Five Year Recidivism after Arrest for Operating

While Intoxicated:
A Large-scale Cohort Study

D. Paul Moberg and Daphne Kuo

UW Population Health Institute

April 2017

Prepared for the

Intoxicated Driver Program

Wisconsin Department of Health Services

Bureau of Prevention Treatment and Recovery

1
 Table of Contents

Executive Summary 3

Introduction and background 8

Methods 11

Results 16

Discussion and Conclusions 37

References 43

Appendices with County Level Data 45

Acknowledgments: We gratefully acknowledge the assistance and guidance from LeeAnn Cooper (Intoxicated
Driver Program, Department of Health Services, Bureau of Prevention, Treatment and Recovery) who commissioned
this study, and Michael Campbell, also from the DHS, who assisted in data access. Review and assistance in
understanding the nuances in the policies and data regarding impaired drivers came from Dale Simon and Vanna
Steffen from the Department of Transportation and Kristi Obmascher from the UW-Extension. We also obtained
useful conceptual review and advice from Donald Lyden, Laura Van de Hey, and Neil May at DOT. We particularly
thank the late Marthinus “Tinus” Taute of DOT for his crucial assistance in accessing data.

2
 Executive Summary
This evaluation addressed questions regarding the effectiveness of Wisconsin’s Intoxicated
Driver Program (IDP). The focus is on the five-year post-arrest recidivism rate for individual OWI
offenders, assessing whether recidivism rates vary by personal characteristics (age group, sex,
race/ethnicity, prior offense) of offenders and by contextual variables (county characteristics). We then
ask whether the IDP process is effective in helping offenders to attain services and to reduce recidivism.
The analysis also examines whether the nature of the driver safety plan (DSP) which recommends
specific levels of service (group dynamics education program, outpatient treatment, etc.) and
subsequent compliance with the DSP have an impact on recidivism.

Methods: We used data on 441,916 OWI arrests in Wisconsin from 2004 to 2014, with links
between arrest, conviction, Wisconsin Assessment of the Impaired Driver (WAID; Jacobson et al., 1979)
assessment, and compliance data in order to follow the 2004-2008 arrest cohort of 175,466 Wisconsin
drivers over the subsequent 60 months. Analysis used descriptive techniques as well as logistic
regression, random effects modeling with covariates, and hierarchical linear modeling to address
county-level variables.

Findings: OWI arrests declined steadily over the 11 years (2004-2014) we analyzed. These rates
are likely affected by individual driver’s behavior, intensity of law enforcement activity, and other issues.
The decline mirrors national trends. Among all of those arrested (n=341,157 individuals with 417,347
arrests) for OWI over the 11 year time period, most (78%) had only one arrest and 17% had two arrests
during the 11 years. A few had three (4%) or four (1%) arrests, while only 0.3% (about 1,000 people)
had five or more arrests.

Taking into account all documented lifetime OWI arrests prior to the study period, based on
417,347 arrests (and not individuals) over the 11 years, 58% were first arrests, 23% were second
arrests, 11% were third, 5% were fourth, 2% were fifth arrests and slightly more than 1% were sixth or
higher. For the most recent year in this data set (2014) there were 17,134 first arrests (58%), 6,416
second arrests (22%), 3,198 third arrests (11%), 1,475 fourth arrests (5%), 707 fifth arrests (2%), and
539 arrests (1.8%) which were 6th or more. 1

We intensely analyzed data from a 5 year cohort of 175,446 individuals with at least one OWI
arrest from 2004 through 2008 (see Figure1). Those with OWI arrests were most often young adult (39%
age 20-29) white (82%) males (77%). Among those with any OWI arrest during this 5 year time period,
31% had at least one prior arrest; those with priors were older, more likely to be male, and
disproportionately more white or American Indian than those with no priors.

1
Note that these are duplicate counts since one individual may have more than one arrest in a given year; also, the
number of arrests we report is likely higher than the official number taken into account for sentencing purposes.

3
 In our cohort, about 88% of those arrested were convicted of an OWI 2 offense. Regression
analysis showed that conviction was most likely among 20-29 year olds; more likely among American
Indians but less likely among Hispanics and African Americans than among Whites; more likely among
females than males; and less likely among those with prior arrests. Conviction rates varied by county,
ranging from 66 to 92 percent of those arrested.

The Intoxicated Driver Program (IDP) includes an assessment for all persons convicted of an
OWI offense using a tool called the Wisconsin Assessment of the Impaired Driver (WAID), the
development of a Driver Safety Plan (DSP) based on the WAID findings, and subsequent
monitoring/certification of compliance with the DSP. In our cohort of arrestees, 74% participated in the
IDP WAID assessment (80% of those who were convicted plus about a third of those not convicted for
the index offense). Those with prior offenses, those under 20 or 40-49, and Black, Hispanic and Native
Americans were significantly less likely to participate in the IDP assessment process than others. IDP
participation also varied by county, ranging from 52% to 79%.

Figure 1: Pathway to Recidivism

2004-08 Wisconsin OWI Arrestees

Arrested
N=175,446

21,843 153,603
N Convicted Y

123,295 30,308
14,741 7,102
N IDP Y Y IDP N

2137 26974
N DSP DSP N

4964 96317

Y Y

Re-Arrested (N=42,157)
5 Year Recidivism: 24%

2
We do not have data on how many of those individuals not convicted of OWI were convicted of a lesser offense,
diverted from prosecution, or found innocent.

4
 Among participants in the IDP, 48% were assessed as “irresponsible use” and 18% as
“irresponsible use-borderline”. Ten percent were classified as “suspected dependency” and 24%
“dependency” (including 7% in remission). Most frequent DSP primary education or treatment
requirements were group dynamics course (46%), outpatient treatment (43%), and multiple offenders
program (6%). Compliance with the Driver Safety Plan (DSP) within two years of the index offense was
reported for78% of those with a DSP. Compliance (among those who had entered the IDP) did not differ
by prior offense, but was lower for those under 30 years old, and lower for minority groups (African
American, Hispanic, and Native American). Compliance by county ranged from 64% to 85% of those
who participated in the IDP. WAID assessment data indicate that those found to have “irresponsible
use” were more likely to comply than other diagnostic groups.

Recidivism (re-arrest for OWI) within 60 months of the index arrest occurred in 24% of this
cohort. This rate did not differ among those who were or were not convicted of the index offense.
However, among those convicted who participated in the IDP and complied with their DSP, only 19%
recidivated, compared to 32% of those who did not comply with a DSP and 34% of those who bypassed
the entire IDP process by failing to comply with the court order for an assessment (see Figure 1).
Regression analysis, including the entire arrest cohort, found significantly higher re-arrest rates for
those with prior offenses, 3 those under age 29, American Indians, and those who were convicted of the
index offense. Females, older drivers, and those who entered the IDP process had significantly lower
recidivism rates. Recidivism varied considerably by county, ranging from 18% to 36%.

Among those who participated in the IDP assessment process, older age groups and females
had lower recidivism rates. Hispanic and American Indians (but not African Americans) who participated
had higher recidivism than other groups, as did those who had been convicted. Those who were
compliant with the DSP had a lower recidivism rate (19%) than those who were non-compliant (32-34%).
Relative to those assessed as irresponsible use, the borderline, suspected dependency and dependent
cases had higher recidivism rates (28% of dependent, 22% of borderline and 16% of irresponsible users),
as did those with prior arrests (unless they complied with the DSP, in which case they did not differ).

Conclusions: What do these findings suggest regarding the IDP as an intervention to prevent
recidivism? Several conclusions can be reached:

• Any policy changes should consider that to improve overall population safety in a
significant manner, a focus on preventing or deterring first offenses is likely to have
more impact than a focus on those with four, five or more offenses (as much recent
legislation has done). Recidivism rates are not much different between those with and
without a prior arrest; research (Rauch et al., 2010) indicates that an individual with a
history of even one OWI arrest is 7 times more likely to be arrested in a given year than
an individual who has never been arrested for OWI.

3
While statistically significant, the compliance and recidivism patterns for those with prior arrests were not much
different from those with no prior arrests; the overall 60 month re-arrest rate for those with priors was 26%,
compared to 23% for those with no priors. For those with prior offenses who entered and complied with the IDP,
there was no difference from the average in recidivism.

5
• Recidivism is considerably lower among those arrestees who participate in the IDP
WAID assessment and subsequently comply with their driver safety plan (DSP). Our
primary hypothesis—that IDP participation and compliance would add significant value
in reducing recidivism—is confirmed. Thus strategies to improve participation and
compliance with the IDP program would likely be productive, even though this finding
undoubtedly reflects some degree of self-selection bias. Tested strategies include
follow-up with OWI offenders who fail to schedule their IDP assessments, and increased
use of motivational interviewing to improve compliance with the driver safety plans
(DSPs) (see DCTS, 2016). Other possible (but untested) strategies to enhance
compliance could include further court intervention with those who fail to comply,
higher penalties for failure to participate, and reduced initial penalties and fees to
remove financial barriers to IDP participation while reserving higher penalties for those
who fail to follow through.

• Preventive interventions should be particularly targeted to young white males. Native

Americans should also be specifically targeted for recidivism prevention. Strategies to
Increase the perception of the likelihood of arrest, increasing individual agreement with
the goals of the policy, and multi-component programs have been found to reduce
alcohol-impaired driving and arrests (Bertelli & Richardson, 2008; Miller et al., 2015;
Shults et al., 2009; UWPHI, 2017). The use of publicly announced sobriety checkpoints,
ignition interlocks for convicted offenders, and mass media campaigns are effective
strategies (Jewett et al., 2015; Guide to Community Preventive Services). Additional
strategies to reduce overall alcohol consumption (such as increased taxation of alcohol,
reduced outlet density, electronic screening and brief intervention, public education
and awareness campaigns), server liability and host training, and policies which reduce
the risk of driving after drinking by considering location of establishments and
transportation options have some evidence of success (Guide to Community Preventive
Services; Scott, 2013; UWPHI, 2017; Voas and Lacey, 2011). Increased penalties are not
effective beyond some minimal threshold (Scott, 2013).

• There are disparities in participation in the IDP program which contribute to

systematic differences in recidivism rates. Participation is lower for those with prior
offenses, those under 20 or in the 40-49 age group, and among African Americans,
Hispanics and Native Americans. Similarly, among African Americans, Hispanics and
Native Americans who do participate in the IDP, compliance with the DSP was
significantly lower. For Hispanics and Native Americans, but not African Americans,
recidivism was also higher than for non-Hispanic Whites. These disparities suggest that
strategies are needed to improve the participation of non-White populations at all
stages of the IDP in order to reduce recidivism. The cost of IDP participation may be a
barrier to participation and compliance, particularly with socio-economically
disadvantaged groups.

6
 • County differences were pronounced in rates of conviction, IDP participation and
recidivism. County level enforcement activity, court procedures, and treatment systems
(not assessed in this analysis) vary considerably and may contribute to these observed
differences. Variation on some of these factors could also be addressed, particularly in
low IDP participation counties with higher recidivism rates.

• This analysis was limited by a lack of access to data on the status of driver’s licenses
(e.g., revocation) at the point of OWI arrest, inability to account for conviction for lesser
offenses, lack of analysis of BAC levels, and data linkage difficulties. Further analyses
could examine to what extent noncompliant drivers with revoked licenses continue to
drive and commit subsequent offenses, the predictive value of BAC level, the role of
specific symptom patterns, factors related to county-level variability and other relevant
questions.

Policy proposals should be clear about the goals of proposed changes, and use existing data to
predict potential outcomes—intended and unintended. Is the objective to deter, punish or to extend
treatment? Particularly for drivers with multiple repeat offenses, OWI is a symptom of a serious
substance use disorder, not necessarily willful or rational criminogenic character where deterrence
approaches are effective. Meta-analytic evidence reported by Miller et al., (2015) suggests that multi-
component interventions, rather than education or treatment alone, have the most promise in reducing
recidivism. Programs and policy should be framed to reduce disparities of access and benefit, while
increasing public safety. Potential unintended negative consequences, including overburdening law
enforcement, courts, and correctional systems, must be weighed relative to the potential public safety
benefits of proposed policy changes. These considerations would seem to argue for a preventive and
treatment focus in any policy changes.

Regardless of its limitations, the analysis presented in this report provides an in-depth source of
data which we hope will inform future policy deliberations to address the declining but still very
prevalent incidence of operating while intoxicated in Wisconsin. It is clear that for offenders who
comply with the IDP system, recidivism rates are lower. Perhaps the critical issue is how to increase
rates of compliance and reduce disparities in benefit from the IDP services. This is likely best
accomplished both by improving and standardizing the community level systems in place and by
working to further motivate offenders to participate in beneficial services.

7
 Introduction and Background
While there is significant attention in the literature to the problem of Operating While
Intoxicated (OWI), few studies have considered long-term recidivism rates. The National Roadside
Survey of Alcohol and Drug use by Drivers (NHTSA, 2015) found a decreasing trend in alcohol use among
weekend nighttime drivers nationally, with intercepted drivers registering > .08 BrAC dropping from
7.5% in 1973 to 1.5% in 2013-14. SAMHSA’s National Survey on Drug Use and Health (NSDUH) also
suggests a national decline in driving under the influence of alcohol and/or illicit drugs decreased
significantly from 2002 to 2012 (Lipari, Hughes and Bose, 2016). The Behavioral Risk Factors
Surveillance Survey (BRFSS) indicates that in Wisconsin, adult binge drinking has been of high prevalence
and stable for a number of years, although self-reported impaired driving decreased over this period
(WiDHS & UWPHI, 2016) These data suggest that prevention efforts resulting in cultural change in
perceptions of acceptable behavior have been effective in reducing the rate of impaired driving
nationally and in Wisconsin.

Nonetheless, Wisconsin in 2014 experienced about 30,000 arrests for OWI offenses, and, over
the years 2006-2009 23.7% of drivers surveyed admitted to driving “under the influence of alcohol”, the
highest of all states in the nation (SAMHSA, 2010). Based on the Behavior Risk Factors Survey, Jewett et
al (2015) estimate that there were 828 impaired driving episodes per 1,000 adults in 2012 in Wisconsin,
compared to 505 nationally. Analysis of WI DOT DMV Driver Record File data as of December 31, 2012
(www.dot.wisconsin.gov/safety/motorist/drunkdriving/arrests.htm; accessed 1/9/2015) found that
612,249 individuals had at least one OWI conviction. Of these, 32% had more than one conviction. The
total number of individuals with one or more convictions represents 14.7% of all licensed drivers in
2012. An analysis of recidivism in Wisconsin with a one year time window focused on assessment
findings relative to recidivism at the county level (Hill, 2013); offenders who were assessed as
dependent or suspected dependent had the highest recidivism while those who were irresponsible users
had lower recidivism in one year. A longer-term assessment of recidivism in Maryland was reported by
Rauch et al. (2010) focusing on the role of number of offenses prior to the five-year study period in
predicting future offenses. This study concluded that the annual rate of an OWI violation among those
drivers with only one prior arrest was seven times higher than for drivers who had never had an arrest.

Most states require screening, assessment and interventions for individuals arrested and
convicted of OWI; the effectiveness of these programs has not been established. In order to address
these concerns, we conducted a study of five-year recidivism after arrest for OWI for a large (n=175,446)
retrospective cohort of offenders in Wisconsin.

This evaluation addressed the following questions:

1. What is the post-arrest five-year recidivism rate for individual OWI offenders?
a. Do these rates vary by personal characteristics (age group, sex, race/ethnicity, prior
offense) of offenders?
b. Do these rates vary by county of residence?

8
 2. Is the Intoxicated Driver Program (IDP) process working to get offenders appropriate services
and to reduce recidivism? Does the IDP process (WAID assessment and driver safety plan—DSP-
- completion):
a. predict recidivism and
b. mediate the baseline prediction of recidivism?

3. Does (a) the nature of the driver safety plan (e.g., group dynamics, outpatient treatment, etc.)
and (b) subsequent compliance with the treatment plan have an impact on recidivism, when
background and WAID findings are taken into account; and (c) to what extent does compliance
mediate the predictive value of the WAID finding?

The framework for the analysis is depicted in the following model (Figure 1), which provides an
overview of our analytic structure.

9
 In this model, we hypothesize that personal background characteristics are directly related to
recidivism, predict the findings from the Wisconsin Assessment of Intoxicated Drivers (WAID), and
predict compliance with the Driver Safety Plan (DSP). Contextual variables related to geographic
location and local variability such as law enforcement saturation, alcohol availability, criminal justice
processing and treatment capacity, are also important. In our analysis, these contextual variables are
measured by a proxy variable--county of residence. The model shows (dark red lines) that we expect
WAID findings and compliance with the IDP (including treatment) to mediate the relationship between
personal background and recidivism, indicating that the IDP program adds value as an intervention to
reduce recidivism. Figure 1 also shows our data structure, in which a cohort of OWI arrestees from
2004 through 2008 are followed for the subsequent 5 years.

