In this discussion, you’ll consider how to identify and prioritize diversion, intervention, or mitigation -related care for individuals who are members of multiple special populations.
After reviewing the video and readings for this module, select a case to discuss. You may choose either (1) the case you are using for your final project, or (2) any of the other cases discussed in this course. For your chosen case, use the file titled “Tool: Risk and Protective Factors” to identify risk and protective factors for the person at the heart of the case. In your primary post for this discussion, summarize your findings around risks and protective factors based on the tool, and identify 1-2 issues you would address first. Justify your choices with research evidence and guidance provided in this module video.
In this discussion, you’ll consider how to identify and prioritize diversion, intervention, or mitigation -related care for individuals who are members of multiple special populations. After reviewing
Original article Risk assessment tools in criminal justice and forensic psychiatry: The need for better data T. Douglas a, J. Pugh a, I. Singh a ,b , J. Savulescu a, S. Fazel b ,c , * aOxford Uehiro Centre for Practical Ethics, Faculty of Philosophy, University of Oxford, Suite 8, Littlegate House, St Ebbes Street, Oxford OX1 1PT, United Kingdom bDepartment of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX, United Kingdom cOxford Health NHS Foundation Trust, Warneford Hospital, Oxford OX3 7JX, United Kingdom There are currently more than 200 structured tools available for assessing risk of violence in forensic psychiatry and criminal justice  . These are widely deployed to inform initial sentencing, parole, and decisions regarding post-release monitoring and rehabilitation. In some jurisdictions, including Canada, New Zealand, and until 2012 in the United Kingdom, risk assessment tools are or were also used to justify indeterminate post-sentence detention. In addition, violence risk assessment tools are used to inform decisions regarding detention, discharge, and patient management in forensic and, increasingly, general psychiatry. This article highlights some potential ethical problems posed by risk assessment tools and argues that better data on predictive accuracy are needed to mitigate these. It focuses on the use of risk assessment tools in forensic psychiatric and criminal justice settings. 1. Professional obligations and competing values In the psychiatric literature, criticism of risk assessment has focused on the possibility that, in deploying risk assessment tools, mental health professionals may fail to fulﬁl their professional obligations to their patients [2,3] . Health professionals are expected to make the care of their patients their ﬁrst concern, to build trust, and to respect patient preferences, and this expectation is reﬂected in professional guidelines  . Some argue that the use of risk assessment tools is unjustiﬁed when it is intended to realise other values, such as justice or public protection, and does not beneﬁt the assessed individual [5– 8] . Buchanan and Grounds hold that ‘‘it is inappropriate to comment on a defendant’s risk unless psychiatric intervention is proposed or other beneﬁt will result’’  . Similarly, Mullen claims that ‘‘[r]isk assessments . . . are the proper concern of health professionals to the extent that they initiate remedial interven- tions that directly or indirectly beneﬁt the person assessed’’  . The use of risk assessment tools is perhaps most clearly at odds with the interests of the assessed individual where the tool is used to inform decisions regarding post-sentence detention. In this context, the default position is that the person will be released; however, if the tool indicates a high risk of violence, detention may be extended. It could be argued that deploying the tool thus runs against the individual’s interest in being released as soon as possible. In some cases, however, the application of a risk assessment tool will beneﬁt the assessed individual. There are at least three European Psychiatry 42 (2017) 134–137 A R T I C L E I N F O Article history: Received 17 September 2016 Received in revised form 4 December 2016 Accepted 11 December 2016 Available online 28 December 2016 Keywords: Violence Forensic psychiatry Ethics and human rights Risk assessment Crime prediction Racial proﬁling A B S T R A C T Violence risk assessment tools are increasingly used within criminal justice and forensic psychiatry, however there is little relevant, reliable and unbiased data regarding their predictive accuracy. We argue that such data are needed to (i) prevent excessive reliance on risk assessment scores, (ii) allow matching of different risk assessment tools to different contexts of application, (iii) protect against problematic forms of discrimination and stigmatisation, and (iv) ensure that contentious demographic variables are not prematurely removed from risk assessment tools. C2016 The Author(s). Published by Elsevier Masson SAS. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). * Corresponding author at: Department of Psychiatry, Medical Sciences Division, University of Oxford, Warneford Hospital, Oxford OX3 7JX, United Kingdom. E-mail address: [email protected] (S. Fazel). Contents lists available at ScienceDirect European Psychiatry jo u rn al h om epag e: h ttp ://ww w.eu ro p s y- jo ur n al.co m http://dx.doi.org/10.1016/j.eurpsy.2016.12.009 0924-9338/ C2016 The Author(s). Published by Elsevier Masson SAS. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ).https://doi.org/10.1016/j.eurpsy.2016.12.009 Published online by Cambridge University Press ways in which it could confer such a beneﬁt. First, the risk assessment may be used to identify beneﬁcial treatments. Second, the use of a risk assessment tool may facilitate an earlier release or discharge. Suppose an individual is being considered for parole or discharge from a secure psychiatric institution, but this is likely to be refused on the basis that there is insufﬁcient evidence for a low risk of violence. In this situation, application of a risk assessment tool may provide the evidence necessary to secure an end to detention. Third, even when a risk assessment results in further detention, it might nevertheless confer a beneﬁt because extended detention is itself in the individual’s best interests. For example, it may prevent re-offending and an even longer period of detention in the future. Moreover, even when mental health professionals administer risk assessments that are against the assessed individual’s best interests, it is not clear they thereby violate a professional obligation, for the view that medical professionals ought never to act against a patient’s best interests can be contested. In the setting of infectious disease control, it would be widely accepted that physicians may sometimes compromise a patient’s best interests in order to promote other values, such as the health of family members and the wider public [9,10]. Similarly, many would hold that an obstetrician may sometimes act to protect a future child, even if this comes at some cost to the patient—that is, the prospective mother . It can be argued that a parallel point holds in relation to forensic psychiatry: professionals in this ﬁeld may sometimes give precedence to values besides the welfare of their own patients . Those who hold that risk assessment tools should be used only when they beneﬁt the patient may thus be overstating the ethical difﬁculties created by such tools. Nevertheless, the presence of competing values in risk assessment does create a potential ethical problem: it is possible that some values will be unjustiﬁably sacriﬁced for the sake of others. For example, there is a risk that the interests of individual patients or prisoners will be unjustiﬁably compromised in the name of public protection, or the reverse. We will argue that a lack of high quality data on predictive accuracy compounds this ethical risk. 2. Predictive accuracy Existing data suggest that most risk assessment tools have poor to moderate accuracy in most applications. Typically, more than half of individuals judged by tools as high risk are incorrectly classiﬁed—they will not go on to offend . These persons may be detained unnecessarily. False positives may be especially common in minority ethnic groups [14,15]. Rates of false negatives are usually much lower. Nevertheless, in typical cases around 9% of those classed as low risk will go on to offend . These individuals may be released or discharged too early, posing excessive risk to the public. Such failures of negative prediction are frequently associated with signiﬁcant controversy and outrage, as reactions to recent high proﬁle cases demonstrate . The prevalence of prediction errors does not entirely undermine the rationale for deploying risk assessment tools. To balance risk to the public against the interests of the assessed individual, some method for assessing risk is required, and risk assessment tools, even if limited in accuracy, may be the best option available. However, to mitigate the possibility of inadequate or excessive detention, the limitations of risk assessment tools need to be well understood and factored into clinical and criminal justice responses. Unfortunately, published validation ﬁndings for the most widely used tools, which allow for predictive accuracy to beestimated in advance, frequently present a misleading picture . First, though there are exceptions, most tools have not been externally validated outside of their derivation sample [18,19]. Of particular concern, few validation studies have been conducted in women, ethnic minority populations, and individuals motivated by religious or political extremism [14,15,17]. Consequently, it is unclear how far reported accuracy ﬁndings