Identification and management of domestic violence: a randomized

advertisement
Identification and Management of Domestic Violence
A Randomized Trial
Robert S. Thompson, MD, Frederick P. Rivara, MD, MPH, Diane C. Thompson, MS, William E. Barlow, PhD,
Nancy K. Sugg, MD, MPH, Roland D. Maiuro, PhD, David M. Rubanowice, BS
Background: Diagnosis of domestic violence (DV) in primary care is low compared to its prevalence.
Care for patients is deficient. Over a 1-year period, we tested the effectiveness of an
intensive intervention to improve asking about DV, case finding, and management in
primary care. The intervention included skill training for providers, environmental
orchestration (posters in clinical areas, DV questions on health questionnaires), and
measurement and feedback.
Methods:
We conducted a group-randomized controlled trial in five primary care clinics of a large
health maintenance organization (HMO). Outcomes were assessed at baseline and
follow-up by survey, medical record review, and qualitative means.
Results:
Improved provider self-efficacy, decreased fear of offense and safety concerns, and
increased perceived asking about DV were documented at 9 months, and also at 21 months
(except for perceived asking) after intervention initiation. Documented asking about DV
was increased by 14.3% with a 3.9-fold relative increase at 9 months in intervention clinics
compared to controls. Case finding increased 1.3-fold (95%, confidence interval 0.67–2.7).
Conclusions: The intervention improved documented asking about DV in practice up to 9 months later.
This was mainly because of the routine use of health questionnaires containing DV
questions at physical examination visits and the placement of DV posters in clinical areas.
A small increase in case finding also resulted. System changes appear to be a cost-effective
method to increase DV asking and identification.
Medical Subject Headings (MeSH): domestic violence, primary health care, intervention
studies, patient care team (Am J Prev Med 2000;19(4):253–263) © 2000 American Journal
of Preventive Medicine
Introduction
D
omestic violence (DV) affects up to16% of U.S.
couples per year.1 The spectrum of DV manifestations, as encountered in primary care settings, is broad and includes injury; gastrointestinal,
gynecologic, and somatic symptoms; sexually transmitted diseases; and psychological problems.2–16 Estimated
prevalence of domestic violence (physical and/or sexual abuse) within the last year is between 3.9% and
23%3,17–19 in medical practice.
From the Department of Preventive Care and the Center for Health
Studies, Group Health Cooperative of Puget Sound (RS Thompson),
Harborview Injury Prevention and Research Center, University of
Washington (Rivara, DC Thompson), Center for Health Studies,
Group Health Cooperative of Puget Sound (Barlow, Rubanowice),
Pioneer Square Clinic, University of Washington (Sugg), and Harborview Medical Center, University of Washington (Maiuro), Seattle,
Washington
Address all correspondence and reprint requests to: Robert S.
Thompson, MD, Group Health Cooperative of Puget Sound, Department of Preventive Care, 1730 Minor Avenue, Suite 1600, Seattle, WA
98101-1448. E-mail: Thompson.rs@ghc.org.
The full text of this article is available via AJPM Online at
http://www.elsevier.comlocate/ajpmonline.
Reportedly, only 7% to 25% of DV presenting in
practice is identified,6,20 –23 and queries occur in only
1% to 15% of encounters.3,17,18,24 –26 Interventions are
inadequate in 60% to 90% of cases.17,18,27–29
Barriers to action in primary care have been described for United States, Canadian, and Australian
physicians. These include fear of offending patients, a
sense of futility, and lack of training and time.30 –33
Increased pressure for improved identification and
management of DV22,23,34 –38 has resulted in protocol
development and use in emergency departments,20,21,39 – 46 prenatal care settings,47–54 and some
hospital settings,55,56 as well as mandatory reporting
requirements in some states.57,58 Reports of primary
care efforts are more limited.59 Practice improvements
in most settings have tended to be short-lived.1,46,59
We conducted a group-randomized trial of an intensive intervention directed to primary care practice
teams to improve identification of, and assistance for,
DV victims. We assessed provider knowledge, attitudes,
and beliefs; rates of asking; case finding; and quality of
assistance provided as outcomes.
Am J Prev Med 2000;19(4)
0749-3797/00/$–see front matter
© 2000 American Journal of Preventive Medicine • Published by Elsevier Science Inc.
PII S0749-3797(00)00231-2
253
Figure 1. Experimental design.
Methods
Definition
We defined DV as violence between current or former
intimate partners, or between a parent and an adult child.
Individuals aged ⱖ18, both genders, and heterosexual or
homosexual couples were included. Physical or sexual abuse,
threats of violence, or clear-cut controlling behavior were
included.
Setting
Five primary care clinics from the Group Health Cooperative
(GHC) of Puget Sound, a large HMO, volunteered to participate. The patient population served is demographically
comparable to the metropolitan area, except that GHC
members are more highly educated.
254
Members of adult care teams practicing in the five clinics at
least half-time were targeted in the intervention. Pediatricians, receptionists, and personnel not located in practice
clusters were excluded. Teams included physicians, nurse
practitioners (NPs), physician assistants (PAs), registered
nurses (RNs), licensed practical nurses (LPNs), and medical
assistants (MAs).
Experimental Design
After stratification into two groups based on member characteristics, two of the five clinics were randomly selected to be
intervention clinics while three served as controls (Figure 1).
Provider knowledge, attitudes, and beliefs were surveyed at
baseline (T1), at 9 –10 months (T2), and 21–23 months (T3)
following the initial training sessions. Care was measured by
record reviews at baseline and 9 months.
American Journal of Preventive Medicine, Volume 19, Number 4
Approval for the project was obtained through GHC’s
Institutional Review Board. Informed consent from all providers completing the surveys was subsequently obtained.
The Intervention
The conceptual basis for the intervention was the Precede/
Proceed model for behavior change.60 – 64 It focuses on
changing practitioner predisposing factors (knowledge, attitudes, beliefs, and barriers), enabling factors (environmental
and infrastructure processes supporting the intervention),
and reinforcing factors (the use of feedback and other means
to amplify the process). We assessed barriers at baseline in
this project and have published these results.33 We implemented and sustained specific intervention components from
September 1995 to August 1996.65–70
The intervention, detailed elsewhere,71 began with two
separate, half-day training sessions attended by 84% of all
experimental clinic practice teams. Our DV protocol was based
on the literature up through mid-1996.18,21,37,39 – 44,72–74 It
shares strong similarities to the protocol developed by Ambuel et al.,75 emphasizing screening high-risk patients in
primary care and systems approaches. It is similar to the
Family Violence Prevention Fund (FVPF) protocol,76 which
has been tested in an emergency department setting.77
The training was focused on skill building and empowering
practice teams to verbally ask about DV in high-risk situations
(i.e., an injury)16,71,75; use a general questionnaire with two
DV questions (“Within the last 12 months, have you experienced any uncomfortable touching or forced sexual contact?”
and, “Within the last 12 months, have you been in a relationship in which threats, pushing, grabbing, hitting, kicking,
breaking things, or other hurting was used by someone?”)48
at all physical exam visits; and to become familiar with
other environmental enablers (i.e., cue cards for providers,
wall posters, and DV brochures in restrooms). Questionnaires administered by mail or completed on site were used
for physical exam patients only. The wide range of topics
covered and large number of questions (60) provided
patient security.
