Predicting violent behavior among inmates: Washington Correctional Institute's development of a risk protection tool.
The prediction of violent behavior among inmate populations is of obvious value not only for prison officials, but also for those who will supervise offenders once they are released into the community. It is corrections officials' primary duty to provide public safety, as well as a safe environment for inmates and staff inside correctional facilities. In order to succeed in this important mission, correctional staff need ways to determine the prior history of inmates and predict the likelihood of future violent and criminal behaviors. Researchers Paul Gendreau, Claire Goggin and Thomas Little (1) noted that efficient management of prisons is contingent on knowledge about the predictors of criminal behavior.
The development of risk prediction tools for specialized populations recently has become an area of intensive study, with tools being developed for various populations including adults, children, sex offenders, the mentally ill, the mentally retarded and those who have suicidal ideation. In August 2002, permission was requested from and granted by the administration of the Washington Correctional Institute (WCI) and the Louisiana Department of Public Safety and Corrections to conduct a research study that would help develop a violence risk prediction tool to use with the inmate population.
The proposed research project involved only a record search on a randomly selected group of inmates. Research data were intended only to develop the tool. There was no identified danger, harm or risk to inmates from participation in the study. WCI is a medium-security state prison in southeast Louisiana housing adult male inmates. As the state prison closest to New Orleans, and with a largely urban inmate population, the facility's population is representative of the incarcerated population in Louisiana as a whole.
A risk prediction tool for the adult incarcerated population has been developed from the above mentioned study by the author of this article and has been remarkably accurate in predicting which inmates will have disciplinary problems. It is useful for prison administrators, classification staff and others who make important decisions about housing and work assignments, and which inmates will be made trustees.
The tool determines if inmates have a low, medium or high risk of behavioral issues. A study at WCI of more than a year revealed that high-risk inmates received more than 86 percent of rule violation reports, while low-risk and medium-risk inmates received 4 percent and 9 percent of rule violation reports respectively.
RISK PREDICTION HISTORY
Although the field of risk prediction is new, it is quickly evolving. The history of risk prediction is fraught with errors and misconceptions about what factors make one person more likely to be at risk for violence. Early efforts toward risk prediction relied solely on clinical judgment of professional treatment staff. However, this approach has been largely discredited, as clinical judgment alone has proved to have an accuracy rate of less than 50 percent. The problem seemed to be based on misunderstanding the motivations of committing acts of violence and violence itself. Clinicians were forced to rely almost entirely on information that was self-reported by an offender--an approach almost doomed to fail among a population with a vested interest in appearing as low risk.
Once the failure of clinical judgment became evident, second-generation risk prediction developed. It involved exclusively relying on static, historical or unchangeable information such as inmates records. This approach was centered on the concept that in order to predict future behavior, past behavior should be expected to continue. While these tools were somewhat better than clinical judgment, they still fell short of the needs of the correctional environment. Predictive rates of between 50 percent and 70 percent were common. However, this rate meant that the correct prediction of violence was somewhere between flipping a coin to slightly better.
Researchers studying the field finally turned to statistical analysis to determine the best way to predict violence. A landmark meta-analysis published by Karl Hansen and Monique Bussiere (2) reviewed numerous studies on sexual recidivism to find correlations between risk factors and risk of recidivism. Gendreau, Goggin and Little (3) proposed that any measurement tool should include both static and dynamic predictors. This approach was echoed by Andrews and Bonta, and Hansen and Bussiere. (4)
Third generation prediction instruments have added dynamic changing factors to the static factors previously examined and give a clearer picture of the offender's actual behavior history and future. They also provide much better predictive accuracy and validity. A number of these tools are under development at various places for specific applications, including sex offense, domestic violence, recidivism and juvenile violence prediction.
One area that has not generally been addressed is the development of a predictive instrument for correctional staff. Corrections officials make many management decisions about offenders in custody, however, they are not able to predict violent behavior. This point was driven home in an incident that occurred at a Louisiana state prison when an inmate who had been made a trustee attacked an officer with a hammer, escaped and eventually committed suicide. Could such a tragedy have been avoided?
