoot. Both psychiatrists testified, based on a hypothetical fact situation, that the individual described would be a threat to society. The judge sentenced Barefoot to death. The U.S. Supreme Court rejected the defendant’s challenge that psychiatrists were unable to make sufficient accurate predictions of violence and ruled that the use of hypothetical questions to establish future dangerousness was admissible. The court concluded that mental health professionals’ predictions were “not always wrong . . . only most of the time” (p. 901). Canadian courts have also supported the role of mental health professionals in the prediction of violent behaviour (R. Moore v. the Queen, 1984). For example, in a dangerous offender case, the issue of whether psychiatric testimony should be admitted as evidence was evaluated. The court concluded, “The test for admissibility is relevance, not infallibility . . . psychiatric evidence is clearly relevant to the issue whether a person is likely to behave in a certain way” (R. v. Lyons, 1987, para. 97). Methodological Issues Risk assessment assumes that risk can be measured. Measurement, in turn, assumes that an instrument exists for the measurement of risk. What would be the ideal way to evaluate an instrument designed to measure risk? The way to proceed would be to assess a large number of offenders and then, regardless of their risk level, release them into the community. The offenders would then be tracked to see if they commit another criminal act. This way, the risk instrument could be evaluated to determine if it could accurately predict future criminal acts. However, although this is an ideal scenario from a research perspective, it is not ethically feasible to release high-risk individuals into the community. In reality, the sample available for evaluating a risk-assessment instrument is limited to those with a relatively low risk of reoffending. This constrains the kinds of conclusions that can be drawn when risk assessment is evaluated in the real world. Monahan and Steadman (1994) identified three main weaknesses of research on the prediction of violence. The first issue concerns the limited number of risk factors being studied. Violent behaviour is due to a complex interaction between individual dispositions and situational factors. In other words, people engage in violence for many different reasons. Thus, many risk factors are likely involved, including the person’s background, social situation, and biological and psychological features. Many studies have focused on only a limited number of risk factors. Assessment of risk may be improved by measuring more of the reasons why people engage in violence. Yang and Mulvey (2012) recommended researchers pay more attention to understanding how an individual’s subjective state leads an individual to commit violence. The second issue concerns how the criterion variable (the variable you are trying to measure) is measured. Researchers have often used official criminal records as their criterion measure. However, many crimes may never be reported to police. Thus, many false positives may be undiscovered true positives. Even violent crimes may go undiscovered and many violent sexual crimes are recorded as simply violent in nature. In short, use of official records underestimates violence. When official records are 296 M10_POZZ8067_05_SE_C10.indd 296 Chapter 10 28/09/16 7:23 PM combined with interviews with patients or offenders and with collateral reports (information from people or agencies who know the patient or offender), the rate of violence increases. The MacArthur Violence Risk Assessment Study (Steadman et al., 1998) illustrates the effect of using different measures. Using official agency records, the base rate for violence was 4.5%, but when patient and collateral reports were added, the base rate increased to 27.5%, a rate of violence six times higher than the original base rate. Finally, how the criterion variable is defined is a concern. In some studies, researchers will classify their participants as having either engaged in violence or not. Monahan and Steadman (1994) recommended that researchers expand this coding to include the severity of violence (threatened violence versus severe violence), types of violence (spousal violence versus sexual violence), targets of violence (family versus stranger), location (institutions versus community), and motivation (reactive [unplanned violence in response to a provocation] versus instrumental [violence used as an instrument in the pursuit of some goal]). It is likely that some risk factors will be associated with certain forms of violence; for example, a history of sexual offences may predict future sexual offences but not future bank robberies. Judgment Error and Biases How do psychologists make decisions when conducting risk assessments? Researchers have identified the typical errors and biases