Opinion leaders67,68 recruited from clinic staff received
three extra training sessions. On four occasions, a newsletter
titled, “No Excuse,” was circulated to all teams to serve as a
reinforcer.69 –71 Four additional educational sessions on skill
improvement, community resources, and early results, were
conducted for the entire clinic staff. System support (environmental enablers) included posters in waiting areas, cue cards
for providers, and the two questions on physical exam
questionnaires.
In 1993, GHC developed a DV manual. At study initiation,
staff knowledge of the manual was limited. An article on DV
appeared in an internal publication at project initiation. We
have no information suggesting differential use of these
materials. No other system-wide training for DV occurred in
control clinics.
Outcome Measures
Study outcomes were assessed via provider survey and record
review at baseline and subsequently (Figure 1).
Provider Survey
A validated, Likert-scaled provider survey, detailed elsewhere,78 was used to ascertain providers’ knowledge, attitudes, and beliefs (KAB) about DV and to assess barriers. To
develop the final questionnaire, precursors were administered successively to 139 and 246 separate providers not
included in the study. Our final factor analyses resulted in six
separate domains (provider self-efficacy, perceived system
support, blaming the victim, fear of offense, safety concerns,
and perceived frequency of asking about domestic violence)
with Chronbach alphas of .73–.91 for each and an overall
alpha of .88 for the 39-item final scale. In addition, 32
non– domain-associated questions were used, so that the final
questionnaire contained 71 questions.
Outcome variables on the provider survey are described
below. We considered self-efficacy to be the most critical of
the six outcome domain variables. Ultimately, there were
seven questions in this domain. There were four questions in
the system support domain, such as “I feel that MSW personnel can help manage DV patients.” The blaming-the-victim
domain contained seven questions such as, “People are only
victims if they choose to be.” In the fear of offense/role
resistance domain, there were seven questions, such as, “I am
afraid of offending if I ask about DV.” There were eight
questions in the victim/provider safety domain covering
issues of provider safety from batterers and issues of patient
safety. There were six questions in the perceived frequency of
asking domain, such as, “In the last three months, when
seeing someone with injuries, how often have you asked the
patient about the possibility of domestic violence?” Complete
descriptions of all domain questions appear elsewhere.78
Covariates used in our analyses were provider’s graduate
degree, years of experience, DV training outside the project
or reading in the last year, and membership in professional
organizations focused on DV.
KAB information for the six separate domains was assessed
at baseline (T1) and two subsequent occasions (T2, 9 months;
T3, 21 months). Confidential surveys were mailed to all adult
care personnel at the five clinics. Nonrespondents received
three follow-up phone calls and one clinic visit by study
personnel.
Medical Record Review
From the literature on high-risk indicators for DV,16,71 we
selected some as “sentinel diagnoses” for measurement purposes: injuries; chronic pelvic pain to represent somatic
problems; depression to reflect psychological conditions; and
physical examination visits as a time for general review and
questioning about the complete medical history. Individuals
had to be aged ⱖ18 years and continuously enrolled at GHC
during the study period; both genders were sampled except
for chronic pelvic pain.
Patients with at least one clinic visit for one or more of the
sentinel diagnoses over the 12-month baseline sampling
window (T1: July 1993–June 1994) were selected by stratified
random sampling. The sampling fractions were 100% of
patients with visits for chronic pelvic pain (CPP); 10% with
depression visits but no CPP; and a 10% sample of patients
with injury visits but no CPP or depression. In addition, 100
Am J Prev Med 2000;19(4)
255
patients per clinic with physical examination but no other
sentinel diagnosis visits were selected for chart review. This
sampling approach meant that some of the people sampled
had more than one sentinel visit type.
After noting no seasonal trends in the T1 data, an identical
sample for the T2 chart review utilized a 6-month sample
period (1 February 1996 –31 July 1996). For CPP, a less
common diagnosis, we kept the sample period at 12 months.
The final sample totaled 3795 patients at T1 (469 pelvic
pain, 992 depression, 1620 injury, and 714 physical examination) and 3392 at T2 (485 pelvic pain, 969 depression, 1490
injury, and 448 physical examination). We estimated that our
study power was 80% to detect an absolute increase in
questioning of 14% to 16%, depending on risk category.18,79
Chart abstractors, blinded to intervention status, ascertained any mention of possible DV in the records for 12
months after the last sentinel visit. Examples of “any mention”
included the following: “husband threatens her”; “controls
her every move”; “hits her”; “keeps all the money”; or patient
answered “yes” to a DV question on the physical examination
questionnaire.
All records mentioning DV were then independently reviewed by two study physicians (blinded to each other’s review
results) to confirm: if DV were asked about or diagnosed, the
time of occurrence (past history or presently active), and
quality of assistance rendered. A patient was considered to
have been “asked” if a physical examination questionnaire
with the two DV questions on it was completed or if a provider
made an indicatory chart note (i.e., “denies DV,” patient says
“husband hit her,” “was pushed downstairs,” and says husband
“threatens to hurt her”). The quality of the assistance provided to victims was scored on a five-point unweighted scale
indicating whether the patient was asked about DV; DV was
acknowledged; DV was documented; a safety plan was
present; and an offer of resources, referral, and follow-up was
made.71 Initial agreement between physician reviewers
ranged from 0.80 to 0.93 for asking, diagnosis of DV, and
presence of a management plan. Discrepancies were resolved
by consensus.
An outcome variable for case finding was constructed. Case
finding was defined as “the application of tests or screening
procedures amongst patients who are seeking healthcare for
reasons which may be related or unrelated to the reasons for
screening.”80,81 From this definition, the numerator was
designated as the number of cases found in each of the
sentinel diagnosis categories, while the denominator was
designated as the total number of patients sampled with that
sentinel diagnosis at T1 or T2.
Qualitative Information
Study team personnel were in the clinics frequently during
the year-long intervention period to distribute provider questionnaires, to stock and replace DV wall posters and DV
brochures for restrooms, and to conduct and assist with
follow-up training. Through this informal process, some
important observations were forthcoming and are incorporated in results reporting.