The risk prediction screening tool that has been developed at WCI is designed to address this specific need. It examines both static and dynamic risk factors associated with risk of violence and recidivism, adding in some factors that specifically pertain to the incarcerated. At the same time, such a prediction tool must be simple and efficient so institutional staff who have minimum training can use it. Further, the tool needs to give consistent and accurate results. The design of the tool relies on a selected group of factors to make predictions.
WCI's RISK PROTECTION TOOL STUDY
A random sample of identification numbers was pulled from the WCI inmate population in August 2002. For the purposes of this screening, 50 identification numbers were selected. Of the original sample of 50, one was a duplicate number. This left 49 in the sample population. From a population of 1,132, the 49 inmates represented 4.3 percent of the prison population. As is expected, the sample size diminished as some of the original sample group were discharged or transferred from WCI.
The group sampled was screened February 2003, approximately six months into the study, and a year later. At the time of the final screening, 25 members of the original sample were still housed at WCI. The results from those 25 individuals, as well as the other results from the first screening, are considered in the study results.
The screening instrument consists of 15 questions that could be answered through a review of the subjects' records. The questions are divided into five basic areas, which include criminal history, social skills, activities/companions, substance abuse and emotional/personal attitudes. These areas are supported in research as the main areas of concern for risk of violence.
Each of the questions include a scoring guide. Adding the scoring numbers gives a risk rating, which can be compared with a suggested risk scale to determine a general level of risk for violence or behavioral problems. The scale consists of three levels: low risk, medium risk and high risk. There also is a space for other factors that can be considered in assigning a level of risk for a particular individual. The questions are as follows:
Does the offender have a history of criminal offenses in his record? The first question examines the number of arrests listed regardless of disposition. The number is easy to find based on criminal record checks, which are included in all prison master records.
Does the offender have a history of convictions for offenses in his record? The second question is related to the first, but considers disposition of the charges filed. Prior history of criminal behavior and violence is considered the highest predictor of recidivism, and so the questions related to prior behaviors are heavily weighted in this study.
Does the offender have a record of juvenile offenses? Juvenile offenses followed by adult offenses generally are accepted as an indicator of long-term behavioral patterns. Negative patterns of behavior developed in adolescence that are not corrected early are highly likely to continue in adulthood. This information can be difficult to access, as juvenile records often are sealed.
Does the offender have a record of escapes from correctional facilities or authorities? Escaping from corrections or law enforcement indicates poor judgment and low self-control and is indicative of antisocial personality traits, as the person attempting to escape is trying to evade responsibility for actions. Caution should be used in searching for escape charges. Charges such as resisting arrest by flight or with force should also be considered.
Does the offender have a history of multiple serious institutional rule violation reports (write-ups) within the last two years in his record? Just as historical factors involving behaviors on the street are indicative of future behavior, so are behaviors during incarceration. However, only multiple serious violations are considered predictive. A one-time violation may be dismissed as an aberration rather than an indication of a pattern of behavior. Also, multiple disciplinary problems during incarceration indicate resistance to changing problematic behaviors and resistance to authority. The information required is available in the inmate's disciplinary record.
Has the offender failed in the past to complete supervision (violated supervision)? Failure to complete treatment or supervision has been closely identified with risk to re-offend in multiple studies, especially those reviewing risk of sexual re-offending. Indeed, the meta-analysis developed by Hansen and Bussiere (1996-2000) found that failure to complete treatment was one of the highest correlating factors for sexual re-offending. The same pattern also is believed to be true for general violence. This information is available either in the inmate's master record or in a treatment record.
Does the offender have a record of assaults or violence involving injury to a victim? Previous violence is considered an excellent predictor of future violence. It typically indicates not only willingness to commit acts of violence and the associated callousness necessary for these acts, but also an antisocial orientation. The question of violence is reflected in the kinds of criminal charges filed as well as behavioral record at the institution.
