A Practical Guide to Risk Assessment

Mary Alice Conroy, Ph.D.

Diplomate in Forensic Psychology

American Board of Professional Psychology

Directions: To receive 1 hour ethics continuing education credit for psychologists, licensed psychological associates, licensed professional counselors and licensed social workers, for this TPA sponsored home study assignment, you must:

1)Read the article in its entirety;

2)Take the test at the end of the article;

Mail or fax the test answers along with $25 (TPA Members) or $50 (Non-TPA Members) to the Texas Psychological Association, PO Box 1930, Cedar Park, TX78630 or (512) 255-1642. Examinations will not be scored without appropriate fee. You must answer correctly 70% or better to pass the examination.

Risk assessment is the contemporary term for an estimation of the level of risk an individual presents for certain types of violent behavior. In Texas, assessment of risk is required for traditional civil commitment, sexual predator evaluations, sentencing phase evaluations in death penalty cases, evaluations for juvenile transfer to adult court, and various decisions regarding detention of those found not restorable to competence for trial and not guilty only by reason of insanity. Risk assessment may also be important in cases involving probation or parole decisions, the use of diversion programs, domestic violence intervention, termination of parental rights, fitness for duty in potentially hazardous occupations, and certain types of malpractice litigation. This chapter is intended to impart very practical information regarding what courts have said about risk assessment, issues in deciding to perform a risk assessment, essential data collection, methodology, and issues of risk management.

The View from the Bench

Courts have generally relied on mental health professionals to provide information regarding the risks individuals pose to society. In 1983, this practice was challenged in the case of Barefoot v. Estelle. Mr. Barefoot, sentenced to death for murder in a Texas court, contended that psychiatrists have no unique tools or ability to predict dangerousness. An amicus curiae brief filed by the American Psychiatric Association supported this position. Data were presented suggesting that these predictions were wrong as often as two thirds of the time. However, the U. S. Supreme Court ruled that not allowing psychiatrists to testify in this regard would be tantamount to “disinventing the wheel.” The justices went so far as to note that psychiatrists were not always wrong – only wrong most of the time.

Since Barefoot, both psychiatrists and psychologists have continued to provide assessments of violence potential to the judiciary. Thankfully, the scientific data relevant to such predictions has grown and the standards for admitting expert testimony have tightened. In the case of Daubert v. Merrill Dow Pharmaceuticals (1993), the U. S. Supreme Court declared the judge to be the “gatekeeper” to assure that experts who testified were, in fact, experts, and that their evidence was scientific. An expert was to be qualified by knowledge, skill, experience, training, or education – and not simply by degree or profession. The Court suggested criteria a judge might apply in assessing the evidence presented and methodology employed by experts:

  1. Does it reflect a theory that has or can be tested?
  2. Has it been subject to peer review or publication?
  3. Has a potential error rate been established?
  4. Has it been accepted by the relevant scientific community?

In 1995, Texas courts embraced the essence of Daubert in the case of E. I. Dupont & Co.v. Robinson. Since that time, so-called “Daubert challenges” have become common in hotly contested cases and “Daubert hearings” are sometimes held before judges prior to allowing the expert to testify before the jury.

In the 1999 case of Kumho v. Carmichael, the U. S. Supreme Court made an additional ruling in this regard. In this case the Court said that not only the presentation of scientific data was at issue, but, if an expert performed some type of evaluation, the methodology employed was subject to scrutiny.

The Decision to Conduct a Risk Assessment

Before agreeing to conduct a risk assessment, several issues should be clarified.

  1. Over what period of time is the risk to be assessed? (This could be a lifetimeor it could be the time an individual will be on bond.)
  2. Is the risk in a structured or unstructured environment? (This is a criticaldistinction in a death penalty evaluation.)
  3. What will be the individual’s likely circumstances?
  4. Is there leeway to alter these circumstances (e.g., can conditions be placedupon the release)?
  5. Is the decision-maker concerned only about a particular type of violence (e.g.,sexual violence, spousal abuse, child abuse)?
  6. Is the issue confined to physical violence or is it intended to mean criminal recidivism in general?
  7. Must the particular risk be linked to a mental illness or mental abnormality? Once circumstances have been fully clarified, the evaluator should give careful consideration to potential ethical pitfalls. First and foremost the mental health professional needs to examine whether he or she has the specific competencies needed for the evaluation. Risk assessment is becoming very specialized and an extensive literature has developed. (A brief list of major references is included in Appendix A.) Instruments have been developed to assess risk for specific types of violence in target populations and any evaluator should be familiar with these tools. (A brief catalogue is included in Appendix B for reference.) Cultural competence should be considered in deciding whether the clinician is sufficiently experienced with the population represented. If the mental health professional is uncertain regarding his or her expertise, consultation is strongly recommended.

Ethical guidelines require that evaluators avoid a conflict of interest or a potentially harmful dual relationship. One of the most common problems in this area stems from mental health professionals serving as forensic evaluators while also functioning as the subject’s therapist. Forensic risk assessments generally should not be performed by the treating professional. A therapist should maintain an alliance with the patient, while a forensic evaluator should be unbiased.