Data Analysis
We compared changes over time between intervention and
control clinics for recorded asking about DV. All data forms
256
were coded and double-key entered. The SAS and Stata
statistical packages were used for analysis.82,83
Paired t-tests and analysis of variance (ANOVA) were used
to compare provider survey results. Regression analyses were
conducted using a general linear model85 adjusting for
baseline domain scores and with and without adjusting for
provider-specific confounders such as age, gender, education,
and experience. There was no interaction effect between
selected provider characteristics and the DV training
intervention.
Rates and type of asking (questionnaire or provider notation in the record) about DV were calculated for intervention
and control clinics at baseline and follow-up. Since the unit of
randomization was the clinic, and the analysis was conducted
at the level of the provider teams, or the individual patient,
the effect of within-group correlation was adjusted for by
using a GEE model.84,85 Rate differences and odds ratios
(logistic regression analysis)86 were calculated to show
changes between intervention and control clinics over
time.
Results
Characteristics of Clinics at Baseline
Characteristics of clinic personnel responding
(N⫽179) and clinic membership at baseline are displayed in Table 1. There were 66 physicians, 13 physician assistants, 4 nurse practitioners, 44 nurses, and 52
other members of the health care teams. Over 30% of
team members had been at GHC for ⱖ15 years. Intervention and control groups at baseline did not differ by
gender, length of service, or job type. Provider survey
nonrespondents did not differ from respondents by
clinic, gender, or job type.
Because of the large numbers, there were some
statistically significant differences in the characteristics
of intervention and control clinic members. These were
adjusted for in our analyses.
Provider Survey
There were 208 providers for adults in the five study
clinics at baseline with 179 (86%) responding to the
survey. These clinics serve 84,091 GHC members aged
⬎20 years. Attrition resulted in 190 providers being
employed in the clinics at the 9-month (T2) follow-up
and 171 at the 21-month (T3) follow-up. Using these
denominators, response rates were 79% at T2 and 82%
at T3. For the analysis, we used providers who responded to both T1 and T2 (N⫽128), and T1 and T3
(N⫽102) surveys, and answered at least one question in
the domain being analyzed.
After intervention initiation, four of the six domain
scores improved significantly (self-efficacy, fear of offense, safety concerns, and perceived asking about DV)
from baseline to 9 months in intervention compared to
control clinics, adjusted for baseline (Table 2). For
example, as a result of the intervention, self-efficacy
American Journal of Preventive Medicine, Volume 19, Number 4
Table 1. Characteristics of intervention and control clinics at baseline
Characteristics
Medical care personnela,b
Total
Prescribing practitioners (MDs, PAs, NPs)
Nonprescriptive staff
Gender of providers
Male
Female
Provider years at GHC
1–5
6–10
11–15
⬎15
Number of clinic members
Age distribution of membersc
20–34
35–49
50–64
65–79
80⫹
Number of respondents to routine GHC membership surveys
Race/ethnicityd
White or Caucasian
African American
Asian or Pacific Islander
Native American or Alaskan Native
Other/Hispanice
Educationf
⬍HS
HS graduate
College, no degree
College degree
Some graduate studies
Advanced degree
Total
provider
group
(Nⴝ179)
Intervention clinics
%
n
Control clinics
%
n
p value
2
(␹MH
)
179
83
96
91
36
55
50.8
39.7
60.4
88
47
41
49.2
53.4
46.6
53
126
26
65
28.6
71.4
27
61
30.7
69.3
0.757
42
49
27
61
19
29
12
31
20.9
31.9
13.2
34.1
23
20
15
30
26.1
22.7
17.1
34.1
0.506
21.7
34.9
21.4
16.5
5.5
⬍0.001
39,860
9,940
14,131
8,796
5,484
1,509
0.063
44,213
24.9
35.4
22.1
13.8
3.8
5,108
9,611
15,437
9,430
7,308
2,427
6,656
4,291
174
398
56
266
84.0
3.4
7.8
1.1
5.2
5,773
153
524
69
231
86.7
2.3
7.9
1.0
3.5
⬍0.001
1,058
1,103
2,043
787
51
66
20.7
21.6
40.0
15.4
1.0
1.3
1,693
1,520
2,403
874
93
73
25.5
22.8
36.1
13.1
1.4
1.1
⬍0.001
Survey respondents from target group of medical care personnel. No receptionists, no pediatricians, work ⱖ50% time, work in a practice cluster.
MD, PA, NP vs nonprescriptive staff. Intervention clinics: Physician (MD) ⫽ 29, 31.9%; physician assistant (PA) ⫽ 7, 7.7%; nurse practitioner
(NP) ⫽ 0; registered nurse (RN) ⫽ 26, 28.6%; licensed practical nurse (LPN), medical assistant (MA) LPN/MA ⫽ 29, 31.9%. Control clinics:
MD ⫽ 37, 42%; PA ⫽ 6, 6.8%; NP ⫽ 4, 4.5%; RN ⫽ 18, 20.5%; LPN/MA ⫽ 23, 26.1%.
c
Age distribution of members aged 20 as of June 1995.
d
Race/ethnicity from 1998 GHC membership surveys.
e
Hispanics (N⫽217) fall into the other and Caucasian categories. Our data do not permit perfect separation from these categories. It is estimated
that Hispanics comprise approximately 4% of the population.
f
Education from 1998 GHC randomized mailed membership surveys.
a
b
improved nearly half a point (0.47) on a five-point
Likert scale. We found no significant change in the
intervention effect when adjusting for provider characteristics, nor did provider characteristics modify the
interaction effect. As shown in Table 2, three of the
positive intervention effects (self-efficacy [0.41 Likert
units], fear of offense [0.24 Likert units], and safety
concerns [0.17 Likert units]) remained significant at
the 21–23-month, post-intervention initiation.
At 9 months, 69.8% of providers in the intervention
group were aware of the DV guidelines, an absolute
change of 58.6% from baseline, compared to a change
of 11.2% for control clinic providers. In analyses adjusted for covariates and clustering, the intervention
increased the rating of the guidelines as useful by 0.70
Likert scale units, 95% confidence interval (CI) (0.18,
1.22). Intervention effects on the provider beliefs of not
knowing how to ask (0.70 Likert units, 95% CI [0.43,
0.98]), and not knowing what to do (0.41 Likert units,
95% CI [0.03, 0.78]) were significant and indicated
provider effects of the intervention.
