The Validity of the VRAG with Female Inmates

Download Report

Transcript The Validity of the VRAG with Female Inmates

The Validity of the Violence
Risk Appraisal Guide (VRAG)
with Female Inmates
Mark E. Hastings, Ph.D., CSOTP
Loudoun County Mental Health Center
George Mason University
Private Practice
Shilpa Krishnan, M.A.
Jeff Stuewig, Ph.D.
June Tangney, Ph.D.
George Mason University
Paper presented March 8, 2008 at the annual meeting of the American PsychologyLaw Society in Jacksonville, FL
Overview



Violence risk assessment has become a staple
of forensic psychological practice (Monahan,
1996).
Females represent one of the fastest growing
segments of the inmate population (Bureau of
Justice Statistics, 2006).
Despite the proliferation of actuarial risk
assessment instruments for men, none have
been validated for use with females.
Research To Date

Only two studies to date have examined the use of
VRAG with females.

Harris et al. (2002)


VRAG scores were unrelated to violent recidivism in a
small sample (n=59) of female forensic inpatients.
Harris et al. (2004)

A modified version of the VRAG was equally predictive of
future community violence in a large sample of civillycommitted male (n=423) and female (n=318) patients.
Research To Date (cont.)

Recent incremental validity study of the VRAG.

Edens et al. (2006)


Analyzed data used in Harris et al. (2004) study and found
that the VRAG did not account for any unique variance in
violent outcomes beyond that predicted by the PCL:SV.
No study to date has looked at the incremental or
predictive validity of the VRAG in a female inmate
sample.
Research Questions
Are there gender differences in the:

VRAG item and total score means?

Correlations between the VRAG and other
measures of antisociality and aggression?

VRAG’s prediction of institutional misconduct and
recidivism?

Incremental validity of the VRAG in relation to the
Hare Psychopathy Checklist (PCL:SV)?
Study Sample




473 inmates charged with a felony enrolled in
a longitudinal criminal recidivism study at a
suburban adult detention center.
344 male inmates and 129 female inmates.
Most of sample was being detained pre-trial
or was serving jail sentence at time of
admission to study.
Approximately half of sample was sentenced
to state prison time.
Measures

Violence Risk Appraisal Guide (VRAG; Quinsey
et al, 1998).


Hare Psychopathy Checklist: Screening Version
(PCL:SV; Hart et al. 1995).


Inter-rater reliability coefficient for 52 cases (.90).
Inter-rater reliability coefficient for 54 cases (.85).
Personality Assessment Inventory (PAI; Morey,
1991).

ANT, AGG, AGG-P, and Violence Potential Index (VPI)
were utilized in concurrent validity analyses.
Outcome Measures

Institutional Misconduct

Was coded using jail records obtained between time of enrollment in
the study and release date. Three types were used in analyses:




Number of Incidents
Number of Formal Charges
Number of Charges for Violence
One-Year Post-Release Recidivism

Self-report of arrests and “undetected” recidivism. Four types were
used in analyses:




