Funded by the National Institute on Alcohol Abuse and Alcoholism (R37- AA10908)
Population Attributable Risk, Intimate Partner Violence and Drinking
Raul Caetano, M.D., Ph.D.
Susie Mikler, M.P.H., Ph.D.
University of Texas School of Public Health
Objectives of this Presentation
Briefly discuss the application of the population attributable risk in assessing population level alcohol control policies to prevent intimate partner violence.
Prevalence of IPV
Male to Female Partner
Violence (MFPV)
Female to Male Partner
Violence (FMPV)
1975: 12% 1975: 11.6%
1985: 11.6% 1985: 12.4%
1992: 9% (minor) 1992: 9% (minor)
1992: 2% (severe) 1992: 4.5% (severe)
Prevalence of Violence: 1995 National Study of Couples
79%
13%
8%
No IPV Moderate Severe
Violence, drinking and alcohol policy
• As suggested by Room and Rossow (2000) it is important to consider different types of violence: domestic, bar and street, collective (e.g., war atrocities).
• The association between alcohol and IPV in general population data are not strong and not always consistent.
• This is probably due to the nature of IPV (mostly moderate) and drinking (mostly light) in these data.
• The effect of alcohol-related policies on IPV is difficult to estimate because the Population Attributable Risk (PAR) is dependent on the definition of exposure.
Population Attributable Risk
PAR = It – Io/It x 100
It: Incidence among the exposedIo: Incidence among the non-exposed
Rate of disease in the total population that is attributable to the exposure
Estimates in the Literature
• 33% of assaults (Lenke)
• 50% of convicted assaults (Skog & Bjorks, 1988)
• 69% of homicides, 47% of assaults (Norstrom, 1998)
• Positive relationship with per capita consumption in aggregate studies
Population Attributable Risk %
IPV MFPV FMPV
Drinking
Perp/Victim 19 25 0
Perp 28 0
Binge
Perp/Victim 27 32 17
Perp 27 4
If Exposure Is Drinking in the Event
• Alcohol exposed IPV cases are those among drinkers drinking in the event.
• IPV cases among drinkers not drinking in the event are not exposed. In U.S. data, about 70% of MFPV and 80% of FMPV events by drinkers had no drinking in the event.
Other Considerations
• Violence literature has not considered a measure such as “drink 6 hours before” the event, which has been used in the injury field.
• Consider culture: expectancies about alcohol effects.• Consider drinking patterns: how much binge and
intoxication?• Consider population assumptions about the relationship
between drinking and intimate partner violence.• Consider then the scientific and political uses of the PAR.
Other Considerations
• Consider that PAR may vary across population subgroups because strength of association alcohol/IPV varies.
• There are no specific alcohol-control policies to prevent intimate partner violence (or domestic violence).
• Policies should not be justified by IPV prevention alone, but should based on the general prevention of violence and other problems.
• Policy effectiveness may have to be assessed by aggregate-level studies.
The End
Alcohol and Intimate Partner Violence
Issues for consideration: Is this association causal and if so which mechanisms
underlie it.Many aggressive events are not associated with alcohol. It is difficult to establish a temporal relationship
between drinking and aggression.Alcohol and aggression may be both associated with a
third factor (e.g., Impulsivity).The association may be due to expectancies about the
effect of alcohol.
Overall Design
1995 National Survey of Couples
1635 Couples Interviewed
555 Whites, 358 Black, 527 Hispanics
85% Response rate
2000 Follow-up National Survey of Couples
1392 couples reinterviewed,
1136 intact couples.
406 White, 232 Black and 387 Hispanic couples
72% Response rate
Other Methodological Features
Multistage area household probability sample.
Both partners interviewed separately.
Male-to-Female-Partner Violence (MFPV) and female-to-male partner violence (FMPV) measured with the Conflict Tactics Scale.
Upper-bound estimate used; violence considered to have occurred if either partner reports event.