As another way of indicating the magnitude of
change, Figure 2 displays the percentage of providers
with high self-efficacy for each question in that domain
at baseline (T1) and at 9 months (T2) by intervention
or control clinic status. These crude data indicate T2
effects favoring the intervention group for all seven
domain questions. These effects were most proAm J Prev Med 2000;19(4)
257
Table 2. Effects of interventiona on clinic staff over time after baseline (T1) measurement
Domain
Perceived self efficacy
Perceived system support
Blaming the victim
Fear of offending
Safety concerns
Perceived asking about DVg
Baseline
mean
Likert
scoreb
3.03 (I)
2.87 (C)
3.46 (I)
3.15 (C)
3.72 (I)
3.76 (C)
3.80 (I)
3.96 (C)
3.27 (I)
3.18 (C)
1.90 (I)
1.96 (C)
9-month follow-up (T2)
Nc
Parameter
estimatedd
95% CI
21-month follow-up (T3)
p value
Ne
Parameter
estimatedf
⬍0.001
102
0.41
0.20–0.62
95% CI
p value
128
0.47
0.25–0.69
0.0002
115
0.21
⫺0.28–0.71
0.4
96
0.14
⫺0.11–0.39
0.26
126
⫺0.05
⫺0.18–0.09
0.5
102
0.06
⫺0.12–0.23
0.52
126
0.13
0.003–0.26
0.05
102
0.24
0.07–0.40
0.005
118
0.13
0.002–0.26
0.05
99
0.17
0.04–0.31
0.01
88
0.30
0.02–0.58
0.04
75
0.24
⫺0.07–0.54
0.12
a
Adjusted for baseline values and clustering effects.
Mean baseline values in Likert units where 1 ⫽ strongly disagree and 5 ⫽ strongly agree; polarity of some questions was flipped.
Staff present at both baseline (T1) and 9 months post intervention initiation (T2) who answered at least one question in the specific domain
being analyzed.
d
This is the difference in clinic staff responses, expressed in Likert scale units, between the 9 months, post intervention and baseline scores in
the intervention clinics minus the control clinic scores. It summarizes the intervention effect. For example, at the nine month follow-up,
perceived self-efficacy was positively shifted 0.47 Likert scale units on a scale of 5.
e
Staff present at both baseline (T1) and 21 month post intervention (T3) who answered at least one question in the specific domain being
analyzed.
f
This is the difference in clinic staff responses, expressed in Likert scale units, between the 21 month post intervention and baseline scores in
the intervention clinic minus the control clinic scores.
g
Measures how often the staff perceived asking about the possibility of DV. Staff were instructed not to answer this quesiton if they did not
consider this part of their responsibilities or hadn’t seen the specific condition within the last 3 months.
I ⫽ intervention clinic; C ⫽ control clinic; DV, domestic violence; CI, confidence interval
b
c
nounced for access to DV management information,
confidence in referring victims or batterers, and having
strategies for helping DV victims change their situation.
Control group responses to five of the seven domain
questions moved toward lower self-efficacy during the
T1 to T2 period.
Figure 2. Perceived self-efficacy of providers over time
I, intervention; C, control; T1, baseline; T2, 9 months
258
American Journal of Preventive Medicine, Volume 19, Number 4
Table 3. Effect of intervention on medical record documented questioning about domestic violence (DV)
Time 2 (9 months
post-intervention
initiation)
Time 1 (baseline)
Intervention
N/n
Total sample
Female
Male
Asked about DV
Total subject
Female
Male
Method of askingb
Provider notation
Questionnaire
Index diagnosis
Depression
Injury
Pelvic pain
Physical
examination
Control
%
N/n
1590
1089
501
Intervention
%
N/n
2205
1499
706
%
Control
N/n
1372
886
486
Absolute difference:
intervention ⴚ control
at Time 2
%
Adjusted ORa
%
difference
95% CI
OR
95% CI
2020
1354
666
56
53
3
3.5
4.9
0.6
77
71
6
3.5
4.7
0.8
281
195
86
20.5
22.0
17.7
125
108
17
6.2
8.0
2.6
14.3
14.0
15.1
11.9–16.7
0.9–17.1
11.5–18.7
3.9
3.2
11.7
2.5–5.9
2.0–4.9
2.6–51.8
26
31
1.6
1.9
32
46
1.5
7.1
31
256
2.3
18.7
37
92
1.8
4.6
0.5
14.1
⫺0.6–1.4
11.9–16.4
1.1
5.2
0.5–2.2
3.1–8.7
16
13
22
5
4.0
2.0
8.3
1.8
23
18
24
12
3.9
1.8
11.8
2.8
63
67
40
111
15.9
11.2
20.9
59.4
41
42
30
12
7.1
4.7
10.2
4.6
8.8
6.5
10.7
54.8
4.6–13.0
3.6–9.4
4.0–17.5
47.3–62.2
2.4
2.3
3.4
47.1
1.1–5.2
1.0–5.3
1.6–7.6
13.7–162.5
a
Odds ratio (OR) compares intervention clinic changes to changes in control clinics adjusting for baseline rates.
Categories not mutually exclusive; chart review indicated that only 0.5% of patients were asked about DV by both questionnaire and provider.
CI, confidence interval
b
Medical Record Review
mainly to changes in asking among patients who had a
physical examination as the initial sentinel visit or in
the follow-up period. Accordingly, asking rates at physical examination sentinel visits to intervention clinics
were 55% greater (absolute) than at control clinics. In
intervention clinics at T2, DV asking rates at visits for
the other three sentinel visits were 15.9%, 11.2%, and
20.9%, respectively, for depression, injury, and chronic
pelvic pain patients. The relative and absolute rate
changes, mainly because of questionnaire use, were as
follows: depression (increased 2.4-fold, absolute 8.8%),
Intervention impact on care measures over time is
shown in Tables 3 and 4. Recorded asking about DV at
baseline was very low (3.5%) in both intervention and
control clinics overall, and among patients with each
sentinel diagnosis. At follow-up, overall asking in intervention clinics was 20.5%, which was fourfold greater
(odds ratio [OR] 3.9, 95% CI [2.5, 5.9]) than in control
clinics. The absolute increase was 14.3%. No significant
increase in provider initiated asking as written in the
record was seen overall. The positive effects were due
Table 4. Domestic violence case finding and quality of assistance to victims
Time 2 (9 months
post-intervention
initiation)
Time 1 (baseline)
Intervention
n
Case findingb
Total
Female total
Male total
Case finding by diagnosis
Depression
Injury
Pelvic pain
Physical examination
Recorded quality of assistancec
Good or excellent
%
Control
n
%
Intervention
n
%
Absolute difference:
intervention ⴚ control
at Time 2
Control %
%
difference
95% CI
Adjusted ORa
OR
95% CI
n
27
25
2
1.7
2.3
0.4
32
29
3
1.5 37
1.9 35
0.4 2
2.7
4.0
0.4
35
30
5
1.7
1.0
2.2
1.7
0.8 ⫺34
11
5
10
1
2.8
0.8
3.8
0.4
12
7
11
2
2.0 10
0.7 8
5.4 16
0.5 3
2.5
1.3
8.4
1.6
15
8
10
2
2.6
0.9
3.4
0.8
14
66.7
16 64.8 14
63.6
13 61.9
⫺0.08
0.4
5.0
0.8
9.4
⫺1–2.0 1.3
0.2–32.0 1.5
⫺1.2–5.0 0.6
⫺2.1–1.9
⫺0.7–1.5
0.5–9.4
⫺1.3–2.9
0.67–2.69
0.73–3.17
0.05–6.62
0.71 0.22–2.25
1.4 0.31–6.33
3.8
1.1–12.5
2.7 0.13–54.7
⫺27.2–30.6 0.96 0.17–5.4
a
Odds ratio (OR) compared intervention clinic changes to changes in control clinics adjusting for baseline rate.