can be extrapolated to new settings and populations . Second, there is strong evidence that conﬂicts of interest are often not disclosed in this ﬁeld, and some evidence of publication and authorship bias . (Author- ship bias occurs when research on tools tends to be published by the authors of those tools, who typically ﬁnd better performance.) Third, published studies frequently present only a small number of performance measures that do not provide a full picture of predictive accuracy . Thus, not only is the predictive accuracy of risk assessment tools imperfect, it is also imperfectly presented in the literature. This limited and skewed evidence base creates a risk that decision makers will rely more heavily on risk assessment scores than their accuracy warrants. To mitigate this risk, there is a need for better quality data covering more subpopulations. Validation studies should include more than just one or two performance statistics, and data on the numbers of true and false positives and negatives should be clearly presented. Conﬂicts of interests need to be disclosed, and reviews by authors with ﬁnancial conﬂicts of interests should be treated with caution. In addition to risking over-reliance on risk assessment scores, deﬁciencies in the evidence base also generate at least three more speciﬁc problems, which we explain below: they (i) thwart attempts to match risk assessment tools to different contexts of application, (ii) complicate efforts to determine whether risk assessment tools are unjustiﬁably discriminatory or stigmatising, and thereby (iii) contribute to the possibility that contentious demographic variables will be prematurely eliminated from assessment tools. 3. The right tool for the context Selecting the optimal risk assessment tool for a given application requires trade-offs to be made between false negatives and false positives; attempts to reduce the number of false positives will increase the number of false negatives . Tools with a low rate of false negatives (due to high sensitivity) will be most effective at protecting the public, and may garner most political support, while tools with a low rate of false positives (due to high speciﬁcity) will best protect the rights and interests of prisoners and psychiatric patients. The optimal balance between false positives and false negatives is an ethical issue and will depend on the social and political context in which the tool is to be used . For example, avoidance of false positives may be more important in jurisdictions with less humane detention practices than in jurisdictions with more humane practices, since the less humane the conditions of detention, the greater the harm false positives will tend to impose on the assessed individual . The appropriate balance between false positives and false negatives will also depend on the stage in the criminal justice process or patient pathway at which the tool will be deployed. For instance, suppose that a risk assessment tool is used to inform decisions about post-sentence detention in a setting where an individual’s initial sentence is proportionate to their degree of responsibility and the seriousness of the crime. Detaining the individual beyond the end of the initial sentence thus involves imposing a disproportionately long period of detention. In this context, special care should be taken to avoid false positives, and T. Douglas et al. / European Psychiatry 42 (2017) 134–137 135https://doi.org/10.1016/j.eurpsy.2016.12.009 Published online by Cambridge University Press there may be grounds to prefer a tool with a very low false positive rate to one that is overall more accurate. However, the situation is different when a tool is used to inform parole decisions. In this context, false positives may lead to refusal of parole and an unnecessarily long period of incarceration from the point of view of public protection. Yet if we assume that the initial sentences are themselves proportionate, then the overall period of detention for ‘false positive’ individuals will remain within the upper limit set by considerations of proportionality. In this context it may be more important to avoid false negatives. Matching risk assessment tools to different contexts of application thus requires trade-offs between positive and negative predictive accuracy. For each context, we must ﬁrst decide which type of accuracy to prioritise to which degree, and then select a tool that reﬂects this priority. Unfortunately, in the absence of reliable data, it is not possible to make the latter decision conﬁdently. There is a need for studies using representative samples for relevant subpopulations, avoiding highly selected samples, and presenting performance measures that allow false negative and false positive rates to be reliably estimated for a particular application. 4. Discrimination and stigmatisation Some argue that singling out individuals for unfavourable treatment on the basis of their demographic characteristics amounts to unjustiﬁed discrimination. This criticism is often levelled at racial proﬁling by police and airport security . A similar concern might be raised regarding risk assessment tools that take into account an individual’s demographic characteristics such as ethnicity, age, immigration status and gender. It has been suggested that risk assessment tools should employ only ‘individu- alised’ information, such as information about declared plans and desires based on face to face interviews [17,27], though, even then, judgments may be subject to implicit biases based on the demographic characteristics of the individual being assessed . However, the requirement to utilise only individualised information is overly restrictive. Many would argue that demo- graphic proﬁling is discriminatory, or problematically so, only when the demographic variables used are recognised social groups (such as ethnic or gender groups) , or certain kinds of recognised social groups, for instance, those whose membership is unchosen , or that have historically been subject to oppression . Risk assessment tools could theoretically exclude such variables. In reply, it might be argued that exclusion of such variables is insufﬁcient to avoid moral concerns. First, even if the problematic demographic variables are formally excluded from the analysis, they may continue to exert an inﬂuence; there remains the potential for implicit bias in the application of risk assessment tools and interpretation of risk scores [14,15,17]. Second, even if the problematic demographic variables are formally excluded from the analysis and there is no implicit bias in applying the tools, there may still be a correlation between membership of certain demographic groups and risk score. For example, members of a particular ethnic group may be more likely than average to receive high risk scores. Some may hold that such a correlation is problematic, especially if it is due to past wrongdoing against members of the demographic group in question (e.g., members of the ethnic group are indeed more likely to offend, but only because they are victims of unjust social exclusion), if the correlation does not reﬂect a true difference in risk (e.g., false positives occur more frequently than average in the minority ethnic group), or if the correlation is likely to lead to stigmatisation of the group deemed to be higher risk. However, even if the use of risk assessment tools does involve a problematic form of discrimination or stigmatisation, it could nevertheless be justiﬁed if the case in favour of using theinformation is powerful enough. The parallel with racial proﬁling in airport screening is instructive here. Airport screening is a limited resource and there are reasons to deploy it to detect the maximum number of would-be terrorists. If proﬁling enables a far greater number of terrorist attacks to be prevented with the resources available than any other policy, and if the cost to those proﬁled is low, then it is arguably justiﬁed even if somewhat problematic, for example, because discriminatory or stigmatising. Similarly, the resources available for the prevention of violence are limited, and if deploying a risk assessment tool prevents far more violence than could otherwise be prevented with the resources available, it might be justiﬁed even if it does raise some concerns about discrimination and stigmatisation. Nevertheless, it is important that risk assessment tools deploy the most speciﬁc predictive information available. Arguably, what is most objectionable about some forms of racial proﬁling is that they deploy racial appearance as a predictor when more speciﬁc predictors of security threat are available and, were these predictors used, racial appearance would add no further predictive value [32,33]. In such circumstances, use of racial appearance seems unnecessary. Similarly, it may be problematic to use demographic predictors in risk assessment tools when more speciﬁc predictors of future offending are available and these predictors would render the use of demographic categories redundant. Unfortunately, the lack of good evidence on accuracy makes it difﬁcult to ascertain whether existing tools do use the most speciﬁc predictors available. To determine this, we would need to be able to compare the accuracy of more speciﬁc and less speciﬁc tools using relevant, reliable and unbiased data on accuracy. Currently deployed tools frequently do use demographic factors such as age and immigration status as predictors, and although recent evidence suggests that including such demographic factors improves predictive accuracy [34,35], further data are needed to conﬁrm this. In the absence of these data, there are two risks. On the one hand, mental health professionals may continue to employ coarse demographic variables that result in unnecessary discrimination or stigmatisation. On the other, given growing public concern regarding the use of such variables [36,37], professionals or policy makers may prematurely remove them from risk assessment tools . Before variables are removed because they are potentially contentious, high quality research that uses transparent methods and presents all relevant outcomes should investigate whether the demographic factors included in current tools add incremental validity to tool performance . Funding This work was supported by grants from the Wellcome Trust (100705/Z/12/Z, WT086041/Z/08/Z, #095806, WT104848/Z/14/Z), and the Uehiro Foundation on Ethics and Education. Disclosure of interest SF has published research on risk assessment, including as part of a team that has derived and validated one tool for prisoners with psychiatric disorders. References  Singh JP, Desmarais SL, Hurducas C, Arbach-Lucioni K, Condemarin C, Dean K, et al. International perspectives on the practical application of violence risk assessment: a global survey of 44 countries. Int J Forensic Mental Health 2014;13:193–206. T. Douglas et al. / European Psychiatry 42 (2017) 134–137 136https://doi.org/10.1016/j.eurpsy.2016.12.009 Published online by Cambridge University Press  Eastman N, Adshead G, Fox S, Latham R, Whyte S. Professional duties and personal integrity and conﬂicting ethical values. In: Forensic psychiatry. Oxford University Press; 2012.  