Other Methodological Features
1995: Interviews of about 1 hour with main respondents and of about 20 minutes with partner. 2000: Interviews of about 1 hour with both
Interviews conducted by trained interviewers. Standardized questionnaire used (close ended). Bilingual interviewers (Spanish/English) available
on request. Spanish questionnaire available on request.
Items in the Conflict Tactics Scale
Threw something (at him/her) (moderate) Pushed, grabbed or shoved (moderate) Slapped (moderate) Kicked, bit or hit (severe) Hit or tried to hit with something (severe) Beat up (severe) Choked (severe) Burned or scalded (severe) Forced sex (severe) Threatened with a knife or gun (severe) Used a knife or gun (severe)
Data Coverage
Alcohol consumption was assessed with a series of questions covering quantity and frequency of drinking wine, beer and spirits in the past 12 months.
Sociodemographic factors: Gender, age, income, education, marital status, place of birth, religion, employment status, occupation, number of children <17 at home, relationship length.
Data Coverage
Psychosocial factors: Childhood history of parent-perpetrated violence, witnessing violence between parents, approval of marital aggression, impulsivity, risk taking.
Acculturation. Drug use. Residential addresses matched to Census Tract
number. 1990 Census data appended to the geocoded sample in 1995.
Cross-Sectional Results
Overall Rates of Male to Female Partner Violence
21
12
16
23
30
17
0
5
10
15
20
25
30
35
40
45
Any MFPV* Any FMPV
White
Black
Hispanic
Chi-square across ethnic groups: * p < .01
*
*
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
WhiteBlack
Hispanic
* Chi-square across ethic groups: p < .05
Type of Male to Female Partner Violence
Push Slap Kick BurnBeatHitWith
ChokeThrow Threat/Knife or
Gun
Sex Use/Knife or
Gun
*
*
* *
Moderate 10% Severe 4%
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
WhiteBlack
Hispanic
* Chi-square across ethnic groups: p < .05
Type of Female to Male Partner Violence
Push Slap Kick BurnBeatHitWith
ChokeThrow Threat/Knife or
Gun
Sex Use/Knife or
Gun
* *
*
* *
Moderate 11% Severe 7%
Unidirectional and Bidirectional Intimate Partner Violence
0
10
20
30
40
50
60
70
MFPV FMPV Both
WhiteBlackHispanic
Male to Female Partner Violenceby Drinking Patterns
6
10
14
11
16
31
41
15
19
24
18
1317
7
17
0
5
10
15
20
25
30
35
40
45
Abstain Infrequent Less Frequent Frequent Frequent HeavyDrinker
White
Black
Hispanic
Chi-square: p:ns
Female to Male Partner Violenceby Drinking Patterns
12
19
5051
79
3229
20
33
1515
24
27
2517
0
10
20
30
40
50
60
70
80
90
Abstain Infrequent Less Frequent Frequent Frequent HeavyDrinker
White
Black
Hispanic
Chi-square: Black - p<.001, Hispanic - p<.05
Odds Ratios from Multiple Logistic Regression on Male to Female Partner
Violence
White Black Hispanic
Male Alcohol Problems
- 10.33 (4.44-24.03) -
Female Alcohol Problems
- 5.40 (1.97-14.81) -
Variables in the analysis: Income, education, employment, age, marital status, number of children, relationship length, impulsivity, volume of alcohol consumption.
Odds Ratios from Multiple Logistic Regression on Female to Male Partner
Violence
White Black Hispanic
Male AlcoholProblems
- 2.84 (1.17-6.89) -
Female AlcoholProblems
6.57 (1.87-23.06) 6.61 (2.50-17.51) -
Variables in the analysis: Income, education, employment, age, maritalstatus, impulsivity, volume of alcohol consumption.
Other factors associated with Intimate Partner Violence
Risk factors for MFPV: Young age, lower income, unemployment, childhood physical abuse, approval of aggression, alcohol problems, neighborhood poverty.