10
 Methods
Data

Our analytical data were extracted and linked from records of the Wisconsin Department of
Transportation, the IDP program, and the Department of Justice. There were four types of data: 2004-
2014 all OWI arrest records, all convicted OWI records from 2004 to 2015, WAID/DSP records 2004-2015
from DHS/DOT, and county characteristics downloaded from UWPHI’s County Health Rankings and
Roadmaps.

The arrest data included all arrests for OWI offenses (multiple charge codes) from 2004 to 2014.
Characteristics of arrested drivers, including sex, race/ethnicity, county of residence, age, date of birth,
and number of priors, were obtained from the driver license files and included in the arrest data 4.
Conviction data included 2004-2015/6 OWI related convictions. The OWI convictions, collected by the
WI courts and DOJ, were also obtained from DOT. Two data files can be matched by a unique DOT
internal customer identification number and the time and date of OWI violation.

WAID/DSP data included four files: a master file including DOT unique ID, IDP enrollment
incident identification number and time of enrollment; and three files including WAID findings,
symptoms, individual DSP (driving safety plan) plans, and completion time. The data were collected by
local assessment agencies and reported to DOT. WAID/DSP data files could be merged into one by
linking with the incident number assigned by DMV when the patient/offender first enrolled into IDP.
The incident number is person-incident specific. Thus, the numbers of DMV incident ID’s for a driver
indicated how many times s/he enrolled into IDP programs in the data. Though the DOT customer
numbers were also given in the IDP data, the date and time of violation were not included. Additionally,
the DMV incident numbers used to link WAID or DSP related files together were not used in arrest or
conviction data from DOT. For confidentiality, these data were provided for analysis in the form of a
limited data set without individual driver’s names, addresses or license numbers; DOT customer ID
numbers and DMV assigned incident numbers were used to match records.

Because our data were limited to arrests occurred between 2004 and 2014 and the first OWI
violation in our data might occur at various time point, the length of time exposed to the risk of getting
rearrested varies individually. We thus limited our analytical sample, i.e., focal cohort, to drivers
arrested during the calendar years 2004-2008 in order to follow each arrestee in the sample equally for
a 60-months recidivism period (i.e., subsequent arrests through 2014). For example, for those with the
first OWI offense in January 2004, another arrest before January 2009 is logged as recidivism; for those
with the first OWI in December 2008, another arrest before December 2013 is recidivism. Within the
2004-2008 cohort of arrests, we defined the first arrest in the data (i.e., the first arrest in 2004-08) as
the “index arrest”. Note that approximately 26% of all arrestees during that time period had one or

4
Comparison of the data set we were provided to other data on OWI offenses from DOT reports and websites
indicates that our data set may not have included arrests for “prohibited alcohol concentration” (PAC) offenses
where there was no other OWI charge for the same offense. Omission of these cases should not change the
overall conclusions and findings in this report, given the large numbers of cases in our analytic file.

11
more prior OWI arrests on their record. Prior arrest was included as a moderator (background) variable
in the analysis.

For drivers in the focal cohort, we searched and identified whether they were rearrested within
60 months of the index arrest. Only being rearrested within 60 months is considered as recidivism. Note
that conviction was used as a mediating variable in this analysis, with arrest as the qualifying index
event. We do not use convicted drivers as our main analytical sample because conviction is an
intermediate study outcome and conviction is not required for participation in the Intoxicated Driver
Program (IDP).

We first selected drivers arrested between 2004 and 2008. We then identified total numbers of
arrests in the subsequent 60 months of the index arrest for each driver. The arrests after the first re-
arrest were not analyzed. Then the index arrest and the conviction records were matched in order to
create the variable of conviction for the index arrest. The file of 2004-08 drivers with their re-arrest,
conviction status and their background information were then matched with the (separate)
IDP/WAID/DSP files. About 85% of arrest-conviction records produced a unique and “valid” match
record in WAID/DSP file. A valid IDP enrollment met two criteria: 1) the IDP enrollment date was AFTER
the violation date of the index arrest; and 2) it was BEFORE the end of 24 months after the index arrest
or before a re-arrest date, whichever date comes first. Note that while most eventually matched
arrestees had the first WAID-IDP enrollment within 18 months after the violation, some of them (3%)
first enrolled into the IDP program after more than 24 months without any re-arrest. Additionally, few
arrestees had more than one valid enrollments without any recidivism within the observation of 60
months. And some have a re-arrest before or on the date of the first enrollment. All of above
conditions would be considered as no enrollment for the index arrest, whether they were subsequently
convicted or not.

In addition to those assumptions, two minor assumptions were necessary in creating our
analytic files. We assumed that multiple OWI-related charges on the same day or within 24 hours of
each other are a single offense. Less than 0.1% of arrest records were within 24 hours of each other.
Second, our analysis assumes that arrest, rather than conviction, is the better indicator of impaired
driving for public health and policy purposes (see Rauch et al., 2010).

Variables

The following variables were used in our analysis.

A. Offender characteristics
• Age
• Sex
• Race/ethnicity
• County of residence
• Number of documented prior OWI arrests

12
 B. Offense variables (index OWI and subsequent offenses)
• Offense date and time
• Offense location (county)
• Conviction status for each offense
C. Intoxicated Driver Program (IDP) Variables
• Participation in the IDP (i.e., is there a record that the driver presented for a WAID
assessment?)
• Wisconsin Assessment of the Impaired Driver (WAID; Jacobson et al, 1979) interview
data including:
o Symptoms and pattern of substance abuse as documented by assessors
o Alcohol only and/or other drug involvement
o Finding regarding severity of substance use
• Driver Safety Plan (DSP) recommendations/requirements
• Data showing subsequent compliance (date) with the DSP

Age in the analysis is defined as age at the index arrest (first OWI arrest between 2004 and
2008). Race/ethnicity has 6 categories: White, Hispanic, Black, American Indian, Asian American, and
others. We also use the number of prior arrests documented in the driver record at the index arrest.
Prior arrest is defined dichotomously (yes or no) for the primary analysis while the number of priors is
also explored descriptively. Our main analyses are also limited to Wisconsin residents. Given that only a
handful of arrests have more than one record within a day, multiple records within a day are considered
as the same arrest. Arrest dates are also used to calculate the date of recidivism, if ever occurred.
Conviction status is a dichotomous variable created by matching conviction records with arrest records;
the matched records are given one conviction status (convicted or not) for the OWI offense. Note that a
conviction for a lesser offense is not counted as an OWI conviction.

Symptoms reported on the WAID were documented and entered by the assessors. In addition to
the current direct assessment of the offender by the assessor, some symptoms may have been retrieved
from previous IDP records or other sources obtained by the assessor. Symptoms documented on the
DOT form DSC756 include self-concern about substance use, self-described impaired use, relief use,
cognitive deficits, unexplained social functioning changes, social functioning complaints, others
concerned about use, family disruption, social role changes, defensive, exaggerated negative feelings,
depression, self-described severe emotional problems, rapid use, concealed use, morning use, stop
attempts, blatant use, decline in social or financial functioning, excuses for missing work, multiple
substance use, substance using peers, loss of interest in usual activities, employment facilitates use,
traffic crashes, legal problems, family emotional problems, change of residence, anxiety contacts,
emotional outbursts, prior treatment, use despite medical contraindications, use despite social
functioning contraindications, loss of control, blackouts, tolerance, moderate withdrawal and severe
withdrawal.

Findings include severity, type of substance, chronicity and pattern. Five levels of severity are,
from most to the least severe, dependency, dependency in remission, suspected dependency

13
irresponsible use of alcohol-borderline, and irresponsible use of alcohol. This assessment was originally
developed by Jacobson et al. (1979), applying the criteria for the diagnosis of alcoholism from the
National Council on Alcoholism. A handful of drivers have more than one severity level noted in a single
IDP participation. They are assigned to the highest level of severity coded on the DOT form. Substances
can be documented to include alcohol and/or other drugs. One case might have both. Chronicity
includes early stage, far advanced, and moderately advanced. Pattern includes intermittent, recurring,
and steady. While severity and substance are recorded for all drivers, chronicity and pattern are not,
and are not currently used in the assessment algorithm. Our analysis thus focuses on severity and
substance.

Possible recommendations include both education and treatment programs. They are
Abstinence, Aftercare, Alcohol Education, Alcoholics Anonymous, Antabuse, Biomarker, Day Care
Treatment, Detoxification, DWI Clinic, Evening Care Treatment, Exit Interview, Fu Appointment With
Agency, Fu With Agency + Withdrawal, Group Dynamics, Group Therapy, Halfway House, In Patient
Treatment, Level I Alcohol Education, Level II Alcohol Education, Medical Exam, Multiple Offender
Program, Narcotics Anonymous, Night Care Treatment, Out Patient Treatment, Out-Of-State Alcohol
Education, Out-Of-State Plan, Psychiatric Exam, Random Urine Screens, Reassessment After GD/MOP,
Residential, Care Treatment, Traffic Safety School, Transitional Living, Tribal Treatment, Victims Panel,
and 1 Day Alcohol Education. Some treatment/education programs are considered as primary; the other
are secondary. Each recommendation plan includes at least one primary program and some might have
more than one. Drivers assessed as “irresponsible use” are usually referred to an education program
(group dynamics). Those drivers with dependency would have been assigned at least one treatment
program. In the recommendation file, each recommendation comes with a completion date entered
upon follow-up by the assessment agency. To be considered compliant, one has to complete all
recommended primary programs.

The arrest data contained up to 5 charges per arrest. Unfortunately, non-OWI-related charges
were not in the algorithm of data extraction and thus the information on those charges is not available
for us. Those omitted charges of possible interests included driving without a license and speeding. The
data file did not include individuals arrested only for “Prohibited Alcohol Concentration” (PAC), although
on further consideration these charges are often considered a category of OWI.

Analytical Plan

The analysis began by examining descriptive data including characteristics of the cohort of
drivers, number of arrests, changes in number of arrests and convictions over time, and a summary of
the assessment findings, recommendations and compliance from the WAID and IDP. For policy and local
evaluation purposes, we provide data on county variation in arrest, conviction, WAID and IDP
participation and compliance (in the Appendix). We then explored pathways to recidivism. In the
analysis, we summarized the flow of drivers transitioning from index arrest between 2004 and 2008, to
conviction, to IDP program participation, to DSP compliance, and finally to re-arrest within the
observation period of 60 months. We also calculated those transitions for non-convicted drivers, and
non-IDP participants. We additionally looked at predictors of conviction, compliance with IDP findings,

14
and overall recidivism, comparing offenders with and without prior offenses, largely using cross-tabular
and mean difference analysis. Statistical significance estimation for descriptive data is not important in
these analyses, since we have a very large and complete population of arrestees and even small
differences are statistically significant.
After the descriptive analysis, multi-variate regressions on conviction, program participation,
compliance and recidivism were estimated to account for inter-correlations between the personal
predictors and outcomes. Next we to assessed the extent to which program-related variables, including
program participation, assessment results and compliance mediated the relationship between personal
predictors and recidivism. Finally, multilevel regression (county and county variables at level 2) was used
to account for clustering of offenders within counties, and take account of the differences between
counties.

15
 RESULTS
A. Cohort Description

Table 1 summarizes our data set of Wisconsin OWI-related arrests and convictions from 2004
through 2014. The first column shows the numbers of unique drivers in each year in our data set. They
might or might not have another OWI offense in subsequent years. The third column shows numbers of
OWI records in each year. Except year 2008, OWI arrests and arrestees have decreased gradually over
time. The 2nd column shows the number of drivers entering the pool of first time arrests between 2004
and 2014. For example, in 2007, there were 43,438 unique drivers arrested for OWI. In total, there were
45,974 arrests that year. Of those drivers, 35,637 had no arrest records between 2004 and 2006. Of
35,637 arrested drivers, 29,879 were convicted. In sum, from 2004 to 2014, there were 441,916 OWI
arrests for 341,157 drivers. Of these arrested individuals, 287,268 (81.5%) persons were convicted of
OWI. Note that this included 24,676 arrests from 23,792 out of state drivers.

Table 1. Records and Persons in Data 2004-2014

Arrests Convictions
Persons2
Persons1 (non- Records3 Persons
(duplicate) duplicate) (unique)
2004 44,979 44,979 47,972 37,855
2005 43,898 40,679 46,701 34,849
2006 44,028 38,254 46,677 33,135
2007 43,438 35,637 45,974 29,879
2008 32,562 26,474 33,539 26,789
2009 43,050 33,436 45,342 27,719
2010 38,210 28,534 40,166 24,049
2011 34,259 25,075 35,802 20,899
2012 33,912 24,506 35,471 20,356
2013 31,322 22,268 32,757 18,550
2014 30,116 21,315 31,515 13,188
Total 419,774 341,157 441,916 287,268
1. Number of unique persons in each year’s dataset.
2. Number of unique persons added each year to the merged data set.
3. Number of unique arrest records (i.e., the same violation date) for each
year.
Table includes 24,676 arrests for 23,792 out of state drivers.

16
 Figure 2 (see next page) provides a graphic depiction of these numbers on a monthly basis,
showing an overall decline in the total number of arrests for OWI in the years since 2004 and in particular
since 2009. The figure also shows fewer convictions than arrests in a consistent pattern, and seasonality
of arrests with more OWI arrests occurring in the spring and summer months. The apparently high (100
percent) conviction rates during 2008 reflects missing arrest data on cases without a conviction—which
also artificially reduces the total number of arrests for that year.

While public policy has concentrated on penalties for drivers with fourth, fifth and higher
numbers of arrests/convictions for OWI offenses, Table 2 shows that, over our 11-year time frame,
77.7% of the individuals arrested were arrested only once during that time period. Only 1.4% of those
arrested during this 11-year period were arrested 4 or more times for OWI. Similarly, looking at all
arrests (22% of individuals have more than one arrest) during this time frame, 77.2% are first arrests
during this time window, and only 1.3% represent fourth or higher arrests. Note, however, that based
on data in the records, 31% of the individuals arrested during this time window did have a prior OWI
arrest before 2004 (although statutorily these may or may not have been considered in sentencing).
Later our advanced analysis of recidivism rates takes prior arrest record into account.

Table 2. All OWI Arrests 2004-14 by Number of Arrests*

No. of Arrests Arrestees Arrests

1 265,188 77.7% 341,157 77.2%
2 57,068 16.7% 75,969 17.2%
3 14,296 4.2% 18,901 4.3%
4 3,586 1.1% 4,605 1.0%
5 805 0.2% 1,019 0.2%
6 171 0.1% 214 0.1%
7 36 0.0% 43 0.0%
8 6 0.0% 7 0.0%
9 1 0.0% 1 0.0%
Total 341,157 441,916

*Does not include arrests prior to 2004.

Our analysis used the cohort of individuals arrested during the period of 2004 to 2008 to allow
for a five year (60 month) recidivism assessment. We eliminated 10,577 out of state offenders from the
subsequent analysis, leaving a cohort of 175,446 individuals with at least one OWI arrest between 2004
and 2008.

Table 3 summarizes demographic data on these offenders at their first arrest during 2004 to
2008. Overall, 31% had a prior offense before this index period, and were examined separately from the
first offenders in many of our analyses. The age group with the largest number of offenders was age 20-
29 (39.1% of offenders, and 44.5% of those with no priors), followed by ages 30-39 and 40-49. Those in
the 30+ age groups were much more likely than those under 30 to have prior arrests. Males accounted

17
18
for 76.9% of the individuals in our cohort, and 83.4% of those in the group with prior arrests. White
individuals accounted for 82.3% of our index cohort, and (disproportionately) 86.7% of all those with
prior arrests. Asian and Pacific Islanders, African Americans, and Hispanics represented a smaller
proportion of arrests than their proportions in the Wisconsin population, while Native American Indians
were over-represented in total and also over-represented among those with prior arrests. Appendix 1
shows the number of individuals included in this sample by county of residence at index arrest; the
range of individuals with priors by county is from a low of 25% to a high of 41%.