Appelbaum PS. Dangerous severe personality disorders: England’s experiment in using psychiatry for public protection. Psychiatr Serv 2005;56(4):397–9.  General Medical Council (GMC). Good medical practice: guidance for doctors – the duties of a doctor; 2013, http://www.gmc-uk.org/guidance/ good_medical_practice/duties_of_a_doctor.asp.  Eastman N. The ethics of clinical risk assessment and management: developing law and the role of mental health professionals. In: Gray N, Laing J, Noaks L, editors. Criminal justice, mental health and the politics of risk. Cavendish Publishing; 2002.  Buchanan A, Grounds A. Forensic psychiatry and public protection. Br J Psychiatry 2011;198(6):420–3.  Gunn J. Future directions for treatment in forensic psychiatry. Br J Psychiatry 2000;176(4):332–8.  Mullen PE. Forensic mental health. Br J Psychiatry 2000;176(4):307–11.  Gostin LO. Public health law: power, duty, restraint. University of California Press; 2001.  Wilkinson TM. Contagious disease and self-defence. Res Publica 2007;13:339–59.  Savulescu J. Future people, involuntary medical treatment in pregnancy and the duty of easy rescue. Utilitas 2007;19:1–20.  Appelbaum PS. A theory of ethics for forensic psychiatry. J Am Acad Psychiatry Law 1997;25:233–47.  Fazel S, Singh JP, Doll H, Grann M. Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24 827 peo- ple: systematic review and meta-analysis. BMJ 2012;345:e4692.  Shepherd SM. Violence risk instruments may be culturally unsafe for use with indigenous patients. Australas Psychiatry 2016;24:565–7.  Shepherd SM, Lewis-Fernandez R. Forensic risk assessment and cultural diversity – contemporary challenges and future directions. Psychol Public Policy Law 2016;22:427–38.  Parry H. Rapist released halfway through sentence went on to attack three more while on parole including a schoolgirl who was raped in front of her boyfriend. Daily Mail 2015. http://www.dailymail.co.uk/news/ article-3131895/Rapist-released-halfway-sentence-went-attack-three- parole-including-two-schoolgirls-raped-boys-with.html.  Shepherd SM, Sullivan D. Covert and implicit inﬂuences on the interpretation of violence risk instruments. Psychiatry Psychol Law 2016. http://dx.doi.org/ 10.1080/13218719.2016.1197817.  Siontis GCM, Tzoulaki I, Castaldi PJ, Ioannidis JPA. External validation of new risk prediction models is infrequent and reveals worse prognostic discrimi- nation. J Clin Epidemiol 2015;68:25–34.  Siontis and collaborators found that only 16% of new prediction models are validated by different authors within ﬁve years of their ﬁrst use, and when external validation of tools did occur, predictive accuracy was generally reduced, Siontis GCM, et al. External validation of new risk prediction models is infrequent and reveals worse prognostic discrimination. J Clin Epidemiol 2015;68(1):25–34. A recent Chinese review found that instruments developed in the West had signiﬁcantly lower predictive accuracy when used in China compared to that reported for Western populations. See Zhou J, Witt K, Xiang Y, Zhu X, Wang X, Fazel S. Violence risk assessment in psychiatric patients in China: a systematic review. Aust N Z J Psychiatry 2016;50:33–45.  Singh JP, Grann M, Fazel S. Authorship bias in violence risk assessment? A systematic review and meta-analysis. PLoS ONE 2013;8:e72484.  Singh JP. Predictive validity performance indicators in violence risk assess- ment: a methodological primer. Behav Sci Law 2013;31:8–22.  Walker N. Dangerousness and mental disorder. R Inst Philos Suppl 1994;37:179–90. http://dx.doi.org/10.1017/S1358246100010055. p. at 182.  Sinnott-Armstrong W. Neural lie detection in courts. In: Buzzi E, Hyman SE, Raichle ME, Kanwisher N, Phelps EA, Morse SJ, et al., editors. Using imaging to identify deceit: scientiﬁc and ethical questions. American Academy of Arts and Sciences; 2009. p. 35–9.  Ryberg J. Racial proﬁling and criminal justice. J Ethics 2010;15:79–88.  Lever A. Why racial proﬁling is hard to justify: a response to Risse and Zeckhauser. Philos Public Affairs 2005;33:94–110.  Nadelhoffer T, Bibas S, Grafton S, Kiehl KA, Mansﬁeld A, Sinnott-Armstrong W, et al. Neuroprediction, violence, and the law: setting the stage. Neuroethics 2012;5:67–99.  Amodio DM, Devine PG. Stereotyping and evaluation in implicit race bias: evidence for independent constructs and unique effects on behavior. J Pers Soc Psychol 2006;91:652–61.  Lippert-Rasmussen K. Born free and equal? A philosophical inquiry into the nature of discrimination.. Oxford University Press; 2013. p. 336.  Kahlenberg RD. The remedy: class, race, and afﬁrmative action. Basic Books; 1997. p. 386.  Scanlon T. Moral dimensions: permissibility, meaning. Blame: Belknap Press; 2008. p. 247.  Lippert-Rasmussen K. Nothing personal: on statistical discrimination. J Polit Philos 2007;15:385–403.  Lippert-Rasmussen K. ‘‘We are all different’’: statistical discrimination and the right to be treated as an individual. J Ethics 2010;15:47–59.  Fazel S, Chang Z, Fanshawe T, La˚ ngstro¨ m N, Lichtenstein P, Larsson H, et al. Prediction of violent reoffending on release from prison: derivation and external validation of a scalable tool. Lancet Psychiatry 2016;3(6):535–43.  Skeem JL, Lowenkamp CT. Risk, race, and recidivism: predictive bias and disparate impact. Soc Sci Res Netw 2016. http://dx.doi.org/10.2139/ ssrn.2687339.  Angwin J, Larson J, Mattu S, Kirchner L. Machine Bias: there’s software used across the country to predict future criminals. And it’s biased against blacks. ProPublica 2016. https://www.propublica.org/article/machine-bias-risk-assessments-in- criminal-sentencing [accessed 02.12.16].  Barry-Jester AM, Casselman B, Goldstein D. Should prison sentences be based on crimes that haven’t been committed yet? FiveThirtyEight 2015. http:// ﬁvethirtyeight.com/features/prison-reform-risk-assessment/ [accessed 02.12.16].  Monahan J, Skeem JL. Risk assessment in criminal sentencing. Annu Rev Clin Psychol 2016;12:489–513. T. Douglas et al. / European Psychiatry 42 (2017) 134–137 137https://doi.org/10.1016/j.eurpsy.2016.12.009 Published online by Cambridge University Press
In this discussion, you’ll consider how to identify and prioritize diversion, intervention, or mitigation -related care for individuals who are members of multiple special populations. After reviewing
Key Risk Factors for Relapse and Rearrest Among Substance Use Treatment Patients Involved in the Criminal Justice System Albert M. Kopak 1&Norman G. Hoffmann 2,3 & Steven L. Proctor 4 Received: 17 December 2015 / Accepted: 17 December 2015 / Published online: 7 January 2016 #Southern Criminal Justice Association 2016 Abstract Substance use treatment programs for criminal justice populations have great potential for crime reduction, if they can effectively manage patients ’risk for relapse and rearrest. The current study used data drawn from the Comprehensive Assessment and Treatment Outcome Research (CATOR) system, a national registry of substance use treatment programs, which collected patient outcome data at 6- and 12-month intervals following discharge from treatment. The primary objective was to examine sets of factors that may compromise relapse and rearrest outcomes among patients who were court mandated to participate in treatment. Findings demonstrated that patients ’ clinical severity of substance use was associated with relapse, which also significantly increased the probability of post-treatment arrest. Adolescent risk behaviors represented another set of risk factors, particularly among patients who experienced the most severe pattern of relapse and arrest outcomes. Additionally, demographic risk factors, includ- ing age, marital status (i.e., single or unmarried relative to married), employment (i.e., Am J Crim Just (2016) 41:14 –30 DOI 10.1007/s12103-015-9330-6 * Albert M. Kopak [email protected] Norman G. Hoffmann [email protected] Steven L. Proctor [email protected] 1 Department of Criminology & Criminal Justice, Western Carolina University, 1 University Drive, Belk 410, Cullowhee, NC 28723, USA 2 Department of Psychology, Western Carolina University, Cullowhee, NC 28723, USA 3 Evince Clinical Assessments, Waynesville, NC, USA 4 Department of Psychology, Albizu University –Miami Campus, 2173 NW 99th Avenue, Miami, FL 33172-2209, USA being unemployed compared to employed), and lower educational attainment were consistently linked to higher probabilities of relapse and rearrest. Treatment programs for criminal justice populations should