in clinical decision making (Elbogen, 2002). The shortcuts people use to help to make decisions are called heuristics (Tversky & Kahneman, 1981). Some of these heuristics lead to inaccurate decisions. Clinicians may make several types of decision errors by including traits they intuitively believe to be important or assume to be associated with the risk but that actually are not (Odeh, Zeiss, & Huss, 2006). Chapman and Chapman (1967) defined an illusory correlation as the belief that a correlation exists between two events that in reality are either not correlated or are correlated to a much lesser degree than believed. For example, a clinician might assume a strong correlation between a diagnosis of mental disorder and high risk for violent behaviour. Although some forms of mental disorder are related to an increased risk, a relationship has not been consistently found (Monahan & Steadman, 1994). For example, Bonta, Law, and Hanson (1998) reported that offenders without a mental disorder were more likely to recidivate than those with a mental disorder. Whereas Bonta, Blais, and Wilson (2014) found no significant difference in recidivism between the mentally disordered group and the non-mentally disordered group. Clinicians also tend to ignore base rates of violence (Monahan, 1981), where clinicians working in prisons or forensic psychiatric facilities may not be aware of how often individuals with specific characteristics act violently. For example, the base rate for recidivism in homicide offenders is extremely low, however, given the nature of their crime, they might be perceived as high risk. Other investigators (Borum, Otto, & Golding, 1993) have noted the tendency to rely on highly salient or unique cues, such as bizarre delusions. In general, people tend to be overconfident in their judgments (see Kahneman & Tversky, 1982). Clinicians who are very confident in their risk assessments will be more likely to recommend and implement intervention strategies. However, while Risk Assessment M10_POZZ8067_05_SE_C10.indd 297 Illusory correlation: Belief that a correlation exists between two events that in reality are either not correlated or correlated to a much lesser degree 297 28/09/16 7:23 PM people can be very confident in their risk assessments, they may not be accurate. Desmarais, Nicholls, Read, and Brink (2010) investigated the association between clinicians’ confidence and accuracy of predicting short-term in-patient violence. Clinicians completed a structured professional judgment measure designed to assess the likelihood of violent behaviour (e.g., verbal and physical aggression, self-harm) and indicated on a five-point scale their level of confidence. Most clinicians were highly confident; however, the association between confidence and accuracy was minimal. This pattern of findings suggested clinicians tended to have an overconfidence bias. Approaches to The Assessment of Risk Unstructured clinical judgment: Decisions characterized by a substantial amount of professional discretion and lack of guidelines Actuarial prediction: Decisions are based on risk factors that are selected and combined based on their empirical or statistical association with a specific outcome 298 M10_POZZ8067_05_SE_C10.indd 298 What are the existing methods of risk assessment? Three methods of risk assessment are most commonly described. Unstructured clinical judgment is characterized by a substantial amount of professional discretion and lack of guidelines. There are no predefined rules about what risk factors should be considered, what sources of information should be used, or how the risk factors should be combined to make a decision about risk. Thus, risk factors considered vary across clinicians and vary across cases (Grove & Meehl, 1996; Grove, Zald, Lebow, Snitz, & Nelson, 2000). Grove and Meehl (1996) described this type of risk assessment as relying on an “informal, ‘in the head,’ subjective, impressionistic, subjective conclusion, reached (somehow) by a human clinical judge” (p. 294). See Box 10.1 for an example of a professional using this type of risk assessment. In contrast, mechanical prediction involves predefined rules about what risk factors to consider, how information should be collected, and how information should be combined to make a risk decision. Thus, risk factors do not vary as a function of the clinician and the same risk factors are considered for each case. A common type of mechanical prediction is called actuarial prediction. With actuarial prediction, the risk factors used have been selected and combined based on their empirical or statistical association with a specific outcome (Grove & Meehl, 1996; Grove et al., 2000). In other words, a study has been done in which a number of risk factors have been measured, a sample of offenders have been followed for a specific period, and only those risk factors that were