Table 3. Sociodemographic Characteristics of 2004-8 arrestees*

(Wisconsin Drivers)
No Prior Prior Total
N Row% Col% N Row% Col% N Col%
Total 121,823 69.4 53,623 30.6 175,446
Age
< 20 19,649 96.9 16.1 637 3.1 1.2 20,286 11.6
20-29 54,266 79.0 44.5 14,397 21.0 26.9 68,663 39.1
30-39 20,503 55.6 16.8 16,369 44.4 30.5 36,872 21.0
40-49 17,281 52.9 14.2 15,368 47.1 28.7 32,649 18.6
50-59 7,524 57.9 6.2 5,476 42.1 10.2 13,000 7.4
60+ 2,600 65.4 2.1 1,376 34.6 2.6 3,976 2.3
Gender
Male 90,235 66.9 74.1 44,699 33.1 83.4 134,934 76.9
Female 31,588 78.0 25.9 8,924 22.0 16.6 40,512 23.1
Race/Ethnicity
Asian
Am. 1,461 85.7 1.2 244 14.3 0.5 1,705 1.0
Black 6,315 76.5 5.2 1,940 23.5 3.6 8,255 4.7
Hispanic 9,514 77.8 7.8 2,709 22.2 5.1 12,223 7.0
Am.
Indian 2,433 55.8 2.0 1,924 44.2 3.6 4,357 2.5
White 97,834 67.8 80.3 46,488 32.2 86.7 144,322 82.3
Missing 4,266 93.1 3.5 318 6.9 0.6 4,584 2.6

*Unduplicated data showing characteristics at first entry into arrest cohort.

The Wisconsin Intoxicated Driver Program (IDP) includes several measures intended to reduce
recidivism. All drivers convicted of OWI offenses are required to obtain and pay for an assessment by a
trained IDP assessor. Failure to participate in an assessment and/or to comply with the
recommendations may lead to suspension of driving privileges. The assessors use the Wisconsin
Assessment of Intoxicated Driver (WAID) tool originally developed and validated by Jacobson et al.
(1979) to explore the driver’s substance use history and problems, issue a finding and develop a Driver
Safety Plan (DSP). The WAID findings are on a continuum from irresponsible use to substance
dependence (and dependence in remission). Chronicity, pattern and substance (alcohol and/or other
drugs) are noted. Resultant DSP options and recommendations for the offender range from
participation in a “group dynamics” education session to inpatient rehabilitation; most offenders have

19
several recommendations. Within a year, compliance with the DSP is reported by the local AODA
agency online to the Department of Transportation and entered into the driver’s record. Since
compliance may require up to two years, our assessment of compliance was done with a period of 24
months (2 years) following the arrest.

Table 4 provides the data on seventeen symptoms with more than 10% of drivers coded in the
WAID assessment for our 2004-2008 index cohort of OWI offenders. Data on symptoms was available
for 117,368 participants in the IDP. Most frequent symptoms of substance use disorders (SUDs)
recorded by the assessors were legal problems (62% of those with symptoms documented), use of
substances—alcohol and/or other drugs—for relief (43%), peers with heavy substance use (43%), self
concern about use (36%); others concerned (36%); and tolerance (34%). Of 38 symptoms in total, the
remaining 21 symptoms not in the table were reported for less than 10% of drivers assessed. (These
other symptoms are: self-described impaired use, use despite social functioning contraindications.
severe withdrawal, use despite medical contraindications, multiple substance use, moderate
withdrawal, blatant use, morning use, defensive, traffic crashes, employment facilitates use, loss of
interest in usual activities, social functioning complaints, decline in social or financial functioning,
excuses for missing work, unexplained social functioning changes, anxiety contacts, exaggerated
negative feelings, self-described severe emotional problems, change of residence, and cognitive
deficits.)

A driver usually has more than one symptoms submitted and recorded by the assessors.

Table 4. Symptoms Reported by DSP Assessors for 2004-08 Cohort

(N=117,368 Individuals Assessed)
Symptom codes and descriptions N %
26 legal problems 72,387 62%
3 relief use 50,849 43%
22 substance using peers 50,626 43%
7 others concerned about use 42,710 36%
1 self-concern about substance use 41,717 36%
36 Change in tolerance 39,501 34%
27 family emotional problems 35,759 30%
31 prior treatment 33,626 29%
35 blackouts 30,357 26%
17 stop attempts 29,611 25%
34 loss of control 25,941 22%
9 social role changes 19,339 16%
8 family disruption 18,283 16%
14 rapid use 15,839 13%
12 depression 15,513 13%
15 concealed use 12,400 11%
30 emotional outbursts 12,197 10%

20
 Assessment Findings, i.e., the highest severity level, are presented in Table 5. Approximately
48% of our index offender cohort who entered IDP were initially assessed as “Irresponsible Use” of
substances. “Use-Borderline” accounted for 18% of those drivers, “Suspected Dependent” for 10%, and
“Dependent” including those in remission for 24%. Note that these categories were developed under
the older National Council on Alcoholism guidelines for the diagnosis of alcoholism as operationalized by
the WAID (Jacobson et al., 1979), and do not reflect the current nomenclature of DSM V. While not
shown in the table, alcohol was coded as involved for 98.7% of the offenders, and other drugs for 11.1%
(totals greater than 100% since most offenders coded for drug use also were coded for alcohol).

Table 5. IDP Assessment Findings,

2004-08 arrestees in IDP (N=130397)
N %
Highest Severity
Irresponsible Use (IU) 61,883 47.5%
IU-Borderline (IUB) 23,963 18.4%
Susp. Dependent (SD) 13,674 10.5%
Dependent (incl in remission) 30,877 23.7%
-Dependent in remission 9,111 7.0%

The Driver Safety Plan (DSP) includes a number of possible recommendations to ameliorate the
individual offender’s substance use problem and improve driver safety. Table 6 summarizes the data on
these recommendations. There were 34 programs recommended to the IDP drivers. The education or
treatments listed in table 6 are primary programs. Most IDP drivers were recommended to at least one
primary program. Only a small handful of drivers were recommended to secondary programs only. The
group dynamics program was most frequently recommended—appearing in 45% of all DSPs. Next most
frequent was Outpatient Counseling/ Treatment (43%). The most frequent secondary recommendations
were abstinence from substances (23%) and participation in Victim Impact Panels (14%). The complete
list of recommendations is shown in the methods section.

B. Outcomes for 2004-2008 Cohort

In this section we provide descriptive data on outcomes from 2004-2008. The outcomes of
interest include conviction for the index OWI offense, participation in the Intoxicated Driver Program
(IDP) by presenting for an assessment, compliance with the resultant Driver Safety Plan (DSP), and, of
primary interest, recidivism as defined by a subsequent OWI offense within 60 months (5 years) of the
index offense. The unit of analysis is the individual offender, tracking his or her subsequent arrests
following the index offense—the first OWI from 2004 through 2008. We consider those with, and
without, prior arrests separately in most of the analyses.

Figure 3 provides an overview of the transitions from arrest to rearrest. Overall, 42,157 (24%)
drivers in our arrest cohort (of 175,446 arrested drivers) had at least one subsequent arrest during the
60 months after the index arrest. Of the 175,446 arrestees, 153,603 (87.6%) were convicted, and
123,295 (80.3% of those convicted) subsequently participated in an IDP assessment. Of those assessed,

21
 Table 6. Recommended Primary Treatment Programs/plans (N=130,397)
DSP Recommendation codes and descriptions N %
GRD Group dynamics 60,046 46.0%
OPT Outpatient treatment 55,789 42.8%
MOP Multiple offender program 8,176 6.3%
AFT Aftercare 7,093 5.4%
OAE Out-of-state alcohol education 3,699 2.8%
AE Alcohol education 2,614 2.0%
AOP Out of state plan 1,628 1.2%
IP In patient treatment 925 0.7%
RC Residential care 501 0.4%
DCT Day care treatment 366 0.3%
TAE Level II Alcohol Education 316 0.2%
GT Group therapy 287 0.2%
OAE Level I Alcohol Education 217 0.2%
DAE One day AE 183 0.1%
TS Transitional Living 136 0.1%
DWI DWI clinic 84 0.1%
HH Halfway house 65 0.0%
NCT Night care treatment 3 0.0%
There are 34 programs in the data. The other 16 programs/treatments are considered
as secondary programs, which are prescribed along with these primary programs.
.

96,317 (78.1%) were coded as compliant with their DSP, i.e., completion of all recommendations.
Among those convicted, assessed, and compliant, only 19% were re-arrested, compared to 32% of
those not compliant with the DSP, and 34% of those who did not participate at all in the IDP.

Note also the left hand side of Figure 3, which shows a similar re-arrest rate of 24% (5,151 of
21,843) among those who were not convicted. Of those 21,843 arrestees who were not convicted,
32.5% (N=7102) nonetheless participated in the IDP. Of 4964 drivers who were compliant with the DSP,
15% (733) had a subsequent arrest, while of 2137 drivers who participate IDP but did not comply with
the recommendations, 21% (429) had a subsequent arrest. Of 14,741 non-convicted and not IDP
participant drivers, 27% (3989) had a subsequent arrest, while the recidivism is 34% among those who
were convicted of the index offense but did not participate IDP. Overall, the 24% rate of re-arrest in 5
years did not differ between those who were and were not convicted of the index offense. A closer
examination, however, shows that all convicted subgroups have higher recidivism rate than the
corresponding non-convicted subgroups. Depending on the IDP participation and DSP compliance,
recidivism rates vary significantly between subgroups. The compliant groups are the least likely to be re-
arrested.

22
 Figure 3. Pathway to Recidivism
2004-08 Wisconsin OWI Arrestees

Arrested
N=175,446

21,843 153,603
N Convicted Y

123,295 30,308
14,741 7,102
N IDP Y Y IDP N

2137 26974
N DSP DSP N

4964 96317

Y Y

Re-Arrested (N=42,157)
Recidivism: 24%

23
 Prior arrest status made a difference in these transitions. Figure 3a and 3b (next page) show the
transitions for drivers with priors and without priors. Those with past OWI arrests were equally as likely
to be convicted as those without priors, 86.8% (46,570 out of 53,623 drivers with priors) vs. 87.1%
(107,033 out of 122,823 without priors). Those with priors were significantly less likely to enter IDP than
those without priors, 73.5% (34,215 out of 46,570) vs. 83.2% (89,080 out of 107,033); and once they
entered IDP, less likely to complete the program than those without priors, 74.7% (25,566 out of 34,215)
vs. 79.4% (70,751 out of 89,080). Arrestees with priors tend to be rearrested slightly more often (26%)
than those without (23%). However, the rates vary by whether they were convicted and entered the
IDP. If they were not convicted, regardless of whether they entered or completed the program, or if
they were convicted but did not enter IDP, recidivism was significantly higher for those with priors than
their counterparts without priors.

The transition data on IDP entry and DSP compliance are also provided by county of residence in
Appendix 2. Appendix 7 includes data on conviction and recidivism rates by county. There is fairly wide
variation by county on each of these outcome measures.

While overall recidivism is 24%, there is variation in the total number of re-arrests (Table 7).
Among the 42,157 individuals (24% of index cohort arrestees) re-arrested, 9,062 had more than one
subsequent re-arrest. Seventy-six percent had only one arrest with no recidivism in the subsequent 60
months. About 18% had one re-arrest for OWI within 5 years, 4% had two re-arrests, and 1% (1519
individuals) had 3 additional arrests. Only 0.3%, or 320 individuals had 4 or more subsequent arrests.

Table 7: 60 month Recidivism for 2004-08 cohort1

Arrestees2 Arrests3
Number of Arrests N % N %
1 133,287 76.0 175,446 76.6
2 33,095 18.9 42,198 18.4
3 7,225 4.1 9,088 4.0
4 1,519 0.9 1,841 0.8
5 271 0.2 320 0.1
6 39 0.0 49 0.0
7 8 0.0 10 0.0
8 1 0.0 2 0.0
9 1 0.0 1 0.0
Total 175,446 228,955

1. The sample consists of OWI arrest records for 2004-08 Wisconsin drivers over the 60
months after their first OWI arrest during this period.
2. The number of unique persons who were arrested by their total number of OWIs over 60
months.
3. The total number of arrests for 1st, 2nd, 3rd, ….9th offense during the corresponding
period--a duplicated count showing 228,955 arrests for the 175,446 individuals.

24
 Figure 3a. Pathway to Recidivism
2004-08 Wisconsin Arrestees with Priors

Arrested with Prior

(N=53,623)

7053 46570
N Convicted Y

34215 1235
4707 2346
N IDP Y Y IDP N

728 8649
N DSP DSP N
1618 25566

Y Y

Re-Arrested (N=14,131)
Recidivism: 26%

25
 Figure 3b. Pathway to Recidivism
2004-08 Wisconsin Arrestees without Priors
Arrested w/o Prior
(N=122,823)

14790 107033
N Convicted Y

89080 17953
10034 4756
N IDP Y Y IDP N

1409 18325
N DSP DSP N
3346 70751

Y Y

Re-Arrested (N=28,026)
Recidivism: 23%

26
 These data are further elaborated on in Appendix 3a and 3b, which include all arrestees from
2004 to 2014. Note that the total number of cases with more than one prior is quite small (< 10%),
although the tables do not emphasize that since the metric is percent of group. Appendix 3a suggests a
pattern of recidivism for the full sample of 2004-14 arrestees by prior arrests that is similar to that for
our focal cohort of 2004-2008 cases. Appendix 3b provides an annualized assessment which could help
in projecting the annual volume of cases by the number of priors, potentially of use when legislative
policy change based on number of prior offenses is considered.

Time to recidivism was also explored by the status of priors. Survival curves without a
subsequent arrest show that rates of recidivism are gradual and tend to level out after five to six years,
particularly for those with no priors. Figure 4 shows the survival curve for our cohort of arrestees,
comparing those without any prior arrests to those with priors; rates of recidivism are higher for those
with priors but the curve is similar. Since we have OWI arrest data through 2014, the curve has been
plotted out past the 60 months that are the focus of our analysis. Here we see that the recidivism rate
increasingly flattens after 5-6 years, with a widening gap between those with no priors and those with 2
or more prior arrests.

Figure 4. Survival to Re-arrest for WI 2004-08 OWI Arrestees by Prior

100%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
101
103
105
107
109
111
113
115
117
119
121
123
125
127
129
131
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
65
67
69
71
73
75
77
79
81
83
85
87
89
91
93
95
97
99
1
3
5
7
9

All No Prior 1 2+
*60 months is the minimum length of observation for all 04-08 arrestees.

Appendix 4 provides crude recidivism rates by demographic characteristics of the arrested

drivers. Males recidivated at 25% if convicted, and 24% if not convicted, while female recidivated at
20% if convicted and 21% if not convicted. White offenders had a 24% recidivism rate whether they
were convicted or not, compared to 34% (convicted) and 40% (not convicted) for American Indians, 24%
(convicted) and 22% (not convicted) for African Americans, 27% and 18% for Hispanics, and 23% and
22% for Asians. Younger drivers were more likely to have repeat subsequent arrests than older drivers

27
whether or not they were convicted of the index arrest. For example, among convicted drivers, 33% of
under age 20, 26% of ages 20-29, but 16% of those ages 50-59 and 12% of those over 60 recidivated.
Among not convicted drivers, the age-specific recidivism rates are 30%, 23%, 23%, 23%, 18% and 12%
respectively. Table 8 further shows crude recidivism rates by IDP participation, assessment (highest

severity) and compliance status. Overall, the 60 month recidivism rate is 24% for all 04-08 arrestees.
The recidivism rate is 32% for those non-IDP drivers and 21% for IDP drivers. Nineteen percent of
compliant IDP drivers were rearrested within 60 months while 31% of non-compliant IDP drivers were
rearrested.

Table 8. Recidivism by Compliance and Highest Severity, 2004-08 Arrestees

Not rearrested Rearrested Total
All Arrestees 133,287 76% 42,159 24% 175,446
Arrestees no valid IDP 30,771 68% 14,278 32% 45,049
Irresponsible Use (IU) 51,751 84% 10,132 16% 61,883
IU Borderline (IUB) 18,666 78% 5,297 22% 23,963
Suspected Dependence 9,846 72% 3,828 28% 13,674
Dependence 22,253 72% 8,624 28% 30,877
Arrestees in IDP (total) 102,516 79% 27,881 21% 130,397
Compliant (by Finding):
IU 45,254 85% 8,005 15% 53,259
IUB 14,870 80% 3,757 20% 18,627
SD 7,389 76% 2,343 24% 9,732
D 14,965 76% 4,703 24% 19,668
All Compliant 82,478 81% 18,808 19% 101,286
Non-compliant (by Finding):
IU 6,497 75% 2,127 25% 8,624
IUB 3,796 71% 1,540 29% 5,336
SD 2,457 62% 1,485 38% 3,942
D 7,288 65% 3,921 35% 11,209
All Non-Compliant 20,038 69% 9,073 31% 29,111

A general glance at Table 8 further suggests that recidivism is positively associated with severity.
That is, drivers assessed as irresponsible use of alcohol (IU) are less likely to be rearrested than those
who are irresponsible use-borderline (IUB); those assessed IUB are less likely to be rearrested than those
who have suspected alcohol dependency (SD). Those with SD are less likely to be rearrested than those
with alcohol dependence or dependence in remission (D). (Recidivism outcomes were very similar for
dependent drivers whether or not assessed as in remission so we combined these groups in most
analyses.) For example, if the drivers complied with the DSP plan, the recidivism rates are 15% for IU,
20% for IUB, and 24% for both SD and D drivers. If the drivers did not comply with the plan, the rates
are 25%, 29%, 38% and 35% respectively. (For more details on compliance and recidivism of each
recommendation by severity, see Appendix 5.)