Does the offender have trouble getting along with inmates? Those with antisocial personalities and those who oppose authority often have serious problems dealing with others. Further, problems relating even to other inmates is indicative of a lack of any adequate support system that might otherwise provide some motivation and support for behavior change. Multiple protection requests, fights, requests for housing changes or transfers would be considered indicative of problems relating to fellow inmates for the purposes of this tool.
Does the offender have trouble getting along with officers? Some offenders only have problems getting along with authority figures. This is problematic because while incarcerated, the offender will be required to submit to the authority of officers and other institutional staff. An inmate who has great difficulty getting along with these authority figures may turn to violence against either the authority figure or other inmates to relieve perceived stress. The inmate's disciplinary record will reveal these kinds of problems.
Has the offender completed any treatment programs while incarcerated? As mentioned earlier, completion of treatment programs is an excellent predictor of reduction in the risk of continued inappropriate behaviors, including violence. This is one of only two questions in this survey in which the subject can actually lower his score by proactively engaging in appropriate behaviors. The information requested here is generally found either in the inmate's master record or in treatment records. Prior successful completion of treatment programs on the street might also be considered in scoring this item.
Has the offender been involved in any organized groups or activities while incarcerated? Treatment programs that are self-help programs also can be considered a developer of an appropriate support system that can help deter violence by modeling and instill values that support pro-social behaviors and problem solving rather than violence.
Does the offender have alcohol or drug problems? The use of illegal substances is widely associated with criminal behavior and violence. The research is replete with studies linking the two. The inmate's criminal record and intake screening will generally reveal substance abuse-related problems.
Does the offender have indicated emotional or psychological problems? While relatively a lower predictor of violence or recidivism, there are links between mental illness or depression and self-control and judgment. Antisocial personality disorder specifically is identified as a common attribute among violent people. The information is normally available as part of the inmate's intake screening. Care should be taken to ensure that the sections of the inmate record, which are protected by confidentiality, are actually protected.
Does the offender have attitudes generally supporting of crime or a criminal lifestyles? The antisocial personality is considered the most dangerous of the personality disorders due to specific risk of violence and behaviors that are injurious to others. The determination of attitudinal orientation is best made by treatment staff. However, other documentation in the inmate's record may reveal such attitudes, including blame-avoidant statements and behaviors.
Does the offender have a poor attitude toward supervision? Similar to the question about problems with authority figures, this item examines the inmate's cooperativeness with supervising officials. Does the inmate go to work as ordered? Does the inmate complain constantly about officers in general? Does the inmate try to get away with or justify inappropriate behaviors? Has the inmate violated his supervision in the past?
The other factors category is used to address other factors that are evident but not specifically mentioned in the instrument. These are too varied to list individually, but generally include anything that the tool administrator feels is significant. Examples might include antisocial personality disorder or behavior patterns that create problems but that do not rise to the level requiring disciplinary action.
The individual questions are scored the following way:
Does the offender have a history of criminal offenses in his record? One to two indicated arrests is worth one point. Three to five arrests is worth three points. Six or more arrests is worth five points. No arrests is worth zero points.
Does the offender have a history of convictions for offenses in his record? One to two convictions is worth one point. Three to five convictions is worth three points. Six or more convictions is worth five points. No convictions is worth zero points.
Does the offender have a record of juvenile offenses? A "yes" answer is worth two points, while a "no" answer is scored zero.
Does the offender have a record of escapes from correctional facilities or authorities? A "yes" answer is worth five points, while a "no" answer is worth zero points. Any charged offense of escape is counted, whether from correctional institutions or from law enforcement.
Does the offender have a history of multiple serious institutional rule violation reports within the last two years in his record? "Yes" answers are worth two points, while "no" answers are worth zero points. In general, only the more serious offenses are counted, unless there are many less serious offenses indicated in the offender's record.
Has the offender failed in the past to complete supervision (violated supervision)? A "yes" answer is worth two points, while a "no" answer is worth zero points.
Does the offender have a record of assaults or violence involving injury to a victim? One victim of a violent offense indicated is worth two points. Two or more victims of violence is worth five points. No violent offenses indicated is worth zero points.
