Moving from Principle to
Execution
Applications of the Risk-Dosage
Relationship
Kimberly Sperber, PhD
Support for the Risk Principle
• Hundreds of primary studies
• 7 meta-analyses
• Men, women, juveniles, violent offenders, sex
offenders
• Programs that target higher risk offenders are
more effective
• Reductions in recidivism are greatest for higher
risk offenders
• Intensive interventions can harm low risk
offenders
2002 UC Ohio Study of HH’s and CBCF’s
20
-6 -5
-10
-18
-20
5
3
3
3
2
9 10 10
8
8
7
6
10
15
12 12 12 13 13 13
0
-2 -2
-15 -14
Probability of Reincarceration
2002 UC Ohio Study of HH’s and CBCF’s
Treatment Effects For High Risk Offenders
40
21 22
24 25
27
30
34
32
30
-30
-40
-34
n
tio
ia
y
oc
nt
ss
A
ou
ns y
C C ng C
tio rm
m
EO oni A rec n A
ra
ah VO or tio
og
C
M o
Pr
va
y
l
d
y
le nit Sa
nc
To mu ght
de
an
m i
m
en
L
Co or g eek eek ep
b u B r lD
C
ar r
H p D use m ica
m Ho lu em
Co ert se A Ch
lb o u A C
Ta s H VO MR
i
i
lv at T
A nn se
i
u
nc o s
Ci na H gram
ns
ria ro ty io
O ll P oun nsit
a C ra
Sm klin ty T
an ni
Fr mu
m
C o TA
P
ty
SE r un um
tle Co ing
r
te
Bu mit usk
en
m M
/
tC
Su ing ies
en
ck lit
tm
Li aci
ea
F
Tr
ll
A CC nty nity
C u u
o
SR C m ies
s
m
t
ca Co ili
l
al
Lu on Fac
H
nt F a ng
Ca BC din nni
C e
u
ll
A in/M se D IP
e
r a ou R
ov all
r
Lo H se
H
G
is u
lv Ho
n g on
i
)
h
r
A
a
a
’s
p
n
en
ria P S cM
O CA use M
(M
o
EO H OA
am
N rt
V
gr
e
ro
lb ti
a
tP
Ta inn
en e
nc
sm n
C i R TH
es to
ss ers
O
W day ty A orn
on ni C
M mu use ncy
m Ho ge
C o er t e A
lb iv
Ta rnat t
lte ar
A h St
es ity
Fr r C
ve
Ri
Challenges for Practitioners
• We understand more services/supervision for
high risk and less services/supervision for low
risk
• Conceptual understanding of the risk principle
versus operationalization of the risk principle
in real world community settings to achieve
maximum outcome
• “Can we quantify how much more service to
provide high risk offenders?”
Limited Guidance
• Prior Literature:
– Lipsey (1999)
• Meta-analysis of 200 studies
• Serious juvenile offenders
– Bourgon and Armstrong (2005)
• Prison study on adult males
• More Recent Literature:
– Sperber, Makarios, and Latessa (2013)
• Community-based setting
Identifying Effective Dosage Levels in
a Community-Based Setting
• Sperber, Latessa, & Makarios (2013):
– 100-bed CBCF for adult male felons
– Sample size = 689 clients
– Clients successfully discharged between 8/30/06 and
8/30/09
– Excluded sex offenders
– Dosage defined as number of group hours per client
– Recidivism defined as new sentence to prison
– All offenders out of program minimum of 12 months
Unanswered Questions
(Sperber, Makarios, & Latessa, 2013)
1.
2.
3.
4.
Defining dosage
What counts as dosage?
