Recidivism Study of New Mexicos Convicted DWI Offenders 2000 2003

1
Recidivism Study of New Mexicos Convicted DWI
Offenders 2000 2003
Judith S. Harmon, M.A., Brian WoodsSubstance
Abuse Epidemiology UnitInjury and Behavioral
Epidemiology BureauEpidemiology and Response
Division
International Traffic Records Forum Buffalo,
NY August 3, 2005
2
Purpose

• Demonstrate how convicted DWI offenders are
  screened and tracked in New Mexico
• Describe how data analysis is achieved through
  data linkage of databases for convicted DWI
  offenders in New Mexico
• Describe recidivism of convicted DWI offenders

3
Background

• In 1993 the New Mexico Legislature enacted DWI
  laws to enhance serious problem of DWI
• Increased the rate of liquor excise tax
• Required all convicted DWI offenders be screened
  for alcohol and/or substance abuse
• Created the Local DWI Grant Program (LDWI) funded
  through the Local Government Division of the
  Department of Finance and Administration
• All 33 counties in New Mexico have LDWI Programs

4
Total Alcohol-Involved Crash Deaths and
Legislative and Administrative Actions in New
Mexico, 1980 2003
5
New Mexico has contracted with ADE Incorporated,
Clarkston, Michigan to

• provide a computerized alcohol and substance
  abuse assessment (NEEDS) and client tracking
  program (CTP) to standardize screening and
  tracking of DWI offenders for the state of New
  Mexico

6
Screening Process
7
Methods

• There are four processes for this project
• Collection and cleaning of DWI data from 33
  county programs
• Data preparation
• Data linkage between DWI county program data and
  Motor Vehicle Division DWI citations (DWICTS)
• Data analysis for demographics, sanction
  completion and survival analysis to show
  recidivism

8
The Datasets

• The NEEDS and CTP data
• collected at the county level between 2000 - 2003
  on DWI offenders
• Data collected quarterly from each county to
  address data quality to insure end of year data
  collection is accurate
• All data sent to the State DOH where it is
  cleaned and combined into one statewide dataset
• Contains Demographic information NEEDs, a
  standardized alcohol and substance abuse
  assessment instrument to establish treatment
  need Tracking information on the sanctions
  assigned (CTP) Court information
• Type of sanction with start, completion, and
  termination dates Agency where sanction was
  carried out

9

• DWICTS
• Information on all DWI Citations since 1984 is
  tracked by the Motor Vehicle Division in the DWI
  Citation Tracking System.
• Contains personal identifiers, DWI arrests,
  administrative actions and court information.
• Pulled biannually for traffic safety analysis
  purposes by Division of Government Research at
  UNM for the Traffic Safety Bureau, New Mexico
  Department of Transportation

10
Number of DWI Convictions and Screenings in New
Mexico, 2000 - 2003 
11
Data Preparation

• The NEEDS and CTP data
• Originally MS-Access relational database.
• Each table imported into SAS.
• The tables are merged and one client level
  dataset is created.
• Cleaned up name, residence city and SSN.
• Standardized matching variables with variables in
  DWICTS.

12
Data Preparation

• DWICTS Data
• Fixed common errors in name and city fields.
• Unduplicated the DWICTS so that there was one
  record for each person.
• Standardized the matching variables with the
  variables in the NEEDS and CTP data

13
Data Linkage Why Link?To Get More Complete
Information

• To get DWI arrest and conviction histories on
  individuals in the NEEDS and CTP data
• To get an estimate of the proportion of
  individuals convicted of DWI who are being
  screened and tracked in each county
• To get information on recidivism
• As time goes by more complete information on
  recidivism will be available

14
Probabilistic Data Linkage

• Ideal for linking data with imperfect personal
  identifiers
• Creates probabilities from the frequency
  distribution of matching variables
• Produces weighted matched criteria based on the
  sum of the matching variables probabilities

15
Linkage Process

• All linkage carried out using the Automatch
• Three matching passes carried out
• Blocking variables
• Pass 1 SSN
• Pass 2 Date of Birth
• Pass 3 First and Last Name
• Matching variables
• SSN, date of birth, first name, last name, middle
  initial, residence city, reporting county and
  residence county, residence zipcode, residence
  state, sex, evaluation date, start dates for DWI
  school, probation, community service and court
  dates

16
Variable Comparisons
17
Match Rates

• 96 of the individuals in the NEEDS and CTP data
  linked to the DWICTS data
• 63 of the individuals in the DWICTS file with a
  DWI conviction on a court date from 2000 - 2003
  linked to the NEEDS and CTP data
• This excludes three counties data due to poor
  quality
• Other factors are being investigated

18
Survival Analysis

• Exposure-Time to repeated DWI event with complete
  and incomplete observations
• Mean-Time to repeated DWI following conviction
• Risk of repeated DWI, hazard and survival
• Comparisons by factors to repeated DWI

19
(No Transcript)
20
Cumulative Re-Arrest Rates by Race/Ethnicity
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
21
(No Transcript)
22
Cumulative Re-Arrest Rates by Gender
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
23
Cumulative Re-Arrest Rates by Age
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
24
Cumulative Re-Arrest Rates by BAC Levels
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
25
Cumulative Re-Arrest Rates by Prior Convictions
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
26
Cumulative Re-Arrest Rates by Source of Alcohol
Before Arrest
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
27
Cumulative Re-Arrest Rates by Type of Sanction
Source Linked NM Screening and Tracking System
and CTS convicted DWI offenders, 2000 - 2003
28
Major Findings 2003

•     The median blood alcohol level for convicted
  DWI offenders is 0.16g/dl excluding offenders
  with missing blood alcohol levels. Including
  offenders missing blood alcohol levels the median
  is 0.14g/dl. The legal limit in New Mexico is
  0.08g/dl.
• Outpatient treatment is recommended by the CTP
  for 73 of convicted DWI offenders
• 50.9 of convicted DWI offenders consumed
  alcohol obtained directly from restaurants, bars,
  or stores where liquor is sold just prior to
  their DWI arrest.
• Native Americans have the highest re-arrest
  rates
•    The younger a convicted DWI offender is, the
  higher the re-arrest rate
•   10.2 of convicted DWI offenders with one
  prior conviction have a re-arrest

29
Future Directions for Datasets and Analysis

• Further analysis of linked datasets to examine
  sentence completion and recidivism, as well as
  successful sanction combinations to reduce
  recidivism
• Data linkage expanded to include New Mexico crash
  data, Behavioral Health Services Treatment data
  as well as NEEDS, CTP and DWICTS data
• A web-based NEEDS and CTP has been initiated and
  will be operational in all 33 counties in New
  Mexico by the end of August 2005.
• All data from 1999 to present has been uploaded
  into web-based system
• Additional data fields to track offenders from
  arrest to final disposition of sentence are
  included in web-based system

30
Ways to Improve the Linkage

• Include Violation Date on the CTP to enable
  linkage at the citation level instead of
  offenders.
• Soundex algorithms to overcome variable
  spellings, e.g. C de Baca or Cedebaca
• Incorporate recognition supplemental file to
  include aliases and other used identifiers.

31
Future Challenges

• CTP data shows possible lapse of 10 years between
  violation (arrest) and conviction
• Improve data from counties with poor data quality
  by further training of county personnel
• 38 of Motor Vehicle Division convictions not
  shown in CTP data collected by counties
• As a result of our finding, the New Mexico DWI
  Czar is investigating this further
• In the past Ignition Interlock data has been too
  sparse for any noticeable comparison
• Evaluate recidivism dynamics over time

32
Declining Recidivism Rates?