Po ur aseres0" lda DSM-5 Antisocial Personality Disorder: Predictive Validity in a Prison Sample John F. Edens and Shannon E. Kelley ‘Texas ASM University Scott O, Lilienfeld Emory University Jennifer L. Skeem Kevin S. Douglas University of Colifornia-Berkeley Simon Fraser University and Mil-Sweden University Symptoms of antisocial personality disorder (ASPD), particularly remorselessess, ae ffequenty into ‘The American Psychiatrie Association's Diagnastic and Stati © teal Manual of Mental Disorders (DSM) diagnosis of antisocial 1994) to DSM-S, Although the DSM-5 (as well a earlier editions) contains cautionary language indicating that psychiatric diagnoses personality disorder (ASPD) is characterized hy an enduring pat- tem of unlawful behavior, aggressiveness, deceitfulness, impulsiv= ity, iresponsibily, reckless disregard for the welfare of others, andéor remorselessness manifest during adulthood, as well as ev~ ‘idence of conduct disorder (CD) in childhood or adolescence Despite a radical proposal by the DSIM-S (APA, 2013) Personality, Disorder Workgroup to transform the personality disorders into dimensionally informed personality prototypes, the ASPD nostic criteria have survived unchanged from DSM-IV (APA, Jon F. Edens and Shannon E. Kelly, Department of Psychology, Texas A&M University: Scott ©, Lilienfeld, Deparment of Payehology, Emory Univecsiy,Jemifer L. Skeem, School of Social Welfare, University of Califrnin-Berksley: Kevin 8. Doula, Deparment of Payshology, Si mon Fraser University and Mi-Swedes Universi ‘Comespondence concerning this arse shold be addressed to John dens, Deparment of Psychology, 4235 TAMU, Texas A&M University, College Staton, TX 77483, Femail: jonedensiatams el were not developed to meet the needs of the courts oF legal professionals, diagnoses of ASPD have been applied to address Pragmatically important legal questions. ‘The ASPD diagnosis is frequently used for forensic purposes in capital murder trials support the argument that defendants diagnosed with ASPD remain a continuing threat to society and will commit criminal acts of violence in the future if they are not put to death (DeMatteo, Mure, Anumba, & Keesler, 2011; Edens & Cox, 2012). For example, in one well-known case psychiatrist testified that the defendant's ASPD was “extremely severe” and. that he was “one hundred pereent certain that he is and will continue t0 be a threat no matter where he is" (Cook v. State, 1991), Ironically, this defendant ultimately was exonerated and released—after 22 years on death row (Cook, 2008). Such cases are not isolated events: Recent estimates suggest that at least one in every 25 death-sentenced inmates in the United States has been \wronghilly convicted (Gross, O’Brien, Iu, & Kennedy, 2014). Although the exact frequency with which diagnoses of ASPD continue to be used by prosecution experts to justify death sex= ‘ences is unknown, a recent survey of [egal professionals has2 [EDENS, KELLEY, LILIENFELD, SKEEM, AND DOUGLAS suggested that it is common and has considerable impact in these ceases (Edens & Cox, 2012). Despite ASPD's apparent widespread use in the legal the reliability and validity of this diagnosis has repeatediy been called into question (Cunningham & Reidy, 1998; Freedman etl. 2013; Morey & Ochoa, 1989). For example, the DSW-5 (APA, 2013) field trials indicated that ASPD had one of the poorest diagnostic interrater relisilties (Freedman et al, 2013), exacer- baling previous concerns that this diagnosis can be unstable across examiners (Cunningham & Reidy, 1998). Poor reliability may stem fiom the fact that clinicians are prone fo inappropriately assign an ASPD diggnosis when patients do not mest sufficient diagnostic criteria, but do have histories of criminal behavior, exploitation of others, vandalism, or school expulsion (Morey & (Ochoa, 1989). In forensic and correctional settings the accuracy of the ASPD diagnosis may also suffer from an excessive emphasis fon isolated events (i.e, specific erimes) rather than enduring pallens of behavior (Cunningham & Reidy, 1998). Confidence in ASPD diagnoses is further undermined by evidence of temporal Instability, significant erteria overlap with sequelae of substance abuse, and heterogeneous symptom constellations. Related to the latter point, the likelihood of separable subtypes of ASPD (Poyth= ress, Edensm etal, 2010) may attenuate the value of the broader diagnosis in forecasting important outcomes, ‘Compounding these concems is a dearth of research examining the predictive validity of ASPD. Although there is evidence that ASPD is frequently comorbid with substance abuse and suicidality (Glenn, Johnso, & Raine, 2013), surprisingly litle is known, bout the diagnosis's prognostic value for behavioral outcomes. such as aggression and violence. Despite the Frequency with which this disorder is introduced as evidence of future dangerousness at capital sentencing, there is scant evidence that ASPD diagnoses identify offenders who will engage in serious violence while incarcerated (Edens, Buffington-Vollum, Keilen, Roskamp, & An~ thony, 2005). A few recent studies have identified a relationship between categorical and dimensional measures of ASPD and self reported institational violence (Warren & Bumette, 2013; Warren ct al, 2002) and the disorder has been linked to disciplinary infractions