Collecting Research and Collateral Data

Prior to contact with the person to be evaluated, both nomothetic and idiographic data can be assembled. This should include relevant research data on factors predictive of the particular type of risk being assessed in the particular population represented by the individual. Factors with protective value should not be overlooked. The data base on risk assessment is exploding exponentially, and the clinician should actively seek the most current literature or risk serious embarrassment in the courtroom. Websites are now available devoted to various areas of risk assessment. Key sites to explore would include, but are not limited to:

  1. (the mentally ill offender)
  2. (the psychopathic offender)
  3. (the sex offender)

Exploring idiographic data involves seeking out collateral information about the individual to be assessed. Collateral information in performing a risk assessment is not simply helpful – it is essential! Much of the risk assessment research has been conducted in Canada – one reason being that the Canadians have developed an extensive record keeping system. Securing records in Texas may be more challenging. To perform a valid assessment there must be sufficient information to verify key facts. Official documents may contain errors and contradictions should be clarified. Attorneys may wish to be helpful, but are frequently unaware of what is needed. Making specific requests for documents and suggesting sources can be productive. Examples of records and possible sources would include:

  1. Presentence investigations and reports of prior functioning while undersupervision (probation officer)
  2. Information, indictment, police reports, witness statements for current andpast offenses (district attorney)
  3. Reports of institutional behavior (TDCJ, TYC, county jail)
  4. Past criminal record (district attorney, probation officer)
  5. Report of past mental health treatment or evaluations (defense attorney,hospitals, clinicians)
  6. Reports of special programs attended, such as substance abuse or sex offendertreatment (defense attorney, TDCJ, TYC, program providers)
  7. School records, especially in the case of youthful offenders (defense attorney)
  8. Collateral interviews with family members or care-givers may be helpful (generally check with the attorney before proceeding)

Methodology

Clinical and Actuarial Methods

A debate continues to rage in the field between those who favor a strictly actuarial approach and those who defend the role of clinical judgment. Vernon Quinsey and his colleagues, who have done some of the major research in the development of actuarial approaches, have argued that clinical judgment should be completely supplanted by actuarial instruments, because actuarial approaches are too good to risk contaminating them with unreliable clinical judgment (Quinsey, Harris, Rice, & Cormier, 1998). Onthe other hand, Thomas Litwack has argued vociferously that the day will probably never come when actuarial instruments can be applied to single individuals without the careful application of the clinician’s judgment (Litwack, 2001). Karl Hanson (1998) holds amore conciliatory view, suggesting the application of a guided actuarial approach, adjusting the weight given to both as the research becomes more sophisticated. John Monahan and his colleagues in the MacArthur risk assessment studies have generally supported this view, arguing that it clearly cannot be said that all of the variables potentially related to risk have been uncovered nor that they should be given the same weight for every individual (Monahan et al., 2001). It is even possible that well researched variables involving violence may have paradoxical effects for a specific person.

Beyond Traditional Approaches

Consent or disclosure. In clinical work, the informed consent of the patient orguardian is often the first essential step. However, if an evaluation is court-ordered or required by statute, the individual may have few choices – the evaluation must be completed with or without consent. In such cases an evaluation disclosure may be more honest and appropriate than an informed consent. If, on the other hand, the evaluation is being conducted at the request of an attorney or probation officer, the person’s informed consent will be needed. In either case, the individual under evaluation should be informed of the identity and professional affiliation of the evaluator, the purpose of the evaluation, the limits of confidentiality, who is likely to have access to the evaluation results, the potential consequences of the evaluation, and given a clear statement that the evaluator is not a treating clinician.

The interview. Ordinarily, ethical standards require that an interview of theindividual under evaluation be conducted as part of the assessment. It is certainly themost desirable and defensible approach. However, there are circumstances when aninterview is not a viable option – either because the individual is incapable of participating, is uncooperative, or is advised by his or her attorney not to speak. Extensive, reliable records may nonetheless make it possible to conduct a thorough risk assessment. However, the reverse is not the case – one cannot complete a risk assessment relying only on the individual’s self-report. If an interview is not conducted this should be clearly explained in subsequent reports or testimony.

Traditional psychological test batteries. The traditional psychological test battery, often composed of personality and cognitive testing, may be of limited value in a risk assessment. Such instruments may be helpful in confirming a particular diagnosis, if one is required. However, broad, general personality inventories (e.g., the MMPI-2, the MCMI-III, the PAI) have not been validated to predict violence, either general or specific. Problems arise when forensic evaluators confirm long-held “myths” about test results or suggest illusory correlations.

Specialized risk assessment instruments. Recently, some instruments have been developed specifically designed to predict violence risk. These are generally constructed in one of three formats: 1) the structured interview, 2) an actuarial device consisting of variables chosen based upon regression analysis, and 3) the decision-tree approach. (See Appendix B.)

Instruments of this type have some distinct advantages in the courtroom. They can provide an anchor onto which an evaluator may graft more idiographic data. Such instruments have almost invariably been published and peer reviewed, which can be helpful if faced with a Daubert challenge. It is often possible to grossly compare offenders by level of risk using these methods. For example, the Violence Risk Appraisal Guide (VRAG) divides the population into nine risk levels ranging from the highest (in which virtually the entire sample re-offended) to the lowest (in which almost none of the norm group re-offended.)

