Context Approximately 1.5 million US women experience intimate partner violence
annually. Approximately 20% of these women obtain civil protection orders,
but the effectiveness of such orders in preventing future violence is unclear.
Objective To assess associations between obtaining a protection order and risk
of subsequent police-reported intimate partner violence.
Design, Setting, and Subjects Retrospective cohort study of 2691 adult female residents of Seattle,
Wash, with an incident of male intimate partner violence reported to the Seattle
Police Department between August 1, 1998, and December 31, 1999.
Main Outcome Measure Relative risk (RR) of police-reported physical and psychological abuse
in the 12 months following the index incident according to protection order
status (temporary protection order, usually in effect for 2 weeks; permanent
protection order, usually in effect for 12 months; or no protection order).
Results Overall rates of police-reported physical and psychological abuse in
the 12 months of follow-up were 13.5 per 100 person-years and 12.3 per 100
person-years, respectively. After controlling for cohabitation at time of
index incident and index incident offense type, women with temporary protection
orders in effect were more likely than women without protection orders to
be psychologically abused (RR in the first 6 months after the index incident,
4.0; 95% confidence interval [CI], 2.2-7.2; RR in the entire 12 months after
the index incident, 4.9; 95% CI, 2.8-8.6), while women with permanent protection
orders in effect were less likely than those without orders to be physically
abused (RR in the first 6 months, 0.4; 95% CI, 0.1-1.1; RR in the entire 12
months, 0.2; 95% CI, 0.1-0.8).
Conclusions Permanent, but not temporary, protection orders are associated with
a significant decrease in risk of police-reported violence against women by
their male intimate partners.
Intimate partner violence (IPV) is a frequent occurrence in the United
States, with nearly 5 million physical or sexual assaults by intimate partners
experienced by approximately 1.5 million women annually.1
In addition to injury-related visits, abused women have high frequencies of
emergency department visits and hospitalizations for somatic and psychiatric
diagnoses related to stress, including functional gastrointestinal disorders,
loss of appetite, chest pain, headaches, anxiety, insomnia, alcohol abuse
or dependence, post-traumatic stress disorder, depression, and suicide attempts.2-5
Several strategies can be used by abused women in an attempt to deter
future violence, but limited financial and community resources such as battered
women's shelters may restrict women's options. One widely available option
is to obtain a civil protection order, a legally binding court order that
restrains an individual who has committed an act of violence against a person
from further acts against that person.6 Specifically,
a protection order can prohibit the abuser from committing acts of violence;
exclude the abuser from the residence shared by the petitioner and abuser;
prohibit the abuser from harassing or contacting the petitioner by mail, telephone,
or in person; award temporary custody of minor children; establish temporary
visitation and restrain the abuser from interfering with custody; prohibit
the abuser from removing the children from the jurisdiction of the court;
and order the abuser to participate in treatment or counseling. Although approximately
20% of US women experiencing IPV obtain civil protection orders, their effectiveness
in preventing IPV recurrence is unclear, and it has been suggested that they
may in fact aggravate violence under certain conditions.7-10
While many case series have described the experiences of women with protection
orders, only 1 published study has investigated protection order effectiveness
by comparing abused women with and without protection orders. This interview-based
study reported that violence frequency was not significantly affected by the
presence of a civil restraining order, but the study's small size, low response
rate, and short follow-up period limit this interpretation.11
The current study addresses this issue using linked data from a large population
of women in an entire US city on whom criminal justice system information
was available.
In this retrospective cohort study, subjects were all 2691 female residents
of Seattle, Wash, who had a police-reported episode of IPV inflicted by a
male former or current intimate partner between August 1, 1998, and December
31, 1999, and who had not obtained a permanent protection order in the prior
12 months. We obtained names of abused women from the Seattle Police Department
Domestic Violence Unit database of all IPV incident reports and ascertained
protection order status using information from the King County, Washington
District and Superior Court records of filings for civil protection order.
In King County, women seeking protection orders petition (free of charge)
first for a temporary protection order, which is granted by a judge or commissioner
for a period of 2 weeks. During these 2 weeks the abuser is served with both
the petition and the temporary order, with notice of the date set for a hearing
(approximately 2 weeks after the initial petition), at which time the court
grants or denies a "permanent" protection order effective for 1 year or more.