in prison (Toch & Adams, 1986). However, these investigations assessed institutional misconduct retrospectively rather thas. prospectively and the assoeiations reported may be autfacts of labeling problemetic inmates as “antisocial.” That is, interview-based diagnoses of ASPD in these instances may reflect 2 history of disruptive Behavior while incarcerated rather than indicate future risk for aggression or violence. Indeed, there are strong reasons to doubt a potent association between ASPD and violence. Prevalence rates of ASPD among, prisoners are high, typically’ ranging anywhere between S0%- 180%, depending on the individual study, In sharp contrast, base rates of serious violence in prisons are low (Cunningham & Reidy, 2002; Cunningham, Reidy, & Sorensen, 2008; Cunningham & Sorensen, 2007; Edens eta, 2005). Even if ASPD were a signif ‘cant predictor of aggressive behavior during incarceration, such a combination invariably compromises predictive utility because of an exceedingly high false-positive rate: The majority of individu als with the disorder will not engage in the behavior of intrest ‘To our knowledge, only one prospective study has examined the relationship between ASPD and prison misconduct, Walters and Knight (2010) assessed newly admitted mae inmates for adult childhood symptoms of ASPD using a structured interview and conducted a brief follow-up (0.1 t0 9.5 months) eonceming receipt of disciplinary infraction reports while they were incarcerated ‘Adult antisocial symptoms were associated with greater levels of ‘general institutional misconduct only when two or more symptoms ‘of antecedent CD were also endorsed Infact, iamates only meet ing adult ASPD criteria were indistinguishable from non-ASPD. inmates in terms of the likelihood of receiving a disciplinary inaction, suggesting that historical information detail childhood conduct is salient componeat in predi ‘ional adjustment. Also of note, hase rates of violent behavior in this study were too low to perform meaningful analyses—consis- tent with other research indicating that violent infractions in prison settings are relatively rare phenomena (Cunningham & Reidy 2002; Cunningham et al, 2008; Cunninghars & Sorensen, 2007; Cunningham et al, 2005: Edens eta, 2005), ‘When attempting to forecast institutional misconduct, focus on ASPD is also problematic in that it may detract attention from, known predictors of prison violence (see DeMatteo etal, 2011, ‘Chapter 8), which include such factors as young, age, low educa ‘on, low IQ, prior prison terms, shor-tem sentences, and prison- gang affiliation (Cunningham & Sorensen, 2006, 2007; Cunning hham, Sorensen, & Reidy, 2008; Diamond, Moris, & Barnes, 2012). These findings also suggest that violence risk could be confounded with ASPD diggnoses, given that both are clevated among men, young adults, African Americans, and less educated individuals (Cunsingham & Reidy, 1998; Grant etal, 2004; Hueb- ner, 2003) Similaly, a spurious eorrelation—reflecting an admix- ture of two distributions—berween ASPD and prison violence may be present in samples that include both female and male jnmates. ASPD diagnoses are relatively uncommon among, women, and women generally commit less (and less severe) vio Tenee than men while ineareerated (Harer & Langan, 2001), whieh creates the appearance of an association between the two variables that is in reality due to gender differences. To address the limitations of previous research examining the relationship between ASPD and institutional behavior, the present study examined the utility of ASPD diagnoses in prospectively predicting general, aggressive, and violent forms of misconduct among male and female prisoners incarcerated across the United States, Newly admitted inmates were the focus of our investi ‘ion, as institutional misconduct tends to occur more frequently within the fist few years of confinement, prior to enviommental, and social adgptation (Flanagan, 1980), Although the main focus of this research is to address the predictive validity of ASPD. diagnoses, we also examined three related potential predictors, First, given evidence that ASPD is a dimensional rather than categorical construct (Marcus, Lilienfeld, Edens, & Poythress, 2006), we examined symptom counts to determine whether those ‘who met relatively more ofthe criteria were more likely to engage in fature misconduct than those who met fewer of the criteria Second, because mental health experts often lack high-quality information about whether examinees met CD criteria daring thet, youth, we examined separately the predictive utlty of the adult, ‘riteria forthe disorder (absent consideration of the CD eritetion) We likewise examined the predictive validity of CD symptom, counts in isolation, because item-response theory analyses have suggested thatthe CD criterion for ASPD is most strongly related to the latent antisocial trait tapped by the diagnosis Jane, Olt-ASPD AND PRISON MISCONDUCT 3 rmanns, South, & Turkheimer, 2007). Finally, because (a) expert testimony conceming a defendants alleged remorselessness is frequently provided to support an argument that he or she is, particularly dangerous (Weisman, 2007), and (b) remorsclessness Js a specific symplom within the ASPD criteria se, we examined its independent utility in predicting prison