consider incorporating appropriate clinical risk assessment measures, behavioral risk assessments, and appropriate employment inter- ventions into standard treatment programming in an effort to improve outcomes. KeywordsTreatment outcomes . Relapse . Recidivism . Drug dependence . Crime prevention Introduction Criminal justice approaches to substance use in the U.S. can be traced back to the passage of the Harrison Narcotics Tax Act in 1914 and the subsequent establishment of hospitals designed to treat opiate and cocaine dependence (Musto, 1999). A century later, the available substance dependence treatment programming options have sub- stantially expanded and formal recognition of the need to provide these services to criminal justice populations has become a core principle of the National Drug Control Strategy (Office of National Drug Control Policy [ONDCP], 2012). Despite these efforts, however, substance dependence continues to be a major issue embedded in our criminal justice system. Recent estimates show that over one-third of individuals who have come into contact with the criminal justice system reported use of illicit drugs in the past 12 months (ONDCP, 2011). Estimates also suggest that upward of 1.47 million arrestees are at-risk for drug dependence and likely require treatment services when they enter the criminal justice system (Bhati & Roman, 2010). Interestingly, more than one-in-five arrestees reported a lifetime history of prior substance use treatment (ONDCP, 2011). Considering that such a large number of criminal justice-involved individuals have a history of problematic substance use, many are eligible for treatment services, and a sizable proportion have already received treatment in the past but have been arrested in the time since that treatment experience, further work is warranted to better refine the identification of treatment needs and optimization of services for this vulnerable population. These observations also beg the question, what are some of the key risk factors associated with post-treatment relapse and arrest experienced within this highly vul- nerable population? There are a substantial number of criminal justice-involved indi- viduals and policymakers alike who have a lot to gain from a clear and concise answer. The key objective of the current study was to pinpoint these risk factors so that treatment providers and policy makers can better address the needs of high-risk substance use treatment patients. Recent Research on Substance use Treatment Outcomes in Criminal Justice Populations The phrase Bdrug treatment programs work ^for drug-involved criminal justice popu- lations is commonly stated (e.g., Holloway, Bennett, & Farrington, 2006), but there is a Am J Crim Just (2016) 41:14 –30 15 significant amount of mixed evidence for treatment, particularly when concentrating on specific outcomes. A reduction in post-treatment contact with the criminal justice system is one of these questionable outcomes that seem to be achieved in some circumstances, but not in others. Evidence supporting the premise that substance use treatment can reduce rearrest was found in an analysis of drug-involved probationers in the state of Florida (Lattimore, Krebs, Koetse, Lindquist, & Cowell,2005). This study showed that participation in outpatient treatment reduced the anticipated number of probationers who were rearrested and also decreased the number of probationers ’ subsequent arrests. Additional evidence from a study of low-risk, non-violent, and short-term inmates in Monroe County, New York also demonstrated the effectiveness of drug treatment (Turley, Thornton, Johnson, & Azzolino, 2004). Inmates who participated in the Monroe County drug treatment program experienced significant reductions in their likelihood of post-treatment arrest, as well as the number of times inmates were re-arrested after their release. In contrast to these favorable results, there is also a fair amount of divergent evidence which shows that treatment has not played a significant role in reducing reoffending. For instance, the National Institute of Drug Abuse (NIDA)-funded Drug Abuse Treatment Outcome Study (DATOS) serves as one example in its investigation of the relationship between the length of drug treatment and illegal activity. The primary finding from this federally-funded project was that treatment participation was not associated with a significant decrease in the likelihood of illegal activity among patients (Hubbard, Craddock, Flynn, Anderson, & Etheridge, 1997). Similarly, in their recent study of a mandatory drug treatment program in Kansas, Rengifo and Stemen ( 2013 ) demonstrated that offenders ’participation in mandated drug treatment (com- pared to participation in other community corrections programs) was no more likely to reduce future criminal activity. Overall, some research favors the use of drug treatment programs for criminal justice populations as a viable method to reduce prospective criminal activity, but others have failed to find similar support, complete with contradictory findings presented within a single study (e.g., Listwan, Sundt, Holsinger, & Latessa, 2003). The inconsistency in empirical support may be, in part, related to variations in treatment services provided by individual programs, differences in the assessments of reoffending, as well as the demographic and clinical composition of the study samples. Most importantly, a substantial portion of this research has not taken into account the risk factors associated with negative treatment outcomes that are most prevalent in criminal justice populations. Potential Risk Factors for Relapse and Post-Treatment Arrest Research has identified key criminogenic risk factors, and evidence suggests that these need to be considered as patients enter treatment vis-à-vis the criminal justice system. Findings from a prospective study of California ’s Proposition 36 program —which was designed to provide drug treatment for criminal justice-involved offenders —revealed that those who experienced the greatest rearrest risk were most likely to be younger, male, and have greater pre-treatment contact with the criminal justice system (Evans, Huang, & Hser, 2011). Similar risk factors were also observed among drug-involved probationers who had failed to complete their mandated treatment programs and were 16 Am J Crim Just (2016) 41:14–30 rearrested in that they were also younger, unemployed, unmarried, and less educated (Huebner & Cobbina,2007). Unemployment is a particularly important issue to consider with respect to substance use treatment outcomes and arrest. Not only have positive treatment outcomes been associated with employment immediately preceding treatment entry (Wickizer et al., 1994), but stable employment has also been identified as an important protective factor with resp ect to post-treatment substance use and criminal activity (Henderson, 2001; Tripodi, Kim, & Bender, 2010). Criminal justice-involved substance use treatmen t patients face significant challenges in obtaining and maintaining employment (Staton et al., 2001), but those who do experience more stable social circumstances may be more likely to achieve better treatment outcomes. Another factor that is likely to influenc e treatment outcomes, particularly for drug-involved offenders who may be mandated to certain types of treatment programs by judges, is past criminal ju stice contact. The age at which first criminal justice contact is made can hav e a significant impact on the extent to which offenders are considered for drug treatment, especially after being arrested or incarcerated several times. Research has shown that adolescent problem behavior is associated with ad ult criminal activity among criminal justice-involved drug treatment patients (Piquero et al., 2012). Participants in the Drug Treatment Alternative to Prison (DTAP) program in New York, for example, were more likely to be rearrested if they were younger and had more juvenile arrests, despite the fact they all had successfully completed the re- quirements of the treatment program (Sung & Belenko, 2005). In addition to these risk factors, there are also important prognostic indica- tors which need to be considered as criminal justice-involved individuals enter treatment. Some patients may need treat ment for initial signs of substance misuse (i.e. non-medical and problematic substance use that has the potential to lead to social, psychological, or medi cal consequences (World Health Orga- nization (WHO), 2006), for instance. In comparison, others may suffer from fully developed severe substance use disorders, which can include a conglom- eration of physiological (e.g., withdrawal and tolerance) and behavioral (e.g., neglect of responsibilities) problems (A merican Psychiatric Association [APA], 2013 ). Consideration of clinical severity is crucial to maximizing treatment effects, especially to reduce the likelihood of relapse and rearrest, given the evidence that shows more severe substance use can increase the probability of both of these negative outcomes (Hubicka, Laurell, & Bergman, 2010;Robles, Huang, Simpson, & McMillan, 2011). In order to optimize substance use tre atment services for criminal justice populations, full consideration needs to be given to the primary risk factors that are likely to increase the potential for rel apse and rearrest. The greatest threats to successful treatment need to be immediately addressed to promote an effec- tive recovery process, reduce substance use, and avoid future contact with the criminal justice system. The goal of thi s study was to identify these prognostic indicators in an effort to move toward mat ching patients mandated to treatment programs to the best services so they may achieve the lowest possible risk for relapse and post-treatment arrest. Am J Crim Just (2016) 41:14 –30 17 