Presentation is loading. Please wait.

Presentation is loading. Please wait.

Relative Efficacy of the PAI, PCL:SV, and VRAG in Predicting Institutional Misconduct and Short- term Recidivism Mark E. Hastings, Ph.D. Loudoun County.

Similar presentations


Presentation on theme: "Relative Efficacy of the PAI, PCL:SV, and VRAG in Predicting Institutional Misconduct and Short- term Recidivism Mark E. Hastings, Ph.D. Loudoun County."— Presentation transcript:

1 Relative Efficacy of the PAI, PCL:SV, and VRAG in Predicting Institutional Misconduct and Short- term Recidivism Mark E. Hastings, Ph.D. Loudoun County Mental Health Center George Mason University & Jeff Stuewig, Ph.D. June Tangney, Ph.D. George Mason University Paper presented March 2, 2006 at the annual meeting of the American Psychology-Law Society in St. Petersburg, FL

2 Main Study Questions How well do various PAI scales predict institutional misconduct? –Previous research shows ANT and AGG significantly correlate with institutional misconduct (Buffington- Vollum et al., 2002; Edens et al., 2001; Walters, Duncan, & Geyer, 2003) How well do various PAI scales predict short-term recidivism? –Two prior studies show ANT and AGG significantly correlate with recidivism in female inmates and male inmates referred for forensic evaluation in federal prison system (Salekin et al., 1998; Walters & Duncan, 2005).

3 Violence Potential Index (VPI) The VPI consists of 20 features of the PAI profile that are congruent with research on the assessment of violence (e.g., impulsivity, agitation, lack of empathy, history of antisocial behavior). Wang et al. (1997) – VPI significantly correlated with staff ratings of aggression on the Overt Aggression Scale (OAS). Caperton et al. (2004) – VPI significantly correlated with any and verbal disciplinary infractions. No study to date has examined the VPI and prediction of recidivism.

4 Study Participants N=326 male inmates incarcerated at large urban jail. Age = 31 (s.d.= 9.7; range= 18 to 69) Race = 44.4% African-American, 33.9% Caucasian, 9.3% Mexican American/Other Hispanic, 3.7% Asian/Pacific Islander, 3.7% Mixed, 2.8% Other, & 1.2% Middle Eastern Wonderlic IQ Score = 93.19 (s.d.= 13.77, range= 67 to 138) Education Level = 11.62 years (s.d.= 2.18, range= 3 to 19) WRAT Reading Standard Score = 91.54 (s.d.= 16.68, range= 44 to 120) PCL:SV Total Score = 12.88 (s.d.= 4.96, range= 1 to 22) VRAG Score = +7.62 (s.d.= 8.07, range= -18 to +25) Violence Potential Index = 5.87 (s.d. = 4.1, range 0 to 19).

5 Practical Application Of “Touch Screen Tablet” For Standardized Interview Audio and visual presentation accommodates participants with minimal reading ability Touch-screen response mode does not require familiarity with computers Circumvents social desirability demands of face-to-face interviews

6 PAI Correlations SCALESPCL:SV Part 1 PCL:SV Part 2 PCL:SV Total VRAG ANT.28**.49**.44**.53** AGG.21**.47**.39**.50** MAN.24**.31**.32**.43** PAR.18**.28**.26**.39** BOR.10.36**.27**.38** DOM.24**.20**.26** VPI.25**.45**.41**.50** Note: N=326; p <.05* p <.005**

7 Jail Behavior Institutional misconduct data were collected from official jail records and were classified into four categories: –Physical Acts (e.g., assaults, setting fires, etc.) Base Rate = 6% –Verbal Acts (e.g., threats, curse and abuse, etc.) Base Rate = 5% –Defiance (e.g., refuse order, contraband, etc.) Base Rate = 25% –Other (e.g., self-mutilation, banging on cell door, etc.) Base Rate = 4%

8 Predicting Jail Misconduct Note: N = 326; p <.05* p <.005** Scales/Jail BehaviorPhysical ActsVerbal ActsDefianceOther PCL:SV Total.08.13*.23**.12* PCL:SV Part 1.03.10.16**.10 PCL:SV Part 2.10.13*.24**.11* VRAG.12*.03.24**.14* VPI.11*.08.21*.07 ANT.09.06.23**.05 AGG.12*-.02.14*.07 MAN.13**.11.24**.07 PAR.18**.04.17**.00 BOR.06 -.04.10

9 Receiver Operating Characteristics (ROC)

10 Receiver Operating Characteristics (ROC) for Any Disciplinary Infraction Scale/ROC DataAUCSESign.95% CI PCL:SV Total.614.035.002.544-.683 PCL:SV Part 1.576.037.504 -.648 PCL:SV Part 2.613.033.002.548 -.677 VRAG.654.038.000.579 -.728 VPI.625.036.001.555 -.695 ANT.619.036.001.548 -.690 AGG.598.036.007.527 -.669

11 Recidivism 1-year Post Release Participants were contacted either by phone or in person one year after their release from incarceration. Participants were asked about whether they had been formally arrested for or engaged in any of several types of criminal behavior in the previous year.