misconduct, Method Participants Participants were 353 male (n = 298) and female (w = $5) incarcerated offenders who were interviewed using the Structured Clinical Interview for DSM-IV Axis II Personality Disorders (SCID-U; Firs, Gibbon, Spitzer, Williams, & Benjamin, 1997) as part of larger study funded by the United States National Institute of Mental Health (NIMH) that examined antisocial and psyeho- Pathic personality traits among prisoners and substance abusers, Detailed information regarding this larger sample and addtional ‘measures administered has been deseribed elsewhere oythres Edens, eta, 2010; Poythress, Lilienfeld, et al, 2010), Patieipants ‘were generl-population inmates recruited from multiple medium- and maximum-security prison sites in Florida, Nevada, Oregon, and Utah, Recruitment targeted newly admitted inmates for whom official records of institutional misconduct were obtaized aller a follow-up period of 1 year. The mean age ofthe sample was 30.43 ‘years (SD = 6:82), Most partcipans self-identified as Caucasian (66.7%; 41.1% African American; 2.2% not indicated; 6.5% of participants additionally selC-identified as ethnically Hispanic). Measures SCIDAII, The ASPD module ofthe SCID-I! (First et, 1997) is a semistructured interview that assesses adult and childhood, DSM-IV (APA, 1994) criteria for ASPD. This measure is a widely used diagnostic instrument and exhibits adequate levels of inter- rater teliabilty for total symptom count and categorical ratings of ASPD (Mate ct al, 1997), Intorater reliability for this research project was established through observation of randomly selected ‘SCID-II interviews conducted by researc assistants, in addition to requisite file reviews, All observations were performed by the project manager (KSD), whose SCID-I scoring was treated a the criterion ageinst which research assistant scoring was compared, ‘The project manager traveled to each ste approximately every 6 months and observed two eases per visit, for a maximum of six visits, Concordance for this project was good for ASPD diggnoses («= 74,2 = 50) and total symptom count (intraclass correlations [icc] = 86, n = 46). Research assistants scored SCID-I items dichotomously as ci- ther present or absent, based on information extacted during a clinical interview and from a detailed review of panicipans* institutional files. A slight majority of participants inthis sample met criteria for ASPD (52.10%). Consistent with previous liter ture, the prevalence was much higher among men (56.9%) than women (25.5%). Rates ofthe ASPD diggnosis also varied across race, with more Aftican Americans (61.1%) mesting criteria than Caucasians (45.5%). The average number of adult symptoms met was 3.67, with considerable variability evident within our sample (SD = 197, range = 0 w 7). Institutional misconduct. Disciplinary records were obtained ftom participants’ respective prisons covering a period of I year alter study enrollment. Disciplinary infraction policies were ob- tained from each state's department of correction and compared to develop a standard coding system for categorizing documented misconduct. To be consistent with prior mets-analyte research on Institutional infactions (eg, Guy, Edens, Amhony, & Douglas, 2008), we coded infractions into three hierarchical outcome vari~ ables: (a) general infractions of any type (ext, possession of contraband), (b) aggressive ingfactions (including both physical and verbal aggression), and (©) physically violent inaction (eg, assault or battery with a deadly weapon). ‘The prevalence of misconduct ver the course of the study was relatively low, with the mean (SD) numberof infractions being 1.33 (3.33) for general, 73 (2.58) for aggressive, and 0S (27) for physically violent, and the modal number being 0 for each category. As such, the outcome ‘measures were dichotomized (0 = no infractions; 1 = one or more infractions). Over the l-year follow-up period, 40.2% of our sample incurred atleast one general iftaction and 25.5% incurred at least one aggressive inffacton, The base rate for physically violent infractions was low (4.2%), which is in Keeping with other research in this area (eg, Walters & Knight, 2010). As such, the ageressive infraction category primarily represents acts of verbal ‘aggression, though not exclusively Procedure Participants at each site were recruited randomly from lists of _zeneral-population inmates meeting basi inclusion criteria, which Fequired that participants speak English and sel@identify as Afri can American of Caucasian, We elected not to recruit individuals who self-identified primarily as members of other racial groups, given the paucity of evidence regarding the validity of ASPD ai Psychopathy in these groups. In addition, individuals currently receiving psychotropic medication for active symptoms of psycho- sis and those with an estimated 1Q sereening of less than 70 (Ammons & Ammons, 1962) were excluded. The study protocol was administered by research ssistants who were advanced eli jeal psychology or social work graduate students and who were extensively trained on data-collection procedures by the senior investigators. After a complete description of the study, written informed consent was obtained using procedures approved by university institutional review boards. Approximately 85% of i= dividuals approached