b
Denominator was all patients sampled. Numerator was the number noted to be a victim of domestic violence (DV).
c
Good or excellent management on a 5-point Likert scale among patients with a need for ongoing assistance as determined by blinded physician
reviewers.
Am J Prev Med 2000;19(4)
259
injury (increased 2.3-fold, absolute 6.5%), and chronic
pelvic pain (increased 3.4-fold, absolute 10.7%). Asking
increased more for male patients than for female,
probably because of the very low rate (⬍1%) of asking
men at baseline.
As shown in Table 4, there was a small, relative (OR
1.3, 95% CI 0.67–2.69) increase in overall case finding
and an absolute increase of 1%. There was a relative
increase of 1.5-fold, with an absolute increase of 1.7%
in women; no effect was observed in men. The overall
case-finding rate in intervention clinics at T2 was 2.7%.
Case finding for sentinel diagnoses (Table 4) showed
positive changes in all except depression; however, only
for patients with chronic pelvic pain was this significant
(OR 3.8, 95% CI 1.1–12.5, with a T2 absolute difference
of 5%).
Quality of management of DV cases was judged to be
good or excellent in the majority of patients at pre- and
post-intervention (Table 4).
Qualitative Information on
Intervention Components
Several important qualitative themes emerged. Wall
posters about DV were a useful environmental resource: Six DV victims self-identified after seeing the
posters. DV brochures placed in clinic restrooms were
taken regularly by patients. Clinicians reported that the
DV questions on health questionnaires used at physical
examination were useful.
Discussion
This effectiveness trial tested an intervention to improve DV identification and care in primary care practice. We documented positive provider KAB outcomes
up to 21 months after program initiation, and process
of care (asking) outcomes at 9 months. In prior work,
effects usually dissipated within 3 to 6 months of
program initiation.21,46,59
Sustained positive impacts on provider self-efficacy,
which generally correlates with action,87,88 were documented as were effects on fear of offense, safety concerns, and perceived asking about DV at 9-month and
21-month follow-ups except for perceived asking. At 21
months the latter result was not maintained; however,
the change pattern was clearly sustained.
In our work, the linkage of provider self-efficacy to
action was limited. While overall case finding increased
by 30% (OR 1.3), this was not a statistically significant
finding. For the subset of patients with chronic pelvic
pain, case finding increased 3.8-fold adjusted (95% CI
1.1, 12.5), a significant finding. Overall asking about
DV increased by 14.3% (absolute), nearly all of which
was because of the use of screening questions on
physical exam questionnaires. Recorded quality of DV
patient assistance did not change.
260
Our results are similar to the only other published
evaluation of a protocol in primary care by Harwell et
al.,59 a pre-post evaluation of routine screening, for
women only, in 12 community health centers in Philadelphia. Screening rates at the 6-month post-training
increased by 20% (absolute). The case finding increase
(3%) was not significant. This crude rate is similar to
the adjusted rate of 1.7% for women in our work.
Campbell et al.77 completed a group-randomized
trial in six hospital emergency departments to test a
systems-change approach to implementing a protocol,
for women only, developed by the FVPF.76 Length of
follow-up and results were quite similar to ours. Staff
knowledge and attitudes, and system-change indicators
(the use of protocols, brochures, posters, and intervention checklists) were positively impacted. There was no
significant change in case finding or the recorded
quality of care provided.
Self-efficacy is postulated to have four components:
(1) vicarious experiences, (2) verbal persuasion (least
important in improving self-efficacy), (3) performance
accomplishment (realizing you can do the action most
important), and (4) tasks that are emotionally draining
(which create a negative feedback loop).87,88 Our results imply that we were not able to completely overcome component four. We feel that we addressed
components one through three quite adequately.
The correlation between what occurs in a medical
encounter and what is recorded in the chart is imperfect. For example, 78% of a random sample of smokers
from GHC (1997) reported that they were counseled
about smoking when they were exit-interviewed; however, counseling was documented in the records of only
48%. This type of phenomenon may have played a role
in our study.
In retrospect, the use of the medical record to
determine the most important study endpoints—asking, case finding, and quality of documentation of
care—is open to question. The literature at the time of
our grant application (1993–1994) provided no solid
basis for power calculations based on primary care
records. This is still the case. In the final analysis, our
study power was limited. The number of clinics (five)
was small, and the total number (131) of active DV
cases from T1 (N⫽64) and T2 (N⫽67) was not large.
However, the case finding rates we describe are comparable to those that may be estimated from the
literature by combining prevalence rates for DV,3,17–19
and screening rates in practice.3,17,18,24 –26,44 – 46,48 –54
Our medical record reviews demonstrated that DV is
frequently a chronic, recurring problem, and difficult
to quantify by episode number and time course. Cessation of DV is particularly difficult to ascertain. In
conclusion, the use of the medical record in DV
outcomes work is difficult.
Could bias be a partial explanation for our results?
Intervention occurred during a period of intense mar-
American Journal of Preventive Medicine, Volume 19, Number 4
ket competition for GHC. Personnel were laid off
shortly after the beginning of the study and providers
felt productivity pressures. Our linked analysis included
personnel who had been in the clinics for the duration,
because we believed that this was the group for whom
intervention effects would be expected. To assess possible bias because of the exclusions used, we compared
characteristics of survey respondents included and excluded from the linked analysis. There were no significant differences for the proportion of prescribing
(MD, PA, and NP) versus nonprescribing staff, years of
experience, attendance at DV workshops, and reading
journal articles on DV. Finally, we compared overall
intragroup (intervention or control) mean total respondent-provider scores at T1 and T2 in a nonlinked
analysis. No significant control clinic differences were
observed for the six domains at T2, while three of six, as
opposed to four of six in the linked analysis, showed
significant changes in the intervention clinics. The
fourth domain, perceived asking, was borderline significant (p⫽0.06, Wilcoxon test). These results indicate
that the provider-selection criteria used for the linked
analysis did not bias the conclusions drawn.