actually related to reoffending in this sample are selected (for an example of an actuarial scale, see the Violence Risk Appraisal Guide described later in this chapter). A debate in the literature exists concerning the comparative accuracy of unstructured clinical versus actuarial prediction. In a review of 20 studies, Paul Meehl (1954) concluded that actuarial prediction was equal to or better than unstructured clinical judgment in all cases. A similar conclusion was reached almost 50 years later, when Meehl and his colleagues (Grove et al., 2000) conducted a meta-analysis of prediction studies for human health and behaviour (including criminal behaviour). In sum, the weight of the evidence clearly favours actuarial assessments of risk (Ægisdóttir et al., 2006; Mossman, 1994), even with samples of offenders with mental disorders (Bonta, Law, & Hanson, 1998; Phillips et al., 2005) and sex offenders (Hanson & MortonBourgon, 2009). A criticism of many actuarial assessments has been their sole reliance on static risk factors, which do not permit measuring changes in risk over time or provide information relevant for intervention (Wong & Gordon, 2006). Chapter 10 28/09/16 7:23 PM Box 10.1 Forensic Psychology in the Spotlight Dr. Death: A Legendary (Notorious) Forensic Psychiatrist Dr. James Grigson was a Dallas psychiatrist who earned the nicknames “Dr. Death” and “the Hanging Shrink” because of his effectiveness at testifying for the prosecution in death-penalty cases. For nearly three decades, Dr. Grigson testified in death-penalty cases in Texas. Death-penalty trials are divided into two phases. First, the defendant’s guilt is decided. Next, if the defendant is guilty of a serious crime, the same judge and jury decide whether to impose life in prison or to sentence the defendant to die. For example, under the Texas Penal Code, one of the issues the jurors must decide on is “whether there is a probability that the defendant would commit criminal acts of violence that would constitute a continuing threat to society.” Psychiatrists and psychologists are often hired to testify about the likelihood of future violence. Dr. Grigson’s testimony was very effective. He often diagnosed defendants as being sociopaths and stated with 100% certainty that they would kill again. For example, in Estelle v. Smith (1981), Dr. Grigson testified on the basis of a brief examination that the defendant Smith was a “very severe sociopath,” who, if given the opportunity, would commit another criminal act. The diagnosis of sociopath appears to have been based on the sole fact that Smith “lacked remorse.” Dr. Grigson has been proven wrong. In the case of Randall Dale Adams (documentarian Errol Morris made a movie about Adams’s story in 1988, called The Thin Blue Line, which helped to get the case reopened), Dr. Grigson testified that Randall Adams was a “very extreme” sociopath and would continue to be a threat to society even if kept locked in prison. Dr. Grigson based his assessment on a 15-minute interview in which he asked about Adams’s family background, had Adams complete a few items from a neuropsychological test designed to measure visual-motor functioning (Bender Gestalt Test), and asked Adams the meaning of two proverbs: “A rolling stone gathers no moss” and “A bird in the hand is worth two in a bush.” Randall Adams was sentenced to death. However, after he spent 12 years on death row, his conviction was overturned and he was released (another inmate confessed to the murder Adams had been charged for). It has been 13 years since Randall Adams was released. He is now married, employed, and living a nonviolent life. Dr. Grigson was wrong in this case—and potentially in how many others? In 1995, Dr. Grigson was expelled from the American Psychiatric Association (APA) for ethical violations. He was disqualified for claiming he could predict with 100% certainty that a defendant would commit another violent act (and, on at least one occasion, testifying that the defendant had a “1000%” chance of committing another violent act). The APA was also concerned that Dr. Grigson often testified in court based on hypothetical situations and diagnosed an individual without even examining the defendant. Dr. Grigson often diagnosed defendants as sociopaths on the basis of his own clinical opinion and not on any structured assessment procedures. Dr. Grigson was also involved in the death-penalty case of Canadian Joseph Stanley Faulder, who was convicted and sentenced to death for the robbery and murder of Inez Phillips. Dr. Grigson testified that Stanley Faulder was an “extremely severe sociopath,” that there was no cure, and that he would certainly kill again. We will never assess the accuracy of Dr. Grigson’s predictions, since on June 17, 1999, after spending 22 years on death row, Stanley Faulder was executed. When Dr. Grigson died in 2004 at the age of 72, he had