to participate in the larger study chose to do s0. The protocol took on average 4.5 hours to complete (typically, ‘over two or three sessions), Participants received $20 as compen sation for their tim. Analytic Plan In recent years, the area under the curve (AUC) derived from receiver-operating-characteristc (ROC) curve analyses has gained ascendency as the effect sizeof choice in rsk-assessment research, (Mossman, 1994), Although AUC values ate informative regard ing the performance of an instrument or predictive tool in genera, ‘they usually do not directly address the question of how accurate predictions of a criterion are when they are provided, because ‘AUC values typically are not derived from any one partieular cut-off score Instead, they provide a global indicator of the per4 [EDENS, KELLEY, LILIENFELD, SKEEM, AND DOUGLAS formance of the predictor variable across all possible values ofthat predictor variable, Although not commonly reported, AUCS can be ‘computed for individual cut-off scores, in which ease they repre sent the midpoint hetween sensitivity and specificity for that particular score. ‘A more important concer about AUC values is that, when autempting t0 predict low base-rate events, high values ean be obtained even when predictions ofa eriterion occurring are wrong, ‘most of the time, due to the disproportionate number of false positives evident among these predicted to engage in the outcome. of interest (eg, future violent behavior; Steiner, 2003), As sue, 2 consideration of classical test-heary statistics (eg, Sensitivity, specificity), particularly those that are influenced by base rates (a, postive predictive power [PPP], negative predictive power INPP)), is also necessary when judging the accuracy of predictions ‘of misconduct among prisoners. In particular, rates of PPP quan- tify how frequently predictions offre violence are correct when based on the presence of an ASPD diagnosis Results Table 1 summarizes the performance of ASPD diagnoses as predictors of the thee types of disciplinary infractions. Across all three types of misconduct, diagnoses did not signifiantly predict these outcomes, with global performance indicators (phi, kappa) being close to zero in absolute magnitude, The differenti! rates of PP and NPP reported across the criterion measures are chiefly a function of the differing base rates of the three criterion measures, under investigation. Although ROC curves most commonly are used to examine predictive validity across an entire range of test scores oF diagnostic symptom counts, we computed these values for the ASPD diognostic cut-off point and obteined similarly negligible results: AUC (SE) = 5 (any ~ 03); 52 (oparessive 04); 51 (viotent = 08) ‘Table 1 provides results basod on diagnostic status independent of the presencelabsence ofthe CD criterion for ASPD. Disregard ing the need to meet this criterion resulted in higher prevalence of ASPD diagnoses (68.6%), which resulted in some minor dif- ferences in terms of classification statistics. Such differences are essentially moot, however, given that none ofthe global statistics was satstically significant, ROC results were similar to the neg ligible phi aad kappa results reported in Table 1: AUC (SE) = 52 (03, any); 53 (04, aggressive); and 49 (08, violent Next, we addressed whether predictive validity might be im- proved by examining symptom counts rather than dichotomous [ASPD diagnoses. For the adult-criteria symptom count, bserial correlations withthe three outcome measures were r = 10 (any), 04 (aggressive), and 01 (violent, all p values nonsignificant [AUC (SE) values from ROC analyses were .S5 (03) for any, .52 (03) for aggressive, and .52 (.07) for violent infractions, again all, nonsignificant, We next examined the childhood symptom counts of CD. Unlike the adult criteria, these demonstrated modest evi dence of predictive utility, with correlations of r = «16, p < .01 (any), 12, p< .05, (aggressive), 11, p< 05 (violent. Corre sponding AUC (SE) values were: .60 (03), p < 01 (any); 9 (04), p = .05 (aggressive); .67 (05), p = 05 (violent). We then examined the uty ofa total symptom count that combined both CCD and adult criteria into a single variable, This resulted in effet sizes that were virtually identical to (or slighty lower than) those achieved by the childhood symptom coust in isolation, ‘Next, we examined the lack of remorse criterion in relation to seh outcome. Phi coefficients ranged between —04 and 00, as, did kappa values (all p values nonsignifican), Similarly, AUC values were all nonsignificant, ranging from 48 to 51 Because our sample included a relatively small percentage of female inmates, we reconducted the analyses after dropping women from our sample. Among male inmates, results were virtually identical to those reported for the fll sample. Other than, weak trends for childhood-crteria symptom counts to relate (0 misconduct (r values between -10 and .14; AUC values between 57 and 68), ASPD was nota significant predictor of misconduct cross our thres outcome measure. [Because dichotomizing our criterion measures could have re- sulted in a loss of statistical power, we also conducted the preced- ing set of analyses using both negative binomial and Poisson, regression methods, which are appropriate when outcome mea sures are count data. Results