28
 Recommendations for IUB drivers is a concern because they are between irresponsible use and
dependency. The recommendation guidelines for IUB drivers allows either outpatient treatment or
education, or both. Appendix 6 addresses this concern. We found that those recommended for both
treatment and education programs tended to be less compliant than those recommended for treatment
or education alone. The compliance rates are 32% for both, 81% for education only, and 80% for
treatment only. Of all compliant IUB drivers (3rd column), recidivism rates are 12% for both, 18% for
education only, and 22% for treatment only. Among all IUB drivers, these rates are 15%, 19%, and 25%
correspondingly. Consistent with many other tables in this report, we again see that compliance is
important to reduce recidivism. Additionally, it is intriguing that those who were recommended for
both programs, though less compliant than others, were re-arrested the least whether they complied or
not. Alternatively, it could be interpreted that about 25% (i.e., 35/135) of the rearrested IUB drivers with
both programs had complied, 76% with education only had complied, and 70% with treatment only had
complied. Several possible reasons to interpret the results. For example, a selection effect of the
recommendation assignment during the assessment pre-determined the assignment; being assigned to
two types of programs works better than single type; or the finding and recommendation were
mismatched. In all, those findings warranted further analysis on the assessment process.

While we have described conviction rates, participation in IDP, DSP compliance and recidivism
rates, the analysis thus far has only been descriptive and bivariate. Our descriptive findings confirm that
recidivism varies by personal characteristics, court decision, assessment and compliance as suggested by
our conceptual framework in Figure 1. These various outcomes are inter-related and are also potentially
predicted by local contextual variables and personal factors. The role of the IDP is one factor among
many that may contribute to recidivism or its prevention. To sort this out, multi-variate regression
models were estimated.

Multi-variate Analysis

In this section, we take into account the personal characteristics and assessment findings to
analyze (a) conviction for the index offense, (b) participation in an IDP assessment, (c) compliance with
the subsequent driver safety plan (DSP), and (d) 60-month recidivism. Personal predictors (individual
background) include prior arrest status, age group, race/ethnicity, and gender, as explained in the
method section. For those participating in the IDP we also have data on assessment findings, DSP
recommendations and compliance. Based on the multi-variate regression at individual level, we
believed that contextual variables, that is, county characteristics, where socioeconomic composition,
drinking culture, alcohol outlet density (exposure), traffic and road conditions, law enforcement, court
decision, and human services which vary through the state may influence the outcomes of interest.
Since we do not have exact measures of these variations, we used county level data from the UWPHI’s
County Health Rankings (http://www.countyhealthrankings.org/app/Wisconsin/2016/downloads) as
proxies for county environment. By also entering an indicator variable for each county, control for
variation in other county-level factors in which offenders are clustered is accomplished.

We started the multivariate analysis using regression equations to estimate the effects of
personal characteristics on conviction, IDP participation, DSP compliance, and recidivism. Then

29
depending on the outcomes, nesting on the first set of personal characteristics equations, step by step,
we added the transitions, or results from the transition such as assessment findings, into the equation as
the intermediate variables to explain the subsequent outcomes. As described above, outcomes varied
by county. Due to a lack of ideal explanatory variables at the county level, we used two kinds of
multilevel analysis. We first used a fixed effect model. That is, there is a general effect for the
individuals nested in the same county and then the county is used as the indicator to remove all
potential unobserved confounding variables at the county level from the analysis to improve the
estimates of individual characteristics. In a second modeling approach, we followed the traditional
hierarchical linear model by adding county characteristics to explain some contextual differences. In
addition to recidivism, multilevel models were also applied to intermediate outcomes, i.e., conviction,
participation in IDP, compliance with DSP. Since our county level measures to explain the county
variation were less than ideal, the hierarchical approach did not fit much different from fixed effect
models. That is, by adding county level variables, the models did not improve significantly from the less
complex fixed effect model. Thus the fixed effect models are our preferred models. To simplify this
report, in tables of findings, we only report the estimates and odds ratios from the fixed effects models.
However, a few interesting findings from the HLM are also mentioned in the narrative. We also
calculated the probability of an outcome for specific subsets of arrestees based on the regression

models; these estimated probabilities are provided in the narrative when appropriate.

Table 9 provides the results from modeling the likelihood of conviction for the index arrest.
The individual background variables were all significantly (i.e., t-test > 1.96) related to conviction. Those
with priors were less likely to be convicted (odds ratio 0.83). Using the model, the predicted average
probability of conviction is 0.86 (SE=.006) for those without priors and 0.84 (SE=0.007) for those with

Table 9. Predicting Conviction, OWI offenders first arrested in 2004-08

B T-test OR
Individual Covariates
20-29 (Ref) 1.000
<20 -0.609 -26.13 0.544
30-39 -0.067 -3.15 0.936
40-49 -0.097 -4.41 0.908
50-59 -0.152 -5.08 0.859
60+ -0.121 -2.39 0.886
White (Ref) 1.000
Asian Am. -0.213 -2.90 0.808
Black -0.248 -7.16 0.780
Hispanic -0.380 -13.79 0.684
Am. Indian 0.116 2.21 1.123
Female 0.176 9.36 1.192
With Prior -0.183 -10.62 0.833
County latent factor 1.000 2.718
Variance(County) 0.175 5.68 1.192

30
priors. All age groups were less likely to be convicted when compared to the reference age group of 20-
29, which had the highest conviction rate. Relative to the White reference group, Asian, Black and
Hispanic individuals were less likely to be convicted. Of all three groups, the OR for Hispanic, the lowest
one, is 0.68. Native American Indians were more likely than other groups to be convicted, with
OR=1.123. Females had a higher conviction rate than males (OR=1.19).

Entering county had an overall significant effect in these models, whether entered as a random
effect (shown in Table 9) or as a hierarchical second level with covariates at the county level. Conviction
rates varied by county, ranging from 66 to 92 percent of those arrested. Among the variables used in
hierarchical modeling from the County Health Rankings, conviction rates were lower in counties with
higher numbers of liquor outlet density per capita, but higher in counties with higher rates of per capita
smoking.

Table 10 shows the results from modeling of the likelihood of participating in the IDP
assessment for the index arrest. Both the full arrestee and the convicted samples are reported. The full
arrestee sample is important because it provides a general estimation for drivers once arrested during
2004-2008. As shown in Figure 3, we found that entering IDP also made a differences in reducing
recidivism of non-convicted drivers.

The model for all arrested, which includes conviction as a variable, shows a huge effect of
conviction on participation (OR=11.74). Also, on average those arrestees with priors are slightly less
likely to enter the IDP (OR=0.96) than those without. Interactions become complex here. Compared to
those without priors and not convicted (NP/NC), while everything else is assumed equal, the odds ratios
are 11.74 for no prior and convicted (NP/C), 0.96 for prior and not convicted (P/NC), and 5.414 for prior
and convicted (P/C). The average predicted probability that an offender would enter the IDP/WAID are
0.34 for NP/NC, 0.84 for NP/C, 0.33 for P/NC and 0.72 for P/C. Another striking contrast is between
those underage drinker drivers and the rest of drivers. The OR is 0.66 for underage drinker drivers in
comparison with the drivers in their 20’s. Minority drivers (except Asians), including Black, Hispanic, and
American Indian, are also significantly less likely to enter IDP, i.e., OR=0.31, 0.36, and 0.38 respectively.

The coefficients (or OR’s) were similar between the equations for the full sample and the
convicted sample for racial/ethnic groups. However, the age variation between the youngest group and
the 20’s group increased from the full sample (OR=0.66) to the convicted sample (OR=0.80). Part of the
improvement may be due to the fact that the underage drivers are much less likely to be convicted
(OR=0.544, see Table 9) than adult drivers. So for the full sample, the participation OR for those under
age is also relatively smaller than for the reference group (in their 20’s). In short, conviction increases
IDP participation. The IDP participation OR for the underage thus improves as well.

Convicted drivers with priors were less likely to participate (OR= 0.46 in the second equation).
The average predicted probability of participation for those with priors is 0.72 while for those without
priors it is 0.84. Females were significantly more likely than males to participate (OR=1.25) in IDP
assessment.

31
 Entering county had an overall significant effect in these models, whether entered as a random
effect or as a hierarchical second level with covariates at the county level. IDP participation also varied
by county, ranging from 52% to 79%. Among the variables used from the County Health Rankings,
participation was significantly higher in counties with higher percent of individuals with some college
(OR=1.13), lower in counties with higher rates of smoking (OR= 0.49), and higher in counties with higher
overall ratings on the CHR health factor.

Table 10. Predicting IDP/WAID Participation, OWI offender arrested in 2004-08

Arrestee Sample Convicted Sample
(N=170862) (N=149935)
B T-test OR B T-test OR
Ref: Age 20's
Age < 20 -0.410 -19.38 0.66 -0.225 -9.20 0.80
Age 30's -0.053 -3.07 0.95 -0.026 -1.39 0.97
Age 40's -0.115 -6.44 0.89 -0.091 -4.63 0.91
Age 50's -0.037 -1.47 0.96 0.027 0.94 1.03
Age 60 + -0.187 -4.49 0.83 -0.075 -1.59 0.93
Ref: White
Asian -0.036 -0.55 0.96 -0.130 -1.78 0.88
Black -1.165 -43.49 0.31 -1.299 -46.31 0.27
Hispanic -1.032 -46.13 0.36 -1.155 -48.81 0.32
American Indian -0.972 -26.39 0.38 -1.037 -26.88 0.35
Female 0.204 13.01 1.23 0.222 12.75 1.25
With prior -0.043 -1.32 0.96 -0.767 -50.30 0.46
Convicted 2.463 119.30 11.74
With prior X Convicted -0.731 -21.08 0.48
County latent factor 1.000 2.72 1.000 2.72
Variance(County) 0.044 5.14 1.05 0.048 5.19 1.05

Table 11 shows the results from modeling compliance with the driver safety plan (DSP)
associated with the index arrest. To estimate compliance, only those participating in IDP and thus with a
DSP are relevant. We have two samples: the sample for any IDP participants and that for the convicted
participants. Most explanatory variables, including personal background, except Asian and female, and
assessment findings (severity and substance) are significantly associated with the DSP compliance. Age
shows an almost linear direct relationship with compliance; the OR’s are 0.84, 1, 1.25, 1.53, 1.85 and
1.83 for the six age groups. Relative to White (i.e., the reference group) and Asian, Black, Hispanic and
Native American Indians were far less likely to comply with their DSP (ORs of 0.56, 0.74 and 0.65,
respectively). Since the DSP was based on the IDP assessment, we are able to examine influences of
assessment in this model. The highest severity level is inversely related to compliance. Relative to
offenders assessed as irresponsible use, those with each of the more serious findings are significantly
less likely to comply with their plan (ORs of 0.55, 0.38 and 0.23). Those coded as having other drug
involvement also had lower rates of compliance (OR= 0.74). Note that there were only small proportion
of drivers coded with drug problem.

32
 Table 11. Predicting WAID Compliance for Index Arrest,
OWI offenders arrested in 2004-08
IDP/WAID sample Convicted IDP/WAID
(N=129134) (N=122129)
B T-test OR B T-test OR
Ref: Age 20's
Age < 20 -0.179 -7.753 0.836 -0.190 -7.983 0.827
Age 30's 0.225 11.660 1.252 0.225 11.220 1.252
Age 40's 0.425 20.370 1.530 0.431 19.870 1.539
Age 50's 0.616 20.040 1.852 0.626 19.570 1.870
Age 60 + 0.603 11.150 1.827 0.603 10.820 1.827
Ref: White
Asian -0.072 -0.999 0.930 -0.090 -1.192 0.914
Black -0.581 -16.780 0.559 -0.594 -16.440 0.552
Hispanic -0.300 -10.120 0.741 -0.316 -10.150 0.729
American Indian -0.436 -9.317 0.647 -0.430 -8.913 0.651
Female -0.019 -1.135 0.981 -0.024 -1.418 0.976
Convicted 0.514 14.640 1.672
Ref: Irresponsible User
IU borderline -0.604 -26.360 0.547 -0.604 -25.140 0.546
Suspected dependency -0.977 -34.110 0.377 -1.022 -34.610 0.360
Dep./Dep in remission -1.466 -58.880 0.231 -1.505 -58.920 0.222
Drug user -0.292 -11.130 0.747 -0.295 -11.040 0.744
With prior -0.175 -2.638 0.840 -0.414 -10.650 0.661
Interactions
With prior x conviction -0.244 -4.144 0.783
With prior x IU borderline 0.405 7.955 1.499 0.396 7.416 1.485
With prior x SD 0.518 9.513 1.679 0.548 9.656 1.729
With prior x D/DR 0.658 14.300 1.931 0.660 13.810 1.935
With prior x Drug -0.064 -1.411 0.939 -0.059 -1.280 0.943
County latent factor 1.000 2.718 1.000 2.718
Variance(County) 0.034 4.907 1.035 0.039 4.934 1.040

The estimates are very similar between the two equations except for the influence of prior
arrest status. That suggests that the estimated average effects of age, race/ethnicity, severity, and
substance did not vary by conviction. Of all IDP participants, however, those with priors are significantly
less likely to comply with the DSP (OR= 0.84) than those without priors. Among the convicted IDP
participants, the odds ratio for those with priors decreases to 0.66. After taking into account the
interactions between convicted and prior status, in contrast to the no prior and no conviction
participants (NP/NC), the odds ratios are 0.84 for those with prior and not convicted, 1.67 for those with
no prior but convicted, and 1.11 for those with priors and convicted. Alternatively, the average
predicted probability of compliance among all IDP participants are 0.68 for NP/NC, 0.77 for NP/C, 0.71
for P/NC, and 0.76 for P/C.

33
 Note that the interaction effects of prior status and severity are also fairly similar between the
equations for all IDP participants and convicted participants. Complex interactions must be taken into
account in interpreting the lower part of Table 11. The odds ratios contrast the reference group of IU
without priors to IUB without prior, SD without prior, D/DR without prior, IU with prior, IUB with prior,
SD with priors and D/DR with priors. For the convicted IDP participants, the calculated odds ratios are
0.82, 0.63, 0.44, 0.83, 0.69, 0.53 and 0.36 correspondingly. The differences between those subgroups of
severity and prior mainly are due to the differences in effects of prior status between convicted and not
convicted IDP participants.

Entering county had an overall significant effect in these models, whether entered as a random
effect or as a hierarchical second level with covariates at the county level. Compliance by county ranged
from 64% to 85% of those who participated in the IDP. Among the variables used from the County
Health Rankings, compliance was higher in counties with higher percent of individuals with some college
(OR=1.07), and lower in counties with higher rates of low income (OR= 0.29).

Recidivism in 60 months was modeled using the same basic modeling estimation structure.
Recidivism was estimated for the full sample of arrestees and the sub-sample of IDP participants in
Table 12. Analysis of full sample helps us to summarize recidivism for the arrestee population; the IDP
sample gives us some insights on the effects of assessments findings and compliance with the DSP plan.

A. Full Sample Recidivism:

Most individual background variables were significantly (t-test > 1.96) related to recidivism, with
the exception of Black and Asian race/ethnicity. Age was inversely associated with recidivism. The
younger the driver was at the index arrest, the more likely s/he would be rearrested in 60 months. Since
we have limited the length of observation to 60 months for each person, age is not confounded with the
length of the period. The association between age and recidivism is not mediated by IDP participation,
since the coefficients are nearly identical for the sample of all arrestees and the IDP participating
sample.

In terms of race and ethnic groups, relative to White drivers, Native American Indians are the
only group who are more likely to be rearrested regardless of their IDP participation, that is, OR=1.45 for
all and 1.38 for IDP participants. Among all IDP participants, Asian and Black drivers do not differ
significantly from similar white drivers in re-arrest rates, whether or not they participate in the IDP
program. Hispanic drivers in the overall sample also do not have different re-arrest rates than the
White reference group (OR=0.968); however, Hispanics who participate in the IDP are more likely to be
rearrested in 60 months (OR=1.19). Female drivers have significantly lower recidivism rates than males
regardless of IDP participation (OR= 0.78).