We used the Precede/Proceed planning model60,61,64 to guide intervention development. Our intervention focused on changes in the practice environment to support provider behavior change. We believe
that these environmental enablers were largely responsible for the positive process of care results achieved.
This feature distinguishes our intervention from most
prior work. The inclusion of two screening questions
about DV on physical examination questionnaires was
an enabling factor, which significantly increased
screening for DV. The effectiveness of written screening questions has been reported89; however, the issue
of written versus verbal screening remains controversial.16,90 Placing posters in patient care areas and DV
brochures in restrooms emerged as important intervention components. These results are congruent with our
experiences in orchestrating a wide range of clinical
preventive services from tobacco cessation, to breast
cancer screening, to childhood immunizations.62,63
The intervention was an intense effort and the effects
were modest. Were such modest effects “worth” it? Our
original intention was to estimate the incremental costs
for additional cases found, recurrence rates, and time
to recurrence. Given the small numbers of cases and
the difficulties in assessing when DV ceases, we addressed only the question of whether asking and case
finding could be increased in a sustained fashion.
This study provides the following information to
improve future interventions for DV and their evaluation in health care settings:
1. The use of medical records as a source document
for screening behavior should be questioned. Exit
interviews of patients may hold more promise.
Record review seemed adequate for case documentation, but it likely underestimates the true prevalence of DV and is inadequate to determine the
time course of DV.
Battering, as distinguished from acute physical
assault, is chronic and continuous by nature.91,92
Our record reviews demonstrated that it is difficult
to ascertain DV time course by this means.
2. Environmental enabling factors such as we employed are relatively easy to initiate and are proven
to increase inquiries about DV.
We believe that the findings from our study
should translate into systematic use of screening
questions on periodic health questionnaires in all
core venues where adults are seen, placement and
replacement over time of posters and brochures in
patient care areas, provision of support services for
providers, and assistance to patients.a At this point
in our understanding, these steps should be undertaken because “it’s the right thing to do.” We can
collectively learn from future experience. While
such approaches have been espoused, our empirical experience indicates that they are far from
routine and systemized at present.
3. We need head-to-head comparisons of direct verbal
questioning and screening questionnaires for DV
case finding. Such evaluations should take into
account the impact of these approaches on the
population.93
4. We are early in the process of demonstrating
ongoing effectiveness of various screening approaches in multiple practice settings. We need to
know what works for detection. Beyond this is a
huge void of information on the longitudinal impacts of DV and of its identification and management. What are the costs and effects of intervening?
In conclusion, we are on the verge of breakthroughs for
this very important societal, cultural, legal, public
health, and clinical problem. The use of standardized
definitions for DV/intimate partner violence is indicated for future work.94 In our opinion, work on the
burden of disease in medical practice, and the longitudinal impacts of DV over time are necessary groundwork for accelerating progress. Interventions that are
multifactorial and theory based seem to hold the most
promise. A particular focus on orchestrating environmental enablers, such as questionnaires, seems indicated. Further work on standardizing, delineating, and
communicating attitudinal, process, and health outcomes for the evaluation of intervention programs will
a
A complete training package, including a video, may be obtained
through Lori Fleming at GHC’s Center for Health Promotion (Ph:
206-287-4320).
Am J Prev Med 2000;19(4)
261
facilitate comparisons across levels of care and different
geographic locations. Finally, we believe that the concept of intervention impact (efficacy or effectiveness ⫻
reach)93 should be kept firmly in mind in planning
interventions and evaluating them.
Our project was funded by the Agency for Health Care Policy
& Research (grant #HS07568-03) and The Group Health
Foundation. We are grateful to Eve Adams for manuscript
preparation; the DV project study team, comprised of Barbara
Meyer, Kathy Smith-DiJulio, Madlen Caplow, Ben Givens,
Michael Shafer, Colleen McBride, Lori Fleming, and Delores
Meyer; and the five Group Health Cooperative study clinics
for their dedication and willingness to participate.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
References
1. Straus MA, Gelles RJ. Societal change and change in family violence from
1975 to 1985 as revealed by two national surveys. J Marr Fam 1986;48:465–
79.
2. Talley NJ, Zinsmeister AR, Melton III LJ. Gastrointestinal tract symptoms
and self-reported abuse: a population-based study. Gastroenterology 1994;
107:1040 –9.
3. McCauley J, Kern DE, Kolodner K, et al. The “battering syndrome”:
prevalence and clinical characteristics of domestic violence in primary care
internal medicine practices. Ann Intern Med 1995;123:737– 46.
4. Domino JV, Haber JD. Prior physical and sexual abuse in women with
chronic headache: clinical correlates. Headache 1987;27:310 – 4.
5. Drossman DA, Leserman J, Nachman G, et al. Sexual and physical abuse in
women with functional or organic gastrointestinal disorders. Ann Intern
Med 1990;113:828 –33.
6. Drossman DA, Talley NJ, Leserman J, Olden KW, Barreiro MA. Sexual and
physical abuse and gastrointestinal illness. Ann Intern Med 1995;123:782–
94.
7. Walker E, Unutzer J, Rutter C, et al. Costs of health care use by women
HMO members with a history of childhood abuse and neglect. Arch Gen
Psychiatry 1999;56:609 –13.
8. Post R. A preliminary report on the prevalence of domestic violence among
psychiatric inpatients. Am J Psychiatry 1980;137:974 –5.
9. Stark E, Flitcraft A. Violence among intimates: an epidemiological review.
In: Van Hasselt VB, Morrison RL, Bellack AS, Hersen M, eds. Handbook of
family violence. New York: Plenum Press, 1988:293–317.
10. Schei B, Bakketeig LS. Gynaecological impact of sexual and physical abuse
by spouse: a study of a random sample of Norwegian women. Br J Obstet
Gynaecol 1989;96:1379 – 83.
11. Schei B. Psycho-social factors in pelvic pain: a controlled study of women
living in physically abusive relationships. Acta Obstet Gynecol Scand
1990;69:67–71.
12. Jacobsen A, Richardson S. Assault experiences of 100 psychiatric inpatients:
evidence of the need for routine inquiry. Am J Psychiatry 1988;144:903–13.
13. Longstreth GF, Wolde-Tsadik G. Irritable bowel-type symptoms in HMO
examinees. Dig Dis Sci 1993;38:1581–9.
14. McCauley J, Kern DE, Kolodner K, et al. Clinical characteristics of women
with a history of childhood abuse: unhealed wounds. JAMA 1997;277:
1362– 8.
15. Eby KK, Campbell JC, Sullivan CM, Davidson II WS. Health effects of
experiences of sexual violence for women with abusive partners. Health
Care Women Int 1995;16:563–76.