Risk factors for FMPV: Young age, childhood physical abuse, alcohol, number of children, approval of aggression, alcohol problems, neighborhood poverty.
Longitudinal Results
Initial Research Questions
What is the prevalence of IPV across ethnic groups in 1995 and 2000?
What is the stability, remission and incidence of IPV across ethnic groups?
What is the course of IPV (no violence to moderate to severe violence) across ethnic groups?
What are the predictors of stability, remission and incidence?
Non-Response Analysis
Non-respondents were more likely to be: Younger men (18-29), unemployed men, women 40 to 49.
Women who experienced abuse during their childhood were less (OR=0.6) likely to be among non-respondents.
Gender-specific multivariate logistic regression models accounted for only 5% of the variance in survey participation.
Prevalence of Intimate Partner Violence: 1995 and 2000 (1)
17
10
27
22
2624
0
5
10
15
20
25
30
1995 2000
White
Black
Hispanic
(1) Intact Couples Only
* *
* Chi-square: p>.01
Stability and Incidence of Intimate Partner Violence (1)
38
6
51
14
57
14
0
10
20
30
40
50
60
Stability Incidence
White
Black
Hispanic
(1) Intact Couples Only
*
**
•*Chi-square: Hispanic vs. White, p>.01•**chi-square: p>.01
Odds Ratios from Multinomial Logistic Regression: Alcohol Problems a
Incidence
Re: No violence
Recurrence
Re: No violence
Male Alcohol
Problems
1.2 (0.2-5.4) 1.8 (0.6-5.9)
Female Alcohol Problems
0.1 (0.01-.6)* 1.7 (0.6-5.2)
* Significant at p<.05a Also controlling for: drug use, childhood victim., parental violence, impulsivity, attit. toward violence, age, income, education, emplm. status, marital status, length of relationship, ethnicity, weekly N drks, 5 or +.
2000 Status of 1995 Non-Violent Couples
94
5 1
87
121
86
112
0102030405060708090
100
No Violence Moderate SevereViolence
WhiteBlackHispanic
66
26
8
59
30
10
58
37
4
0
10
20
30
40
50
60
70
No Violence Moderate SevereViolence
WhiteBlackHispanic
2000 Status of 1995 Couples with Moderate Violence
2000 Status of 1995 Couples with Severe Violence
52
43
5
39
27
35
28
49
23
0
10
20
30
40
50
60
No Violence ModerateViolence
SevereViolence
WhiteBlackHispanic*
*chi-square: p>.03
Odds Ratios from Logistic Regression: 1995 Alcohol-Related Predictors of 2000 MFPV
White Black HispanicMen N drinks ** .9 (.8 - 1.1) 1.0 (.9 - 1.2)
Women N drinks .3 (.1 - .7)* 1.1 (.9 - 1.4) **
Men 5+ .4 (.1 – 1.3) .6 (.2 – 1.8) 1.2 (.6 – 2.7)
Women 5+ 3.6 (1.1 –12.1)* 2.7 (.5 – 15.7) .6 (.3 – 1.5)
Men’s Problems 3.0 (.7 – 12.4) 2.9 (.4 – 20.2) 1.2 (.5 – 3.3)
Women’s Problems ** 3.6 (.6 – 21.1) 2.0 (.4 – 10.5)
Also controlling for: childhood victim., parental violence, impulsivity, attit. toward violence, age, income, education, marital status, MFPV and FMPV in 95.