Those who have been convicted of the index arrest are more likely to have a subsequent arrest
than those who were not convicted regardless of IDP participation, i.e., OR=1.56 for all arrestees and
1.781 for IDP participants. The status of priors, like conviction, also has a direct relationship with
increased recidivism regardless of IDP participation (OR=1.82 and 1.71). Interaction are again complex
here and calculated values are necessary. In contrast to drivers without prior and not convicted, overall,

34
those convicted with priors are even more likely to be rearrested (OR=2.14). Among IDP participants,
the odds ratios are 1.71 for convicted, 1.78 for with prior, and, thus, 1.82 for convicted with priors.
However, IDP participation is inversely associated with recidivism (OR=0.56). Drivers participating in
the IDP and with priors are less likely to be rearrested than those not participating IDP and without
priors (OR=0.87).

Table 12. Predicting Recidivism, OWI offenders arrested in 2004-08

All arrestees IDP/WAID Sample
(N=170,862) (N=129,134)
VARIABLES B T-test OR B T-test OR
Ref: Age 20's
Age < 20 0.369 20.320 1.447 0.383 17.410 1.467
Age 30's -0.259 -16.230 0.772 -0.262 -13.450 0.770
Age 40's -0.369 -21.660 0.692 -0.377 -18.050 0.686
Age 50's -0.666 -25.700 0.514 -0.694 -21.650 0.500
Age 60 + -1.012 -20.290 0.363 -0.972 -15.830 0.378
Ref: White
Asian -0.103 -1.757 0.902 0.006 0.092 1.006
Black 0.009 0.311 1.009 0.038 0.976 1.038
Hispanic -0.032 -1.415 0.968 0.178 5.997 1.194
American Indian 0.371 10.530 1.449 0.319 6.758 1.375
Female -0.246 -17.130 0.782 -0.247 -14.520 0.781
Convicted 0.446 18.240 1.563 0.577 13.140 1.781
Entering IDP/WAID -0.579 -32.100 0.560
Compliant with DSP -0.507 -26.090 0.603
Ref: Irresponsible User
IU borderline 0.411 18.500 1.509
Suspected dependency 0.785 27.680 2.192
Dep./Dep in remission 0.851 32.790 2.341
Drug user -0.349 -12.380 0.706
With prior 0.599 16.850 1.821 0.534 6.692 1.706
Interactions
With prior x conviction -0.286 -7.336 0.751 -0.513 -7.413 0.599
With prior x IDP/WAID -0.155 -5.548 0.856
With prior x compliance -0.075 -2.199 0.928
With prior x IUB -0.225 -4.285 0.799
With prior x SD -0.305 -5.456 0.737
With prior x D/DR -0.237 -4.906 0.789
With prior x Drug use 0.338 6.890 1.402
County latent factor 1.000 2.718 1.000 2.718
Variance (County) 0.012 4.505 1.012 0.019 4.359 1.019

In summary, once arrested for an OWI, the average predicted probability of re-arrest within 60
months is 0.234 for those without priors, 0.278 for those with priors, 0.198 if not convicted, 0.256 if
convicted, 0.34 for those not participating in the IDP, and 0.218 for those who participate in the IDP. If a
driver was first time arrested between 2004 and 2008 and not convicted, the probability of re-arrest in
60 months is 0.173. For a driver who was arrested with a prior and did not participate in the IDP, the
probability of re-arrest in 60 months is 0.393; if s/he participated in the IDP, the probability would be

35
reduced to 0.240. (These numbers differ from those in the flow diagrams since the probabilities are
adjusted statistically to control for individual differences between groups.)

B. IDP Participant Sample Recidivism:

Since the DSP follows from an assessment, we were able to examine the results of assessment
findings and compliance only in models that excluded cases who did not participate in the IDP
assessment. In the 2nd equation in Table 12 (right hand panel), recidivism was estimated based on the
same set of variables as those in the first model, but dropping IDP participation as a variable since all
cases were participants, and adding the results of the assessment and DSP compliance. The reference
group for the measure of severity (WAID finding) is irresponsible use of substance. Relative to offenders
assessed as irresponsible use, those with the more serious findings were significantly more likely to be
re-arrested. The odds ratios are 1.5 for those assessed as borderline irresponsible use, 2.19 for those
with suspected dependency, and 2.34 for those with dependency (including in remission). Those coded
for drug use in addition to the alcohol were less likely to be re-arrested (OR=0.71). Compliance with the
DSP is also an important predictor of reduced recidivism (OR=0.603).

Taking into consideration priors (OR=1.706), assessed severity level (1.509/2.192/2.341) and
their interaction (0.799/0.737/0.789), the calculated odds ratios for recidivism are 2.05 for borderline
irresponsible use with priors, 2.72 for suspected dependency with priors, and 3.16 for dependency/in
remission with priors, when compared to the reference group of drivers assessed with irresponsible use
without priors.

Entering county had an overall significant effect in these models, whether entered as a random
effect or as a hierarchical second level with covariates at the county level. Recidivism varied
considerably by county, ranging from 18% to 36%.

We explored the county effects on a set of logistic regressions to predict the conviction and
recidivism rates. These are reported in Appendix 8. In the table, both crude (unadjusted) conviction and
recidivism rates and predicted (adjusted) rates are presented for each county. In the predicted rates,
personal background characteristics, including prior OWI status, used in above multi-level models are
also included in the adjustment. As seen in the table, the predicted and unadjusted rates were fairly
similar. That is, the population (arrestee) composition does not influence or explain the large variation
in county conviction and recidivism rates. Given the significant variations from county to county, further
analysis of the role of county characteristics should be explored with additional county-level variables.

36
 Discussion and Conclusions
These data yield important results for consideration in program and policy development related
to operating while intoxicated in Wisconsin (and potentially nationally). In this section we recap the
most significant or salient findings, and suggest some actionable options for discussion among policy
and program stakeholders.

OWI arrests declined steadily over the 11 years (2004-2014) we analyzed. These rates are likely
affected by individual driver’s behavior, intensity of law enforcement activity, and other issues. The
decline mirrors national trends over the years (2004-2014) we analyzed. The National Roadside Survey
of Alcohol and Drug use by Drivers (NHTSA, 2015) found a decreasing trend in alcohol use among
weekend nighttime drivers nationally, with intercepted drivers registering > .08 BrAC dropping from
7.5% in 1973 to 1.5% in 2013-14. The National Survey on Drug Use and Health (NSDUH) also has found
a decrease in driving under the influence of alcohol and illicit drugs during this period; they also point
out that in 2014 approximately 27.7 million people aged 16 or older (11.1 percent of the US population)
drove under the influence of alcohol, and 10.1 million under the influence of other drugs (Lipari, Hughes
and Bose, 2016). In spite of an apparent decrease in impaired driving over time, analysis of the 2012
BRFSS estimates that nationally there are 121 million episodes of alcohol-impaired driving and a rate of
505 episodes per 1,000 population annually; the Wisconsin rate is 828 episodes per 1,000 (Jewett et al.,
2015). While impaired driving may have decreased, the Behavior Risk Factors Surveillance Survey
(BRFSS) indicates that in Wisconsin adult binge drinking and heavy drinking over this period did not
change much, with 22% to 25% of all adults reporting binge drinking (WiDHS & UWPHI, 2016). These
data suggest that prevention efforts resulting in cultural change in perceptions of acceptable driving
behavior may have been effective in reducing the rate of impaired driving both nationally and perhaps in
Wisconsin, but not the rates of heavy or binge drinking. In spite of these changes, Wisconsin in 2014
experienced about 30,000 arrests for OWI offenses, and, over the years 2006-2009 23.7% of drivers
surveyed admitted to driving “under the influence of alcohol”, the highest of all states in the nation
(SAMHSA, 2010).

Based on our analysis of the Wisconsin OWI arrest data, among all of those arrested (n=341,157
individuals with 417,347 arrests) for OWI over an 11 year time period, most (78%) had only one arrest
and 17% had two arrests during the 11 years. A few had three (4%) or four (1%) arrests, while only
0.3% (about 1,000 people) had five or more arrests. Taking into account all documented lifetime OWI
arrests prior to the study period, based on 417,347 arrests (and not individuals) over the 11 years, 58%
of arrests were first arrests, 23% were second arrests, 11% were third, 5% were fourth, 2% were fifth
arrests and slightly more than 1% were sixth or higher. For the most recent year in this data set (2014)
there were 17,134 first arrests (58%), 6,416 second arrests (22%), 3,198 third arrests (11%), 1,475 fourth
arrests (5%), 707 fifth arrests (2%), and 539 arrests (1.8%) which were 6th or more. 5

5
Note that these are duplicate counts since one individual may have more than one arrest in a given year; also, the
number of arrests we report is likely higher than the statutory number taken into account for sentencing purposes.

37
 We intensely analyzed data from a 5-year cohort of 175,446 individuals with at least one OWI
arrest from 2004 through 2008. Those with OWI arrests were most often young adult (39% age 20-29),
white (82%), males (77%). Among those with any OWI arrest during this 5 year time period, 31% had at
least one prior arrest; those with priors were older, more likely to be male, and disproportionately more
white or American Indian than those with no priors. Thus preventive interventions, particular those for
first offenses, should be particularly targeted to young white males. Native Americans could also be
specifically targeted for recidivism prevention.

In our cohort, about 88% of those arrested were convicted of an OWI 6 offense. Regression
analysis showed that conviction was most likely among 20-29 year olds; more likely among American
Indians but less likely among Hispanic and African Americans than among Whites; more likely among
females than males; and less likely among those with prior arrests. Conviction rates varied by county,
ranging from 66 to 92 percent of those arrested. When specific county characteristics were included in
the analysis, we found that liquor license density was associated with lower conviction rates. Further
investigation would be needed to analyze these geographic and personal characteristic differences in
conviction rates.

The Intoxicated Driver Program (IDP) includes an assessment for all persons convicted of an
OWI offense using a tool called the Wisconsin Assessment of Intoxicated Drivers (WAID), the
development of a Driver Safety Plan (DSP) based on the WAID findings, and subsequent
monitoring/certification of compliance with the DSP. In our cohort of arrestees, 74% participated in the
IDP WAID assessment (80% of those who were convicted plus about a third of those not convicted for
the index offense). Regression analysis indicated that those with prior offenses, those under 20 or 40-
49, and Black, Hispanic and Native Americans were significantly less likely to participate in the IDP
assessment process. Participation also varied by county, ranging from 52% to 79%. Measured county
level characteristics that influenced this participation rate included percent with at least some college
education, liquor licenses per capita, and percent smokers in the county, factors that again need further
exploration.

Among participants in the IDP, 48% were assessed as “irresponsible use” and 18% as
“irresponsible use-borderline”. Ten percent had “suspected dependency” and 24% “dependency”
(including 7% “in remission”). Most frequent DSP primary education or treatment requirements were
group dynamics course (46%), outpatient treatment (43%), and multiple offenders program (6%).

Compliance with the Driver Safety Plan (DSP) within two years of the index offense was
recorded for 78% of those with a DSP. Compliance (among those who had entered the IDP) did not
differ by prior offenses, but was lower for those under 30 years old, and lower for minority groups
(African American, Hispanic, and Native American). WAID assessment data, available on all those who
had a DSP, indicate that those found to have “irresponsible use” were more likely to comply than other
diagnostic groups.

6
We do not have data on how many of those individuals not convicted of OWI were convicted of a lesser offense,
diverted from prosecution, or found innocent.

38
 Recidivism (re-arrest for OWI) within 60 months of the index arrest occurred in 24% of this
cohort. This rate did not differ among those who were or were not convicted of the index offense.
However, among those convicted who participated in the IDP and complied with their DSP, only 19%
recidivated, compared to 32% of those who did not comply with a DSP and 34% of those who bypassed
the entire IDP process. Regression analysis, including the entire arrest cohort, found significantly
higher re-arrest rates for those with prior offenses, 7 those under age 29, American Indians, and those
who were convicted of the index offense. Females, older drivers, and those who entered the IDP
program had significantly lower recidivism rates. Recidivism varied considerably by county, ranging
from 18% to 36%. Several local (county) characteristics were also significant in exploratory analysis.

Among those who participated in the IDP assessment process, older age groups and females
had lower recidivism rates. Hispanic and American Indians (but not African Americans) who participated
had higher recidivism than other groups, as did those who had been convicted. Those who were
compliant with the DSP had a lower recidivism rate (19%) than those who were non-compliant (32-34%).
Relative to those assessed as irresponsible use, the borderline, suspected dependency and dependent
cases had higher recidivism rates (28% of dependent, 22% of borderline and 16% of irresponsible users),
as did those with prior arrests (unless they complied with the DSP, in which case they did not differ).

With statistical adjustments for background variables, our results indicate that, once arrested
for an OWI, the average predicted likelihood of re-arrest within 60 months is:

• 23.4% for those without priors, 27.8% for those with priors,
• 19.8% if not convicted, 25.6% if convicted,
• 34.0% for those not participating in the IDP, 21.8% for those who participate in the IDP.
• If first time arrested between 2004 and 2008 and not convicted, the likelihood of re-
arrest in 60 months is 17.3%.
• If arrested with a prior and did not participate in the IDP, the likelihood of re-arrest in
60 months is 39.3%; if the driver participated in the IDP, average recidivism is reduced
to 24.0%.

Limitations:

The most serious limitation of this analysis is the potential role of self- selection bias. The data
conclusively indicate that, holding other variables constant, participation in the IDP and compliance with
the DSP are positively associated with reduced recidivism. However, this is a non-experimental analysis
which may reflect that those who self-select to participate in the IDP, and to comply with DSP, are
inherently different from those who do not choose to participate. If so, the difference in outcomes may
be partially due to those self-selection factors rather than an intervention effect. However, self

7
While statistically significant, the compliance and recidivism patterns for those with prior arrests were not much
different from those with no prior arrests; the overall 60 month unadjusted re-arrest rate for those with priors was
26%, compared to 23% for those with no priors. For those with prior offenses who entered and complied with the
IDP, there was no difference from the average in recidivism.

39
selection is only part of the problem; systemic barriers also clearly serve to reduce participation and
compliance rates, independent of individual factors.

Further, there were significant difficulties with data linkage, and at times missing information, in
particular details regarding the WAID assessment findings. We also were not able to obtain data on
driver’s license revocation status and the role of revocation in compliance and recidivism. The data on
symptoms from the WAID may not be reliably recorded, in particular information on other drug
involvement. The WAID itself is a dated instrument that does not reflect current diagnostic criteria for
substance use disorders. The data set we worked with apparently did not include all OWI-related
arrests in the state; those arrested/charged solely with “prohibited alcohol concentration” (PAC) were
excluded. However, with our very large sample and careful matching procedures and protocols, these
problems are likely to have had only marginal effects on the validity of our conclusions.

There are further limitations in the modeling. We did not build in as many interaction terms as
perhaps are warranted, such as assessment finding by treatment approach and compliance. We were
not able to model the relationship of consistency of symptoms with findings, treatment
recommendations and outcomes. These limitations could be considered as specification errors, as data
limitations, or as the result of time constraints as we refined the variables and models. Model
refinement can proceed for extended amounts of time as new specifications and coding of variables are
tried, interaction terms introduced, and additional variables added. Further work with this data would
provide additional insights of potential value, including better assessment and analysis of county-level
variables and testing of the results during the more recent, post-ignition interlock time frame.

Finally, we used county-level data from a readily available data base (county health rankings).
To understand in depth the variability by county would require far more investigation. For example, it
would be advantageous to incorporate data on local treatment capacity and utilization, population
density, variability in law enforcement intensity, rurality and size of law enforcement pool.

Conclusions:

What do these findings suggest regarding the IDP as an intervention to prevent recidivism? Several
conclusions can be reached:

• Any policy changes should consider that to improve overall population safety in a
significant manner, a focus on preventing or deterring first offenses is likely to have
more impact than a focus on those with four, five or more offenses (as much recent
legislation has done). Recidivism rates are not much different between those with and
without a prior arrest; recent research (Rauch et al., 2010) indicates that an individual
with a history of even one OWI arrest is 7 times more likely to be arrested in a given
year than an individual who has never been arrested for OWI.

• Recidivism is considerably lower among those arrestees who participate in the IDP
WAID assessment and subsequently comply with their driver safety plan (DSP). Our

40
 primary hypothesis—that IDP participation and compliance would add significant value
in reducing recidivism—is confirmed. Thus strategies to improve participation and
compliance with the IDP program would likely be productive, even though this finding
undoubtedly reflects some degree of self-selection bias. Tested strategies include
follow-up with OWI offenders who fail to schedule their IDP assessments, and increased
use of motivational interviewing to improve compliance with the driver safety plans
(DSPs) (see DCTS, 2016). Other possible (but untested) strategies to enhance
compliance could include further court intervention with those who fail to comply,
higher penalties for failure to participate, and reduced initial penalties and fees to
remove financial barriers to participation while reserving higher penalties for those who
fail to follow through.