testified in 167 trials. How many of these defendants fell victim to Dr. Grigson and his misguided attempt to protect society is unknown. Sources: Based on The Washington Times. (Dec 20, 2003) Texas ‘Dr. Death’ retires after 167 capital case trials. Retrieved from http://www.washingtontimes. com/news/2003/dec/20/20031220-113219-5189r/?page=all; Gross, A. (2004). Dangerous predictions: The case of Randall Dale Adams. The Forensic Examiner, 13(4). Retrieved from http://www.biomedsearch.com/article/ Dangerous-predictions-case-Randall-Dale/125957151.html; Amnesty International. (undated). USA: Adding insult to injury: The case of Joseph Stanley Faulder. Retrieved from http://www.amnesty.org/en/library/asset/ AMR51/086/1998/en/c619ed89-d9a2-11dd-af2b-b1f6023af0c5/ amr510861998en.html. Risk Assessment M10_POZZ8067_05_SE_C10.indd 299 299 28/09/16 7:23 PM Structured professional judgment: Decisions are guided by a predetermined list of risk factors that have been selected from the research and professional literature. Judgment of risk level is based on the evaluator’s professional judgment Arising from the limitations associated with unstructured clinical judgment and concern that the actuarial method did not allow for individualized risk appraisal or for consideration of the impact of situational factors to modify risk level, a new approach to risk assessment has emerged—structured professional judgment (SPJ) (Borum, 1996; Webster, Douglas, Eaves, & Hart, 1997). According to this method, the professional (the term professional is used to acknowledge that it is not only clinicians who make evaluations of risk but a diverse group, including law enforcement officers, probation officers, and social workers) is guided by a predetermined list of risk factors that have been selected from the research and professional literature. The professional considers the presence and severity of each risk factor, but the final judgment of risk level is based on the evaluator’s professional judgment. The reliability and predictive utility of these risk summary judgments are only beginning to be assessed. Skeem and Monahan (2011) described violence-risk-assessment approaches as having four components. Not all risk approaches include all these components. These components include “(a) identifying empirically valid risk factors, (b) determining a method for measuring (or ‘scoring’) these risk factors, (c) establishing a procedure for combining scores on the risk factors, and (d) producing an estimate of violence risk” (p. 39). Table 10.2 provides a summary of which risk-assessment approaches include these components. Some structured professional judgment measures, such as the Level of Service/Case Management Inventory, include all components, whereas others, such as the HCR-20, do not. Dr. R. Karl Hanson, the Canadian researcher profiled in Box 10.2, has done extensive research on the predictive validity of actuarial risk-assessment measures. Types of Risk Factors The risk assessments used by clinicians and researchers use various risk factors to predict antisocial and violent behaviour. A risk factor is a measurable feature of an individual that predicts the behaviour of interest, such as violence. Traditionally, risk factors were divided into two main types: static and dynamic. Table 10.2 Components Used across Risk-Assessment Approaches Components Unstructured Clinical Judgment Actuarial Structured Professional Judgment Identify risk factors No Yes Yes Measure risk factors No Yes Yes Combine risk factors No Yes Varies Produce risk estimate No Yes Varies Source: Based on Skeem and Monahan, 2011. © 2011. 300 M10_POZZ8067_05_SE_C10.indd 300 Chapter 10 25/11/16 11:50 AM Box 10.2 Canadian Researcher Profile: Dr. R. Karl Hanson Courtesy of Dr. R. Karl Hanson Dr. Karl Hanson’s research has focused on the assessment and treatment of serious offenders, particularly sexual offenders. Originally trained as a clinical psychologist at the University of Waterloo (under Donald Meichenbaum), Dr. Hanson’s early experience working in conventional mental health settings convinced him that much of what we label as psychopathology was intimately connected with how we have been treated by others. Individuals presenting with intense personal suffering often recounted harrowing stories of abuse, maltreatment, and victimization. Dr. Hanson quickly realized that “what happens to us matters.” He strongly believes if we are going to alleviate the burden of victimization, we need to understand why we do things that leave others with lasting psychic scars. When Dr. Hanson was forming his research agenda, the issue of sexual abuse was just entering public discourse in Canada. Whereas textbooks in the 1970s identified incestuous abuse as very rare (one in a million), surveys studies in the early 1980s found that it was actually common: rates of sexual victimization of girls during childhood