16. Eisenstat SA, Bancroft L. Primary care: domestic violence. New Engl J Med
1999;341:886 –92 (review article).
17. Gin NE, Rucker L, Frayne S, Cygan R, Hubbell FA. Prevalence of domestic
violence among patients in three ambulatory care internal medicine
clinics. J Gen Intern Med 1991;6:317–22.
18. Hamberger LK, Saunders DG, Hovey M. Prevalence of domestic violence in
community practice and rate of physician inquiry. Fam Med 1992;24:283–7.
19. Gazmararian JA, Lazorick S, Spitz AM, Ballard TJ, Saltzman LE, Marks JS.
Prevalence of violence against pregnant women. JAMA 1996;275:1915–20.
20. McLeer SV, Anwar RAH. A study of battered women presenting in an
emergency department. Am J Public Health 1989;79:65– 6.
21. McLeer SV, Anwar RAH., Herman S, Maquiling K. Education is not
262
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
enough: a systems failure in protecting battered women. Ann Emerg Med
1989;18:651–3.
Randall T. Domestic violence begets other problems of which physicians
must be aware to be effective. JAMA 1990;264:940 – 4.
Randall T. Domestic violence intervention calls for more than treating
injuries. JAMA 1990;264:939 – 40.
Gelles RJ, Straus M. Intimate violence. New York: Simon & Schuster; 1988.
Straus MA, Gelles RJ, Steinmetz SK. Behind closed doors: violence in the
American family. New York: Doubleday, 1980.
Straus MA, Smith C. Family patterns and primary prevention of family
violence. Trends Health Care Law Ethics 1993;8:17–26.
Randall T. AMA, Joint Commission urges physicians to become part of
solution to family violence epidemic. JAMA 1991;266:2524.
Kurz D. Emergency department responses to battered women: resistance to
medicalization. Soc Probl 1987;34:69 – 81.
Warshaw C. Limitations of the medical model in the care of battered
women. Gender Soc 1989;3:506 –17.
Sugg NK, Innui T. Primary care physicians’ response to domestic violence:
opening Pandora’s box. JAMA 1992;267:3157– 60.
Ferris LE, Tudiver F. Family physicians’ approach to wife abuse: a study of
Ontario, Canada, practices. Fam Med 1992;27:276 – 82.
Easteal PW, Easteal S. Attitudes and practices of doctors toward spouse
assault victims: an Australian study. Violence Vict 1992;7:217–28.
Sugg NK, Thompson RS, Thompson DC, Maiuro R, Rivara FP. Domestic
violence and primary care: attitudes, practices and beliefs. Arch Fam Med
1999;8:301– 6.
Flitcraft A. From public health to personal health: violence against women
across the life span. Ann Int Med 1995;123:800 –1.
Alpert EJ. Violence in intimate relationships and the practicing internist:
new “disease” or new agenda? Ann Int Med 1995;123:774 – 81.
Flitcraft A. Learning from the paradoxes of domestic violence. JAMA
1997;277:1400 –1.
AMA Council on Scientific Affairs. Violence against women: relevance for
medical practitioners. JAMA 1992;267:3184 –9.
Flitcraft A. Physicians and domestic violence: challenges for prevention.
Health Aff 1993;12:154 – 61.
Reece RM, Grodin MA. Recognition of nonaccidental injury. Ped Clin
North Am 1985;32:41– 60.
Klingbeil KS, Boyd VD. Emergency room intervention: detection, assessment, and treatment. In: Roberts, AR, ed. Battered women and their
families. New York: Springer Publishing Co., 1985:7–32.
Isaac NE, Sanchez RL. Emergency department response to battered women
in Massachusetts. Ann Emerg Med 1994;23:855– 8.
Snyder JA. Emergency department protocols for domestic violence.
J Emerg Nurs 1994;20:65– 8.
Waller AE, Hohenhaus SM, Shah PJ, Stern EA. Development and validation
of an emergency department screening and referral protocol for victims of
domestic violence. Ann Intern Med 1996;27:754 – 60.
Olsen L, Anctil C, Fullerton L, Brillman J, Arbuckle J, Sklar D. Increasing
emergency physician recognition of domestic violence. Ann Intern Med
1996;27:741– 6.
Tilden VP, Sheperd P. Increasing the rate of identification of battered
women in an emergency department: use of a nursing protocol. Res Nurs
Health 1987;10:209 –15.
Fanslow JL, Norton RN, Robinson EM, Spinola CG. Outcome evaluation of
an emergency department protocol of care on partner abuse. Aust NZ J
Public Health. 1998;22:598 – 603.
McFarlane J, Parker B, Soeken K, Silva C, Reel S. Safety behaviors of abused
women after an intervention during pregnancy. J Obstet Gynecol Neonatal
Nurs 1998;27:64 –9.
McFarlane J, Parker B, Soeken K, Bullock L. Assessing for abuse during
pregnancy: severity and frequency of injuries and associated entry into
prenatal care. JAMA 1992;267:3176 – 8.
Parker B, McFarlane J. Identifying and helping battered pregnant women.
Matern Child Nurs J 1991;16:161– 4.
Parker B, McFarlane J, Soeken K, Silva C, Reel S. Testing an intervention to
prevent further abuse to pregnant women. Res Nurs Health 1999;22:59 –
66.
McFarlane J, Parker B. Preventing abuse during pregnancy: an assessment
and intervention protocol. Matern Child Nurs J 1994;19:321– 4.
McFarlane J, Gondolf E. Preventing abuse. Matern Child Nurs J 1998;23:
22–7.
McFarlane J, Wiist W. Preventing abuse to pregnant women: implementation of a “mentor mother” advocacy model. J Commun Health Nurs
1997;14:237– 49.
American Journal of Preventive Medicine, Volume 19, Number 4
54. Wiist WH, McFarlane J. The effectiveness of an abuse assessment protocol
in public health prenatal clinics. Am J Public Health 1999;89:1217–21.
55. Hadley SM, Short LM, Lezin N, Zook E. WomanKind: an innovative model
of health care response to domestic abuse. Womens Health Issues 1995;5:
189 –98.
56. Short LM, Hadley SM, Bates B. Evaluation of the WomanKind program:
support systems for battered women. Washington, DC: U.S. Department of
Human and Health Services, Office of the Assistant Secretary for Planning
and Evaluation, 1999.
57. Rodriguez MA, Craig AM, Mooney DR, Bauer HM. Patient attitudes about
mandatory reporting of domestic violence: implications for health care
professionals. West J Med 1998;169:337– 41.
58. Rodriguez MA, Bauer HM, McLoughlin E, Grumbach K. Screening intervention for intimate partner abuse: practices and attitudes of primary care
physicians. JAMA 1999;282:468 –74.