ofthese analyses were highly similar to those reported for the dichotomous outcome measures. Further details regarding these indings are avilable from the frst author. Supplemental Analyses Given the overall weak results and lack of uility ofthe remorse- lessness criterion, we conducted exploratory analyses on the re ‘maining adult erteria to investigate whether individual symptoms might be of some use in forecasting instittional misconduct, The Diagnostic Efficiency Statistis for ASPD Diagnoses in Predicting Institutional Misconduct Table 1 SPD. Statistic Any Anrenive Pep 45 22 Nop. oes 563 Sensitivity $77 555 Speciicty 517 90 Overall 459 3 Phi ot Kenna ® 08 Note, PPP = postive prodicive power, NFP = ASPD without CD eequitement Ray Angee Violin 43 a7 269 at 983 et ns oss 533 a 22 667 79 m2 nt S14 518 56 572 an o ts is “1 00 os 03 o epave prdicive power, Overall = overall comeet slassicaton rate. None ofthe inferential atstical values repore is sgfcet at p= 05,ASPD AND PRISON MISCONDUCT 5 only individual criterion demonstrating any utility was “failure to conform to social norms with respect to lawful behaviors,” which yielded signitieant phi values for any and aggressive infactions (18 and .16, respectively, both p values = 01), but only mang ally significant AUC values (56, SE = .03, ps < .10). The base rate for this criterion in our sample was .8, however, whieh, not surprisingly, resulted in PPP values of only 44 for any inactions and 28 for aggressive infractions. In essence, in the rare cases in Which this criterion was not present, ils absence was @ good indicator that an infaetion would not accur (corresponding NPP. values were 84 and 93), but the presence ofthis exiterion Was an accurate predictor of misconduct only 44% of the time for any infraction and only 28% of the time for an aggressive infraction. Finally, given the generally poor performance of ASPD nostic information to prediet misconduct, one might question whether the criterion measures themselves might be problematic joperationalizations of institutional adjustment problems. To inves tigate this issue, we examined the association between our oul ‘come measures and correlates of institutional adjustment problems. ‘entified in earlier large-scale studies and meta-analyses (Cun- ningham & Sorensen, 2007; Cunningham eta, 2005; Diamond et al, 2012). These variables were age, education level, 1Q sereening, scores, ender, and race. Although we do aot wish to suggest that any of these variables should be used for rsk-assessment purposes. in practice, they are demonstrated correlates of disciplinary intr tions in prisons and thereby provide an opportunity to examine whether our erterion measures display a similar pattera of asto- ciations with other well-established predictor variables. As ean be scon in Table 2, several significant, modest effect-size estimates were obtained withthe three indicators of misconduct, indicating that our criterion measures possess sufficient enough reliability and validity tobe statistically predicted by established risk factors ‘We also examined these predictor variables using both negative binomial and Poisson regression, which again demonstrated results very similar to when the dichotomous criterion measures were Table 2 Predictive Validity for Known Correlates of Institutional Misconduct “Type of misconduct AUC GE) Vario aay “Aaaresaive Viste Age ass7 (09 = 08162 (08) Edocation level ~.11%/56 (03)" ~.147/59(03)"-.07/59(.08) 10 09133 (03) 1S 30 (04) 12165 C08)" PhirKappe (SE) ay ‘Aaeresive Voter Gender 08/05 (03), = 9300 01), Race 487.08 (00) 1510201)" [Nove Due to some missing dass point, sample sizes for dese analyses range fom 34 to 383, ention level was coded es { = no high school {iplome, 2 = completed GED, = high Schoo! diploma, 4 = some college mpletcd, 5 = completed college. Gender was coded 1 = female, 2 tral. Race was coded as 1 = Caucasian, 2 ~ Asean Ameria Spas. “pel examined, In fact, effects wore somewhat stronger than those ‘Table 2 for the dichotomized criterion measures, Fur~ is available from the first author upon request, Discussion ‘The use of psychiatric diagnoses to influence the disposition of legal cases has a long history. In particular, diagnoses related to snlisocial or psychopathic traits have been used to justify punitive sanctions (e.g. indeterminate commitment, capital punishment) for offenders, based in part on arguments that these disgnoses foretell violence end other criminal bchavior (Cunningham & Reidy, 1998; DeMatteo et al, 2011; Sevilla, 1999), We sought to address whether an ASPD diagnosis identified offenders newly admitted to prison who would engage in acts of institutional misconduct during their fist year of incarceration, Such findings, can inform the justice system about the utility of this diagnosis in dentifying high-risk inmates, which bears on sentencing decisions and security classifications. ‘Our results offer essentially no support fr the claim that ASPD Aiagnoses provide meaningful information conceming the instita= tional maladjustment of prisoners. The adult eriteria were not useful fa identifying who would commit diseipinary infactions over out 1-year follow-up. Remorselessnes in particular showed no relevance to identifying those who would engage in problematic bchavior over this time period, despite being commonly cited