were more like 1 in 10 or even 1 in 4. So, why was it so common? And what can we do about it? Dr. Hanson’s answers to these questions were deeply influenced by sexual offender researchers, such as Vernon Quinsey and William Marshall, was well as leaders in correctional rehabilitation, including Donald Andrews, Jim Bonta, and Paul Gendreau. Jim Bonta’s influence, in particular, was important (and unavoidable) given that Jim was his boss during the more than 20 years they both worked as researchers for the Government of Canada (Solicitor General/Public Safety Canada). Dr. Hanson’s most significant contributions have been in sexual offender risk assessment. His work has helped evaluators focus on risk relevant factors and combine these factors into an overall evaluation of risk. His STATIC suite of risk scales (e.g., Static-99R, Static-2002R) are by far the most commonly used sexual offender risk tools in the world. He has also helped shaped the research methods used in the field by teaching and providing accessible examples of useful statistics (e.g., meta-analysis, survival analysis) that had previously been rarely taught in psychology courses. Dr. Hanson is heartened by the substantial progress in the field of forensic and correctional psychology during the past 30 years. Whereas his clinical supervisors in the 1980s warned him that working with offenders was beyond the scope of reputable clinical practice (“nothing we can do for antisocial personality disorder”), there is now strong evidence that psychologists can identify and address the risk relevant propensities of individuals prone to crime and violence. “With the current generation building on what we have learned, and many highly talented individuals entering the field, I am confident that we can help more people stop hurting others”. Static risk factors are factors that do not fluctuate over time and are not changed by treatment. Age at first arrest is an example of a static risk factor, since no amount of time or treatment will change this risk factor. Static risk factors have also been called historical risk factors. Dynamic risk factors fluctuate over time and are amenable to change. An antisocial attitude is an example of a dynamic risk factor, since it is possible that treatment could modify this variable. Dynamic risk factors have also been called criminogenic needs (see Chapter 9 for a discussion). Risk Assessment M10_POZZ8067_05_SE_C10.indd 301 Static risk factor: Risk factors that do not fluctuate over time and are not amenable to change (e.g., criminal history) Dynamic risk factor: Risk factors that fluctuate over time and are amenable to change (e.g., antisocial attitude) 301 17/11/16 2:08 PM More recently, correctional researchers have begun to conceptualize risk factors as a continuous construct (Douglas & Skeem, 2005; Grann, Belfrage, & Tengström, 2000; Zamble & Quinsey, 1997). At one end of the continuum are the static risk factors described above. At the other end are acute dynamic risk factors. These risk factors change rapidly within days, hours, or minutes and often occur just prior to an offence. Factors at this end of the continuum include variables such as negative mood and level of intoxication. In the middle of the continuum are stable dynamic risk factors. These risk factors change but only over long periods of time, such as months or years, and are variables that should be targeted for treatment. These factors include criminal attitudes, coping ability, and impulse control. Recent research has found that dynamic risk factors are related to the imminence of engaging in violent behaviour. Quinsey, Jones, Book, and Barr (2006) had staff make monthly ratings of dynamic risk factors in a large sample of forensic psychiatric patients. Changes in dynamic risk factors were related to the occurrence of violent behaviours. Jones, Brown, and Zamble (2010) compared the predictive accuracy of risk ratings by researchers and parole officers in released offenders across three different time intervals (i.e., one, three, and six months). Both parole officers and researchers were moderately accurate, but it was the combination of time-dependent dynamic factors with static factors that showed the strongest predictive accuracy. Important Risk Factors Historical risk factors: Risk factors that refer to events that have been experienced in the past (e.g., age at first arrest). Also known as static risk factors Static risk factors: Risk factors that do not fluctuate over time and are not changed by treatment (e.g., age at first arrest). Also known as historical risk factors Dispositional risk factors: Risk factors that reflect the individual’s traits, tendencies, or styles (e.g., negative attitudes) Clinical risk factors: Types and symptoms of mental disorders (e.g., substance abuse) Contextual risk factors: Risk factors that refer to aspects of the current environment (e.g., access to victims or weapons). Sometimes called situational risk factors Situational risk factors: Risk factors that refer to aspects of the current environment (e.g., access to victims or weapons). Sometimes called contextual risk factors 302 M10_POZZ8067_05_SE_C10.indd 302 Since the late 1980s, a great deal of research has investigated what factors are associated with future violence. These can be classified into historical, dispositional, clinical, and contextual risk factors. Historical risk factors (sometimes called static risk factors) are events experienced in the past and include general social history and specific criminal history variables, such as employment problems and a history of violence. Dispositional risk factors are those that reflect the person’s traits, tendencies, or style and include demographic, attitudinal, and personality variables, such as gender, age, criminal attitudes, and psychopathy. Clinical risk factors are the symptoms of mental disorders that can contribute to violence, such as substance abuse or major psychoses. Contextual risk factors (sometimes referred to as situational risk factors) are aspects of the individual’s current environment that can elevate the risk, such as access to victims or weapons, lack of social supports, and perceived stress. Some of these factors are likely relevant to risk assessment only, while others are relevant to both risk assessment and risk management. These factors vary in terms of how much they are subject to change. For example, some are fixed (e.g., gender), some cannot be undone (e.g., age of onset of criminal behaviour), and some may be resistant to change (e.g., psychopathy), whereas others (e.g., social support or negative attitudes) may be subject to intervention or may vary across time. Several meta-analytic reviews have examined the predictors of general and violent recidivism in adult offenders, sexual offenders, and patients with mental disorders (Bonta et al., 1998; Gendreau, Little, & Goggin, 1996; Hanson & Morton-Bourgon, 2005). Two key findings have emerged. First, factors that predict general recidivism also predict violent or sexual recidivism. Second, predictors of Chapter 10 28/09/16 7:23 PM recidivism in offenders with mental disorders overlap considerably with predictors found among offenders who do not have a mental disorder. More recently, attempts have been made to determine if there are unique risk factors for various sub-populations including women offenders, aboriginal offenders, and terrorists. See Box 10.3 for a description of risk factors regarding violent extremism. Dispositional Factors Demographics Researchers in the 1970s identified young age as a risk factor for violence (Steadman & Cocozza, 1974): The younger the person is at the time of his or her first offence, the greater the likelihood that the person will engage in criminal behaviour and violence. Dozens of studies have firmly established age of first offence as a risk factor for both general and violent recidivism in both offenders with mental disorders (Bonta et al., 1998) and offenders without mental disorders (Gendreau et al., 1996). Offenders who are arrested prior to age 14 tend to have more serious and more extensive criminal careers than those who are first arrested after age 14 (DeLisi, 2006; Piquero & Chung, 2001). Males are at higher risk than females for general offending (Cottle, Lee, & Heilbrun, 2001; Gendreau et al., 1996). Notably, males engage in more serious violent acts, such as sexual assaults, homicides, and assaults causing bodily harm (Odgers & Moretti, 2002). Some studies using self-report measures have found that females engage in similar or even higher rates of less serious violence (Nichols, Graber, Brooks-Gunn, & Botvin, 2006; Steadman et al., 1994). Personality Characteristics Two personality characteristics have been extensively examined: impulsiveness and psychopathy. Not being able to regulate behaviour in response to impulses or thoughts increases the likelihood of engaging in crime and violence (Webster & Jackson, 1997). Lifestyle impulsivity (being impulsive in most areas of life) distinguishes recidivistic rapists from non-recidivistic rapists (Prentky, Knight, Lee, & Cerce, 1995). Psychopathy is a personality disorder defined as a callous and unemotional interpersonal style characterized by grandiosity, manipulation, lack of remorse, impulsivity, and irresponsibility (see Chapter 11 for more information on psychopaths and the Hare Psychopathy Checklist–Revised [PCL-R], the most widely used measure of psychopathy). Given these features, it is not surprising that psychopathic individuals engage in diverse and chronic criminal behaviours. A recent meta-analysis found that psychopathy is moderately related to general and violent recidivism (Leistico, Salekin, DeCoster, & Rogers, 2008) and moderately related to violence in a prison setting (Guy, Edens, Anthony, & Douglas, 2005). Psychopathy predicts reoffending across different countries, such as Canada, the United States, the United Kingdom, Risk Assessment M10_POZZ8067_05_SE_C10.indd 303 Psychologist conducting a risk-assessment interview with a young offender Lisa F. Young/Shutterstock 303 28/09/16 7:23 PM Box 10.3 Forensic Psychology in the Spotlight Predicting Terrorism: Are There Unique Risk Factors? Although there have been tremendous advances in predicting general violence, the attempt to develop risk-assessment instruments for terrorism is still in its infancy. Whether or not the same risk factors identified for general violence are also valid risk factors for terrorism remains unknown. In a review of the literature, Monahan (2012) identified the following risk factors for politically motivated violence: ■■ Age: The average age of violent terrorists is between 20 and 29. ■■ Gender: Most terrorists are male; however, there is gender variation across terrorist groups. For example, few al-Qaeda terrorists are female, but about half of Chechen and Kurd terrorists are female. Female terrorists are often able to get closer to targets than male terrorists can (Berko, Erez, & Globokar, 2010). ■■ Marital status: Most terrorists are single. This is especially true of suicide terrorists (Merari, 2010). ■■ Social class: Terrorists are not primarily from lower social classes, but rather are representative of the local population they come from. One study found that suicide terrorists are more highly educated compared to the population they come from (Merari, 2010). ■■ Prior crime: Many terrorists do not have a record of past violent criminal behaviour (Merari, 2010). Whether or not past violent terrorism predicts future terrorism is unknown since so few terrorists have been released. According to Bergen, Tiedemann, and Lebovich (2011) only 6% of released Guantanamo detainees committed or were suspected of committing terrorism. ■■ Suicidality: Although only a small number of terrorists commit suicide terrorism, these individuals differ substantially from people who commit suicide in the general population (Merari, 2010; Townsend, 2007). ■■ Major mental illness: There are very low rates of major mental illness such as bipolar disorder or schizophrenia in terrorists. ■■ Substance abuse: Although terrorist organizations often engage in drug trafficking to help finance their illegal activities, substance abuse in terrorists is very rare (Merari, 2010). ■■ Psychopathy: The rate of psychopathy among terrorists appears to be relatively low (Borum, 2011; Merari, 2010). Most psychopaths show little commitment to ideology and would be unlikely to sacrifice themselves for others. 304 M10_POZZ8067_05_SE_C10.indd 304 However, many individuals who commit general violence are young, male, and single; this leaves no risk factors that may uniquely contribute to risk amongst violent extremists. As Monahan (2012) stated, “Terrorists in general tend not to be impoverished or mentally ill or substance abusers or psychopaths or otherwise criminals; suicidal terrorists tend not to be clinically suicidal” (p. 179). In other words, violent extremists do not tend to have the same risk factors as the general offender population. Indeed, it has been indicated that commonly used risk factors are equally present as not present among those who have committed acts of terrorism (e.g., educated versus non-educated; de Mesquita, 2005). Recently, researchers have begun to identify risk factors that might uniquely apply to this sub-population. Malik, Sandholzer, Khan, and Akbar (2015) surveyed Pakistani security officials in order to establish risk factors related to terrorism, and identified 13 critical factors falling within five categories: 1. Social wellbeing (e.g., inequality, non-availability of basic facilities, and a gathered population) 2. Economic indicators (e.g., unemployment and a higher consumer price index) 3. Governance (e.g., dishonest leadership, unjust or unfair accountability system, corruption, improper political process development, underprivileged state of sovereignty, and non-inclusiveness of public wishes in policy making) 4. Law enforcement (e.g., poor judicial system and poor state of national forces) 5. Armed conflict (e.g., unobserved presence of noncombatant foreigners) Furthermore, sympathies and justifications for violent extremism have been noted amongst those who are politically influential and amongst those who have experienced acts of discrimination (Bhui, Warfa, & Jones, 2014; Victoroff, Adelman, & Matthews, 2012). While research is beginning to emerge on risk factors that might contribute to acts of terror, there are still considerable gaps in the literature that must be addressed before we are accurately able to predict risk for violent extremism. It is important to understand the underlying factors that contribute to