59. Harwell TS, Casten RJ, Armstrong KA, Demsey S, Coons HL, Davis M.
Results of a domestic violence training program offered to the staff of
urban community health centers. Am J Prev Med 1998;15:235– 42.
60. Green L, Kreuter M. Application of Precede/Proceed in community
settings. Health promotion planning: an educational and environmental
approach. Mountain View, CA: Mayfield, 1991.
61. Green LW, Eriksen MP, Schor E. Preventive practices by physicians:
behavioral determinants and potential interventions implementing preventive services. Am J Prev Med 1988;4:101–7.
62. Thompson RS, Taplin SH, McAfee TA, Mandelson MT, Smith AE. Primary
and secondary prevention services in clinical practice: twenty years’ experience in development, implementation and evaluation. JAMA 1995;273:
1130 –5.
63. Thompson RS. What have HMOs learned about clinical prevention services: an examination of the experience at Group Health Cooperative of
Puget Sound. Milbank Q 1996;74:469 –509.
64. Green LW, Kreuter MW. Health promotion planning: an educational and
ecological approach. 3rd ed. Mountain View, CA: Mayfield, 1999.
65. Davis D, Haynes RB, Chambers L, Neufield VR, McKibbon A, Tugwell P.
The impact of CME evaluation. Eval Health Prof 1984;7:251– 83.
66. Davis DA, Thomson MA, Oxman AD, Haynes RB. Evidence for the
effectiveness of CME: a review of 50 randomized controlled trials. JAMA
1992;268:1111–7.
67. Davis DA, Thomson MA, Oxman AD, Haynes RB. Changing physician
performance: a systematic review of the effect of continuing medical
education strategies. JAMA 1995;274:700 –5.
68. Thomson MA, Oxman AD, Haynes RB, Davis DA, Freemantle N, Harvey
EL. Local opinion leaders to improve health professional practice and
health care outcomes. In: The Cochrane Library, Issue 2. Oxford: Update
Software, 1998 (Cochrane Review; updated quarterly).
69. Thomson MA, Oxman AD, Davis DA, Haynes RB, Freemantle N, Harvey
EL. Audit and feedback to improve health professional practice and health
care outcomes. In: The Cochrane Library, Issue 2. Oxford: Update
Software, 1998 (Cochrane Review; updated quarterly).
70. Thomson MA, Oxman AD, Davis DA, Haynes RB, Freemantle N, Harvey
EL. Audit and feedback to improve health professional practice and health
care outcomes (part II). In: The Cochrane Library, Issue 2. Oxford: Update
Software, 1998 (Cochrane Review; updated quarterly).
71. Thompson RS, Meyer BA, Smith-DiJulio K, et al. A training program to
improve domestic violence identification and management in primary
care: preliminary results. Violence Vict 1998;13:395– 410.
72. Klingbeil KS, Boyd VD. Harborview Medex training protocol; Harborview
social work emergency protocol. Seattle, WA: University of Washington,
Harborview Medical Center: Psychiatry and Behavioral Sciences, 1988.
73. Sassetti MR. Battered women. In: Hendricks-Matthews M, ed. Violence
education: toward a solution. Kansas City, MO: Society of Teachers of
Family Medicine, 1992:31–54.
74. Massachusetts Medical Society and Ad Hoc Committee on Domestic
Violence. Partner violence: how to recognize and treat victims of abuse.
Waltham, MA: The Massachusetts Medical Society, 1992.
75. Ambuel B, Hamberger LK, Lahti J. Partner violence: a systematic approach
to identification and intervention in outpatient health care. Wisconsin Med
J 1996;95:292–7.
76. Warshaw C, Ganley AL. Improving the health care response to domestic
violence: a resource manual for health care providers. San Francisco, CA:
Family Violence Prevention Fund, 1997.
77. Campbell JC, Coben JH, McLoughlin E, et al. An evaluation of a system
change training model to improve emergency department response to
battered women. Acad Emerg Med 2000. In press.
78. Maiuro RD, Vitaliano PP, Sugg N, et al. The development of a health care
provider survey for domestic violence: psychometric properties. Am J Prev
Med 2000;19:245–52.
79. Friedman LS, Samet JH, Robert MS, Hudlin M, Hans P. Inquiry about
victimization experiences: a survey of patient preferences and physician
practices. Arch Intern Med 1992;152:1186 –90.
80. Sackett DL, Holland WW. Controversy in the detection of disease.
J Chronic Dis 1975; 33:51.
81. Abramson JH. Cross-sectional studies. In: Detels R, Holland WW, McEwen
J, Omenn GS, eds. Oxford textbook of public health. 3rd ed., vol. 2.
London: Oxford University Press, 1997:517–38.
82. SAS Institute Inc. SAS, release 6.12. Cary, NC: SAS Institute Inc., 1997.
83. Stata Corporation. Stata Statistical Software, Release 5.0. College Station,
TX: Stata Corporation, 1997.
84. Murray DM. Design and analysis of group-randomized trials. New York:
Oxford University Press, 1998.
85. Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear
models. Biometrika 1986;73:13–22.
86. Hosmer Jr DW, Lemeshow S. Applied logistic regression. New York: John
Wiley & Sons, 1989.
87. Bandura A. Self-efficacy in changing societies. In: Bandura A, ed. Exercise
of personal and collective efficacy in changing societies. New York:
Cambridge University Press, 1995:1– 45.
88. Bandura A. Self-efficacy: the exercise of control. New York: Freeman & Co.,
1997.
89. Furbee PM, Sikora R, Williams JM, Derk SJ. Comparison of domestic
violence screening methods: a pilot study. Ann Emerg Med 1998;31:495–
501.
90. McFarlane J, Christoffel K, Bateman L, Miller V, Bullock L. Assessing for
abuse: self-report versus nurse interview. Public Health Nurs 1991;8:245–
50.
91. Weiss E. Surviving domestic violence: voices of women who broke free.
Sandy, UT: Agreka Books, 2000.
92. Coker Al, Smith PH, McKeown RE, King MJ. Frequency and correlates of
intimate partner violence by type: physical, sexual and psychological
battering. Am J Public Health 2000;90:553– 65.
93. Glasgow RE, Vogt TM, Boles SM. Evaluating the public health impact of
health promotion interventions: the RE-AIM framework. Am J Public
Health 1999;89:1322–7.
94. Saltzman LE, Fanslow JL, McMahon PM, Shelley GA. Intimate partner
violence surveillance: uniform definitions and recommended data elements, version 1.0. Atlanta, GA: National Center for Injury Prevention and
Control, Centers for Disease Control and Prevention, 1999.
Am J Prev Med 2000;19(4)
263
Download