System Assessment of Risk: Keeping Women and Children Safer

1. System Assessment of Risk: Keeping Women and Children Safer Jacquelyn Campbell PhD RN FAAN Anna D. Wolf Chair and Professor Johns Hopkins University School of Nursing Multi City Intimate Partner Femicide Study Funded by: NIDA/NIAA, NIMH, CDC, NIJ VAWA R01 DA/AA1156

2. INTIMATE PARTER HOMICIDE • In Canada – 30-40% of women killed – Killed by husband, partner, BF or EX – Stats Canada (vs. 7-10% of men) (US – 40-50% vs. 5-8% of men – hand count) • 3:1 Ratio of women killed by intimate partners vs. men in Canada (ranges from 3-5:1 over past 20 years) • Canada – Hx of DV: 62% spousal femicides; 67% male (Stats Canada ‘04) - DV against female partner; 71% Ontario DR • More at risk when leaving or left (Wilson & Daly ‘93; Campbell et. al. ‘01) 79% spousal homicides (Ont DR): during,1st 3 mos, 1st year • Women far more likely to be victims of homicide-suicide (29% vs. .1% in US; 29.3% vs. 2.9% in Canada) • Aboriginal and immigrant women more at risk (NYC) • IPV 21% of aboriginal (M & F) vs. 7% nonaboriginal in Canada) • 44-47% of women killed seen in health care system before killed (Sharps, Campbell ’02; Wadman & Muelleman ‘99)

3. Number of Canadian Individuals Killed 2000 – 2006 Data from Brian Vallee, The War on Women, (2007)

4. 32% Witnessed Femicide; 58% Witnessed Attempts 43% & 37% Found Mother Received Counseling 60% - all children of actuals; only 28% of attempteds 56% & 40% of children who witnessed femicide & attempts 57% & 54% of children who found the body 71% & 76% of mothers abused 22% & 27% threats to take children if she left 20% & 13% threats to harm children if she left 8% fathers reported for child abuse – both actuals & attempteds “He Killed My Mommy” – Lewandowski, Campbell et. al. JFV ’04: kids in the homes of women killed or almost killed

5. DISRUPTION OF CHILDREN AFTER HOMICIDAL EVENT • Moved from home • To mother’s kin • To father’s kin • Split between mother’s and father’s kin • With others (e.g. foster home) Femicide 86% 40% 12% 5% 14% Attempted 25% 16% 6% 0% 2%

6. SERVICES FOR CHILDREN WITNESSING INTIMATE PARTNER HOMICIDE • Vary city to city • Private - Victim Assistance Funds • Public - Specialized City Agency • Services offered but no follow-up • Coordination with school fragmented • Custody issues • Nothing systematic for attempted homicides

7. VICTIMS Police Contacts - 66% of stalked & battered women Any Medical Visit - 56% 27% ED; 25% mental health Shelter Contacts - 4% of battered women Substance abuse Tx - 6% PERPETRATORS Prior Arrest - 56% of batterers (32% of non) Canada – 30% of spousal femicides – prior police contact Mental Health System - 12% but 36% described as having “poor mental health Canada - 17% – “suspected” mental/developmental disorder Substance Abuse Tx - 6% (but 48% with drug &/or alcohol px) Reported Child Abuse – 8.6% MISSED OPPORTUNITIES FOR PREVENTION - 83% of Cases – National Femicide Study (Sharps, Campbell et al, 2001; Stats Canada ’04, ‘05)

8. MISSED OPPORTUNITIES:PERPETRATORS

9. ATTEMPTED & ACTUAL VICTIMS SEEN IN SYSTEM

10. Spousal Homicide in Canada • BF-GF category increasing – rest decreasing • Same sex intimate homicides – 6 since ’97 vs. approx 75 per year heterosexual • Atlantic – lowest rate of spousal homicide in Canada, Alberta, Sascatchewan highest – BC & Manitoba tied for 3rd • Rates: Sascatchewan -.76/100,000 spouses, Alberta .62; Atlantic Region - .35/100,000 spouses; BC – .53 = average of 13 per year ‘99-’03 (10 femicides • If approximately 8-9 attempted femicides for each femicide = 80 – total of at least 90 attempteds & actuals each year • Therefore, approximately 120-150 children each year affected by mother’s murder or attempted murder – prior DV • Ontario Death Review – 21% of spousal homicides – custody or access disputes; overall 10 of 342 women killed were pregnant when killed & 7 of 10 killed by partner or ex