Any Arrest
Any Undetected Recidivism
Any Arrest or Undetected Recidivism
Any Arrest for or Undetected Violent Recidivism
Participants
Mean Score (SD)
Chi-Square
Male
Females
344
129
Age
31.6 (9.7)
35.1 (10.3)
n.s.
Education level
11.5 (2.2)
12.4 (2.1)*
Χ2 = 29.7*
N=473 Urban Jail Inmates
Race
n.s.
Caucasian
32.8%
44.2%
African-American
45.6%
44.2%
Hispanic
10.1%
3.1%
Other
11.5%
8.5%
Wonderlic IQ score
93.2 (13.7)
94.2 (14.2)
n.s.
PCL:SV Total score
12.9 (4.9)
10.5 (4.2)*
Χ2 = 45.1**
* p <.05; ** p <.01
VRAG Total & Item Score Means
VRAG
Mean Score (SD)
Chi-Square
Male
Female
VRAG Total Score
8.0 (8.1)
4.9 (6.7)
n.s.
Item 1: Lived with parents till 16
1.3 (2.4)
1.7 (2.2)
n.s.
Item 2: Elementary school problems
2.1 (2.4)
1.0 (2.3)
Χ2=17.5**
Item 3: History of alcohol problems
0.3 (1.1)
-0.02 (1.0)
Χ2=12.0**
Item 4: Marital status
-1.0 (1.4)
-1.6 (1.1)
Χ2=17.4**
Item 5: Nonviolent criminal history
2.0 (1.9)
2.2 (1.7)
n.s.
Item 6: Failure on conditional release
2.1 (1.4)
2.1 (1.4)
n.s.
Item 7: Age at index offense
-0.7 (2.8)
-1.7 (3.1)
Χ2=17.5**
Item 8: Victim injury
1.9 (0.5)
2.0 (0.1)
n.s.
Item 9: Any female victim
0.7 (0.7)
0.8 (0.6)
Χ2=4.3*
Item 10: DSM-Personality disorder
-1.8 (1.0)
-1.9 (0.8)
n.s.
Item 11: DSM-Schizophrenia
1.0 (0.6)
0.9 (0.5)
n.s.
Item 12: Hare Psychopathy Score
0.3 (2.5)
-0.8 (2.3)
n.s.
* p <.05; ** p <.01
VRAG Concurrent Correlations
VRAG Total Score & Individual Items
PAI: ANT Scale
PAI: AGG Scale
Males
Females
Males
Females
VRAG Total Score
.53**
.34**
.49**
.38**
Item 1: Lived with parents till 16
.14*
-.01
.19**
.18*
Item 2: Elementary school problems
.33**
.26**
.37**
.38**
Item 3: History of alcohol problems
.30**
.10
.31**
.11
Item 4: Marital status
.19**
-.03
.10
-.02
Item 5: Nonviolent criminal history
.23**
.17
.15**
.14
Item 6: Failure on conditional release
.21**
.18*
.18**
.13
Item 7: Age at index offense
.17**
.09
.20**
.09
.00
-.03
-.06
-.08
Item 9: Any female victim
.15**
-.23**
.08
-.14
Item 10: DSM-Personality disorder
.17**
.26**
.14**
.11
-.07
-.04
-.10
-.04
.42**
.35**
.38**
.24**
Item 8: Victim injury
Item 11: DSM-Schizophrenia
Item 12: Hare Psychopathy Score
* p <.05; ** p <.01
VRAG Concurrent Correlations
VRAG Total Score & Individual Items
PAI: AGG-P Scale
PAI: VPI Index
Males
Females
Males
Females
VRAG Total Score
.52**
.35**
.48**
.30**
Item 1: Lived with parents till 16
.17**
.15
.15**
-.02
Item 2: Elementary school problems
.39**
.34**
.29**
.32**
Item 3: History of alcohol problems
.32**
.09
.33**
.20*
Item 4: Marital status
.15**
-.02
.15**
-.16
Item 5: Nonviolent criminal history
.16**
.14
.23**
.15
Item 6: Failure on conditional release
.17**
.13
.25**
.19*
Item 7: Age at index offense
.24**
.12
.12*
-.02
Item 8: Victim injury
-.04
-.05
-.06
.05
Item 9: Any female victim
.10
-.13
.07
-.09
.19**
.08
.13*
.15
-.10
.00
-.14*
-.03
.35**
.19*
.41**
.37**
Item 10: DSM-Personality disorder
Item 11: DSM-Schizophrenia
Item 12: Hare Psychopathy Score
* p <.05; ** p <.01
VRAG Correlations & ROC’s with
Institutional Misconduct
Male Inmates
Female Inmates
(n = 323)
r
(n =110)
AUC
r
(95% CI)
AUC
(95% CI)
Number of Incidents
.28**
-
.04
-
Number of Formal
Charges
.27**
.67**
-.01
.51
Number of Charges for
Violence
.13*
* p <.05; ** p <.01; † p < .10
(.62-.75)
.58†
(.50-.66)
(.40-.63)
-.05
.40
(.26-.53)
VRAG Individual Item Correlations
with Institutional Misconduct