* p<.05; **variable failed criteria for inclusion: p=.25
Odds Ratios from Logistic Regression: 1995 Alcohol-Related Predictors of 2000 FMPV. a
White Black HispanicMen N drinks 1.2 (1.1 – 1.3)* ** **
Women N drinks ** ** **
Men 5+ .2 (.1 – 1.2) 2.6 (.7 – 10.0) .9 (.4 – 1.8)
Women 5+ 3.3 (.6 – 18.3) 1.9 (.4 – 9.7) 1.3 (.6 – 2.6)
Men’s Problems .9 (.3 – 3.0) 1.7 (.3 – 8.4) 1.1 (.4 – 2.9)
Women’s Problems ** 3.0 (.4 – 25.5) 1.7 (.4 – 7.0)
a Also controlling for: childhood victim., parental violence, impulsivity, attit. toward violence, age, income, education, marital status, MFPV and FMPV in 95.
* p<.05; ** variable failed criteria for inclusion: p =.25
Goodness of fit indices
Hu and Bentler (1999)
• Comparative Fit Index (CFI)
– Between .90 and .95< : Acceptable fit
– >.95 : Good fit
• Root Mean Square Error of Approximation (RMSEA)
– RMSEA < 0.05 Close fit
– RMSEA > .05 ≤ 0.08 Reasonable fit
– RMSEA > 0.1 Poor fit
• Ratio of chi-square to degrees of freedom (χ2 / df) = 2
IPV 1995
Male Alcohol Problems
IPV 2000
Female Alcohol Problems
Female 5 plus drinks
Female Alcohol Problems
Female 5 plus drinks
Male Alcohol Problems
Male 5 plus drinks
Figure 1. Paths for the core set of associations in the model – WHITES1995 2000
Female Alc. Vol.Female Alc. Vol.
Male Alc. volume Male Alc. volume Male 5 plus drinks
0.66 *
049 *
0.14
0.35 *
0.38 *
0.41 *
0.71 *
0.42 *
0.10 *
0.42 *
0.05 0.06
0.48 *
0.12 *0.09
0.16 *0.39 *
0.39 *
0.01
- 0.14 *
0.19 *
0.27 *
0.32 *0.15 *
0.07
0.060.31 *
0.13 *
0.13 *
CFI = .879CFI = .879
χχ 2 = 462.2 2 = 462.2
df = 238, df = 238, p=.0000 p=.0000
RMSEA = RMSEA = 0.0510.051
IPV 1995
Male Alcohol Problems
IPV 2000
Female Alcohol Problems
Female 5 plus drinks
Female Alcohol Problems
Female 5 plus drinks
Male Alcohol Problems
Male 5 plus drinks
Figure 2. Paths for the core set of associations in the model – BLACKS1995 2000
Female Alc. Vol.Female Alc. Vol.
Male Alc. volume Male Alc. volume Male 5 plus drinks
0.63 *
0.38*
0.13
0.42 *
0.66*
0.36 *
0.66 *
0.48 *
0.16
0.51 *
0.21*0.03
0.39 *
0.04
0.29 *
0.190.43
0.43 *
- 0.06
- 0.07
0.34 *
0.27 *
0.23 *0.22 *
0.18 *
0.110.08
0.09
0.08 *
CFI = .766CFI = .766
χχ 2 = 592.3 2 = 592.3
df = 238, df = 238, p=.0000 p=.0000
RMSEA = RMSEA = 0.0890.089
IPV 1995
Male Alcohol Problems
IPV 2000
Female Alcohol Problems
Female 5 plus drinks
Female Alcohol Problems
Female 5 plus drinks
Male Alcohol Problems
Male 5 plus drinks
Figure 3. Paths for the core set of associations in the model – HISPANICS1995 2000
Female Alc. Vol.Female Alc. Vol.