Methods Data The current study utilized data drawn from the Comprehensive Assessment and Treatment Outcome Research (CATOR) system. This program was the largest inde- pendent (i.e. non-government funded and not owned by any private individual treat- ment provider) evaluation service and served as one of the most comprehensive assessments of substance use treatment programs across the United States (Proctor & Herschman,2014). The primary objective of CATOR project managers was to evaluate various types of treatment programs for their ability to achieve abstinence from alcohol and drugs while also collecting information on correlates of recovery. All participants included in these analyses are from programs which were monitored by the CATOR system and received a substance dependence diagnosis. Based on this criterion, patients with less severe substance use disorders (e.g., persons with abuse diagnoses only and illicit drug misusers) were excluded from the analyses. The treatment programs were located throughout the United States and included a variety of residential and outpatient (including evening only, day treatment, and intensive outpatients) services. The CATOR system collected information in a longitudinal prospective evaluation design which began with admission to substance use treatment and continued through discharge to include several post-treatment follow-up assessments (Miller, Ninonuevo, Klamen, Hoffmann, & Smith, 1997). Clinicians collected intake information to serve as baseline data, which included information related to pre-treatment substance use including diagnostic indications, treatment history, pre-treatment criminal justice in- volvement, psychosocial factors, and vocational functioning. Follow-up data were collected at 6- and 12-month intervals after discharge through structured telephone interviews conducted by trained technicians. These interviews collected data on sub- sequent substance use, engagement in peer-support groups, post-treatment exposure to psychosocial risk factors for relapse, vocational functioning, and involvement in the criminal justice system. Eligibility criteria for the current study were based primarily on participants ’ entering treatment according to a court mandate. This included receipt of a treatment referral from court as a direct result of an arrest for driving while intoxicated (DWI)/ driving under the influence (DUI) as well as entering treatment as an alternative to incarceration. Among the 13,948 CATOR participants who completed the 6- and 12- month follow-up interviews, 14 % were in treatment due to a court mandate, yielding an eligible sample size of 1980 for the current study. An additional 6 % ( n=123)of cases were excluded due to missing data, which resulted in a net sample of 1857 criminal justice involved substance use treatment patients. Measures Dependent Variables The first dependent variable measured substance use relapse and was designed to assess the pattern of abstinence versus any alcohol or drug use that occurred during the initial 18 Am J Crim Just (2016) 41:14–30 Researchers have recently emphasized the importance of detailed arrest measures (King & Elderbroom, 2014; Prendergast, Hall, & Wexler, 2003). The second dependent variable assessed post-treatment arrest with one of these progressive methods, and was computed according to the same approach as the relapse dependent variable. Four arrest groups were created: one that included patients who were not arrested during the 12 months after treatment discharge (i.e., no arrest (NA)), a second group included patients who were arrested within 6-months of treatment discharge but were not arrested between 6- and 12-months of treatment discharge (i.e., 6-month arrest/12- month no arrest (6A/12NA)), a third group consisting of patients who were not arrested between treatment discharge and the 6-month follow-up interview but were arrested between the 6- and 12-month follow-up interviews (i.e., 6-month no arrest/12-month arrest (6NA/12A)), and a fourth group who reported an arrest during each of the two follow-up intervals (i.e., 6-month arrest/12-month arrest (6A/12A)). Independent Variables The first independent variable examined in this study was an indicator of whether the individual in the treatment program received a current Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; APA, 1994)diagnosisfordrug dependence. Patients who met criteria for dependence on marijuana, cocaine, opiates, prescription drugs, or stimulants were coded B1 ^ and all other patients (i.e., those who received an alcohol dependence diagnosis) were coded B0. ^ The second independent variable examined in this study was a clinical index designed to capture the severity of patients ’substance use. This measure was an additive scale of 8 items, which addressed (1) recent substance use, (2) substance use frequency in the past year, (3) indication of injection drug use, (4) the number of substance dependence diagnoses the patient received, (5) a measure of alcohol use quantity, (6) alcohol withdrawal symptoms, (7) drug withdrawal symptoms, and (8) patients ’substance use during treatment. A complete list of these items is provided in the Appendix. This additive scale ranged from 1 to 13, with larger values representing more severe cases of substance use. The third independent variable examined in this study was an adolescent risk behavior index. This 8-item additive scale was used to assess how many risk behaviors patients had experienced during adolescence, and addressed the following behaviors: (1) skipping school more than 10 times, (2) school suspension or expulsion, (3) arrest, (4) running away from home overnight more than once, (5) vandalism or destruction of Am J Crim Just (2016) 41:14 –30 19 12-months post-treatment discharge. Four groups were created and the first included patients who did not use any alcohol or drugs within 6 months of treatment discharge or within 6 to 12 months of discharge (i.e., complete abstinence (CA)). The second group (i.e., 6-month relapse/12-month abstinent (6R/12A)) used alcohol or drugs during the first 6 months from discharge but remained abstinent between the 6- and 12-month follow-up interviews. The third group (i.e., 6-month abstinent/ 12-month relapse (6A/12R)) remained abstinent between treatment discharge and the 6-month follow-up, but had used alcohol or drugs between the 6- and 12-month follow-up interviews. The fourth group (i.e., 6-month relapse/12-month relapse (6R/12R)) reported post- treatment alcohol or drug use at the 6-month follow-up as well as the 12-month follow-up. property, (6) shoplifting or stealing, (7) sexual intercourse with more than one person, and (8) starting physical fights. Each of these items was codedB0 ^ if patients indicated they had not been involved in the behavior and B1 ^ if they were involved in the behavior. The measure ranged from 0 to 8 with higher values indicative of involvement in a greater number of risk behaviors. The fourth independent variable examined in this study was a demographic risk index previously identified as being related to relapse (Zywiak, Hoffmann, & Floyd, 1999 ). This additive measure assessed four background, educational, and vocational functioning items including (1) age, (2) marital status, (3) highest educational level achieved, and (4) employment status. Following previously established methods for creating additive clinical scales (Kahneman, 2011), patients received a point on the demographic risk index if they were under 25 years of age, had never been married, had not received a high school diploma or GED as their level of education, and were unemployed when entering treatment. Scores on this risk index ranged from 0 to 4 with higher scores representing younger, unattached, less educated, and unemployed patients. Two additional independent variables were examined for their hypothesized influ- ence on post-treatment arrest. The first of which was a measure of the number of times each patient was arrested in the 12 months prior to treatment admission. Patients were asked how many times they were arrested for DWI/DUI, other traffic violations, disorderly conduct, assault or battery, theft, robbery, burglary, prostitution, vandalism or destruction of property, possession of drugs or drug paraphernalia, drug sales, or any other type of criminal offense. Response options for each offense ranged from B0 ^ (not arrested) to B3 ^ (arrested 3 or more times). Responses from all offense categories were summed to indicate the total number of arrests in the past year. The second arrest risk factor was a binary measure of relapse. Patients were coded B0 ^ if they had not used any alcohol or drugs after being discharged from treatment and B1 ^ if they had used alcohol or drugs after discharge from treatment. Control Variables Several control variables were included in the current study. Male patients were coded B 0 ^ and female patients were coded B1. ^Patients ’racial background was also taken into account with patients that self-identified as White coded ‘0 ″ and patients who identified as Asian, Black, Hispanic, Native American, or other coded B1 ″ and classified as non- White. The type of treatment program patients ’participated in was also included in the current study. Patients who received treatment in a residential program were coded B0 ″ and those who received treatment in an outpatient program (i.e., day outpatient, evening outpatient, or combination of day/evening outpatient) were coded B1. ^ Analytic Strategy Analyses for the current study were conducted in several stages. The first stage involved univariate examination of demographic background factors of the participants. Mean levels of the key independent variables, including demographic risk, adolescent risk behaviors, clinical severity, and pre-treatment arrests were also assessed. Bivariate 