12 Percentage of participants self-reporting arrest and/or criminal behavior

13 Percentage of participants that report criminal behavior versus arrest No reports of arrest or offense for robbery, murder, kidnapping, or arson. One report of arrest for a sexual offense. No report of arrest for prostitution

14 Predicting Short-term Recidivism Note N= 121; p <.05* p <.005** Scale/RecidivismSelf- Report Arrest # of Diff. Offenses Undetected Offenses # of Diff. Offenses Any Offense Violent PCL:SV Total.23*.13.31**.28**.32**.17 PCL:SV Part 1.04-.06.16.15.16.03 PCL:SV Part 2.36**.31**.37**.35**.40**.28** VRAG.33**.21**.33**.42**.30**.36** VPI.28**.19**.31**.32**.29**.28** ANT.27**.16*.36**.41**.35**.25** AGG.19*.18*.27**.17.33**

15 Receiver Operating Characteristics (ROC) for Self-Report Arrest Scale/ROC DataAUCSESign.95% CI PCL:SV Total.618.051.026.518 -.717 PCL:SV Part 1.523.053.657.420 -.627 PCL:SV Part 2.705.047.000.614 -.797 VRAG.681.053.002.578 -.785 VPI.670.049.001.575 -.765 ANT.662.049.002.565 -.758 AGG.617.051.026.518 -.717

16 Receiver Operating Characteristics (ROC) for Undetected Offenses Scale/ROC DataAUCSESign.95% CI PCL:SV Total.679.051.001.578 -.780 PCL:SV Part 1.598.054.070.492 -.703 PCL:SV Part 2.705.050.000.607 -.820 VRAG.687.058.002.573 -.802 VPI.687.050.001.588 -.785 ANT.725.049.000.629 -.821 AGG.609.053.044.506 -.712

17 Receiver Operating Characteristics (ROC) for Violent Offense Scale/ROC DataAUCSESign.95% CI PCL:SV Total.630.062.046.508 -.752 PCL:SV Part 1.523.065.720.397 -.650 PCL:SV Part 2.701.057.002.589 -.814 VRAG.741.069.001.606 -.875 VPI.690.055.003.583 -.797 ANT.679.051.006.578 -.780 AGG.720.056.001.611 -.829

18 Overall ROC Performance Order/Type of Recidivism Any Jail Disciplinary Infraction Self-Report Arrest Undetected Offense Violent Offense FirstVRAG (AUC=.654) PCL:SV Part 2 (AUC=.705) ANT (AUC=.725) VRAG (AUC=.741) SecondVPI (AUC=.625) VRAG (AUC=.681) PCL:SV Part 2 (AUC=.705) AGG (AUC=.720) ThirdANT (AUC=.619) VPI (AUC=.670) VPI & VRAG (AUC=.687) PCL:SV Part 2 (AUC=.701)

19 Classification Accuracy for Recommended Cut Scores for the VPI Self-reported arrest –Moderate (VPI ≥ 9): Sensitivity =.228; Specificity =.875 –Marked (VPI ≥ 17): Sensitivity =.018; Specificity = 1.00 Self-reported undetected offenses –Moderate (VPI ≥ 9): Sensitivity =.208; Specificity =.896 –Marked (VPI ≥ 17): Sensitivity =.014; Specificity = 1.00 Self-reported arrest or undetected offenses for violence –Moderate (VPI ≥ 9): Sensitivity =.320; Specificity =.865 –Marked (VPI ≥ 17): Sensitivity =.000; Specificity =.990

20 Conclusions The VPI, ANT, and AGG scales were moderately correlated with the PCL:SV and VRAG. Several PAI scales performed as well or better than the PCL:SV and VRAG in predicting physical acts of aggression and defiance within the jail. However, the correlations for all types of misbehavior were generally small. The VPI, ANT, and AGG scales performed as well or better than the PCL:SV and VRAG in predicting self- reported arrest, undetected offenses, and any violent recidivism.


Download ppt "Relative Efficacy of the PAI, PCL:SV, and VRAG in Predicting Institutional Misconduct and Short- term Recidivism Mark E. Hastings, Ph.D. Loudoun County."

Similar presentations


Ads by Google