Prioritization of criminogenic needs
Counting dosage outside of
treatment environments
Unanswered Questions
5. Sequence of dosage
6. Cumulative impact of dosage
7. Impact of program setting
8. Low risk but high risk for specific
criminogenic need
Unanswered Questions
9. Nature of dosage for special
populations
10.Impact of skill acquisition
11.Identifying moderators of riskdosage relationship
12.Conditions under which dosage
produces minimal or no impact
Treatment Dosage and the Risk Principle: An
Extension and Refinement
Makarios, Sperber, and Latessa(under review)
• Methodology
– 100-bed CBCF for adult male felons
• Sample size = 980 clients
– Clients successfully discharged
between 8/30/06 and 12/31/10
– Excluded sex offenders
– Dosage defined as number of group
hours per client
– Recidivism defined as return to prison
– All offenders out of program
minimum of 12 months
Table 1. Descriptive Statistics for the
Study Sample (n = 941)
Variables
Race
Caucasian
Minority
Risk Level
Low/Moderate
Moderate
High/Moderate
Dosage Categories
0-99
100-149
150-199
200-249
250-299
300+
Return to Prison
Yes
No
Age
Time at Risk
N
%
840
101
89
11
195
587
159
21
62
17
149
190
245
234
90
33
16
20
26
25
10
3
360
581
Mean/SD
31.6/9.2
1344.8/398.8
39
61
Range
18-61
550-2027
Table 2. Cox Regression Predicting
Time to Return to Prison (N = 941)
Variable
Age
Minority
Risk Level
Dosage Categories
Slope (b)
SE
Wald
Exp(B)
-.015**
.006
6.163
.985
.283*
.157
3.265
1.328
-.102***
.044
5.283
.903
.529**
.100
28.285
1.696
Model Chi-Square = 38.760***
Dosage by Risk Level
Categories with less than 20 cases excluded
Summary of Findings
• Overall, increased dosage reduces recidivism
– But not equally for all categories or risk levels
• Low / Moderate and Moderate
– Curvilinear relationship
• Matters at the low ends of dosage, but effects taper off
and eventually reverse as dosage increases
• High / Moderate
– Increases in dosage consistently result in decreases
in recidivism, but
• Saturation effect at high dosage levels
Limitations
• Single study from a CBCF in
Southwestern Ohio
• Male Sample
• Does not consider other potential
moderators
• Lack of “0” Dosage comparison
group
Treatment Dosage and Personality: Examining the
Impact of the Risk-Dosage Relationship on
Neurotic Offenders
Sperber, Makarios, and Latessa
• Research on the risk principle confirms that correctional
practitioners should differentiate services by offender
risk.
• Research also confirms that these services should be
based on a cognitive-behavioral modality.
• At the same time, there is some research to suggest
that offenders with certain personality types (e.g.
neurotics) are higher risk for re-offending and may not
fare as well as other personality types within cognitive
behavioral programs.
• If this is true, increasing cognitive behavioral dosage
for high risk neurotic offenders may have a detrimental
impact on recidivism for those offenders.
• Consequently, this study examines personality type as
a moderator of the risk dosage relationship to
determine the impact on recidivism.
Methodology
– 100-bed CBCF for adult male felons
– Clients successfully discharged
between 8/30/06 and 12/31/10
• 980 offenders total
• 257 neurotic offenders
– Excluded sex offenders
– Dosage defined as number of group
hours per client
– Recidivism defined as return to prison
– All offenders out of program
minimum of 12 months
Personality Types
• Jesness Inventory
• 9 Types collapsed into 4:
– Aggressives
– Neurotics
– Dependents
– Situationals
Table 1: Sample Characteristics
Characteristics
Race
White
Non-White
Age
18 – 30
31 – 40
41 – 50
51+
Mean Age
Risk Level
Low/Low Moderate
Moderate
High/Very High
Mean
Dosage Categories
0 – 99
100-199
200+
Mean Dosage
Recidivism
Yes
No
N
%
236
21
91.8
8.2
111
93
46
7
33.4
43.2
36.2
17.9
2.7
33
155
69
29.8
12.8
60.3
26.8
39
107
111
186.9
15.2
41.6
43.2
111
146
43.2
56.8
Table 2: Cox Regression Predicting Time to Return to Prison
Variable
Slope (b)
SE
Wald
Exp(B)
Age
-.023*
.011
3.984
.978
Moderate Risk
.837*
.404
4.301
2.309
High Risk
1.354**
.442
9.358
3.871
100-199 Hours
-.195
.275
.503
.823
200+ Hours
-.540
.298
3.283
.583
** p < 0.01 *p<.05
Model X2 = 16.10**
Percentage Returned to Prison by Tx Dosage
and Risk Level
Percent Returned to Prison
90
80
70
60
50
0-99 tx hours
40
100-199 tx hours
30
200+ tx hours
20
10
0
low
0-99 hrs
100-199
hrs
200+ hrs
moderate
30
54
15
42
40
high
overall
49
85
45
42
42
Summary
• Pattern for neurotics similar to the
overall sample
• Increasing dosage reduces recidivism
but not equally for all risk levels
• Largest decrease in recidivism was
for the high risk/high dosage group
Limitations
• Single study from a CBCF in
Southwestern Ohio
• Limited number of cases in certain
categories may have limited power of
analysis
• Can’t address issues of service delivery
content that may still be relevant for a
neurotic population
Examining the Risk-Dosage Relationship in
Female Offenders
Spiegel and Sperber
• Studies on the number of treatment hours
necessary to reduce recidivism for high risk
offenders are few in number.