20 Am J Crim Just (2016) 41:14–30 statistics (i.e. ANOVA for continuous measures and Chi-square for categorical mea- sures) were used in the second stage as preliminary assessments of differences in risk factors across relapse and arrest outcome groups. The third and final stage of analyses involved the estimation of multinomial logistic regression models to assess the relative influence of the key risk factors while controlling for confounding variables. These models allow for flexibility in the assignment of a comparison category, which allowed for the comparison of the three relapse groups to the abstinent group as well as the comparison of the three groups which experienced a post-treatment arrest to the group which was not arrested, in their respective models. Multinomial logistic regression results can also be expressed in the form of relative risk (RRR) of one outcome (i.e. relapse or arrest) compared to another (i.e. abstinence or no arrest), which is suitable for the current study (Menard, 2002). All analyses were conducted with Stata 11 (StataCorp, 2009). Results Descriptive Statistics The sample was predominantly White (87 %), male (79 %), and more patients were diagnosed with alcohol dependence (60 %) relative to those who were diagnosed with drug dependence. Patients displayed an average of approximately 1.5 ( M= 1.46, SD = 1.06) demographic risk factors and the average clinical severity level was approximately 3.5 ( M=3.57, SD= 2.08). Patients also reported slightly more than two ( M=2.03, SD= 2.12) adolescent risk behaviors on a scale from 1 to 8 and had an average of 1.61 ( SD= 1.63) arrests prior to entering treatment. More than half (60 %) of patients were treated in residential programs. Slightly less than half (47 %) of the sample experienced alcohol or drug use relapse after discharge from treatment. A much smaller proportion (18 %) of patients reported a post-treatment arrest. Bivariate Results A descriptive summary of the key relapse and arrest risk factors is presented by outcome groups in Tables 1and 2. A preliminary set of ANOVA tests were conducted to assess risk factor levels according to relapse status groups (Table 1). These results indicated that there were significant variations across relapse outcome groups with respect to demographic risk ( F(3, 1853) = 22.28, p< .001), adolescent behavioral risk ( F (3, 1853) = 13.28, p< .001), and clinical severity ( F(3, 1853) = 13.99, p<.001).A chi-square test ( χ 2=12.61,df=3, p= .006) also indicated that there were larger numbers of patients who received a drug dependence diagnosis in the three relapse groups relative to the complete abstinence group. A similar series of ANOVA results provided evidence that there were risk-related differences across the arrest outcome groups (Table 2). The arrest groups significantly varied in their levels of demographic risk ( F(3, 1853) = 21.70, p<.001),adolescent behavioral risk ( F(3, 1853) = 17.50, p< .001), clinical severity ( F(3, 1853) = 10.80, p < .001), and number of pre-treatment arrests ( F(3, 1853) = 14.68, p<.001).An Am J Crim Just (2016) 41:14 –30 21 additional chi-square test (χ 2=39.80,df=3, p< .001) provided preliminary evidence to show there were larger numbers of patients who received a drug dependence diagnosis in the arrest groups relative to the group who was not arrested after treatment. A final test ( χ 2=87.95,df=3, p< .001) indicated the groups who did experience a post-treatment arrest were comprised of significantly more patients who experienced relapse compared to the group that was not arrested. Multivariate Results A series of multinomial regression models were estimated to investigate differences in the relapse and arrest dependent variables while controlling for the different levels of Table 1 Descriptive statistics of criminal justice-involved substance use treatment patients by relapse outcomes Complete Abstinence (CA) (n= 986) 6-month relapse/ 12-month abstinence (6R/12A) ( n= 205) 6-month abstinence/ 12-month relapse (6A/12R) ( n= 238) 6-month relapse/ 12-month relapse (6R/12R) ( n= 428) Variable Percentage/Mean (sd) Female 20 % 22 % 19 % 22 % Non-white 12 % 14 % 16 % 13 % Drug dependence* 36 % 46 % 44 % 43 % Out-patient program 41 % 39 % 37 % 43 % Demographic risk index* 1.28(.97) 1.51(1.10) 1.67(1.08) 1.71(1.15) Adolescent risk index* 1.78(1.96) 1.97(2.13) 2.32(2.31) 2.49(2.28) Clinical severity index* 3.28(1.91) 4.02(2.10) 3.83(2.14) 3.86(2.29) * p <.01 Ta b l e 2 Descriptive statistics of criminal justice-involved substance use treatment patients by arrest outcomes No post-treatment arrest (n=1526) 6-month arrest/ 12-month no arrest ( n=133) 6-month no arrest/ 12-month arrest ( n = 159) 6-month arrest/ 12-month arrest ( n =39) Variable Percentage/Mean (sd) Female 21 % 21 % 14 % 10 % Non-white 13 % 10 % 14 % 23 % Drug dependence* 37 % 53 % 55 % 62 % Outpatient program 41 % 38 % 40 % 31 % Demographic risk index* 1.37(1.03) 1.77(1.16) 1.92(1.10) 2.00(1.05) Adolescent risk index* 1.90(2.06) 2.20(2.12) 2.75(2.29) 3.77(2.65) Clinical severity index* 3.45(2.03) 3.83(1.99) 4.29(2.36) 4.36(2.19) Pre –treatment arrests* 1.51(1.55) 1.85(1.68) 1.92(1.81) 3.00(2.69) Relapse* 42 %67 %72 % 75 % * p <.01 22 Am J Crim Just (2016) 41:14–30 risk experienced by each group. The first multinomial regression model estimated the risk of the three relapse conditions relative to complete abstinence group (Table3). The adolescent behavioral risk index was significantly associated with an increased risk of 6- and 12-month relapse (6R/12R) relative to complete abstinence (CA). In particular, a one-unit increase in the adolescent risk behavior index was associated with an 11 % increase in the likelihood of 6- and 12-month relapse relative to 6- and 12-month abstinence. The clinical severity index was consistently associated with an increased risk of relapse across all three relapse groups. A one-unit increase in the clinical severity of substance dependence was associated with a 21 % increase in the likelihood of 6-month relapse/12-month abstinence (6R/12A) compared to complete abstinence (CA). This same increase in clinical severity risk was also associated with a 12 % increase in the risk of 6-month abstinence/12-month relapse (6A/12R) compared to complete abstinence (CA) and a 15 % increase in the chance of 6-month relapse/12- month relapse (6R/12R) relative to complete abstinence (CA). Demographic risk was also consistently associated with the likelihood of relapse within each of the follow-up intervals compared to 12-month abstinence. As patients ’ demographic risk increased by a factor of one, they were 21 % more likely to fall into the category of 6-month relapse/12-month abstinence (6R/12A) compared to the complete abstinence (CA) category. Likewise, those who experienced an increase in demographic risk were 40 % more likely to report 6-month abstinence/12-month relapse (6A/12R) compared to those who were completely abstinent (CA) during the 12-month follow-up period. Finally, an increase in demographic risk also heightened the risk of 6-month relapse/12-month relapse (6R/12R) by 44 % compared to those who were completely abstinent (CA) during the 12-month follow-up period. The second multinomial regression model estimated the risk of the three arrest conditions compared to no arrest at the 6- and 12-month follow-up interviews (Table 4). Similar to the relapse model, the adolescent risk index was only associated with classification in the most severe arrest group. In other words, an additional adolescent risk behavior was associated with a 23 % increase in the likelihood a patient was arrested before the 6- and 12-month follow-up interviews (6A/12A) compared to patients who were not arrested (NA). An additional pre-treatment arrest was also significantly associated with 6-month arrest/12-month arrest (6A/12A) compared to the group who did not have any follow-up arrests (NA). A drug dependence diagnosis, compared to those who were not diagnosed with drug dependence was associated with an 89 % increase in the likelihood of 6-month arrest/12-month no arrest (6A/12NA) compared to no arrests (NA). The clinical severity index was not associated with any of the arrest conditions relative to the no-arrest outcome. Demographic risk, on the other hand, was consistently associated with arrest outcomes. An additional demographic risk factor was associated with a 26 % increase in 6-month arrest/12-month no arrest (6A/12NA) compared to no arrests (NA) during the 12-month follow-up period. The same relationship was observed between demo- graphic risk and the likelihood of no arrest at the 6-month no arrest/12-month arrest (6NA/12A), which contributed to a 40 % increase in the risk of arrest compared to no arrest (NA). Demographic risk was also associated with a 38 % increase in the likelihood of 6-month arrest/12-month arrest (6A/12A) compared to no arrest (NA). Another significant result was observed in the relationship between relapse and arrest outcomes. Relapse of alcohol or drug use significantly increased the risk of arrest Am J Crim Just (2016) 41:14 –30 23 Ta b l e 3Multinomial regression results predicting relapse outcomes Variable 6-month relapse/12-month abstinence (6R/12A) vs. complete abstinence(CA) 6-month abstinence/12-month relapse (6A/12R) vs. complete abstinence(CA) 6-month relapse/12-month relapse (6R/12R) vs. complete abstinence(CA) Coefficient (se) Relative risk ratio 95 % C.I. C oefficient (se) Relative risk ratio 95 % C.I. Coefficient (se) Relative risk ratio 95 % C.I. Constant 2.57 (.22)** –– 2.42 (.21) –– 2.15 (.18) –– Female .09 (.19) 1.09 0.75–1.58 .10 (.19) 0.91 0.63–1.31 .12 (.15) 1.13 0.84–1.51 Non-white .24 (.23) 1.28 0.82–1.99 .43 (.21)* 1.54 1.03–2.30 .18 (.18) 1.19 0.84–1.69 Drug dependence .07 (.20) 0.93 0.63–1.37 .20(.19) 0.82 0.57–1.18 .27 (.15) 0.76 0.56–1.03 Outpatient program .18 (.17) 1.19 