Approximately 57% of women in King County who file temporary protection orders
against male intimate partners go on to obtain permanent orders. Anytime prior
to the expiration date of a permanent protection order, the petitioner may
return to court to request that the order be modified or terminated. Our study
protocols were approved by the University of Washington Human Subjects Review
Committee and the Washington State Department of Health Human Research Review
Board.
The primary outcome in these analyses, subsequent police-reported abuse
of a study subject by the same abuser, was ascertained from police-reported
incidents of IPV during the 12 months following the initial police-reported
incident (the index incident). Using police reports, we categorized subsequent
IPV incidents as those including physical abuse (assault, reckless endangerment,
or unlawful imprisonment) and those including psychological abuse (harassment,
menacing, stalking, threats, disturbance, criminal trespass, custodial interference,
interfering with IPV reporting, or property damage). Incidents in which the
sole offense was a protection order violation were not included in our analyses.
Additionally, we used Washington State Vital Statistics data to ascertain
deaths during follow-up.
Demographic differences between women who obtained a temporary protection
order at any time in the 12 months following the index incident (without a
subsequent permanent order), women who obtained a permanent protection order
at some time during the 12 months of follow-up, and women who obtained neither
type of order at any time during the follow-up were assessed using χ2 tests, with P<.05 denoting significance.
The relative risk (RR) of subsequent police-reported IPV according to
protection order status was estimated using Cox proportional hazards regression.12 Time to abuse (defined as a police-reported IPV incident
during follow-up) was modeled as a function of time from entry into the cohort
(the date of the index IPV incident). In all models, protection order status
was modeled as a time-dependent variable, allowing subjects to change exposure
categories as protection orders were initiated or terminated. Temporary protection
orders were usually granted for a 2-week period, and permanent protection
orders were for a 12-month period. The time from the filing of a temporary
protection order until the order's typical automatic expiration 2 weeks later
(or until the temporary protection order was rescinded if that came first)
was counted as temporary protection order–exposed time. The time from
the filing of a permanent protection order until the end of that woman's follow-up
period (or until the permanent protection order was rescinded if that came
first) was counted as permanent protection order–exposed time. Time
during which a woman had neither a temporary protection order nor a permanent
protection order in effect was counted as unexposed time. Two comparisons
were made: (1) temporary protection order compared with no protection order,
and (2) permanent protection order compared with no protection order. We allowed
multiple incidents per subject, adjusting the SEs for dependencies between
incident times, using Stata statistical software for all analyses.13 We calculated 2 sets of risk estimates of subsequent
IPV according to protection order status: (1) for the first 6 months of follow-up
after the index incident, and (2) for the entire 12 months of follow-up.
In multivariate models of the effect of protection orders on subsequent
physical and psychological IPV, we considered as potential confounders the
following baseline covariates that we previously found to be related to obtaining
a protection order: subject age, pregnancy status, and alcohol and other drug
use; abuser age and alcohol and other drug use; subject/abuser relationship,
cohabitation at time of index incident; number of police-reported IPV incidents
in the previous 12 months; and the type of offense at the index IPV incident
(threat, weapon threat, physical assault, assault with weapon, sexual assault,
injury).14 Covariates were entered into the
regression models if they changed the risk estimates by 10% or more15; only cohabitation at time of index incident and
index incident offense type met this criterion. We tested for the interaction
of physical abuse during a temporary protection order and the effect of a
permanent protection order using the likelihood ratio test; no significant
(P <.05) interaction was noted.
Study subjects who did not obtain any protection orders, those who obtained
only temporary protection orders, and those who obtained permanent protection
orders during the 12 months following the index incident were similar in terms
of age, pregnancy status, and IPV history with their abusers; but subjects
who did not obtain protection orders during the follow-up were significantly
more likely than other subjects to have used alcohol or other drugs at the
index incident, as were their abusers (Table 1). Additionally, subjects who did not obtain protection orders
were less likely than other women to have ever been married to their abusers
and more likely to be living with them at the time of the index incident.
The use of a weapon in the index incident did not differ by protection order
status, but subjects who did not obtain protection orders were more likely
than other women to have been assaulted or injured during the index incident.
In the first 6 months of follow-up there were 222 incidents of police-reported
physical abuse (16.5 incidents per 100 person-years) and 223 incidents of
police-reported psychological abuse (16.6 incidents per 100 person-years).