11. CANADIAN INTIMATE PARTNER HOMICIDE RATE 1974-06 (’91 – first inclusion of BF-GF) (F 3-5 X M) Rate per million couples – 57% decline *BF-GF increased in ‘04 Source: Homicide Survey, Canadian Centre for Justice Statistics

12. U.S. INTIMATE PARTNER HOMICIDE RATE DECLINE 1976-02 FBI (SHR, 1976-02; BJS ’05, ‘07) FEMALE MALE 2004, 05 - no ex’s

13. INTIMATE PARTNER FEMICIDE BY PERPETRATOR IN TEN CITIES (N= 311) (female partner perpetrator – Glass et al 2004 – Canada: 6 since ‘97) EX-BF OTHER EX-SPOUSE BOYFRIEND SPOUSE

14. Decline in Intimate Partner Homicide and Femicide - US • Decline in male victimization in states where improved DV laws and services - resource availability (Browne & Williams ’89; ‘98) • Exposure reduction - increased female earnings, lower marriage rate, higher divorce rate (Dugan, Nagin & Rosenfeld ‘97) • Gun availability decline (Wilt ‘97; Block ‘95; Kellerman ‘93, ‘97- gun increases risk X3)

15. Rates of Spousal Homicide Using Firearms 40% of female; 21% of male (Stats Canada ‘03)

16. U.S. INTIMATE PARTNER HOMICIDE RATES & DOMESTIC VIOLENCE SERVICES 1976-9 (Resources per 50 million - Dugan, Nagin & Rosenfeld ‘03)

17. Spousal Homicide by Relationship Status – Canada 96-05 (*vs. 14% in all Canada)

18. CANADIAN RATE OF SPOUSAL HOMICIDES IN DIFFERENT TYPES OF RELATIONSHIPS, 96 – 06 Rate per 1,000,000

19. “Prediction is very hard to do - especially if it is about the future” Yogi Berra

20. Overview of Issues • High demand for both lethality & reoffending risk assessment by criminal justice, advocacy, victim service, & health systems • Other risk assessment instruments used for general probation purposes - accuracy with batterers for repeat IPV in dispute • Low base rates • Relatively young science in intimate partner violence & risk assessment particularly • 4 interacting parts to consider - instrument, risk assessor, perpetrator & one specific potential victim (vs. sexual assault or mental health – MacArthur study) • Fears that risk assessment will be used to limit service to victims • High stakes with either false negatives or false positives

21. Overlapping Concerns Similar; Not the same Lethality Assessment Risk Assessment Safety Assessment

22. Risk Prediction 4 Quadrant Model (Webster et. al. ‘94) Combination of Sensitivity & Specificity = ROC Curve (Receiver Operating Curve)

23. ROC Curve Analysis – DA Calibrated on Actual Femicides – Tested w/Attempteds - 90% Campbell et al JIPV ‘08

24. Navy (DoD) FAP Victim & Offender -reoffend Evaluation not finished Risk & Safety & safety (FA, MP, health) (Stith, Milner ’02-’04) DV Mosaic (20) (deBecker) Computerized/Victim (criminal .65 under ROC justice) - lethality risk system severe re-assault DVSI- R (Kirk Williams) Offenders (criminal justice) .71% under ROC short – reoffending (Williams & Grant, ’06) K-SID (Gelles & Lyon) Offender – reoffend (CJ) Some evaluation data O.D.A.R.A. (Z. Hilton) Offender - reoffend – (CJ) 77% under ROC (’04) PSI (Duluth) Victim & offender - both Process evaluation (advocates & criminal justice) (CDC) - no outcomes PAS (D. Dutton) Offender – intervention programs, Cross sectional prevention validation good SARA (Kropp et al) Offender (criminal justice) - Most data published B-SAFER reoffending best if perp. psych exam Danger Assessment Victim- Lethality (Advocates, Health) .69% under ROC (Campbell) reassault; .90% att. hom Existing Risk Assessment Scales

25. Risk Factors for Femicide-Suicide in Abusive Relationships: Results From a Multisite Case Control Study Jane Koziol-McLain, Daniel Webster, Judith McFarlane, Carolyn Rebecca Block, Yvonne Ulrich, Nancy Glass, Jacquelyn C. Campbell Violence & Victims (Springer), Volume 21, Number 1, February 2006

