DOJ Study
DOJ Study
The U.S. Department of Justice has published its including homicide, while the other instruments were
“Intimate Partner Violence Risk Assessment Validation,” designed to identify risk of re-offending in spousal
reporting on the five-year study that assessed the assault cases.” [See Table 1]
following approaches to managing and predicting risk Aiming at assessing the predictive accuracy of the
of future harm or lethality in domestic violence cases: approaches studied, the project notes that: “Although
Domestic Violence MOSAIC (DV-MOSAIC), DV-MOSAIC may enhance predictions made by case-
Danger Assessment (DA), Domestic Violence Screening managers, it is not designed solely for prediction.”
Instrument (DVSI), and the Kingston
Screening Instrument for Domestic Violence How likely partner will be How likely partner will seriously
physically abusive in next year hurt you in next year
(K-SID).
DA Point Score .247 ** .344 **
Researchers had full hands-on access to DV-MOSAIC Rating .450 ** .465 **
every aspect of DV-MOSAIC for rating 1,307 DVSI Point Score .248 ** .297 **
battery cases. Comparing interviews, follow- KSID Risk Score .175 ** .143 **
up interviews, and criminal justice data, Table D8. Correlations between T1 (baseline) risk assessment scores and
victims’ perceived risk of abuse at beginning of interview.
the project found that DV-MOSAIC scores
were significantly associated with the level of
abuse. Still, DV-MOSAIC tested highest on what the
researchers called “sensitivity,” correctly classifying most
The report notes that the four approaches differ
of the women that were indeed re-assaulted.
substantially, mostly in that MOSAIC is a comprehensive
method, not merely a paper instrument: “DV- The study determined that when compared to the
MOSAIC was designed to help professionals organize other approaches, “DV-MOSAIC performed best in
case information and assess the likelihood of escalation, predicting subsequent stalking or threats.”
DV-MOSAIC also had the
Instrument/ Sensitivity Sensitivity & Specificity Specificity &
Assessment
Levels/
(Self Report) CJ Data (Self Report) CJ Data
strongest correlation between the victims’
Ratings perception of risk of re-assault and risk of
Method Any Severe Any Severe Any Severe Any Severe
(Variable) serious harm. [See Table D8]
Increased .917 .975 .892 .921 .219 .211 .209 .201
DA
Severe .704 .775 .683 .730 .492 .523 .486 .472
The project found that DV-
Extreme .477 .538 .460 .494 .684 .672 .679 .666 MOSAIC captured relevant information
(3, 4) equally well with victims of various
DV MOSAIC 5-7 .826 .983 .935 .934 .074 .071 .061 .062 ethnicities. [See Table D4]
8-10 .360 .458 .331 .395 .680 .679 .652 .672
(Low) Excerpts from the National Institute
DVSI High .532 .667 .514 .629 .486 .517 .477 .514 of Justice Report:
(Low)
Moderate .658 .672 .648 .658 .417 .408 .415 • “It is notable that DV-MOSAIC had
K-SID High .316 .281 .296 .263 .759 .740 .752 .775 the highest Wald statistic for predicting
Very High .274 .250 .256 .237 .782 .768 subsequent stalking and threats.”
(Low)
Victim Medium .697 .674 .663 .472 .511 .461 • “Scores on DV-MOSAIC were
High .556 .543 .521 .589 .624 .576 significantly associated with level of
Table 3. Sensitivity and specificity for levels of instruments for any and severe re- abuse at follow-up. Subjects that scored
assault using self report outcomes and self report & criminal justice data.
in the highest rating (8 -10) category on DV- Respondent Characteristic Mean (SD) % < 125 (n)
MOSAIC were twice as likely as those who had All respondents (n=641) 162.3 (16.6) 2.7 (17)
scores of below 8 to experience potentially lethal African American (n=174) 166.0* (13.9) .6 (1)
abuse during the follow-up period. Those with the Latina/Hispanic (n=350) 160.3 (17.6) 4.0 (14)
highest ratings on DV-MOSAIC were only half as Non-Hispanic White (n=60) 161.3 (17.0) 1.7 (1)
Mixed/Other (n=54) 163.7 (15.6) 1.9 (1)
likely to experience no abuse or only verbal abuse.”
[Among the 38 women who were administered the Foreign Born (n=136) 161.2 (17.1) 2.9 (7)
DV-MOSAIC and subsequently experienced very U.S. Born (n=240) 162.9 (16.3) 2.5 (10)
high (and potentially lethal) violence, all 38 had Spanish Interview (n=56) 157.9** (17.5) 5.9 (6)
scores at the highest ratings.] English Interview (n=319) 163.1 (16.3) 2.0 (11)
Instrument/
Information Source/
Assessment Method Administrator Modality System/Setting Purpose
Respondent
(Developed by)
Domestic Violence Probation Officer Offender, Criminal Record In-Person interview, Criminal Justice Level of Community
Screening Inventory (DVSI) Record review Supervision,
(Williams & Houghton) Terms of Probation/
Parole
Kingston Screening Probation Officer, Police Reports, Offender, In-Person interview, Probation Probation Release, Risk
Instrument for Domestic Advocate Victim Record Review of Re-Offending
Violence (K-SID)
(Gelles, Lyon)
DV MOSAIC Police Officer, Victim, Ongoing investigative In Person Interview/ Law Enforcement Immediate Safety
(de Becker) Detective Sources (Criminal Justice Phone, Record Planning, Criminal
Records, Other Informants) review Justice Response, Victim
Awareness, Police Officer
Training
Danger Assessment (DA) Advocate Victim In Person Interview Health Care, Victim Education/
(Campbell) Victim Awareness,
Assistance Advocate Assessment
Table 1. Instruments/Assessment Methods Tested: Intended Administration Method, Setting, Purpose and Use.