Over the entire 12 months of follow-up, there were 363 incidents of police-reported
physical abuse (13.5 per 100 person-years) and 330 incidents of police-reported
psychological abuse (12.3 per 100 person-years).
In the first 6 months of follow-up, the rate of police-reported physical
abuse during times in which no protection order was in effect was 17.2 per
100 person-years, and the corresponding rates for temporary protection order–exposed
time and permanent protection order–exposed time were 14.7 per 100 person-years
and 5.4 per 100 person-years, respectively (Table 2). In time-dependent Cox proportional hazards regression
models controlling for cohabitation and index incident offense type, the RR
of police-reported physical abuse during the first 6 months after the index
incident associated with a temporary protection order was 0.8 (95% confidence
interval [CI], 0.2-3.4), and the risk associated with a permanent protection
order was 0.4 (95% CI, 0.1-1.1) compared with no protection order. In the
first 6 months of follow-up, the rate of police-reported psychological abuse
during times in which no protection order was in effect was 16.0 per 100 person-years,
and the corresponding rates for temporary protection order–exposed time
and permanent protection order–exposed time were 95.6 per 100 person-years
and 16.2 per 100 person-years, respectively. The RR of police-reported psychological
abuse during the first 6 months after the index incident associated with a
temporary protection order was 4.0 (95% CI, 2.2-7.2), and the risk associated
with a permanent protection order was 1.1 (95% CI, 0.5-2.3), compared with
no protection order.
Over the entire 12 months of follow-up, the rate of police-reported
physical abuse during times in which no protection order was in effect was
14.0 per 100 person-years, and the corresponding rates for temporary protection
order–exposed time and permanent protection order–exposed time
were 24.7 per 100 person-years and 2.9 per 100 person-years, respectively
(Table 3). In time-dependent Cox
proportional hazards regression models controlling for cohabitation and index
incident offense type, the RR of police-reported physical abuse during the
12 months after the index incident associated with a temporary protection
order was 1.6 (95% CI, 0.6-4.4), and the risk associated with a permanent
protection order was 0.2 (95% CI, 0.1-0.8) compared with no protection order.
In the entire 12 months of follow-up, the rate of police-reported psychological
abuse during times in which no protection order was in effect was 11.8 per
100 person-years, and the corresponding rates for temporary protection order–exposed
time and permanent protection order–exposed time were 104.9 per 100
person-years and 10.2 per 100 person-years, respectively. The RR of police-reported
psychological abuse during the entire 12 months after the index incident associated
with a temporary protection order was 4.9 (95% CI, 2.8-8.6), and the risk
associated with a permanent protection order was 0.9 (95% CI, 0.5-1.7) compared
with no protection order.
To address the possibility that our results were unduly influenced by
repeated recurrences among a small group of women, we conducted a series of
sub-analyses limited to at most 1 failure per subject. While the risk for
psychological abuse associated with temporary protection order exposure moderated
somewhat in these analyses (RR for 12 months of follow-up, 4.4; 95% CI, 2.3-8.2),
all other findings remained essentially the same.
There were 5 deaths from homicide in the study cohort, for a rate of
1.9 per 1000 person-years. Homicide mortality rates did not differ significantly
by protection order status.
In this population-based cohort of all women with an incident of IPV
reported to Seattle police, the overall rates of police-reported recurrence
of physical and psychological abuse in the following 12 months were 14 per
100 person-years and 12 per 100 person-years, respectively. These numbers
are substantially lower than those reported in studies using convenience samples
of women obtaining protection orders, which have generally found that one
third to one half of abused women self-report physical abuse in follow-up
periods ranging from 4 months to 1 year, and approximately half report psychological
abuse.16-19
The frequency of IPV recurrence we found was more in accord with results from
a study by Carlson et al,20 also based on police
records, in which 23% of women with protection orders reported physical violence
to the police in 2 years of follow-up.
Our findings also differ from those of the only other published study
to compare outcomes for women with and without protection orders. In an interview-based
study of abused women who participated in a family violence demonstration
program, Grau et al11 reported that the likelihood
of any abuse or violence in 4 months of observation did not differ significantly
by protection order status. The study provided follow-up violence information
on only 170 of the 270 participants, however, raising concerns about the adequacy
of the study's power and its internal validity.