26. Femicide – Suicide Cases (32% of femicide cases in study – 29.3% -Canada; 29% -US) • No significant difference in proportion of prior physical DV – 77% w/o suicide (n = 210) ; 72% with suicide (n = 100) & similar characteristics – proportions of forced sex, choking, stalking • < abuse during pregnancy (14% vs. 30%) • 50% illicit drug use – not significantly more than abused • Approximately same proportion victim > 50 yo (w/o suicide 7%; with suicide 11%) • Little evidence of “mercy” killings or suicide pacts – one case with a suicide note suggesting such – family disputed motive • Prior DV – 32% in older women vs. 79% in younger but close to same as in older women killed w/o suicide as with suicide & 49% of older women afraid of partner

27. Femicide – Suicide cases • Significant explanatory power for same femicide – suicide risk factors. • Unemployment – slightly different profile • Partner access to gun – used in 61% of homicide/ suicides vs. 28% w/o • Same proportions in Quebec (Lund & Smorodinsky ‘01) • Threats with a weapon • Forced sex • Step child in the home • Highly controlling perpetrator – especially if estranged • Estrangement • (52% of femicides w/suicide 56% w/o) & leaving “trigger” 31% vs. 33%)

28. Femicide- Suicide • Unique to femicide – suicide: • Partner suicide threats – history of poor mental health – higher proportion to mental health care provider • In another study, majority of married femicide perpetrators failed to receive treatment for depression or received inappropriate anti-anxiety medications (Cohen, 2004). • Married • Somewhat higher education levels (unemployment still a risk factor), more likely to be white, Hispanic, Asian vs. African American • Children more likely to be killed along with female partner

29. Intimate Violence Risk Assessment Validation Study – J. Campbell, C. O’Sullivan & J. Roehl – NIJ #2000WTVX0011 • N = 782 women who had accessed system through calling police, civil court, shelter or hospital ED’s – NY & CA • Relatively severely abused – 43% severe abusive tactic from CTS last 6 mos • 60% retention from Time 1 – more severely abused. employed & Latina less like to return • 38% foreign born; 28% African American; 48% Hispanic/Latina

30. NIJ “RAVE” study • Women randomly assigned to 2 of 4 risk assessment methods • Also CTS, WEB scale, HARASS • Also 40 other items hypothesized to increase risk • Also 2 items to assess victim’s perceived risk • Recontacted 6 – 12 months later – interviewed by phone • Also a criminal justice record check for violent crime & DV offenses • Women – especially those who saw themselves at high risk took many impressive protective actions – all kinds • Their accuracy of perception of re-assault – significantly better than chance but

31. ROC curve analysis (excluding 27 victims w/no exposure to abuser) with potential confounders • Chance - .50 • Any & severe re-assault – all significant at <.01 • DA - .694; .714 • DV-MOSAIC .618; .665 • DVSI - .650; .664 • K-SID - .639; .657 • Victim perception .635; .627 • Instruments/method improved on victim assessment • But none of approaches without serious margin of error

32. Impact of the questionnaire • Repeat of self-perceived risk questions: not significant • Did answering these questions (CTS, 2 risk questionnaires, self-protective questions) change your view or thinking about his behavior? • No effect - 64% • More abusive or dangerous than I thought – 29% • Less abusive or dangerous than I thought – 7% • Do you think you will do anything differently in the future as a result of answering these questions (such as taking more safety precautions, or fewer; spending more or less time with him, etc.)? 61% yes

33. Women’s Statements After Risk Assessment Process (NIJ RAVE study) • “I never knew – this makes me much more resolved to not go back” • “I’m gonna’ go get that permanent thing (PO) – I wasn’t gonna’ go through the hassle before but now I surely will” • “Damn…. He is really dangerous, isn’t he? I keep foolin’ myself about that – now I know I gotta do something” • “I knew he was scary but no one believed me – I’m going to keep pushing now” • From a woman in Alberta after doing the DA “It was like filling in a piece of a puzzle – I could finally see the whole picture"

34. % taking protective actions during follow-up by victim perception of risk of serious harm.