as an important risk {actor for future aggression (Cunningham & Reidy, 2002). Similar o the recent findings of Walters and Knight (2010), the only aspect of ASPD diagnoses that showed any potential utility for forecasting misconduct was CD symptom count, This finding js consistent with research demonstrating an essociation between, ‘more severe childhood antisocial conduct and risk of criminal and violent havior as an adult Farrington, 1991; Mason et al, 2004;, Robins, 1978), even among generally at-risk groups (¢4,patiens with schizophrenia; Swanson et al, 2006). In one study of emo- tional and behavioral problems, childhood conduct problems (e.g attacking people, destroying things) as reported by 10 year-old children and their paren and teachers best predicted a patter of serious violence in early adulthood (Mason otal, 2004). Further, specific constellations of adult antisocial behaviors may corre. spond fo the number of behavioral disturbances evidenced in childhood (Robins, 1978). ‘The overall poor performance of ASPD diagnoses in our study does not bode well regarding their predictive utility in US. pris fons, These Findings are consistent with previous meta-analytic work demonstrating low predictive validity for measures of psy~ chopathy within U.S. prison settings (Guy etal, 2008), although cfc sizes were somewhat stronger in non-US. prisons and forensic facilites, One could argue that our study provided an optimal oppor- tunity to demonstrate a relationship between ASPD diagnoses and prison misconduct, i such a relationship were to exist. Our research assistants conducted extensive interviews and had access to institi= tional file data concerning participants. Our relibility data indicated thatthe SCID-I ASPD module (Fist, Gibbon, Spier, Williams, & Benjamin, 1997) yiekded consistent scores aross examiners—much ‘more so than what bas been demonstrated for this diagnosis in the ‘DSW.5 field wials (APA, 2013; Preedman etal, 2013). Moreover, our criterion measures were based on comprehensive institutional records ‘of misconduct and displayed expected patterns of association with6 [EDENS, KELLEY, LILIENFELD, SKEEM, AND DOUGLAS known predictors of prison misconduct (eg, age, 1Q) that were generally of a magnitude similar to what bas been reported in other large-scale studies. If ASPD diagnoses are not useful indicators of risk in prison settings, what factors should be considered? As noted earlier, numerous individual risk factors have been demonstrated to be relevant to institutional risk, although some (e.g. race/ethnicity) raise important ethical/legal concems in regard 0 their potential use in legal decision-making. Some individual-level risk factors (eg, age, education, prior prison confinement) have een incor Porated into actuarial systems that have been shown in large-scale studies to classify prisoners into meaningful subgroups in tems of ‘isk level (Cunningham & Sorensen, 2006). Further, several struc- tured professional judgment (SPJ) measures have been shown to predict violence within prison and forensic settings (or reviews, see Otto & Douglas, 2010). The key point is that most robust risk assessment approaches do not rely on any single construct, but emp to incorporate un array of risk Factors Although not the focus of this partieuar study, anther impor- tant domain widely known fo influence rates of prison violence is ‘the role of the situational and environmental factors, which almost surely constrains the rates of physical aggression behind bars (Gadon, Johnstone, & Cooke, 2006; Sorensen, Cunningham, Vi- gen, & Woods, 2011; Tewksbury, Connor, & Denney, 2014), Forensic practitioners working in this area might consider the PRISM model (promoting risk interventions by situational man- agement; Cooke & Johnstone, 2010) as a means of moving beyond, simple individual-level risk factors when they engage in the pro- cess of institutional risk assessment and management. Although an exhaustive review goes beyond the scope ofthis aticle, in brief, the PRISM model focuses attention on empirically supported sitwational/environmenta factors, such as structural aspects (@, supervision and security level), staf features (e.g, age, expeti~ ence), temporal aspects (eg, titsing of assaults), location, row ing, organizational management practices and relationships (eg. ‘ationships between administrators and fontline stall), and pro- gram availability for inmates. Models such as PRISM might lead to more nuanced risk-assessment and management strategies that go beyond a primary focus on the individual offender. Although our study benefited fom several methodological strengths, there were aso important limitations we should note that ‘may inform future work in this ares, Fist the rates of physically violent infractions were low, raising concems about the stability of ur resis fortis eriterion measure. Tat said, our study is typical in this regard, in tha the incidence of serious ts of violene in prison tends tobe low, even among inmetes who have committed violent crimes (eg, capitel murder) in the community (Cunningham & Reidy, 2002; Cunningham otal, 2008; Cunningham & Sorensen, 2007; Cunningham etal, 2005; Edens et al, 2005). Second, our sample did not contain @ large sample of female offenders, whieh constrains the generalizability of our resulis. Third, it is unclear