• Preventive interventions should be particularly targeted to young white males. Native

Americans should also be specifically targeted for recidivism prevention. Increases in
the perception of the likelihood of arrest, increasing individual agreement with the
goals of the policy, and multi-component programs have been found to reduce alcohol-
impaired driving and arrests (Bertelli & Richardson, 2008; Miller et al., 2015; Shults et
al., 2009; UWPHI, 2017). The use of publicly announced sobriety checkpoints, ignition
interlocks for convicted offenders, and mass media campaigns are effective strategies
(Jewett et al., 2015; Guide to Community Preventive Services). Additional strategies to
reduce overall alcohol consumption (such as increased taxation of alcohol, reduced
outlet density, electronic screening and brief intervention, public education and
awareness campaigns), server liability and host training, and policies which reduce the
risk of driving after drinking by considering location of establishments and
transportation options have some evidence of success (Guide to Community Preventive
Services; Scott, 2013; UWPHI, 2017; Voas and Lacey, 2011). Increased penalties are not
effective beyond some minimal threshold (Scott, 2013).

• There are differences in participation in the IDP program which contribute to systematic
differences in recidivism rates. Participation is lower for those with prior offenses,
those under 20 or in the 40-49 age group, and among African Americans, Hispanics and
Native Americans. Similarly, among African American, Hispanic and Native Americans
who do participate in the IDP, compliance with the DSP was significantly lower. For
Hispanics and Native Americans, but not African Americans, recidivism was also higher
than for non-Hispanic Whites. These disparities suggest that strategies are needed to
improve the participation of non-White populations at all stages of the IDP in order to
reduce recidivism. The cost of IDP participation may be a barrier to participation and
compliance, particularly with socio-economically disadvantaged groups.

• County differences were pronounced in rates of conviction, IDP participation and

recidivism. County level enforcement activity, court procedures, and treatment systems

41
 (not assessed in this analysis) vary considerably and may contribute to these observed
differences, as was pointed out by Piper and King (1997) 20 years ago. Variation on
some of these factors should also be addressed, particularly in low IDP participation
counties with higher recidivism rates.

• Our analysis was limited by a lack of access to data on the status of driver’s licenses
(e.g., revocation) at the point of OWI arrest, inability to account for conviction for lesser
offenses, lack of analysis of BAC levels, and data linkage difficulties. Further analyses
could examine to what extent noncompliant drivers with revoked licenses continue to
drive and commit subsequent offenses, the predictive value of BAC level, the role of
specific symptom patterns, factors related to county-level variability and other relevant
questions.

Policy proposals should be clear about the goals of proposed changes, and use existing data to
predict potential outcomes—intended and unintended. Is the objective to deter, punish or to extend
treatment? Particularly for drivers with multiple repeat offenses, OWI is a symptom of a serious
substance use disorder, not necessarily willful or rational criminogenic character where deterrence
approaches are effective. Meta-analytic evidence reported by Miller et al. (2015) suggests that multi-
component interventions, rather than education or treatment alone, have the most promise in reducing
recidivism. Programs and policy should be framed to reduce disparities of access and benefit, while
increasing public safety. Potential unintended negative consequences, including overburdening law
enforcement, courts, and correctional systems, must be weighed relative to the potential public safety
benefits of proposed policy changes. These considerations would seem to argue for a preventive and
treatment focus in any policy changes.

Regardless of its limitations, the analysis presented in this report provides an in-depth source of
data which we hope will inform future policy deliberations to address the declining but still very
prevalent incidence of operating while intoxicated in Wisconsin. It is clear that for offenders who
comply with the IDP system, recidivism rates are lower. Perhaps the critical issue is how to increase
rates of compliance and reduce disparities in benefit from the IDP services. This is likely best
accomplished both by improving and standardizing the community level systems in place and by
working to further motivate offenders to participate in beneficial services.

42
 References
Bertelli, AM, Richardson, LE, 2008. “The behavioral impact of drinking and driving laws.” The Policy Studies
Journal 36 (4): 545-569.

Division of Care and Treatment Services, 2016. “Wisconsin Intoxicated Driver Program Noncompliance with
Assessment Survey Results.” Wi Department of Health Services Publication P-01557.

The Guide to Community Preventive Services website, https://www.cdc.gov/epo/communityguide.htm, accessed

April 20, 2017.

Hill, JD, 2013. “Report on OWI Recidivism: An analysis of assessment outcomes and repeat offenses.” Report to
the WI Department of Transportation.

Jacobson, GR, Niles, DH, Moberg, DP, Mandehr, E, Dusso, LN, 1979. “Identifying alcoholics and problem drinking
drivers: Wisconsin’s field test of a modified NCA criteria for the diagnosis of alcoholism.” Pp. 273-293 in M.
Galanter (Ed.) Currents in Alcoholism, Vol. VI. New York: Grune and Stratton.

Jewett, A, Shults, RA, Banerjee, T, Bergen, G. 2015. “Alcohol-impaired driving among adults—United States,
2012.” Morbidity and Mortality Weekly Report, 64 (30): 814-817.

Lipari, RN, Hughes, A, and Bose, J, 2016. “Driving under the influence of alcohol and illicit drugs.” The CBHSQ
Report, Dec. 27. Rockville, MD: Substance Abuse and Mental Health Services Administration.

Miller, PG, Curtis, A, Sonderlund, A, Day, A, Droste, N, 2015. “Effectiveness of interventions for convicted DUI
offenders in reducing recidivism: A systematic review of the peer-reviewed scientific literature.” The American
Journal of Drug and Alcohol Abuse 41 (1):16-29.

National Highway Traffic Safety Administration (NHTSA) 2015. ”Results of the 2013-2014 national roadside survey
of alcohol and drug use by drivers.” Traffic Safety Facts Research Note, (February, 2015). Washington DC: NHTSA
Office of Behavioral Safety Research # DOT HS 812 118..

Piper, DL and King, MJ. 1997. An Evaluation of Wisconsin’s Post-OWI Conviction Process; Assessment and
Educational Rehabilitation. Report to the Wisconsin Dept. Of Transportation. Madison, WI: Pacific Institute for
Research and Evaluation.

Rauch, WR, Zador, PL, Ahlin, EM, Howard, JM, Frissell, KC, Duncan, GD, 2010. “Risk of alcohol impaired driving
recidivism among first offenders and multiple offenders.” American Journal of Public Health 100 (5): 919-924.

Scott, Michael, 2013. “Deterring drunk driving: responses to the problem.” Presented at the UWPHI Evidence-
Based Health Policy Program Legislative Briefing, September 10, 2013.

Shults, RA, Elder, RW, Nichols, JL, Sleet, DA, Compton, R, Chattopadhyay, SK, Task Force on Community Services,
2009. “Effectiveness of multicomponent programs with community mobilization for reducing alcohol-impaired
driving.” American Journal of Preventive Medicine 37 (4): 360-371.

Substance Abuse and Mental Health Services Administration, 2010. The National Survey on Drug Use and Health
(NSDUH) Report: State Estimates of Drunk and Drugged Driving. Rockville, MD: SAMHSA.

43
UW Population Health Institute (UWPHI), 2017. What Works for Health Data Base,
http://whatworksforhealth.wisc.edu/search-results.php, accessed April 15, 2017.

Voas, RB and Lacey, JC, 2011. Alcohol and Highway Safety 2006: A Review of the State of Knowledge. National
Highway Traffic Safety Administration, Report #DOT HS 811 374.

Wisconsin Department of Health Services (WiDHS)and UW Population Health Institute (UWPHI), 2016. Wisconsin
Epidemiological Profile on Alcohol and Other Drugs, 2016. Madison, WI: DHS, Report # P-45718-16.

44
45

Appendix 1. County of residence for 2004-8 arrestees

No prior (N=121823) With prior (N=53623)
Total
(N= 175446)
N % N %
ADAMS 545 60% 364 40% 909
ASHLAND 399 66% 210 34% 609
BARRON 1250 71% 516 29% 1766
BAYFIELD 335 63% 194 37% 529
BROWN 5337 68% 2458 32% 7795
BUFFALO 300 70% 128 30% 428
BURNETT 459 61% 291 39% 750
CALUMET 770 71% 320 29% 1090
CHIPPEWA 1300 64% 736 36% 2036
CLARK 640 68% 297 32% 937
COLUMBIA 1398 65% 753 35% 2151
CRAWFORD 342 67% 167 33% 509
DANE 11188 72% 4288 28% 15476
DODGE 2086 67% 1023 33% 3109
DOOR 784 69% 357 31% 1141
DOUGLAS 1244 69% 561 31% 1805
DUNN 994 70% 430 30% 1424
EAU 2432 69% 1115 31% 3547
FLORENCE 97 72% 37 28% 134
FOND 2236 68% 1056 32% 3292
FOREST 227 59% 160 41% 387
GRANT 977 70% 420 30% 1397
GREEN 697 66% 360 34% 1057
GREEN 393 65% 213 35% 606
IOWA 506 67% 247 33% 753
IRON 166 67% 80 33% 246
JACKSON 547 65% 292 35% 839
JEFFERSON 1901 67% 917 33% 2818
JUNEAU 610 61% 389 39% 999
KENOSHA 3746 75% 1269 25% 5015
KEWAUNEE 360 69% 160 31% 520
LA 2709 70% 1159 30% 3868
LAFAYETTE 320 71% 130 29% 450
LANGLADE 411 65% 220 35% 631
LINCOLN 725 68% 339 32% 1064
MANITOWOC 1576 66% 807 34% 2383
45
46
MARATHON 3025 68% 1407 32% 4432
MARINETTE 972 65% 533 35% 1505
MARQUETTE 352 59% 245 41% 597
MONEMINEE 247 69% 111 31% 358
MILWAUKEE 16378 75% 5484 25% 21862
MONROE 1187 66% 601 34% 1788
OCONTO 631 67% 317 33% 948
ONEIDA 966 66% 491 34% 1457
OUTAGAMIE 3822 68% 1838 32% 5660
OZAUKEE 1459 75% 495 25% 1954
PEPIN 180 68% 84 32% 264
PIERCE 947 73% 346 27% 1293
POLK 919 64% 521 36% 1440
PORTAGE 1400 69% 642 31% 2042
PRICE 272 61% 176 39% 448
RACINE 3973 70% 1669 30% 5642
RICHLAND 405 67% 198 33% 603
ROCK 4226 69% 1897 31% 6123
RUSK 334 67% 161 33% 495
SAINT CROIX 1656 72% 659 28% 2315
SAUK 1855 69% 820 31% 2675
SAWYER 519 61% 338 39% 857
SHAWANO 957 68% 459 32% 1416
SHEBOYGAN 2595 68% 1245 32% 3840
TAYLOR 322 65% 177 35% 499
TREMPEALEAU 732 68% 341 32% 1073
VERNON 454 63% 270 37% 724
VILAS 566 60% 382 40% 948
WALWORTH 2791 70% 1191 30% 3982
WASHBURN 365 66% 190 34% 555
WASHINGTON 2521 69% 1110 31% 3631
WAUKESHA 7590 72% 2971 28% 10561
WAUPACA 1247 65% 664 35% 1911
WAUSHARA 492 63% 291 37% 783
WINNEBAGO 3764 66% 1927 34% 5691
WOOD 1695 65% 909 35% 2604
Out of state 9,447 89.3% 1130 10.7% 10577
Total (including OS) 131270 54753 186023

46
47
Appendix 2. IDP-DSP Compliance by County, for 2004-8 arrestees in 60 months
Arrests Entering IDP-DSP DSP compl fully DSP compl. None
County Index arr.
Total N % N % N %
(person)
ADAMS 1166 909 617 68% 497 81% 70 11%
ASHLAND 746 609 421 69% 336 80% 43 10%
BARRON 2268 1766 1218 69% 928 76% 115 9%
BAYFIELD 651 529 375 71% 299 80% 34 9%
BROWN 9840 7795 5552 71% 4255 77% 702 13%
BUFFALO 538 428 299 70% 245 82% 30 10%
BURNETT 954 750 421 56% 300 71% 61 14%
CALUMET 1329 1090 835 77% 683 82% 72 9%
CHIPPEWA 2623 2036 1489 73% 1168 78% 117 8%
CLARK 1186 937 590 63% 497 84% 52 9%
COLUMBIA 2740 2151 1572 73% 1212 77% 198 13%
CRAWFORD 637 509 332 65% 269 81% 41 12%
DANE 19259 15476 10919 71% 8539 78% 1324 12%
DODGE 3896 3109 2209 71% 1681 76% 245 11%
DOOR 1408 1141 890 78% 662 74% 114 13%
DOUGLAS 2192 1805 1041 58% 746 72% 166 16%
DUNN 1764 1424 961 67% 756 79% 117 12%
EAU CLAIRE 4530 3547 2563 72% 2046 80% 304 12%
FLORENCE 164 134 85 63% 67 79% 10 12%
FDL 4104 3292 2352 71% 1916 81% 225 10%
FOREST 497 387 238 61% 164 69% 32 13%
GRANT 1738 1397 874 63% 718 82% 74 8%
GREEN 1327 1057 690 65% 551 80% 84 12%
GREEN LAKE 764 606 439 72% 339 77% 54 12%
IOWA 948 753 540 72% 439 81% 49 9%
IRON 309 246 128 52% 97 76% 18 14%
JACKSON 1079 839 521 62% 355 68% 87 17%
JEFFERSON 3502 2818 1948 69% 1559 80% 206 11%
JUNEAU 1266 999 640 64% 475 74% 80 13%
KENOSHA 5946 5015 3148 63% 2391 76% 360 11%
KEWAUNEE 628 520 390 75% 283 73% 44 11%
LA CROSSE 4789 3868 2522 65% 1996 79% 269 11%
LAFAYETTE 556 450 311 69% 243 78% 36 12%
LANGLADE 784 631 465 74% 367 79% 51 11%
LINCOLN 1313 1064 769 72% 591 77% 97 13%
MANITOWOC 2927 2383 1738 73% 1440 83% 133 8%
MARATHON 5571 4432 3317 75% 2492 75% 459 14%
MARINETTE 1831 1505 989 66% 707 71% 136 14%
MARQUETTE 732 597 399 67% 312 78% 44 11%
MENOMINEE 459 358 171 48% 110 64% 41 24%
MILWAUKEE 26279 21862 15271 70% 12099 79% 1515 10%
MONROE 2321 1788 1102 62% 805 73% 153 14%
OCONTO 1154 948 704 74% 562 80% 71 10%
ONEIDA 1822 1457 1126 77% 851 76% 116 10%
OUTAGAMIE 7110 5660 4204 74% 3575 85% 332 8%
OZAUKEE 2339 1954 1563 80% 1265 81% 121 8%
PEPIN 318 264 189 72% 151 80% 17 9%
PIERCE 1570 1293 844 65% 653 77% 72 9%
POLK 1774 1440 804 56% 542 67% 88 11%
PORTAGE 2533 2042 1596 78% 1208 76% 171 11%
PRICE 563 448 304 68% 204 67% 41 13%
RACINE 6810 5642 4037 72% 2905 72% 448 11%
47
48
RICHLAND 772 603 384 64% 300 78% 48 13%
ROCK 7676 6123 3740 61% 2818 75% 483 13%
RUSK 612 495 327 66% 255 78% 31 9%
SAINT CROIX 2781 2315 1403 61% 1035 74% 173 12%
SAUK 3423 2675 1867 70% 1488 80% 184 10%
SAWYER 1160 857 457 53% 301 66% 83 18%
SHAWANO 1754 1416 1047 74% 817 78% 108 10%
SHEBOYGAN 4696 3840 2905 76% 2376 82% 218 8%
TAYLOR 621 499 345 69% 248 72% 41 12%
TREMPEALEAU 1356 1073 683 64% 509 75% 99 14%
VERNON 878 724 502 69% 413 82% 47 9%
VILAS 1222 948 589 62% 445 76% 76 13%
WALWORTH 4921 3982 2743 69% 2213 81% 293 11%
WASHBURN 681 555 337 61% 267 79% 30 9%
WASHINGTON 4540 3631 2868 79% 2228 78% 268 9%
WAUKESHA 13005 10561 8368 79% 6648 79% 771 9%
WAUPACA 2388 1911 1324 69% 995 75% 157 12%
WAUSHARA 976 783 581 74% 456 78% 55 9%
WINNEBAGO 7272 5691 4260 75% 3420 80% 372 9%
WOOD 3356 2604 1843 71% 1534 83% 185 10%

48
 49

Appendix 3a. Crude recidivism by number of prior for all 2004-14 arrestees (N=319,729 Persons)
Total Number of priors before 2004
Total
arrests 0 1 2 3 4 5+
2004-14 N % N % N % N % N % N % N %
1 188,213 78.72 33,719 72.76 13,825 71.65 6,010 68.89 2,631 71.05 1,826 70.64 246,224 77.01
2 38,371 16.05 9,245 19.95 4,014 20.80 2,055 23.56 845 22.82 601 23.25 55,131 17.24
3 9,377 3.92 2,514 5.42 1,137 5.89 537 6.16 177 4.78 143 5.53 13,885 4.34
4 2,394 1.00 676 1.46 276 1.43 101 1.16 45 1.22 9 0.35 3,501 1.09
5 570 0.24 137 0.30 40 0.21 20 0.23 4 0.11 6 0.23 777 0.24
6 124 0.05 39 0.08 4 0.02 1 0.01 0 0.00 0 0.00 168 0.05
7 24 0.01 12 0.03 0 0.00 0 0.00 0 0.00 0 0.00 36 0.01
8 4 0.00 2 0.00 0 0.00 0 0.00 0 0.00 0 0.00 6 0.00
9 0 0.00 0 0.00 0 0.00 0 0.00 1 0.03 0 0.00 1 0.00
The sample consists of OWI arrest records for all 2004-14 Wisconsin drivers arrested during this period.