0.86–1.66 .05(.16) 0.95 0.70–1.30 .25 (.13) 1.29 1.00–1.65 Demographic risk index .19 (.08)** 1.21 1.04–1.40 .33(.07)** 1.40 1.22–1.60 .37 (.06)** 1.44 1.29–1.62 Adolescent risk index .02 (.04) 0.98 0.90–1.06 .07(.04) 1.08 1.00–1.15 .11 (.03)** 1.11 1.05–1.18 Clinical severity inde x .19 (.05)** 1.21 1.10–1.32 .12(.04)** 1.12 1.03–1.23 .14 (.04)** 1.15 1.07–1.23 **p <.01;* p<.05 24 Am J Crim Just (2016) 41:14–30 Ta b l e 4Multinomial regression results predicting arrest outcomes Variable 6-month arrest/12-month no arrest (6A/12NA) vs. no arrest (NA) 6-month no arrest/12-month arrest (6NA/12A) vs. no arrest (NA) 6-month arrest/12-month arrest (6A/12A) vs. no arrest (NA) Coefficient (se) Relative risk ratio 95 % C.I. Coefficient (se) Relative risk ratio 95 % C.I. Coefficient (se) Relative risk ratio 95 % C.I. Constant 3.44 (.29)** –– 4.01 (.28)** –– 5.77 (.56) –– Female .13 (.23) 0.87 0.56– 1.37 .64 (.25)** 0.53 0.32– 0.86 .89 (.55) 0.41 0.14– 1.20 Non-white .43 (.31) 0.65 0.35– 1.19 .04 (.25) 1.04 0.63– 1.70 .71 (.41) 2.04 0.92–4.55 Drug dependence .64 (.23)** 1.89 1.21– 2.97 .43 (.22) 1.54 1.00–2.36 .54 (.42) 1.72 0.76–3.92 Out-patient program .03 (.20) 0.97 0.65– 1.44 .15 (.19) 1.17 0.81– 1.69 .23 (.38) 0.80 0.38– 1.67 Demographic risk index .23 (.09)** 1.26 1.06 –1.49 .34 (.08)** 1.40 1.20 –1.64 .32 (.16)** 1.38 1.01– 1.87 Adolescent risk index .00 (.05) 1.00 0.91 –1.09 .07 (.04) 1.08 0.99– 1.17 .21 (.07)** 1.23 1.07–1.42 Clinical severity index .05 (.06) 0.95 0.85– 1.06 .06 (.05) 1.06 0.96– 1.17 .05 (.09) 0.95 0.79– 1.13 Pre-treatment arres ts .06 (.05) 1.06 0.96–1.18 .03 (.05) 1.03 0.93–1.14 .19 (.07)** 1.21 1.06–1.38 Relapse .95 (.20)** 2.59 1.76–3.80 1.06 (.19)** 2.89 1.99–4.19 1.08 (.38)** 2.95 1.39–6.27 ** p<. 01;* p<.05 Am J Crim Just (2016) 41:14 –30 25 across all three outcomes. Patients who relapsed were 2.59 times more likely to be arrested prior to the 6-month follow-up/not arrested at 12-months (6A/12NA) com- pared to those who were not arrested (NA). Patients who relapsed were also 2.89 times more likely to be in the 6-month no arrest/12-month arrest (6NA/12A) group relative to those who were not arrested. Finally, patients who relapsed were 2.95 times more likely to have been arrested at 6-months/arrested at 12-months (6A/12A) post-treatment relative to those who were not arrested. Discussion The primary aim of this study was to identify the most important risk factors related to relapse and arrest among criminal justice involved substance use treatment patients. Several important findings emerged, and the first is that adolescent behavioral problems were most likely to contribute to the worst outcomes: relapse within 6- and 12-month follow-up periods as well as arrest in both of these time periods. The adolescent items in this study relate to the DSM-5 (APA,2013) diagnosis of adolescent conduct disorder. This diagnosis is a prerequisite for the diagnosis of antisocial personality disorder. Thus, the adolescent behavioral risk indications studied here may be a proxy for an underlying problem behavior syndrome among patients that could be addressed during treatment as an adult. One approach might be to incorporate dialectical behavior therapy (DBT) as a method to address the developmental changes experienced from later adolescence into early adulthood (Robins & Chapman, 2004). This therapeutic technique has been successfully applied among substance use patients from various demographic backgrounds in many phases of life and may be further adapted for criminal justice populations (Dimeff & Linehan, 2008; Little, Butler, & Fowler, 2010). The assessment of clinical severity may also indicate the need for enhanced or extended treatment services delivered to those at greater clinical risk. Clinical severity was consistently associated with greater risk for relapse, which may be a proximal indicator that treatment providers did not adequately tailor the treatment program to meet the needs of patients, or reimbursement limitations may have limited the delivery of necessary services. These conditions could have left patients without access to the requisite treatment services given an inability to pay for these services. Another likely underlying issue here is that most substance-involved offenders are mandated to participate in standardized treatment programs, without any distinctions made for issues unique to this population, such as anti-social attitudes and elevated risks for relapse and arrest (U.S. Department of Justice, Federal Bureau of Prisons, 2009). Best practices should include comprehensive clinical assessments to match treatment needs among members of this high risk population with appropriate programming options. Relapse was a key indicator of post-treatment arrest across all arrest outcome groups. This finding has major implications for the treatment of criminal justice involved patients, because avoiding future criminal justice contact is largely dependent on substance abstinence. The benefits of substance use treatment as a crime prevention measure cannot be realized unless treatment programs also effectively address relapse prevention. 26 Am J Crim Just (2016) 41:14–30 Another interesting observation was that a co-occurring drug dependence diagnosis (relative to alcohol dependence only) was associated with greater risk of arrest within between drug dependence and the economic motivation underlying offending behav- iors specifically used to finance drug use (Gudjonsson, Wells, & Young,2011; van der Zanden, Dijkgraaf, Blanken, van Ree, & van den Brink, 2007). Additionally, drug- dependent offenders have been found to commit crime at twice the rate of that as offenders who do not have a dependence diagnosis (Keene, 2005). Drug dependence requires special attention in treatment services among criminal justice populations to prevent future arrests. Demographic risk was the most consistent indicator of both relapse risk and arrest risk for patients who went through treatment. Treatment services need to focus on the special needs of younger, unattached, lower educated, and unemployed patients, especially those who are involved in the criminal justice system. The presence of these demographic risk factors has emerged in what might represent a subgroup of patients who require more intensive treatment services. Criminal justice treatment programs should consider identifying these at-risk patients with a similar composite index to better match them at admission with appropriate programming options. Some of these factors, such as age and marital status, may not be readily changed in this population, but there are intervention points that can be readily acted upon. Recent findings, for example, have emphasized the importance of including employment- focused interventions in drug treatment programs. These employment enhancement programs have the potential to not only increase the likelihood of post-treatment employment, but also to significantly raise the income of treatment patients, which has great relapse prevention potential (Webster, Staton-Tindall, Dickson, Wilson, & Leukefeld, 2014). Incorporation of a vocational rehabilitation component into standard treatment programming has the potential to positively impact the rates of both relapse and criminal re-offense. The CATOR system includes a substantial amount of information related to the correlates of successful substance use treatment outcomes, but there are limitations to the current study that need to be acknowledged. Most of the treatment providers included in the CATOR system were private nonprofit facilities willing to subject themselves to external evaluation of their performance and may not be representative of all centers across the U.S. This may also limit the sample in its representation of patients treated in public sector programs. Other important limitations to note are the fairly small racial and ethnic minority representation included in the study sample, as well as the larger proportion of the sample in treatment for alcohol use disorder relative to those seeking treatment for a drug use disorder. Prospective research is needed address some of these shortcomings. Further inves- tigation is required to determine if these results are generalizable to publicly funded substance use treatment programs. It is possible that these programs differ in their relapse and arrest rates relative to the private non-profit facilities included in the CATOR system. Additional research should also investigate racial and ethnic differ- ences in treatment outcomes, especially considering the impact of disproportionate criminal justice contact among racial and ethnic minority group members. Finally, research also needs to examine relapse and arrest outcomes in greater detail among criminal justice involved substance users. For example, researchers should investigate whether relapse episodes are consistently linked to similar or different types of offenses among criminal justice-involved treatment patients. Treatment prognoses should Am J Crim Just (2016) 41:14 –30 27 6-months of discharge from treatment. This is likely due to the well-established link account for patients’risk for committing similar offenses (e.g. property or drug-related offenses) due to their relapse compared to those who experience escalation in the severity of their offending (e.g. violent offenses) as a result of persistent substance use. This new information about the risk factors associated with relapse and post- treatment criminal justice contact can pay dividends in providing patients with the best possible chances for successful recovery. Mandated treatment programs should focus on patients ’clinical severity, demographic risk, and relapse potential to maximize treatment effects. Treatment may Bwork, ^but now is the time for empirical refinement. Appendix Items included in the clinical severity index. 