However, the limitations of these devices must also be considered. For valid results, they universally require accurate, extensive records. Courts are sometimes confused by statistical information or may put unwarranted faith in numbers. Selection and proper use of a risk assessment instrument requires a thorough understanding of certain statistical concepts (e.g., ROC curves, sensitivity, specificity, positive predictive power.) Each instrument has been normed on a select population that may not match the individual being evaluated. Finally, most rely almost exclusively on static variables (things that are historical and will not change with time.) Thus the instruments themselves tend to be static and are insensitive to change over time or to varying levels of immediate risk.

Diagnosis. A precise diagnosis may be helpful (or even required) in the process of a risk assessment; however, diagnoses have very limited value in actually assessing risk. The limitation of diagnoses include two specific caveats noted in the DSM-IV-TR.1) Clinical diagnoses often do not equate to what the legal community views as mental disorder or mental abnormality, and 2) No diagnosis carries with it specific implications regarding degree of behavioral or volitional control.

Current research does provide support for some basic premises regarding mental disorder and risk for violence. Some relationship between mental disorder and violence has been found to exist (Monahan & Steadman, 1994). Substance abuse is a critical factor to be considered and generally more predictive of violence than major mental illness (Steadman, Mulvey, Monahan, Robbins, Appelbaum, Grisso, Roth, & Silver, 1998). A personality disorder diagnosis is more predictive of violence than any of the major mental illnesses (Hodgins, 2000). Base rates of violence within any diagnostic category tend to be low (Swanson, Borum, Swartz, & Monahan, 1996). Special attention has been given to delusions, particularly those with paranoid content – also referred to as “threat control override” variables (Appelbaum, Robbins, & Monahan, 2000; Appelbaum, Robbins, & Roth, 1999; Monahan & Steadman, 1994). Hallucinations, particularly command hallucinations, have been studied with mixed results (Hersh & Borum, 1998; McNeil, Eisner, & Binder, 2000; Monahan & Steadman, 1994; Rudnick, 1999).

For a diagnosis to be relevant to a risk assessment the connection between the psychopathology and violence displayed by the individual must be explored. Critical questions would include: 1) Was the individual’s behavior consistent with a delusional belief? 2) Is there evidence the person was experiencing command hallucinations concurrent with past violence? 3) Were there obvious motives for the violence unrelated to the mental abnormality? 4) Did all of the individual’s past violence seem to stem from the mental disorder? 5) To what extent did substance abuse contribute to the events?

Psychopathy. Although not a diagnostic category included in the DSM-IV-TR,

the construct of psychopathy has been found to have strong positive predictive power relative to future violent behavior. Psychopathy must be clearly distinguished from Antisocial Personality Disorder. It is a much more restrictive category and goes back to Hervey Cleckley’s conceptualization of the “as if” personality. Robert Hare (1991, 1996, 1999) has devoted much of his long career to refining and clarifying the construct.

In study after study, psychopathy has been found to be among the best predictors of future violent behavior (Hemphill, Hare, & Wong, 1998; Salekin, Rogers, & Sewell, 1996; Serin & Amos, 1995). Correlates of psychopathy include both general and violent recidivism, institutional maladjustment, and poor treatment response. Meta-analyses of recidivism studies have shown psychopathic offenders to be three times as likely to be reconvicted and four times as likely to commit a violent reoffense (Hemphill et al., 1998). The MacRisk Study found psychopathy more predictive of violence than any of the other 133 variables included in the study (Skeem & Mulvey, 2001). Psychopathy has been found to be predictive of violence even when diagnosed in conjunction with substance abuse or a major mental illness (Hill, Rogers, & Bickford, 1996; Rice & Harris, 1992).

The Psychopathy Checklist—Revised (PCL-R), developed by Robert Hare, is currently the “gold standard” in assessing for psychopathy, but care should be exercised in its use. Specialized training is strongly recommended for evaluators using the PCL-R. It is essential that the person being assessed is representative of the populations on which the instrument has been researched. Although predictive of general institutional maladjustment, the PCL-R has not been demonstrated to predict risk for violence within a structured environment. Finally, a low PCL-R score does not indicate low risk – only that it is one risk factor not present.

The situation. Level of risk may vary markedly with the situation. It is important to determine the period of time over which one is expected to assess risk and whether future circumstances are likely to be similar to those in which past violence occurred. The degree of available structure and support should be considered. Access to potential victims may make a difference. Finally, the availability of effective treatment and the likelihood the individual will comply with treatment should be assessed.

Risk Management

Beyond risk assessment, the forensic evaluator may be called upon to establish specific plans for the management of risk. In such cases, the goal of treatment in this context is not reduction of the person’s distress, but rather the reduction of risk. Well applied, this concept stresses the protection of society along side the least restrictive environment for the offender.

The task is challenging, and several principles of sound risk management have been developed to assist the forensic evaluator:

1)Risk factors should be identified and linked specifically to the risk management strategies proposed.