• Studies to date have relied on male samples.
• Cannot assume that a standard number of
treatment hours necessary to reduce
recidivism exists for both men and women.
• Present study examines the impact of varying
levels of treatment dosage by risk for female
offenders in a halfway house setting.
Methodology
• Sample size = 314 clients
• Clients successfully discharged between
10/1/07 and 2/28/10
• Dosage defined as number of group hours per
client
• Recidivism defined as re-arrest
– Checked Hamilton County and referral/home county
websites
– All offenders out of program minimum of 12 months
Table 1: Sample Characteristics
Characteristics
Race
White
Non-White
Age
18 – 30
31 – 40
41 – 50
51+
Mean Age
Risk Level
Low/Moderate
Moderate
High/Moderate
Mean
Dosage Categories
0 – 50
51 – 100
101+
Mean Dosage
Mental Health History
Yes
No
Re-arrest
Yes
No
Time at Risk
N
%
222
91
70.9
29.1
102
95
95
21
36.5
32.6
30.4
30.4
6.7
169
130
14
22.8
54.0
41.5
4.5
174
96
43
60.0
55.6
30.7
13.7
165
148
52.7
47.3
69
244
22.0
78.0
Mean
1188.2 (SD= 442.0)
Range
7 - 1802
Table 2: Cox Regression Predicting Time to Re-arrest
Variable
Race (White=1)
Slope (b)
SE
Wald
Exp(B)
-.607**
.246
6.084
.545
Mental Health History
.604**
.275
4.839
1.830
Risk Level
.055**
.021
6.829
1.056
.003
.209
.999
Dosage Categories
** p < 0.05
Model X2 = 24.276***
-.001
Recidivism Rates by Dosage and
Risk Level
40
35
30
25
20
15
10
5
0
34.1
19.3
18.2
30
27
20.9
16.3
13.5
11.3
Low-Moderate Risk
0-50 Hours
Moderate + Risk
51- 100 Hours
Overall
101+
Summary
• Further evidence of the application of the
risk principle to women
– Over-treating lower risk women can result in
recidivism increase
• Findings suggest a non-linear relationship
for both risk groups
• Initial increases in dosage have positive
impact on recidivism
• Increasing dosage to 101+ hours appears
to result in increases in recidivism
Limitations
• Sample drawn from a single halfway house
with limited geographical region
• Limited risk distribution
• Limited dosage distribution
• Inclusion of drug court clients in sample
• Reliance on public websites for recidivism
checks
• Limited geographical range for recidivism
– Doesn’t account for all Ohio counties
– Doesn’t account for bordering counties of other
states
Forthcoming Studies
• Under Construction:
– Examining the Risk-Dosage
Relationship in Sex Offenders
– The Relative Impact of Role-Play
versus Treatment Hours: Is There a
Trade-Off?
– The Impact of Client Strengths on the
Risk-Dosage Relationship
But What Do We Know?
• Research clearly demonstrates need to vary services
and supervision by risk
• Currently have general evidence-based guidelines
that suggest at least 100 hours for moderate risk and
at least 200 hours for high risk
• Should not misinterpret to imply that 200 hours is
required to have any impact on high risk offenders
• Not likely that there is a one-size-fits-all protocol for
administering dosage
• Practitioners have a responsibility to tailor
interventions to individual’s risk/need profile based
on best available evidence
Practitioner Responsibilities for
Effective Execution of the Risk Principle
• Process for assessing risk for all clients
• Modified policies and curricula that allow for variation
in dosage by risk
– Assess infrastructure and resources
• Definitions of what counts as dosage and mechanism
to measure and track dosage
– Unit of measurement
– Quality versus quantity
• Formal CQI mechanism to:
– Monitor whether clients get appropriate dosage by risk
– Monitor quality of dosage
– Monitor outcomes of clients receiving dosage outside of
evidence-based guidelines
Conclusions
• Corrections has benefitted from a number of
well-established Evidence-Based Guidelines
and Evidence-Based Practices
• Next evolution will focus on bringing a more
nuanced understanding and application of
these EBG’s and EBP’s to the individual client
level
• Practitioner-driven CQI/data needs to intersect
with research to drive this process so that we
continually move the field forward to
maximize public safety outcomes