1. What was your most recent ingestion of alcohol, marijuana, cocaine, stimulants, barbiturates/sedative, opiates, tranquilizers, hallucinogens, narcotic painkillers, other: (0) Did not use substances in the past 24 h, (1) Used one substance in the past 24 h, (2) Used multiple substances in the past 24 h. 2. What was your typical use of alcohol, marijuana, cocaine, stimulants, barbiturates/ sedative, opiates, tranquilizers, hallucinogens, narcotic painkillers, other in the past year? (0) Did not use any substance daily, (1) Used one substance daily, (2) Used multiple substances daily. 3. Have you ever used a needle to inject street drugs? (0) No (1) Yes. 4. Have you ever drank a fifth of liquor, 20 drinks, 3 six-packs of beer, or 3 bottles of wine in one day? (0) No (1) Yes. 5. Have you ever had delirium tremens, fits, seizures, or hallucinations after stopping drinking? (0) No (1) Yes. 6. Have you ever had withdrawal symptoms after stopping drug use? (0) No (1) Yes 7. Patients who met diagnostic criteria for dependence on one substance were coded (1), those who met criteria for dependence on two substances were coded (2), those who met criteria for three were coded (3), up to (6) for those who met criteria for dependence on six substances. 8. Patients who did not use alcohol or drugs during treatment were coded (0) and those who did use alcohol or drugs during treatment were coded (1). References American Psychiatric Association (2013). Diagnostic and statistical manual of mental disorders, fifth edition . Washington, DC: Author. American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed., ). Washington, DC: American Psychiatric Association. Bhati, A. S., & Roman, J. K. (2010). Simulated evidence on the prospects of treating more drug-involved offenders. Journal of Experimental Criminology ,6,1 –33. Dimeff, L. A., & Linehan, M. M. (2008). Dialectical behavior therapy for substance abusers. Addiction Science and Clinical Practice ,4,39 –47. 28 Am J Crim Just (2016) 41:14–30 Evans, E., Huang, D., & Hser, Y. (2011). High-risk offenders participating in court-supervised substance abusetreatment: Characteristics, treatment received, and factors associated with recidivism. The Journal of Behavioral Health Sciences & Research ,38 ,510 –525. Gudjonsson, G. H., Wells, J., & Young, S. (2011). Motivation for offending among prisoners and the relationship with axis I and axis II disorders and ADHD symptoms. Personality and Individual Differences ,50 ,64 –68. Henderson, M. L. (2001). Employment and crime: What is the problem and what can be done about it from the inmate ’s perspective? Corrections Management Quarterly ,5,46 –52. Holloway, K. R., Bennett, T. H., & Farrington, D. P. (2006). The effectiveness of drug treatment programs in reducing criminal behavior: A meta-analysis. Psicotherma,18 ,620 –629. Hubbard, R. L., Craddock, S. G., Flynn, P. M., Anderson, J., & Etheridge, R. M. (1997). Overview of 1-year follow-up outcomes in the drug abuse treatment outcome study (DATOS). Psychology of Addictive Behaviors ,11 ,261 –278. Hubicka, B., Laurell, H., & Bergman, H. (2010). Psychosocial characteristics of drunk drivers assessed by the addiction severity index, prediction of relapse. Scandinavian Journal of Public Health,38,71 –77. Huebner, B., & Cobbina, J. (2007). The effect of drug use, drug treatment participation, and treatment completion on probationer recidivism. Journal of Drug Issues,37 ,619 –642. Kahneman, D. (2011). Thinking, fast and slow . New York, NY: Farrar, Straus, and Giroux. Keene, J. (2005). A case-linkage study of the relationship between drug misuse, crime, and the psychosocial problems in a total criminal justice population. Addiction Research and Theory,13(5), 489 –502. King, R., & Elderbroom, B. (2014). Improving recidivism as a performance measure (justice policy center brief) . Washington, DC: Urban Institute. Lattimore, P. K., Krebs, C. P., Koetse, W., Lindquist, C., & Cowell, A. J. (2005). Predicting the effect of substance abuse treatment on probationer recidivism. Journal of Experimental Criminology,1, 159 –189. Little, L., Butler, L. S., & Fowler, J. (2010). Change from the ground up: Bringing informed dialectical behavior therapy to residential treatment. Residential Treatment for Children & Youth ,27 ,80 –91. Listwan, S. J., Sundt, J. L., Holsinger, A. M., & Latessa, E. J. (2003). The effect of drug court programming on recidivism: The Cincinnati experience. Crime and Delinquency,49,389 –411. Menard, S. (2002). Applied logistic regression analysis (2 nd ed., ). Thousand Oaks, CA: Sage Publications. Miller, N. S., Ninonuevo, F. G., Klamen, D. L., Hoffmann, N. G., & Smith, D. E. (1997). Integration of treatment and posttreatment variables in predicting results of abstinence-based outpatient treatment after one year. JournalofPsychoactiveDrugs ,29 ,239 –248. Musto, D. F. (1999). The American disease: Origins of narcotic control (3rd ed., ). New York: Oxford University Press. Office of National Drug Control Policy. (2011). ADAM II: 2010 Annual Report. Retrieved from http://www. whitehouse.gov/sites/default/files/ondcp/policy-and-research/adam2010.pdf Office of National Drug Control Policy. (2012). National drug control strategy. Retrieved from http://www. whitehouse.gov/sites/default/files/ondcp/2012_ndcs.pdf Piquero, A. R., Farrington, D. P., Fontaine, N. M. G., Vincent, G., Coid, J., & Ullrich, S. (2012). Childhood risk, offending trajectories, and psychopathy at age 48 years in the Cambridge study in delinquent development. Psychology, Public Policy, and Law ,18, 577 –598. Prendergast, M. L., Hall, E. A., & Wexler, H. K. (2003). Multiple measures of outcome in assessing a prison- based drug treatment program. Journal of Offender Rehabilitation ,37 ,65 –94. Proctor, S. L., & Herschman, P. L. (2014). The continuing care model of substance use treatment: What works, and when is Benough, ^Benough? ^. Psychiatry Journal . doi:10.1155/2014/692423 . Rengifo, A. F. & Stemen, D. (2013). The impact of drug treatment on recidivism: Do mandatory programs make a difference? Evidence from Kansas ’s Senate Bill 123. Crime and Delinquency ,930–950. Robins, C. J., & Chapman, A. L. (2004). Dialectical behavior therapy: Current status, recent developments, and future directions. Journal of Personality Disorders ,18,73 –89. Robles, E., Huang, B. E., Simpson, P. M., & McMillan, D. E. (2011). Delay discounting, impulsiveness, and addiction severity in opioid-dependent patients. Journal of Substance Abuse Treatment,41,354 –362. StataCorp (2009). Stata statistical software: Release 11. In College Station.TX:StataCorpLP. Staton, M., Mateyoke, A., Leukefeld, C., Cole, J., Hopper, H., Logan, T. K., & Minton, L. (2001). Employment issues among drug court participants. Journal of Offender Rehabilitation,33 ,73 –85. Sung, H., & Belenko, S. (2005). Failure after success: Correlates of recidivism among individuals who successfully completed coerced drug treatment. Journal of Offender Rehabilitation,42 ,75 –97. Tripodi, S. J., Kim, J. S., & Bender, K. (2010). Is employment associated with reduced recidivism? The complex relationship between employment and crime. International Journal of Offender Therapy and Comparative Criminology ,54 ,706 –720. Am J Crim Just (2016) 41:14 –30 29 Turley, A., Thornton, T., Johnson, C., & Azzolino, S. (2004). Jail drug and alcohol treatment program reducesrecidivism in nonviolent offenders: A longitudinal study of Monroe county, New York ’sjailtreatment drug and alcohol program. International Journal of Offender Therapy and Comparative Criminology ,48 , 721 –728. U.S. Department of Justice, Federal Bureau of Prisons. (2009). Psychology treatment programs. Retrieved from http://www.bop.gov/policy/progstat/5330_011.pdf van der Zanden, B. P., Dijkgraaf, M. G. W., Blanken, P., van Ree, J. M., & van den Brink, W. (2007). Patterns of acquisitive crime during methadone maintenance treatment among patients eligible for heroin assisted treatment. Drug and Alcohol Dependence ,86 ,84 –90. Webster, J. M., Staton-Tindall, M., Dickson, M. F., Wilson, J. F., & Leukefeld, C. G. (2014). Twelve-month employment intervention outcomes for drug-involved offenders. The American Journal of Drug and Alcohol Abuse ,40 ,200 –205. Wickizer, T., Maynard, C., Atherly, A., Frederick, M., Koepsell, T., Krupski, A., & Stark, K. (1994). Completion of clients discharged from drug and alcohol treatment programs in Washington State. American Journal of Public Health ,84, 215 –221. World Health Orgnization (WHO) (2006). Lexicon of alcohol and drug terms Published by the World Health Organization . Retrieved from http://www.who.int/substance_abuse/terminology/who_lexicon/en/ Zywiak, W. H., Hoffmann, N. G., & Floyd, A. S. (1999). Enhancing alcohol treatment outcomes through aftercare and self-help groups. Medicine & Health/Rhode Island ,82(3), 87 –90. Albert M. Kopak , Ph.D. is an Assistant Professor in the Department of Criminology and Criminal Justice at Western Carolina University. He received his Ph.D. in Justice Studies from Arizona State University. His research focuses on a variety of factors related to substance use disorders, such as involvement in the criminal justice system and treatment outcomes. Norman G. Hoffmann , Ph.D. is a clinical psychologist who has evaluated behavioral health programs, provided consultations, and conducted trainings for more than 35 years. He has worked with private organizations and governmental agencies in a variety of countries. Dr. Hoffmann has developed a number of assessment instruments used throughout the USA and has authored or co-authored more than 190 publications. Steven L. Proctor , Ph.D. is an Assistant Professor of Psychology at Albizu University –Miami Campus. His line of research focuses on the evaluation of addictions assessment procedures, instruments, and treatment outcomes. 30 Am J Crim Just (2016) 41:14–30