In the current study we used police reports to ascertain IPV recurrence
to eliminate any bias that may result from the possibly atypical nature of
the subset of abused women who will agree to participate in an interview-based
study. However, in so doing we captured only that portion of IPV that was
police-reported, which crime victim surveys have estimated to be approximately
50% of IPV incidents.21 Use of police-reported
IPV to represent all abuse implicitly assumes that women with and without
protection orders are equally likely to report violence to police if it occurs
and that police are equally likely to respond. Because Seattle police respond
to all calls, and because our data were based on the incident reports completed
after all responses, we have no reason to believe that police response or
recording depended on protection order status. If the completeness of reporting
of IPV incidents to police varied by protection order status, our results
may be biased. While we have no information to indicate that differential
reporting of IPV existed, we can postulate that because they had taken a formal
legal step to acknowledge IPV against them and request its cessation, it is
possible that women with protection orders were more likely to report new
abuse to police. Reporting differences may have been a factor particularly
in incidents involving psychological abuse, for which temporary protection
orders were associated with a quadrupled risk in our study. We found that
the primary psychological abuse offense was more likely to be harassment in
incidents involving women with current protection orders than it was in incidents
involving women without current orders. If psychological abuse was relatively
overreported by women during temporary protection order–exposed times,
our results would be an overestimate of the adverse effect of temporary protection
order on this type of abuse. On the other hand, a temporary protection order
might have restrained the abuser from inflicting physical abuse, with a consequent
increase in psychological abuse. Our finding of a quadrupling of psychological
abuse risk during the time of a temporary protection order indicates that
the time shortly after the index incident, when most temporary protection
orders are issued, may be one of exceptional volatility between the subject
and her abuser. However, that we did not find a parallel increase in risk
of physical abuse with temporary protection order exposure provides some evidence
that prior concerns of increased violence associated with protection order
filing may be unfounded.
We had no contact with the subjects in this study; therefore, it is
possible that, unknown to us, some subjects moved out of the Seattle area
during the 12 months following their index incidents and we were unable to
ascertain their IPV recurrence. We do not know how likely this out-migration
was, but information we collected for another purpose may help us to estimate
the potential magnitude. In our recent interview-based study of protection
order effectiveness among Seattle women with police- or court-reported IPV,
we found that we were able to retain in the study for 12 months 83% of participants
with a protection order, and 74% of participants without a protection order.22 If the participants who were lost to follow-up in
that study left the Seattle area, the difference in out-migration by protection
order status would indicate that our results are an overestimate of the adverse
effects and an underestimate of the beneficial effects of protection orders.
However, given the relatively small numbers, it seems likely that any effect
would be small. A related possibility is that subjects were not exposed to
the potential for recurrence of violence because of their abusers' incarceration
related to the index incident. Because only about 5% of reported IPV results
in conviction and incarceration in King County, we think this is not likely
to be an important factor.
Police data on IPV incidents provide a limited number of demographic
or explanatory variables; therefore, the possibility of incomplete control
for confounding exists. In our analyses we examined as potential confounders
several variables that we previously found to be associated with obtaining
a protection order14; only 2 of these were
confounders of the associations between protection order status and IPV risk.
The lack of confounding may have resulted from our analytic method, which
used a time-dependent exposure variable, allowing subjects to contribute observation
time in each category they experienced during follow-up. For instance, a woman
who obtained a temporary order 2 weeks after the index incident and then a
permanent order 2 weeks later contributed 0.5 person-months of unexposed time,
0.5 person-months of temporary protection order–exposed time, and 11
person-months of permanent protection order–exposed time; and one who
obtained a temporary order 2 weeks after the index incident and then no permanent
order contributed 0.5 person-months of temporary protection order–exposed
time and 11.5 person-months of time without protection order exposure. Therefore,
some potentially confounding variables, such as study subjects' personal characteristics,
may also have been distributed across protection order and nonprotection order
exposure categories.
In this study we found that having a permanent protection order in effect
was associated with a statistically significant 80% reduction in police-reported
physical violence in the 12 months after an IPV incident. We controlled in
our analyses for all variables that we found to be associated with a woman's
likelihood of obtaining a civil protection order as well as the likelihood
of future violence, but we may not have captured important characteristics
that reflect a woman's motivation and ability to initiate and complete the
process of obtaining a protection order as well as her resolve not to be abused
further. Further comparative studies of abused women with and without protection
orders that ascertain the determinants of the decision whether to seek an
order, other concurrent steps taken to prevent violence recurrence, and women's
opinions of the reasons for violence cessation or recurrence may help explain
how to enhance the protective impact of civil protection orders.