to ‘what extent our results would generalize to less secure contents (e., rminimam-security facilites, community corrections). It is conceiv~ able that ASPD might have greater uty in such stings, particularly ifthe base rate ofthe “failure to conform to social norms with respect ‘tw Lawl behaviors” symptom isnt so close to 100%, a itwas in our data, Fourth, given Guy et al.’s (2005) findings of lower predictive validity forthe closely related concept of psychopathy within US. prisons shan non-US prisons, our findings should not be considered tw generalize beyond the U.S. Fifth, our follow-up period was only ‘one year in Length, Akhough results might have differed over a more extended time frame, such an outcome seems unlikely piven that () we captured the (ary) period of maximum risk for prisoners 10 engage in misconduct and (b) risk decreases considerably the longer psoners are incarcerated, Finally, one could argue that our means of ‘operationalizing the severity of ASPD (i, ‘otal symptom counts) does not correspond to what forensic practitioners mean when they describe a defendant as having a “severe case” of this disorder. ‘Although there may be some merit to this eis, the burden Would fall on such practitioners to provide evidence that their altemative means of quantifying severity has predictive uty. In conclusion, the results ofthis study offer litle or no support for the claim that ASPD diagnoses identify inmates who go on to engage in institutional misconduct in prison. In forensic context, the use of this diagnosis to support claims that defendants will be 1 serious threat to others, even if incarcerated in prison, would seem to rest on dubious empirical grounds References American Paychiatic Asociation, (1994), Diagnostic and static man: ual of mental disorders (Ah ed). Washington, DC: Author. American Psychiatrie Associaton, DSS Task Force, (2013). DSM Diggostc and saitical manual of mental disorders (th ed). Wash- ington, DC: American Psyebiaie Pblishing. ‘Ammons, RB, & Ammons, C.H. (1962). The Quick Test (QT): Provi- sional manual, Pochologcal Repors, 1, UL-161 (Cook, K. 2008). Chasing justice: My story of freeing myself after 080 decades on death row for a crime J did't commit. New York, NY: HaspesCotins. Cooke, D. J, & Johnstone, L, (2010). Somewhere over the rainbow Improving violence risk managemcut in institinal stings. The Je ational Journal of Forensic Mental Health, 8.150196, doi: 1080) 4990013.2010.526868 Cook x State, $21 S.W.24 600 (Tex. Crim. App. 199. (Cunningham, MD, & Reidy, T-1.(1998), Antisocial personality dsocde ‘and psychopathy: Diagnostic dilemmas in casing paters of at- Social behave in sentencing evaluations. Behavioral Sciences & the Lays, 16, 333-381, do:10.1002/S1CI)1099-0798(199822)16:3<333: AID-ASLS14=30.002-N (Cunningham, M.D, & Reldy, T 1. (2002). Violence risk assessment a federal capil sentencing: Inviduaization, generalization, relevance, tnd scientific standards. Criminal Justice and Behavior, 29, 512-837, hi10.1177000385402236731 (Cunningham, M.D, Reidy, T. J, & Sorensen, JR (2008) Assertions of “future dangerousness” at feral capital sentencing: Rats and core Ines of subsequent prison misconduct and vielence. Law and Human Behavior 42, 46-63, oi101007S10979-007-9107-7 Cunningham, M.D. & Sorensen, JR. (2006). Actuarial models for sessing priton violence risk: Revisions and extensions of the risk assessment scale for prison (RASP). Assessment, 13, 253-265. do: ot 7H073191106287991 Cunningham, M.D., Soreasen, JR, & Reidy, T 1.2005), Ar actuarial, ‘model for assessment of prison violence among maximum seunty inmates. dssesiment, 12, 40-49, doi:1017HIO7SI9 1104272815 Cunningham, M.D, & Soremen, JR. (2007). Preditive factors for olen misconduet in close custody. The Prison Journal, 87, 241-253. ot0,170032885507305752 DeMatteo, D., Mure, D.C., Anumba, N.M., & Keesler, M. E011 Forensic mena! health asessments i death penalty cases, New York, NY: Oxford Universiy Press, di10.1093/seprof.os780195385809 01.0001,ASPD AND PRISON MISCONDUCT 7 Diamond, 8 Moris, R. G.,& Bares J.C. 2012), dividual and group 1Q predict inmate violence. Inelligence, 40, 115-122. do: 10.1016} sitll 2012.01.010 Edens, J.F., Buffington-Vollm, JK, Keilen, A, Roskamp, P. & An- tory, C2005), Prediions of funize dangerounes in eaptl murder ‘als Is time to “sinvent the whee!” Law and Human Behavior, 22, S586, dO, 1007/1 09790051399. Edens, JF & Cox, 12012). Examining the prevalence, ole and ispact of evidence regarding antisocial persoality, sociopath and payehopse thy in capital cases: A survey of defense team members. Behavioral ‘Sciences & the Law, $0, 238-288, do, 1002/1 2008 Fariagion,D.P. (1991). Chldiood aggression and adult violence: Ely Precursors atrie outcomes. nD. J Peper, &K. Ht. Robin (Ede), The development and reatment of ldhood aggression (gp. 189-197). Hillsdale, NJ Etbaum, Fest MB. Gibbon, M, Spite, RL, Willams J.B. & Beajamin, LS. (0997). eer’ guide forthe Sracared Clinical Interview for DSM-IV Aus I personaly disrders (SCID-I). Washington, DC: American Psychiat Pres Flanagan, T. (1980). Time served and instiutional misconduct Patterns of involvement in disciplinary infactions among lng-term an shorter inmates. Jownal of Criminal Jutice, 8, 387-867. do10.1016'0087- 2352(80)90111-7 Freednan, R, Lewis, D. A. Michels, R, Pine, D. S, Schult, S. K “Tammings, CA... Voge, J 2013). Te nal field was of DSM [New blooms and old thors. The American Jenrnal of Paychatry, 170, S.do10.1176ppiap 2012. 