For male inmates:


Items 1 (Lived with Parents till age 16), 2
(Elementary School Problems), 7 (Age), 10
(Personality Disorder), and 12 (Psychopathy)
were significantly correlated with institutional
misconduct.
For female inmates:

None of the individual VRAG items were
significantly correlated with institutional
misconduct.
VRAG Correlations & ROC’s with
One-Year Recidivism
Male Inmates
Female Inmates
(N = 193)
(N = 54)
r
AUC
r
AUC
Arrest for Anything
.34**
-
.08
-
Undetected Offense
.33**
-
.17
-
Arrest or Undetected
Offense
.34**
.71**
.21
.62
Arrest or Undetected
Violence
.37**
* p <.05; ** p <.01; † p < .10
(.62-.79)
.76**
(.68-.85)
(.47-.78)
.25†
.73†
(.57-.88)
VRAG Individual Item Correlations
with One-Year Recidivism

For male inmates:


Items 2 (Elementary School Problems), 5
(Nonviolent Criminal History), 6 (Failure on
Conditional Release), and 12 (Psychopathy) were
significantly correlated with recidivism.
For female inmates:

None of the individual VRAG items were
significantly correlated with recidivism, except
for Item 12 (Psychopathy) and violent recidivism.
Incremental Validity – Institutional
Misconduct
Number of Institutional Charges for Non-Violence
Male Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.06
.05
SE B
.03
.02
β
.14
.16
Δ R2
.05**
.02*
Female Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.04
-.01
SE B
.04
.03
β
.10
-.03
Δ R2
.01
.00
Number of Institutional Charges for Violence
Male Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.00
.01
SE B
.01
.01
β
.01
.13
Δ R2
.01
.01*
Female Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.00
-.00
SE B
.00
.00
β
-.01
-.05
Δ R2
.00
.00
* p <.05; ** p <.01
Incremental Validity – One-Year
Recidivism
Any Recidivism (Arrest and Undetected Offenses)
Male Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.06
.09
SE B
.03
.02
β
.14
.30
Δ R2
.08**
.07**
Female Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.14
.06
SE B
.07
.06
β
.25
.14
Δ R2
.07†
.02
Any Violent Recidivism (Arrest and Undetected Offenses)
Male Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.01
.02
SE B
.01
.01
β
.09
.32
Δ R2
.06**
.08**
Female Inmates
Step 1: PCL:SV Total
Step 2: Hare-less VRAG
B
.02
.01
SE B
.01
.01
β
.29
.16
Δ R2
.09*
.03
* p <.05; ** p <.01; † p < .10
Summary



Female inmates scored lower on the VRAG overall and on several
individual items (Items 2, 3, 4, 7, and 9).
The VRAG total score evidenced small to moderate correlations with
PAI measures of antisociality and aggression for both genders.
However, aside from Item 2 (Elementary School Problems) and Item
12 (Psychopathy Score), none of the individual VRAG items were
consistently significantly correlated with other measures of aggression
for female inmates.
Contrary to the pattern of results observed with male inmates, for
female inmates:



Individual VRAG items and Total Score were not significantly correlated
with or predictive of any of three different measures of institutional
misconduct
Individual VRAG items and Total Score were not significantly correlated
with or predictive of any of four different measures of one-year postrelease recidivism
The VRAG did not significantly contribute to the prediction of either
institutional misconduct or one-year recidivism above and beyond that
accounted for by psychopathy.
Conclusions

The VRAG seems to be more valid for male inmates
than for female inmates, and some of the individual
VRAG items behave quite differently with female
inmates.

Any relationship between the VRAG total score and
various concurrent or outcome measures for female
inmates appears to be driven primarily, if not solely,
by psychopathy.

The VRAG should not be used with female inmates
at this time.