1.Tjaden P, Thoennes N. Full Report of the Prevalence, Incidence, and Consequences
of Violence Against Women. Washington, DC: US Dept of Justice; 2000.
2.Kernic MA, Wolf ME, Holt VL. Rates and relative risk of hospital admission among women in violent
intimate partner relationships, King County, Washington.
Am J Public Health.2000;90:1416-1420.Google Scholar 3.McCauley J, Kern DE, Kolodner K, Derogatis LR, Bass EB. Relation of low-severity violence to women's health.
J Gen Intern Med.1998;13:687-691.Google Scholar 4.Campbell JC, Lewandowski LA. Mental and physical health effects of intimate partner violence on
women and children.
Psychiatr Clin North Am.1997;20:353-374.Google Scholar 5.Resnick HS, Acierno R, Kilpatrick DG. Health impact of interpersonal violence, 2: medical and mental health
outcomes.
Behav Med.1997;23:65-78.Google Scholar 6.Finn P, Colson S. Civil Protection Orders: Legislation, Current Court
Practice, and Enforcement. Washington, DC: US Dept of Justice; 1990.
7.Tjaden P, Thoennes N. Extent, Nature, and Consequences of Intimate Partner
Violence. Washington, DC: National Institute of Justice; 2000.
8.Glick B, Johnson S, Pham C. 1998 Oregon Domestic Violence Needs Assessment: A
Report to the Oregon Governor's Council on Domestic Violence. Portland: Oregon Health Division and Multnomah County Health Dept;
1999.
9. Use of medical care, police assistance, and restraining orders by women
reporting intimate partner violence—Massachusetts, 1996-1997.
MMWR Morb Mortal Wkly Rep.2000;49:485-488.Google Scholar 10.Crowell NA, Burgess AW. Understanding Violence Against Women. Washington, DC: National Academy Press; 1996.
11.Grau J, Fagan J, Wexler S. Restraining orders for battered women: issues of access and efficacy. In: Schweber C, Feinman C, eds. Criminal Justice
Politics and Women: The Aftermath of Legally Mandated Change. New York,
NY: Haworth Press Inc; 1985:13-28.
12.Cox DR. Regression models and life tables.
J R Stat Soc Serv B.1972;34:187-202.Google Scholar 13.StataCorp. Stata Statistical Software: Release 6.0. College Station, Tex: Stata Corp; 1999.
14.Wolf ME, Holt VL, Kernic MA, Rivara FP. Who gets protection orders for intimate partner violence?
Am J Prev Med.2000;19:286-291.Google Scholar 15.Maldonado G, Greenland S. Simulation study of confounder selection strategies.
Am J Epidemiol.1993;138:923-936.Google Scholar 16.Chaudhuri M, Daly K. Do restraining orders help? battered women's experience with male violence
and legal process. In: Buzawa ES, Buzawa CG, eds. Domestic Violence:
The Changing Criminal Justice Response. Westport, Conn: Auburn House;
1992:227-252.
17.Klein AR. Re-abuse in a population of court-restrained male batterers after two
years: development of a predictive model. In: Buzawa ES, Buzawa CG, eds. Do Arrests and Restraining
Orders Work? Thousand Oaks, Calif: SAGE Publications; 1996:192-213.
18.Harrell A, Smith B, Newmark L. Court Processing and the Effects of Restraining Orders
for Domestic Violence Victims. Washington, DC: The Urban Institute; 1993.
19.Horton AL, Simonidis KM, Simonidis LL. Legal remedies for spousal abuse: victim characteristics, expectations,
and satisfaction.
J Fam Viol.1987;2:265-279.Google Scholar 20.Carlson MJ, Harris SD, Holden GW. Protective orders and domestic violence: risk factors for re-abuse.
J Fam Viol.1999;14:205-226.Google Scholar 21.Bachman R, Saltzman LE. Violence Against Women: Estimates From the Redesigned
Survey. Washington, DC: US Dept of Justice; 1995.
22.Wolf ME, Kernic MA, Holt, VL, Rivara FP. Do protection orders work?
Am J Epidemiol.2000;151(suppl):S53.Google Scholar