acts of terror in order to accurately predict their likelihood. As stated by Borum (2015), “. . . a robust empirical foundation does not yet exist for understanding the risk of terrorism or involvement in violent extremist activity” (p. 1). Chapter 10 28/09/16 7:23 PM Belgium, Germany, the Netherlands, New Zealand, and Sweden (Hare, 2003), in both male and female offenders (Richards, Casey, & Lucente, 2003), offenders with mental disorders (Nicholls, Ogloff, & Douglas, 2004; Steadman et al., 2000; Strand, Belfrage, Fransson, & Levander, 1999), male adolescent offenders (Corrado, Vincent, Hart, & Cohen, 2004; Forth, Hart, & Hare, 1990; Gretton, Hare, & Catchpole, 2004; Murrie, Cornell, Kaplan, McConville, & Levy-Elkon, 2004), and sexual offenders (Barbaree, Seto, Langton, & Peacock, 2001; Rice & Harris, 1997a). However, psychopathy may be weakly related or unrelated to violent reoffending in adolescent females (Odgers, Repucci, & Moretti, 2005; Schmidt, McKinnon, Chattha, & Brownlee, 2006; Vincent, Odgers, McCormick, & Corrado, 2008). Several studies have found that the combination of psychopathy and deviant sexual arousal predicts sexual recidivism (Hildebrand, de Ruiter, & de Vogel, 2004; Olver & Wong, 2006; Rice & Harris, 1997a). Deviant sexual arousal is defined as evidence that the sexual offender shows a relative preference for inappropriate stimuli, such as children or violent nonconsensual sex. For example, Rice and Harris (1997a) found that about 70% of sexual offenders with psychopathic features and evidence of deviant sexual arousal committed a new sexual offence, compared with about 40% of the other offender groups. Historical Factors Past Behaviour The most accurate predictor of future behaviour is past behaviour. Past violent behaviour was first identified as a predictor in the 1960s and 1970s (see Cocozza, Melick, & Steadman, 1978) and has consistently been associated with future violence in diverse samples, including adolescents and adults, correctional offenders, mentally disordered offenders, and civil psychiatric patients (Farrington, 1991; McNiel, Sandberg, & Binder, 1988; Phillips et al., 2005). Interestingly, it is not only past violent behaviour that predicts violence, but also past nonviolent behaviour (Harris, Rice, & Quinsey, 1993; Lipsey & Derzon, 1998). For example, offenders who have a history of break and enter offences are at an increased risk for future violence. Age of Onset As noted earlier, individuals who start their antisocial behaviour at an earlier age are more chronic and serious offenders (Farrington, 1991; Tolan & Thomas, 1995). For example, Farrington (1991) found that 50% of the boys in his study who committed a violent offence prior to age 16 were convicted of a violent offence in early adulthood. In another longitudinal study, Elliott (1994) reported that 50% of male youth who committed their first violent acts prior to age 11 continued their violent behaviour into adulthood, compared with 30% whose first violence was between the ages of 11 and 13, and only 10% of those whose first violent act occurred during adolescence. Age of onset is not as strong a predictor for female offenders (Piquero & Chung, 2001). Childhood History of Maltreatment Having a history of childhood physical abuse or neglect is associated with increased risk for violence (Smith & Thornberry, 1995; Zingraff, Leiter, Johnsen, & Myers, 1994). In a large-scale study of childhood abuse, Widom (1989a) reported that victims of sexual abuse were no more likely than those Risk Assessment M10_POZZ8067_05_SE_C10.indd 305 305 28/09/16 7:23 PM who were not sexually abused to commit delinquent or violent offences. Those who were victims of physical abuse or neglect were much more likely to commit criminal acts as compared with those who were not abused. Being abused in childhood predicts initiation into delinquency, but continued abuse predicts chronic offending (Lemmon, 2006). Physical abuse in adolescence is also directly related to adolescent offending and may be related to some types of offending in adulthood as well (Fagan, 2005; Smith, Ireland, & Thornberry, 2005). Clinical Factors Substance Use Drug and alcohol use has been associated with criminal behaviour and violence. However, the drug–violence link is complex because of both direct effects (e.g., the pharmacological effects of the drugs) and indirect effects (e.g., the use of violence to obtain drugs; Hoaken & Stewart, 2003). The obvious link between drugs and crime is that the use, possession, and sale of illegal drugs are crimes. In some cases, the individual commits offences to support a drug habit (Klassen & O’Connor, 1994). For example, Chaiken and Chaiken (1983) found that severe drug users commit 15 times as many robberies and 20 times as many burglaries as nondrug-using offenders. The drug that has been most associated with crime is heroin (Inciardi, 1986). A l