12091189 (Gadon, L, lostone, , & Cooks, D. 2006), Situational vrabls and institutional violence: A systematic review ofthe literaure. Clinical Paychalogy Review, 2, S1S-S8, dis0.10165 op, 2006.02.02 Glenn, AL, Johnson, A. K, & Raine, A. (2013). Anos personality shsordec: A curent review. Curent Prchitry Reports, 18, AZI=A34 1010781 1920-013.0427-7 (rant, B.F.,Hasn,D.S, Stinson, FS, Dawson, D. A, Chow, PS, Ros, 1.W, & Pickering, RP. (2004), Prevalence, colts, and disability of personality disorders in the United States: Rest from the National Epidemiologic Survey on Alcohol and Related Conditions. Journal of Clinical Pech, 65, 988-958, do10 088ICP ¥6500711 Gross, SR, O'Brien, B, Hu, C, & Keasedy, EH. 2014). Rate of filse convieton of criminal defendants who ae sentenced to death, Proceed Ings of the National Academy of Sciences of the Unted Stats of “America, Advance online publication, doi:10.1078poas. 1306417111 Gay, LS, Edens, JF, Auhoay, C., & Douglas, K. 8. 2005). Does Teychopathy petit intone micondiet among aks? metae tuulyic investigation. Jounal of Consulting and Ciical Paychology 78, 10861068, di:10.1087/0022-006X.73.6.1086 Haret, M.D, & Langan, N.P. (2001) Gender differences in pedis of ‘nton violence: Assessing the predictive validity of anak caesieation| fystem, Crime & Delinguency, 47, 513-536, doi:10.1177 1 11287010¢7o08002 Huvbner,B.M. (2003). Adminitative determinants of inmate violence: A rmullevel analysis. Journal of Criminal Justice, 31, 1OP-UT. do: To.to161S0047-28s2402)00218-0 Jane, JS, Ottmanss,.F., South, S.C, Tuskheimer, FE. 2007). Gonder iss in diagno eiteri for personality disorder: An item response ‘theory analysis. oural of Abnormal Psychology, 116, 165-175. di 10, 103710021-843X.116. 1.165 Maffei, C, Fossa, A, Agostni L, Banaco, A, Bagnate, ML, Deborb 1,» Pera, M. (1997) Interater lability and internal consistency ‘of the Strstured Clinical Interview for DSM-IV Axis I personality tisorders(SCID-ID, Version 2.0, Journal of Personality Divorders, 11 279-284, doi:10.1521/pedi-1997.113279 Marcos, D.K., Lien, $0, Edens, JF, Poythress, N,Q, (2006), 's antsocil personality disorder continuous or ctegvial? A taxome'= rie analysis, Prcholopieal Medicine, 36 ss0s32917oank23s Masur, W.A, Kosoman, R, Hawking, 1, Heres 1, Leng, LI, {& MoCauey, FM. (208), Predicting depression, sail phobia, nd violence ‘realy abutbood lon cined behavior blens Journal of he American Academy of Child & Adolescent Pachiany, #3, 3-815, dot 101097F tooo3s3.200903000.0012 Morey, LC. & Ochoa, ES. (1989). An investgin of adherence to dagasic ei: Cina aos ofthe DSUCAT penealy iors. ural of esol Disorders, 3 180-192 doi 10152Upe 1989.33 180 Mossman, D_ (1983), Asesing prisons of knee: Betyg acer abut sccuncy Jwnal of Consult and Clnal Psychology, @, 78-792 di: 10 71022.006% 52.475, (to, RK, & Dough KS. (Es). 2010, Hino of ence rik ass. rent, New York, NY: Rottedee Poy, NG, Edens JF, Skoum JL, Lie. 0 Douglas KS, Pick, Jy». Wang. 2010) Mesifingsitsypes amimg oes with aso ponaiy derder A caster tl. Journal of mammal Poche, 19, 399-80 100570018611 Poydbess, NG, Lilienfeld 5.0, Shem, JL, Douglas, KS, Edens, F, psc, Me Paik, C.J (2010) Using the PCL-R to elp estimate the ‘aly of wo seteepor measuts of peop wi oflendc Asexmen, 17, 206-219, oc10177OTSI91108381715, Robins, LN. (1978), Sudy childhood pins of ad associ beavir Replcaioas fom lng ses Pyeholgical Mein, 8, 611-622 0 1017S0033291700018821 Sevilla, CM. (199), Antisocial personality oder: station fete death aly? Saal of Conemporiry Legal ss, 10, 247-292 Sore, | R, Cumingham, M.D, Vien, MP, & Woods, S.0. 01D). Seow stale om pron sll A deserve ans. Journal of Criminal sie, 39, 148-150, 10.1016) inj. 2011.01.02 Sesin, D-L, 2003) Diagnosing tes: Using aad misusing dgrosic and sersting ts Journal of Personality Assesment, 81, 209-219. di01207 SIS32T7S2sPA8103 03 Swansn,1.W, Starz M.S, Van Dom, RA. Ebogen, EB, Wagn, H.R, Rove, RUA,» Liberia, J. 2006 Apso stay of viet Thavir in persons with schizmpzeia Aches of General Pacha. 63, 519, do 10 oacpye 53.5490 ‘ewdsbary, R, Cour D. P,& Demy, AS. 2014), Disp inctions ahi hans An exption of perio a! destin theo, rina asice Review, 39, 201-218 di 101 TIOTSHO1GE 9965 Tech, Hl, & Adams, K. (886), Ptology aad diruptiveness among peson iam Joura of Research in Crime and Dele, 23, 7-21, da \0117 tnzaszrans2s001002 Wier, G.D., & Knight, RA. 2010) Anse persalty deer with and wow scent shidhood cond donde: Dawe rake ace? Joural of Personality Disorders, 24, 288-201. da. 1S2ipe 2010242 258, Ware, J 1, Busts, M. 2013} Th maitiezed construct of paypal [Asocition wth APD, cll, an criminal christie among male an file inmate. The Intemational Journal of Forensic Mental Heath 12, 265-173. do 10 108013999013 201585739 Wares 1, Bumste, M, Sob, 8, C, Chain, P, Bale R, & Friend R (G12), Peony dsr and volo among fale piso omits Jou ral of he American Academy ofPychiaty andthe La, 30, $020. ‘Weisman R. (2007). Remorse an pyshoathy a th penal pas ofthe epi til: How payehity' iw oral wary” bd he ease or deh Stules in Law, Polis, and Soci, 41, 187-217. do:10101681059- ssson0007 IsTI-Usel, doist0.1017 Reocived May 30, 2014 Revision received July 24, 2014 ‘Accepted July 24, 2014