35. Effects of protective actions on severere-assault during follow-up

36. Effects of protective actions on minor-moderate re-assault during follow-up

37. Effects of protective actions on stalkingduring follow-up

38. Conclusions re: protective actions’ effects on re-assault • Going to shelter at T1 significantly reduced risk of severe and moderate assaults • No voluntary contact with abuser during follow-up reduced risk of moderately severe assaults • Arrest at T1 reduced severe assaults • OP/RO reduced moderate assaults • Some protective actions at T2 likely to be response to assaults during follow-up • Places she thinks he can’t find her – often after a while, someone tells

39. 75% children in common with abuser 52% - abuser having visits w/ children 51% of visits by court order Unsupervised – 73% Supervised – 27% 24%: Physical abuse, threats or intimidation during exchange for visits Visitation during Follow-Up

40. GENERAL PRINCIPLES FOR RISK ASSESSMENT IN DV • More sources of information the better – “gold standard” for information is victim – without information from victim, cutoffs for lethality risk problematic – criminal record check important • Perpetrators will minimize perpetration • Few independent evaluations of current instruments – use any cutoffs with caution – DVSI or ODARA best if criminal justice records only & SARA with psych evaluation • Instrument improves “expert judgment” – but clinician wisdom important also • Never underestimate victim’s perceptions (Weisz, 2000; Gondolf, 2002) but often minimize victimization – therefore victim assessment of risk not enough if low • Need for DV identification and risk assessment as part of psychiatric practice

41. Implications for Policy & Safety Planning • Clinical assessment (psychiatry, psychology) needs specific DV training • How to ask about DV as part of suicidality assessment • Batterer intervention - she needs to stay gone until he completes & his attendance monitored • Employment issues – especially for African American and aboriginal men • Protective order for stalking - or use stalking laws • Issues with various “risk” lists included in safety planning • Supervised visitation & other means of keeping her safe during visitation exchanges

42. Implications for Policy & Safety Planning • Engage women’s mothering concerns & skills (Henderson & Erikson ’97 ‘93; Humphreys ‘93; Sullivan et. al.‘00) • If she says she’s going to leave, cannot leave face to face • Importance of forced sex & stepchild variables – not on most risk assessment instruments • Make sure she knows entire range of shelter services • Be alert for depressed/suicidal batterer • Batterer intervention programs working with partners

43. Policy/Practice/Research Implications • Need for substance abuse Tx in abusive men – concurrent with batterer intervention? Combination programs? New models needed with rigorous evaluations • Need for collaborations btw. researchers & clinicians in substance abuse, health, criminal justice, forensic psychiatry and advocacy – for advances in risk assessment – research and policy

44. Gun Issues • Get the gun(s) out!!! • Judges need to order removal of all guns – specify in search warrants & PO’s • Canada better about guns in general than US but still issues

45. Future Directions • “Danger Assessment is a Process not a Product” (B. Hart) • Field developing rapidly – watch literature • Differentiating lethality & reoffending risk - different batterer typologies may explain differences (Holtzworth-Munroe) • Strategies for working with victims important – to increase their realistic appraisal and to determine risk factors not available from criminal record checks or from perpetrators never previously arrested – e.g. as part of batterer intervention programs • Assessing safety – protective strategies as well as danger – implications for interventions • Two parallel processes – reoffending risk for criminal justice cases; danger (lethality) – e.g. DVSI risk for victim safety planning

46. Offenders in CJ or BIP Or MH or SA Tx Ideal Process Model First Responders B-SAFER or MD Lethality Assessment Women in Shelters Or Health Care System Risk Assessment Partners of Men in System Danger Assessment & Safety Assessment Judicial System System Safety Audit – Including Fatality Reviews

47. Also with RAVE data – 3 item version of DA – submitted for publication • Outcome of those 666 who took DA & full data on return - N = 400 • Predicting those seriously re-assaulted – life threatening – seriously choked, shot or knifed or head injury with loss of consciousness (15% of total)

48. 3 item version – acceptable validity for severe re-assault (ROC .79) • Violence increasing in severity &/or frequency over past year • Used a weapon on you/threatened with a weapon • Woman believes he is capable of killing her If one of 3 – 94% sensitivity but only 29% specificity If two yeses of 3 – 81% sensitivity & 56% specificity Either way – outperforms women’s perception of risk by itself

49. Policy/Practice Possibilities • Use 3 item version in ED, protective order hearings, child custody etc. • If 2 of 3, do full DA and proceed based on results • If 1 of 3, tell her has one of 3 highly predictive risk factors for serious assault/homicide – highly recommend further immediate advocacy – call with her • If none of 3, proceed with normal referral/ procedural processes

50. Never forget who it’s for - “please don’t let her death be for nothing – please get her story told” (one of the Moms)