Appendix 3b. Annual Arrests by Priors all 2004-2014 OWI Arrestees

Number of Priors
Year No Prior
1 2 3 4 5 6 7 8 9 10+ Total
2004 26,868 10,386 4,708 2,062 947 337 136 45 12 5 4 45,510
59.04 22.82 10.34 4.53 2.08 0.74 0.30 0.10 0.03 0.01 0.01
2005 25,904 10,073 4,617 2,234 947 359 125 56 15 5 5 44,340
58.42 22.72 10.41 5.04 2.14 0.81 0.28 0.13 0.03 0.01 0.01
2006 25,786 10,240 4,647 2,098 963 367 132 42 27 4 7 44,313
58.19 23.11 10.49 4.73 2.17 0.83 0.30 0.09 0.06 0.01 0.02
2007 25,089 10,176 4,686 2,084 967 372 124 49 26 15 3 43,591
57.56 23.34 10.75 4.78 2.22 0.85 0.28 0.11 0.06 0.03 0.01
2008 19,949 6,789 2,882 1,297 489 230 83 24 14 4 1 31,762
62.81 21.37 9.07 4.08 1.54 0.72 0.26 0.08 0.04 0.01 0.00
2009 24,674 9,778 4,650 2,146 1,004 406 155 55 15 12 1 42,896
57.52 22.79 10.84 5.00 2.34 0.95 0.36 0.13 0.03 0.03 0.00
2010 21,645 8,750 4,152 1,886 835 371 145 52 24 9 9 37,878
57.14 23.10 10.96 4.98 2.20 0.98 0.38 0.14 0.06 0.02 0.02
2011 19,496 7,567 3,672 1,658 768 320 161 32 21 10 6 33,711
57.83 22.45 10.89 4.92 2.28 0.95 0.48 0.09 0.06 0.03 0.02
2012 19,158 7,414 3,588 1,719 774 324 150 73 16 11 9 33,236
57.64 22.31 10.80 5.17 2.33 0.97 0.45 0.22 0.05 0.03 0.03
2013 17,836 6,794 3,317 1,498 663 303 141 53 25 6 5 30,641
58.21 22.17 10.83 4.89 2.16 0.99 0.46 0.17 0.08 0.02 0.02
2014 17,134 6,416 3,198 1,475 707 320 123 59 21 10 6 29,469
58.14 21.77 10.85 5.01 2.40 1.09 0.42 0.20 0.07 0.03 0.02
Total 243,539 94,383 44,117 20,157 9,064 3,709 1,475 540 216 91 56 417,347
58.35 22.61 10.57 4.83 2.17 0.89 0.35 0.13 0.05 0.02 0.01

49
 50

Appendix 4. IDP-DSP Compliance and 60-Months Recidivism by Driver's Prior Status, Gender, Race/Ethnicity, and Age for 2004-2008 Arrestees (N=175,446)
Not convicted (N=21,843) Convicted (N=153, 603)
No Valid DSP* Valid DSP Total No Valid DSP* Valid DSP Total
Total Compl Re-Arrested Total Compl Re-Arrested Total Compl Re-Arrested Total Compl Re-Arrested Total Compl Re-Arrested Total Compl Re-Arrested
N %! N %& N %! N %& N %! N %& N %! N %& N %! N %& N %! N %&
Total 14,741 3,989 27% 7,102 1,162 16% 21,843 5,151 24% 30,308 10,289 34% 123,295 26,719 22% 153,603 37,008 24%
Not Entering DSP 10,338 1,093 11% 10,339 1,093 11% 23,314 4,935 21% 23,314 4,987 21%
In DSP 4,403 2,896 66% 7,102 1,162 16% 11,504 4,058 35% 6,994 5,354 77% 123,295 26,719 22% 130,289 32,021 25%
Compl: None 850 19% 478 56% 1,551 22% 322 21% 2,401 21% 800 33% 1,522 22% 1,150 76% 13,261 11% 5,958 45% 14,783 11% 7,108 48%
Compl: Some 491 11% 372 76% 586 8% 107 18% 1,077 9% 479 44% 977 14% 764 78% 13,713 11% 2,686 20% 14,690 11% 3,450 23%
Compl: All 3,062 70% 2,046 67% 4,964 70% 733 15% 8,026 70% 2,779 35% 4,495 64% 3,390 75% 96,317 78% 18,073 19% 100,812 77% 21,463 21%
No Prior
Total 10,034 2,466 25% 4,756 675 14% 14,790 3,141 21% 17,953 5,606 31% 89,080 19,281 22% 107,033 24,887 23%
Not Entering DSP 7,219 586 8% 7,220 586 8% 14,709 2,630 18% 14,709 2,682 18%
In DSP 2,815 1,880 67% 4,756 675 14% 7,570 2,555 34% 3,244 2,976 92% 89,080 19,281 22% 92,324 22,205 24%
Compl: None 534 19% 304 57% 1,076 23% 185 17% 1,610 21% 489 30% 642 20% 588 92% 9,210 10% 4,075 44% 9,852 11% 4,663 47%
Compl: Some 286 10% 231 81% 333 7% 57 17% 619 8% 288 47% 444 14% 415 93% 9,115 10% 1,875 21% 9,559 10% 2,290 24%
Compl: All 1,995 71% 1,345 67% 3,346 70% 433 13% 5,341 71% 1,778 33% 2,158 67% 1,923 89% 70,751 79% 13,329 19% 72,909 79% 15,252 21%
With Prior
Total 4,707 1,523 32% 2,346 487 21% 7,053 2,010 28% 12,355 4,683 38% 34,215 7,438 22% 46,570 12,121 26%
No DSP records 3,119 507 16% 3,119 507 16% 8,605 2,305 27% 8,605 2,305 27%
In DSP 1,588 1,016 64% 2,346 487 21% 3,934 1,503 38% 3,750 2,378 63% 34,215 7,438 22% 37,965 9,816 26%
Compl: None 316 20% 174 55% 475 20% 137 29% 791 20% 311 39% 880 23% 562 64% 4,051 12% 1,883 46% 4,931 13% 2,445 50%
Compl: Some 205 13% 141 69% 253 11% 50 20% 458 12% 191 42% 533 14% 349 65% 4,598 13% 811 18% 5,131 14% 1,160 23%
Compl: All 1,067 67% 701 66% 1,618 69% 300 19% 2,685 68% 1,001 37% 2,337 62% 1,467 63% 25,566 75% 4,744 19% 27,903 73% 6,211 22%
Male
Total 11,940 3,316 28% 5,566 942 17% 17,506 4,258 24% 24,889 8,589 35% 92,539 21,050 23% 117,428 29,639 25%
No DSP records 8,340 940 11% 8,341 940 11% 19,215 4,335 23% 19,220 4,337 23%
In DSP 3,600 2,376 66% 5,566 942 17% 9,165 3,318 36% 5,674 4,254 75% 92,539 21,050 23% 98,208 25,302 26%
Compl: None 693 19% 391 56% 1,228 22% 256 21% 1,921 21% 647 34% 1,264 22% 948 75% 10,183 11% 4,623 45% 11,447 12% 5,571 49%
Compl: Some 407 11% 303 74% 470 8% 87 19% 877 10% 390 44% 786 14% 603 77% 10,344 11% 2,102 20% 11,130 11% 2,705 24%
Compl: All 2,500 69% 1,682 67% 3,867 69% 599 15% 6,367 69% 2,281 36% 3,623 64% 2,703 75% 72,008 78% 14,323 20% 75,631 77% 17,026 23%
Female
Total 2,801 673 24% 1,536 220 14% 4,337 893 21% 5,419 1,700 31% 30,756 5,669 18% 36,175 7,369 20%
No DSP records 1,998 153 8% 1,998 153 8% 4,098 650 16% 4,098 650 16%
In DSP 803 520 65% 1,536 220 14% 2,339 740 32% 1,321 1,050 79% 30,756 5,669 18% 32,077 6,719 21%
Compl: None 157 20% 87 55% 323 21% 66 20% 480 21% 153 32% 258 20% 202 78% 3,078 10% 1,335 43% 3,336 10% 1,537 46%
Compl: Some 84 10% 69 82% 116 8% 20 17% 200 9% 89 45% 191 14% 161 84% 3,369 11% 584 17% 3,560 11% 745 21%
Compl: All 562 70% 364 65% 1,097 71% 134 12% 1,659 71% 498 30% 872 66% 687 79% 24,309 79% 3,750 15% 25,181 79% 4,437 18%
White
Total 10,230 3,214 31% 5,877 945 16% 17,107 4,159 24% 20,064 7,311 36% 107,151 22,852 21% 127,215 30,163 24%
50
 51
No DSP records 7,425 700 9% 7,426 700 9% 14,073 2,771 20% 14,077 2,773 20%
In DSP 2,805 2,514 90% 5,877 945 16% 9,681 3,459 36% 5,991 4,540 76% 107,151 22,852 21% 113,138 27,390 24%
Compl: None 697 25% 383 55% 1,203 20% 254 21% 1,900 20% 637 34% 1,216 20% 915 75% 10,415 10% 4,882 47% 11,631 10% 5,797 50%
Compl: Some 429 15% 325 76% 489 8% 92 19% 918 9% 417 45% 846 14% 657 78% 11,870 11% 2,358 20% 12,716 11% 3,015 24%
Compl: All 2,679 96% 1,806 67% 4,184 71% 599 14% 6,863 71% 2,405 35% 3,929 66% 2,968 76% 84,862 79% 15,610 18% 88,791 78% 18,578 21%
Am. Indian
Total 419 171 41% 122 45 37% 541 216 40% 1,410 557 40% 2,406 754 31% 3,816 1,311 34%
No DSP records 286 78 27% 286 78 27% 1,129 376 33% 1,129 376 33%
In DSP 133 93 70% 122 45 37% 255 138 54% 281 181 64% 2,406 754 31% 2,687 935 35%
Compl: None 37 28% 27 73% 40 33% 16 40% 77 30% 43 56% 85 30% 57 67% 528 22% 265 50% 613 23% 322 53%
Compl: Some 14 11% 10 71% 11 9% 4 36% 25 10% 14 56% 38 14% 32 84% 329 14% 75 23% 367 14% 107 29%
Compl: All 82 62% 56 68% 71 58% 25 35% 153 60% 81 53% 158 56% 92 58% 1,549 64% 414 27% 1,707 64% 506 30%
Black
Total 792 186 23% 355 65 18% 1,147 251 22% 2,607 761 29% 4,501 932 21% 7,108 1,693 24%
No DSP records 632 95 15% 632 95 15% 2,345 557 24% 2,345 557 24%
In DSP 160 91 57% 355 65 18% 515 156 30% 262 204 78% 4,501 932 21% 4,763 1,136 24%
Compl: None 31 19% 12 39% 100 28% 23 23% 131 25% 35 27% 83 32% 66 80% 763 17% 274 36% 846 18% 340 40%
Compl: Some 18 11% 13 72% 37 10% 4 11% 55 11% 17 31% 46 18% 40 87% 654 15% 113 17% 700 15% 153 22%
Compl: All 111 69% 66 59% 218 61% 38 17% 329 64% 104 32% 133 51% 98 74% 3,084 69% 545 18% 3,217 68% 643 20%
Hispanic
Total 1,350 274 20% 560 79 14% 1,910 353 18% 3,495 1,080 31% 6,818 1,706 25% 10,313 2,786 27%
No DSP records 1,151 141 12% 1,151 141 12% 3,145 792 25% 3,145 793 25%
In DSP 199 133 67% 560 79 14% 759 212 28% 350 288 82% 6,818 1,706 25% 7,168 1,993 28%
Compl: None 52 26% 34 65% 154 28% 24 16% 206 27% 58 28% 107 31% 84 79% 1,034 15% 407 39% 1,141 16% 491 43%
Compl: Some 18 9% 14 78% 33 6% 3 9% 51 7% 17 33% 36 10% 27 75% 633 9% 105 17% 669 9% 132 20%
Compl: All 129 65% 85 66% 373 67% 52 14% 502 66% 137 27% 207 59% 176 85% 5,151 76% 1,194 23% 5,358 75% 1,370 26%
Asian American
Total 131 32 24% 91 16 18% 222 48 22% 230 81 35% 1,253 266 21% 1,483 347 23%
No DSP records 91 8 9% 91 8 9% 171 36 21% 171 36 21%
In DSP 40 24 60% 91 16 18% 131 40 31% 59 45 76% 1,253 266 21% 1,312 311 24%
Compl: None 5 13% 4 80% 17 19% 2 12% 22 17% 6 27% 9 15% 7 78% 103 8% 45 44% 112 9% 52 46%
Compl: Some 6 15% 5 83% 6 7% 2 33% 12 9% 7 58% 5 8% 2 40% 131 10% 23 18% 136 10% 25 18%
Compl: All 29 73% 15 52% 68 75% 12 18% 97 74% 27 28% 45 76% 36 80% 1,019 81% 198 19% 1,064 81% 234 22%
Others
Total 819 112 14% 97 12 12% 916 124 14% 2,502 599 24% 1,166 209 18% 3,668 708 19%
No DSP records 753 71 9% 753 71 9% 2,451 452 18% 2,451 452 18%
In DSP 66 41 62% 97 12 12% 163 53 33% 51 147 288% 1,166 209 18% 1,217 256 21%
Compl: None 28 42% 18 64% 37 38% 3 8% 65 40% 21 32% 22 43% 21 95% 418 36% 85 20% 440 36% 106 24%
Compl: Some 6 9% 5 83% 10 10% 2 20% 16 10% 7 44% 6 12% 6 100% 96 8% 12 13% 102 8% 18 18%
Compl: All 32 48% 18 56% 50 52% 7 14% 82 50% 25 30% 23 45% 20 87% 652 56% 112 17% 675 55% 132 20%
Age < 20
Total 2,879 917 32% 647 146 23% 3,526 1,063 30% 3,269 1,260 39% 13,491 4,236 31% 16,760 5,496 33%
No DSP records 1,813 139 8% 1,814 139 8% 2,406 519 22% 2,406 519 22%
51
 52
In DSP 1,066 778 73% 647 146 23% 1,712 924 54% 863 741 86% 13,491 4,236 31% 14,354 4,977 35%
Compl: None 193 18% 127 66% 160 25% 50 31% 353 21% 177 50% 165 19% 147 89% 1,873 14% 940 50% 2,038 14% 1,087 53%
Compl: Some 127 12% 105 83% 53 8% 11 21% 180 11% 116 64% 120 14% 101 84% 1,865 14% 502 27% 1,985 14% 603 30%
Compl: All 746 70% 546 73% 433 67% 85 20% 1,179 69% 631 54% 578 67% 493 85% 9,753 72% 2,794 29% 10,331 72% 3,287 32%
Age 20-29
Total 4,806 1,298 27% 2,876 444 15% 7,582 1,742 23% 11,668 4,189 36% 49,813 11,805 24% 61,081 15,994 26%
No DSP records 3,204 351 11% 3,204 351 11% 8,453 1,978 23% 8,453 1,978 23%
In DSP 1,602 947 59% 2,876 444 15% 4,378 1,391 32% 3,215 2,211 69% 49,813 11,805 24% 52,628 14,016 27%
Compl: None 334 21% 180 54% 705 25% 116 16% 1,039 24% 296 28% 686 21% 538 78% 5,849 12% 2,693 46% 6,535 12% 3,231 49%
Compl: Some 160 10% 128 80% 230 8% 48 21% 390 9% 176 45% 433 13% 348 80% 5,709 11% 1,190 21% 6,142 12% 1,538 25%
Compl: All 1,008 63% 639 63% 1,941 67% 280 14% 2,949 67% 919 31% 1,696 53% 1,325 78% 38,255 77% 7,922 21% 39,951 76% 9,247 23%
Age 30-39
Total 2,961 781 26% 1,546 265 17% 4,507 1,046 23% 7,013 2,322 33% 25,352 5,008 20% 32,365 7,330 23%
No DSP records 2,125 258 12% 2,125 258 12% 5,422 1,230 23% 5,424 1,230 23%
In DSP 836 523 63% 1,546 265 17% 2,382 788 33% 1,591 1,092 69% 25,352 5,008 20% 26,941 6,100 23%
Compl: None 151 18% 73 48% 320 21% 73 23% 471 20% 146 31% 343 22% 224 65% 2,730 11% 1,188 44% 3,073 11% 1,412 46%
Compl: Some 88 11% 61 69% 135 9% 24 18% 223 9% 85 38% 204 13% 149 73% 2,756 11% 445 16% 2,960 11% 594 20%
Compl: All 597 71% 389 65% 1,091 71% 168 15% 1,688 71% 557 33% 1,044 66% 719 69% 19,864 78% 3,375 17% 20,908 78% 4,094 20%
Age 40-49
Total 2,674 701 26% 1,370 229 17% 4,044 930 23% 6,011 1,861 31% 22,594 4,079 18% 28,605 5,940 21%
No DSP records 1,947 227 12% 1,947 227 12% 4,741 952 20% 4,743 954 20%
In DSP 727 474 65% 1,370 229 17% 2,097 703 34% 1,270 909 72% 22,594 4,079 18% 23,862 4,986 21%
Compl: None 130 18% 82 63% 263 19% 67 25% 393 19% 149 38% 248 20% 179 72% 2,004 9% 850 42% 2,252 9% 1,029 46%
Compl: Some 88 12% 61 69% 116 8% 16 14% 204 10% 77 38% 170 13% 126 74% 2,296 10% 399 17% 2,466 10% 525 21%
Compl: All 509 70% 331 65% 991 72% 146 15% 1,500 72% 477 32% 852 67% 604 71% 18,292 81% 2,828 15% 19,144 80% 3,432 18%
Age 50-59
Total 1,139 242 21% 544 68 13% 1,683 310 18% 2,098 536 26% 9,219 1,285 14% 11,317 1,821 16%
No DSP records 912 92 10% 912 92 10% 1,723 251 15% 1,723 251 15%
In DSP 227 150 66% 544 68 13% 771 218 28% 375 285 76% 9,219 1,285 14% 9,594 1,570 16%
Compl: None 34 15% 14 41% 86 16% 14 16% 120 16% 28 23% 69 18% 52 75% 627 7% 236 38% 696 7% 288 41%
Compl: Some 23 10% 13 57% 43 8% 7 16% 66 9% 20 30% 41 11% 32 78% 840 9% 126 15% 881 9% 158 18%
Compl: All 170 75% 123 72% 415 76% 47 11% 585 76% 170 29% 265 71% 201 76% 7,752 84% 923 12% 8,017 84% 1,124 14%
Age 60-69
Total 382 50 13% 119 10 8% 501 60 12% 649 121 19% 2,826 306 11% 3,475 427 12%
No DSP records 337 26 8% 337 26 8% 569 55 10% 569 55 10%
In DSP 45 24 53% 119 10 8% 164 34 21% 80 66 83% 2,826 306 11% 2,906 372 13%
Compl: None 8 18% 2 25% 17 14% 2 12% 25 15% 4 16% 11 14% 10 91% 178 6% 51 29% 189 7% 61 32%
Compl: Some 5 11% 4 80% 9 8% 1 11% 14 9% 5 36% 9 11% 8 89% 247 9% 24 10% 256 9% 32 13%
Compl: All 32 71% 18 56% 93 78% 7 8% 125 76% 25 20% 60 75% 48 80% 2,401 85% 231 10% 2,461 85% 279 11%
* "No valid DSP" includes three groups: 1) no DSP records at all; 2) no DSP between the index OWI violation and the next violation; and 3) no DSP records in 24 months following the
index OWI.
! Number of the compliance category/Total number in DSP
& Number of rearrested / total number of arrested