Male Alc. volume Male Alc. volume Male 5 plus drinks
0.32 *
0.31*
0.29 *
0.46 *
0.17
0.36 *
0.43 *
0.46 *
0.11
0.49 *
0.20*-0.07
0.40 *
0.025
- 0.045
0.260.33 *
0.43 *
0.06
0.01
0.44 *
0.62 *
0.27 *0.19 *
0.11
0.110.12
0.01
0.10 *
CFI = .888, CFI = .888,
χχ 2 = 430.6 2 = 430.6
df = 238, df = 238, p=.0000 p=.0000
RMSEA= RMSEA= 0.0490.049
Cross-sectional Associations
White (n=366) Black (n=190) Hispanic (n=344)
Paths Estimated 1995 2000 1995 2000 1995 2000
Male Volume Alc Problems * * NS * NS *
Male Binge Alc Problems * * * * * *
Male Alc Problems IPV NS * * NS * NS
Male Binge IPV NS NS NS NS NS NS
Female Volume Alc Problems * NS NS NS NS NS
Female Binge Alc Problems * * NS NS * NS
Female Alc Problems IPV NS NS * * NS NS
Female Binge IPV * Negative * NS NS NS NS
Correlations
Male Volume & Binge * * * * * *
Female Volume & Binge * * * * * *
Male and Female Volume * * * * * *
Summary of Longitudinal Results
Most couples reporting IPV at baseline do not report IPV five years later.
The likelihood of reporting IPV five years later is related to the severity of IPV at baseline. This is equally true of MFPV and FMPV.
Summary of Longitudinal Results
Male weekly N. of drinks is associated with incidence.
Male and female (Blacks only) alcohol problems associated with IPV.
Hispanics are more likely than Whites to report incident IPV (OR=2.9, 1.2-7.1).
Blacks are 3 times more likely than Whites to report IPV in both 1995 and 2000 (OR=2.9, 1.2-6.9).
Predictors of Intimate Partner Violence
Moderate
Severe
Sociodem. Char.
Personality Char.
Relationship Char.
Personal History
Situational Factors
Contextual Factors
Cultural Factors
Alcohol Problems
Psychiatric Diagnoses
Alcohol Dependence
Drug Dependence
VIOLENCE
Intimate Partner Violence: A Longitudinal Perspective
Violence ContinuumModerate Severe
No Violence
Time
Couple’s Level of Agreement in Reporting Intimate Partner Violence (Kappa)
Violent Act Male to Female Female to Male
Threw somethingPushed, grabbed, or shovedSlappedKicked, bit, of hitHit or tried to hit with somethingBeat upChokedBurned or scaldedForced sexThreatened with knife or gunUsed knife or gun
Any violence
.36
.49
.38
.35
.23
.22
.30*
.15**
.39
.40
.45
.34
.34
.28
.00
.28
.00
.16
.13*
.36
* Not calculated, 2 or more cells very close to 0.
Rates of Intimate Partner Violence in the U.S.: Lower and Upper Bound Estimates
56
8
14
18
21
0
5
10
15
20
25
MFPV FMPV ANY
Lower Bound
Upper Bound
Two Different Ways to Report the Prevalence of Intimate Partner Violence
MFPV Only
FMPV Only Both
No
MFPV
FMPV
No
Proportional Representation of MFPV Only, FMPV Only and Bidirectional Violence Among IPV Cases
17 919
39
30
35
4461
46
0%10%20%30%40%50%60%70%80%90%
100%
White Black Hispanic
BothFMPVMFPV
Overall Rates of Intimate Partner Violence Among U.S. Hispanics by Acculturation
1413
20
34
18 18
0
5
10
15
20
25
30
35
Any MFPV Any FMPV*
Low
Medium
High
* Chi-square: p <.01
*
Odds ratios and from Multiple Logistic Regression: Acculturation and IPV.
Male to Female Female to Male
(Ref: Low/Low)
Medium/Medium 3.71 (1.42-9.71) 10.92 (3.96 – 30.12)
Low/Medium 3.17 (1.01 – 9.96) 9.14 (2.43 – 34.38)
High/Medium 3.14 (1.06 – 9.31) 4.93 (1.15 – 21.13)
High/High 5.59 (1.29 – 24.36)
Variables in the analysis: Income, education, age, marital status, childhoodvictimization, approval of aggression, drinking patterns, impulsivity.
Association Between IPV and Expectancies
How likely to be aggressive?