Does the offender have trouble getting along with inmates? A "yes" answer is worth two points, while a "no" answer scores zero. Generally, fights or other behavioral problems involving inmates or protection concerns are grounds for a "yes" answer.
Does the offender have trouble getting along with officers? A "yes" answer is worth two points, while a "no" answer is scored zero. Defiance reports or multiple rule violations of any kind indicate problems with officers as would any assault on an officer.
Has the offender completed any treatment programs while incarcerated? A "yes" answer is worth a negative two or removal of two points, while a "no" answer is scored as a three.
Has the offender been involved in any organized groups or activities while incarcerated? A "yes" answer is worth a negative two or removal of two points, while a "no" answer is scored as a two.
Does the offender have alcohol or drug problems? A "yes" answer is worth five points, while a "no" answer is scored zero.
Does the offender have indicated emotional or psychological problems? A "yes" answer is worth two points, while a "no" scores zero.
Does the offender have attitudes generally supporting of crime or criminal lifestyles? A "yes" answer is worth five points, while a "no" is scored zero. This is one of the more difficult items to score and relies on interpretation of various aspects of the offender's record including his disciplinary record, number of offenses and length of criminal record.
Does the offender have a poor attitude toward supervision? A "yes" answer is worth three points, while a no is scored zero. This also is a difficult item to score. It relies on the offender's indicated relations with supervisory staff, including disciplinary reports, any assaults on officers, prior violations of supervision, etc.
Other factors. While this item is not scored numerically, information may be taken into account in an overall determination of the risk of violence or behavioral problems from an individual. The scores are added up and compared to a risk scale in which a score of 0 to 18 is considered low risk, 19 to 31 is considered medium risk and 32 to 55 is considered high risk.
The initial use of this tool supported the use of risk-factor screening as a way to predict behavioral or violence problems with offenders. A review of the results follows:
Of the 22 individuals in the original sample identified as being at high risk, 18 (82 percent) have violent backgrounds. The other four individuals identified as having violent backgrounds have a higher number of rule violation reports for serious offenses than others in their risk class.
Of the three individuals identified as being at low risk for problems, none had any serious rule violations in the last six months (see Table 1).
Of all the individuals in the low- and medium-risk classes added together, there were a total of 57 serious rule violation reports issued during the time this study took place. During the same time period, the high-risk group accounted for 360 serious rule violation reports.
The low-risk group was responsible for 18 total rule violation reports, with 13 of those coming from one individual, who had moved in the screening from medium to low risk. The medium-risk group accounted for 39 rule violation reports. According to the data, those in the high-risk group were six times more likely to receive rule violation reports than those in the low- or medium-risk groups.
In every case in which an individual was assigned to the high-risk group, he had either a violent background or significant behavioral problems as indicated by rule violation reports.
The correlation between these factors is striking and reveals patterns that can be used by correctional staff to make management decisions. It is indicated that the risk screening tool can be a valuable part of the process to predict and manage the possibility of violence or behavioral problems.
TABLE 1. RESULTS OF THE STUDY NUMBER OF VIOLATION REPORTS PERCENTAGE OF TOTAL REPORTS LOW RISK 18 4.3% MEDIUM RISK 39 9.4% HIGH RISK 360 86.3%
(1) Gendreau, P., C. Goggin and T. Little. 1996. Predicting adult offender recidivism: What works. Criminology, (34):575-607.
(2) Hansen, R.K. and M. Bussiere. 1998. Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66(2):348-362.
(3) Gendreau, P., C. Goggin and T. Little. 1996.
(4) Andrews, D.A. and J. Bonta. 1998. The psychology of criminal conduct, second edition. Cincinnati: Anderson.
Hansen, R.K. and M. Bussiere. 1998.
Larry Wagoner, MSW, GSW, CCSW VI, CCM, is the former mental health director at the Washington Correctional Institute in Angie, La. For additional information regarding this study, contact Wagoner at (985) 732-3909, ext. 25.