52
 53

Appendix 5. Compliance and recidivism rates by recommendations, 2004-08 WI arrestees

IU (N=61883) IUB (23963) SD (13674) D/DR (30877)
Recommended
Recom Rearrested Recom Rearrested Recom Rearrested Recom Rearrested
Treatment Complied Complied Complied Complied
N. (complied) N. (complied) N. (complied) N. (complied)

Group Dynamics 51,285 48,457 94% 7,252 15% 8,019 7,495 93% 1,420 19% 307 263 86% 57 22% 435 365 84% 44 12%
Multiple Offender 4,700 4,287 91% 758 18% 3,098 2,817 91% 611 22% 175 148 85% 34 23% 203 167 82% 38 23%
Out patient 2,223 1,996 90% 335 17% 11,313 10,016 89% 2,129 21% 13,137 11,207 85% 2,676 24% 29,116 23,769 82% 5,629 24%
In patient 9 8 89% 2 25% 24 23 96% 5 22% 35 31 89% 10 32% 857 748 87% 188 25%
Alcohol educ. (AE) 1,754 1,459 83% 233 16% 639 524 82% 76 15% 167 130 78% 14 11% 54 42 78% 7 17%
One day AE 169 164 97% 13 8% 9 9 100% 1 11% 1 1 100% 0 0% 4 4 100% 0 0%
Level I AE 168 141 84% 21 15% 46 40 87% 5 13% 1 1 100% 0 0% 2 2 100% 1 50%
Level II AE 167 141 84% 18 13% 132 107 81% 14 13% 7 4 57% 1 25% 10 6 60% 3 50%
Day care treatment 2 2 100% 2 100% 5 4 80% 1 25% 54 45 83% 14 31% 305 240 79% 62 26%
Night care treatment 0 0 1 1 100% 0 0% 2 2 100% 1 50%
Residential care 3 3 0 13 12 92% 6 50% 24 21 88% 5 24% 461 393 85% 85 22%
After care 28 20 71% 6 30% 303 270 89% 72 27% 1,326 1,123 85% 265 24% 5,436 4,399 81% 1,020 23%
Halfway house 0 1 1 100% 0 0% 4 4 100% 3 75% 60 53 88% 9 17%
Transitional living 0 3 2 67% 1 50% 1 1 100% 0 0% 132 111 84% 22 20%
DWI clinic 57 55 96% 4 7% 20 20 100% 3 15% 2 2 100% 1 50% 5 5 100% 2 40%
Group Therapy 8 5 63% 1 20% 43 38 88% 8 21% 48 41 85% 8 20% 188 145 77% 36 25%
Out of state education 1,830 1,414 77% 122 9% 1,325 860 65% 95 11% 425 336 79% 43 13% 119 91 76% 11 12%
Out of state plan 293 207 71% 12 6% 692 461 67% 51 11% 257 173 67% 19 11% 386 255 66% 42 16%
Victim panel 6,384 6,086 95% 947 16% 3,400 3,137 92% 627 20% 2,720 2,355 87% 579 25% 7,300 6,091 83% 1,488 24%
Detoxification 0 0 3 3 100% 0 0% 31 22 71% 4 18%
Medical exam 8 8 100% 1 13% 12 11 92% 0 0% 13 11 85% 1 9% 64 56 88% 11 20%
Psychiatric 23 21 91% 4 19% 17 16 94% 2 13% 14 11 79% 3 27% 147 133 90% 40 30%
AA 18 8 44% 1 13% 100 72 72% 4 6% 84 64 76% 14 22% 903 704 78% 139 20%
Abstinence 2,896 2,703 93% 379 14% 3,658 3,256 89% 675 21% 6,256 5,402 86% 1,320 24% 17,990 14,753 82% 3,438 23%
Follow up agency 26 24 92% 5 21% 97 85 88% 16 19% 191 156 82% 43 28% 653 529 81% 145 27%
Agency follow-up 4 2 50% 2 100% 16 13 81% 0 0% 302 255 84% 56 22% 1,028 842 82% 216 26%
Antabuse 0 0 2 2 100% 1 50% 22 20 91% 3 15%
Biomarker 0 0 3 0 0% 3 0 0%
Exit Interview 39 36 92% 7 19% 65 60 92% 15 25% 71 57 80% 13 23% 471 407 86% 100 25%
Narcotics anonymous 0 1 1 100% 0 0% 0 37 24 65% 4 17%
Reassessment 144 86 60% 24 28% 73 52 71% 16 31% 20 17 85% 6 35% 15 13 87% 3 23%
Random Urine screen 74 68 92% 12 18% 827 703 85% 151 21% 702 601 86% 153 25% 2,596 2,077 80% 407 20%
Noncompliant 4,274 1,466 34% 2,774 1,051 38% 2,066 1,077 52% 5,698 2,686 47%

53
54

Appendix 6. Recidivism and compliance for IUB arrestees in IDP

All IUB Arrestees in IDP Complied IUB
Recommended primary (N=23963) (N=18627)
programs
Complied Rearrested Rearrested
Both primary treatment and 293 137 35
education (N=921) 31.8% 14.9% 12.0%
Primary education only 7,337 1,763 1,347
(N=9106) 80.6% 19.4% 18.4%
Primary treatment only 10,721 3,340 2,346
(N=13447) 79.7% 24.8% 21.9%
No primary programs 276 57 29
(N=489) 56.4% 11.7% 10.5%
* Treatment programs include outpatient treatment, aftercare, in
patient treatment, residential care, daycare treatment, group therapy,
transitional living, DWI clinic, halfway house, and night care treatment;
Educational programs include group dynamics, alcohol education (AE),
out of state AE, one day AE, level I & II AE.

54
55

Appendix 7. OWI conviction and 60 month recidivism rates with predicted

probabilities for 72 Wisconsin counties, 2004-08 arrestees
Convicted Rearrested
N Arrested Not Not
N Predicted* N Predicted*
Residence county adjusted adjusted
ADAMS 909 792 87.1% 87.3% 248 27.3% 27.5%
ASHLAND 609 529 86.9% 87.0% 135 22.2% 22.7%
BARRON 1,766 1,532 86.7% 86.9% 488 27.6% 27.8%
BAYFIELD 529 444 83.9% 84.0% 125 23.6% 23.6%
BROWN 7,795 7,151 91.7% 91.9% 2,021 25.9% 26.0%
BUFFALO 428 346 80.8% 81.7% 107 25.0% 26.1%
BURNETT 750 569 75.9% 76.2% 204 27.2% 27.7%
CALUMET 1,090 996 91.4% 91.4% 245 22.5% 22.6%
CHIPPEWA 2,036 1,802 88.5% 88.5% 567 27.8% 28.0%
CLARK 937 812 86.7% 86.4% 258 27.5% 27.7%
COLUMBIA 2,151 1,930 89.7% 90.0% 576 26.8% 26.9%
CRAWFORD 509 421 82.7% 82.8% 130 25.5% 25.9%
DANE 15,476 13,791 89.1% 89.3% 3,797 24.5% 24.7%
DODGE 3,109 2,743 88.2% 88.3% 794 25.5% 25.8%
DOOR 1,141 1,061 93.0% 93.4% 273 23.9% 24.4%
DOUGLAS 1,805 1,397 77.4% 77.1% 380 21.1% 21.4%
DUNN 1,424 1,196 84.0% 83.8% 342 24.0% 24.8%
EAU CLAIRE 3,547 3,153 88.9% 88.9% 954 26.9% 27.1%
FLORENCE 134 98 73.1% 74.0% 24 17.9% 17.6%
FOND DU LAC 3,292 2,962 90.0% 90.1% 809 24.6% 24.8%
FOREST 387 305 78.8% 78.7% 115 29.7% 29.7%
GRANT 1,397 1,105 79.1% 79.4% 348 24.9% 25.0%
GREEN 1,057 866 81.9% 82.0% 263 24.9% 25.1%
GREEN LAKE 606 543 89.6% 89.6% 150 24.8% 24.7%
IOWA 753 669 88.8% 89.0% 198 26.3% 26.2%
IRON 246 157 63.8% 63.5% 66 26.8% 27.0%
JACKSON 839 715 85.2% 85.1% 233 27.8% 28.1%
JEFFERSON 2,818 2,530 89.8% 90.2% 690 24.5% 24.6%
JUNEAU 999 865 86.6% 86.8% 272 27.2% 27.4%
KENOSHA 5,015 3,911 78.0% 78.5% 948 18.9% 19.1%
KEWAUNEE 520 471 90.6% 90.5% 105 20.2% 19.8%
LA CROSSE 3,868 3,274 84.6% 84.7% 933 24.1% 24.2%
LAFAYETTE 450 368 81.8% 82.3% 104 23.1% 23.7%
LANGLADE 631 568 90.0% 90.2% 162 25.7% 25.5%
LINCOLN 1,064 936 88.0% 88.1% 258 24.2% 24.6%
MANITOWOC 2,383 2,163 90.8% 91.1% 552 23.2% 23.0%
MARATHON 4,432 4,003 90.3% 90.7% 1,130 25.5% 25.6%
MARINETTE 1,505 1,288 85.6% 85.5% 321 21.3% 21.7%
MARQUETTE 597 508 85.1% 85.1% 143 24.0% 24.4%
MENOMINEE 358 300 83.8% 83.5% 101 28.2% 28.9%
MILWAUKEE 21,862 19,174 87.7% 87.9% 4,403 20.1% 20.3%
MONROE 1,788 1,494 83.6% 84.1% 524 29.3% 29.4%
OCONTO 948 854 90.1% 90.1% 224 23.6% 23.4%
ONEIDA 1,457 1,305 89.6% 89.8% 360 24.7% 24.8%
OUTAGAMIE 5,660 5,199 91.9% 91.9% 1,447 25.6% 25.7%
OZAUKEE 1,954 1,749 89.5% 89.4% 402 20.6% 20.5%
PEPIN 264 229 86.7% 87.3% 59 22.3% 23.1%
PIERCE 1,293 1,014 78.4% 79.1% 270 20.9% 21.4%
POLK 1,440 1,030 71.5% 71.6% 359 24.9% 25.4%
PORTAGE 2,042 1,876 91.9% 92.1% 498 24.4% 24.6%

55
56
PRICE 448 391 87.3% 87.1% 110 24.6% 24.7%
RACINE 5,642 4,897 86.8% 87.0% 1,174 20.8% 21.0%
RICHLAND 603 504 83.6% 84.6% 159 26.4% 26.8%
ROCK 6,123 5,243 85.6% 85.9% 1,537 25.1% 25.4%
RUSK 495 420 84.8% 84.9% 120 24.2% 24.7%
SAINT CROIX 2,315 1,699 73.4% 73.8% 461 19.9% 20.4%
SAUK 2,675 2,383 89.1% 89.1% 742 27.7% 28.0%
SAWYER 857 673 78.5% 79.2% 305 35.6% 35.4%
SHAWANO 1,416 1,293 91.3% 91.4% 325 23.0% 23.1%
SHEBOYGAN 3,840 3,488 90.8% 91.0% 851 22.2% 22.3%
TAYLOR 499 431 86.4% 86.7% 126 25.3% 25.9%
TREMPEALEAU 1,073 920 85.7% 84.8% 284 26.5% 26.1%
VERNON 724 628 86.7% 87.0% 160 22.1% 21.7%
VILAS 948 786 82.9% 82.9% 271 28.6% 28.7%
WALWORTH 3,982 3,367 84.6% 85.1% 957 24.0% 24.6%
WASHBURN 555 448 80.7% 80.9% 132 23.8% 23.6%
WASHINGTON 3,631 3,311 91.2% 91.2% 923 25.4% 25.5%
WAUKESHA 10,561 9,540 90.3% 90.5% 2,435 23.1% 23.1%
WAUPACA 1,911 1,661 86.9% 87.1% 496 26.0% 25.9%
WAUSHARA 783 709 90.5% 90.7% 194 24.8% 25.2%
WINNEBAGO 5,691 5,258 92.4% 92.5% 1,536 27.0% 27.0%
WOOD 2,604 2,359 90.6% 90.8% 746 28.6% 28.8%
* The probabilities of conviction and recidivism are predicted with an equation including arrestee's
personal characteristics including gender, age, race/ethnicity and prior OWI.

56