Violence No violence
No/not much chance
16% 13%
Strong/very strong chance
19% 7%
OR (Strong/very strong vs. no: 3.2 (1.3-7.9). Controlling for gender, age, ethnicity, income, attitudes towards IPV, impulsivity, excuses.
malcprob 2000
ipv 1995
malcvol 2000
ipv 2000
falcvol 2000
falcprob 2000
falcvol 1995
falcprob 1995
malcvol 1995
malcprob 1995
mchildabuse (0.042)fchildabuse(0.128) *mattitudes (0.183) *fattitudes (0.010)mimpul (0.128) *fimpul (0.075) freqrelig (0.045) *relength (-0.093) *
mhispus medu incomemarry memploy mimpul
fhispus fedu incomemarry fimpul
faccult 1995
femploy
Hispanic (including birth place and acculturation)
-0.379
4.6400.265
-6.384 *
3.09
1 *
-0.4
10
4.858 *
-0.008
0.045
-0.079
0.19
9
0.002
-0.0
16 *
0.009 *
0.019 *
0.034
0.082
0.238 *
0.379 *
0.00
7 *
0.284 *
0.014 *
0.045
0.410 *
0.000
-0.00
4
0.053
0.37
0 *
0.17
3 0.755 *
-0.127
maccult 2000
maccult 1995
faccult 2000
0.822 *
0.839 *
0.077
0.062
0.117 *
0.03
2
Potential Courses for Violence Type: 1995 to 2000
No
FMPV BothMFPV
1995
2000
Number of Respondents Eligible for Reinterview in 2000 Follow up Eligible Individuals for 2000 Follow-up Survey N = 3,270
Dead or IncapacitatedN = 124
Capable of Re-interviewN = 3,146
InterviewedN = 2,577
Not InterviewedN = 569
Excluded from the Analysis N = 35
Excluded from the Analysis N = 5
Final Respondents included in the Analysis N = 2,542
Final Non-Respondents in the Analysis N = 564
Proportion of Incident and Stable Cases of IPV in 2000 among All IPV Cases by Ethnicity
Black
40
15
4560
Incidence Chronic Severe Moderate
Hispanic
44
12
4456
Incidence Chronic Severe Moderate
White
32
59
27
41
Incidence Chronic Severe Moderate
Prevalence of MFPV and FMPV in 1995 and 2000
118
1310
1821
24 25
17
22 21 20
0
5
10
15
20
25
30
MFPV 95 MFPV 2000 FMPV 95 FMPV 2000
WhiteBlackHispanic
Odds Ratios from Multinomial Logistic Regression: Drinking Variables a
Incidence
Re: No violence
Recurrence
Re: No violence
Male weekly N. of drinks (5)
1.1 (1.0-1.2)* 1.1 (1.0-1.2)
Female 5+ less once a month
1.5 (0.3-7.7) 3.8 (1.3-10.6)*
Female 5+ at least once a month
1.5 (0.1-29.7) .7 (0.1-5.2)
* p<0.05, **p<0.01
a Also controlling for: female and male alc. prob, drug use, childhood victim., parental violence, impulsivity, attit. toward violence, age, income, education, emplm. status, marital status, length of relationship, ethnicity.
Stability and Incidence of MFPV and FMPV 1995 - 2000
39
6
44
6
05
101520253035404550
Stability Incidence
MFPVFMPV
2000 Status of 1995 Non-Violent Couples
94
5 1
94
5 10
102030405060708090
100
No Violence Moderate SevereViolence
MFPVFMPV
2000 Status of 1995 Couples with Moderate Violence
61
36
3
63
31
6
0
10
20
30
40
50
60
70
No Violence Moderate SevereViolence
MFPVFMPV
2000 Status of 1995 Couples with Severe Violence
58
28
14
45
34
21
0
10
20
30
40
50
60
70
No Violence Moderate SevereViolence
MFPVFMPV