Domestic Violence Against Women Systematic Review of Prevalence Studies


Samia Alhabib & Ula Nur & Roger Jones
Published online: 15 December 2009
# Springer Science+Business Media, LLC 2009
Abstract To systematically review the worldwide evidence
on the prevalence of domestic violence against women, to
evaluate the quality of studies, and to account for variation
in prevalence between studies, using consistent definitions
and explicit, rigorous methods. Systematic review of
prevalence studies on domestic violence against women.
Literature searches of 6 databases were undertaken for the
period 1995 to 2006. Medline, Embase, Cinahl, ASSIA,
ISI, and International Bibliography of the Social Sciences
were searched, supplemented by hand searching of the
reference lists from studies retrieved and specialized
interdisciplinary journals on violence. A total of 134 studies
in English on the prevalence of domestic violence against
women, including women aged 18 to 65 years, but
excluding women with specific disabilities or diseases,
containing primary, empirical research data, were included
in the systematic review. Studies were scored on eight predetermined
criteria and stratified according to the total
quality score. The majority of the sudies were conducted in
North America (41%), followed by Europe (20%). 56% of
studies were population-based, and 17% were carried out
either in primary or community health care settings. There
was considerable heterogeneity both between and within
geographical locations, health care settings, and study
quality The prevalence of lifetime domestic violence varies
from 1.9% in Washington, US, to 70% in Hispanic Latinas
in Southeast US. Only 12% scored a maximum of 8 on
our quality criteria, with 27% studies scored 7, and 17%
scored 6. The mean lifetime prevalence of all types of
violence was found to be highest in studies conducted in
psychiatric and obstetric/gynecology clinics. Results of this
review emphasize that violence against women has reached
epidemic proportions in many societies. Accurate measurement
of the prevalence of domestic violence remains problematic
and further culturally sensitive research is required to
develop more effective preventive policies and programs.
Keywords Domestic violence .Women . Prevalence .
Violence against women includes all verbal, physical, and
sexual assaults which violate a woman’s physical body,
sense of self and sense of trust, regardless of age, race,
ethinicity, or country (Campbell 1995). Violence against
women has been identified as a major public health and
human rights issue (Joachim 2000), and has been estimated
by the World Health Organization (WHO) to account for
between 5–20% of healthy years of life lost in women aged
15 to 44 (WHO 1997).
Twenty years ago, violence against women was not
considered an issue worthy of international attention or
concern. This began to change in the 1980s, as women’s
groups were organized locally and internationally to
S. Alhabib (*)
Academic Unit of Primary Health Care, University of Bristol,
25 Belgrave Road,
Bristol BS8 2AA, UK
U. Nur
Cancer Statistics-Cancer Research UK,
London School of Hygiene and Tropical Medicine,
London, UK
R. Jones
Department of General Practice & Primary Care,
King’s College London,
London, UK
J Fam Viol (2010) 25:369–382
DOI 10.1007/s10896-009-9298-4
demand attention to the physical, psychological, and
economic abuse of women. Gradually, violence against
women has come to be recognized as a legitimate human
rights issue and a significant threat to women’s health and
well being (Ellsberg and Heise 2005). The process began in
Europe and North America, but even in the United States,
where this trend was most apparent, it took 20 years for
rising awareness to lead to legislation and to potentially
effective preventive measures. Only in the early 1990s were
comprehensive laws enforced and effective resources
allocated to deal with gender violence (Gelles 1997).
Worldwide, domestic violence is as serious a cause of
death and incapacity among women aged 15–49 years as
cancer, and a greater cause of ill health than traffic
accidents and malaria combined (The World Bank 1993).
In addition to causing injury, violence increases women’s
long-term risks of a number of other health problems,
including chronic pain, physical disability, drug and alcohol
abuse, and depression (Heise et al. 1999). Secondary to the
biopsychosocial effects of battering are the high costs of
such violence. Abused women have more than double the
number of medical visits, an 8-fold greater mental healthcare
usage, and an increased hospitalization rate compared
to non-abused women (Wisner et al. 1999). The WHO
multi-country study on women’s health and domestic
violence has recently confirmed significant associations
between lifetime experiences of partner violence and self
reported poor health (Ellsberg et al. 2008).
Prevalence studies of violence against women report
wide variations in levels of violence within and between
health care settings. The reported lifetime prevalence of
physical or sexual violence, or both, varied from 15% to
71% among the countries studied in the WHO multi-country
study (Garcia-Moreno et al. 2006). Few studies have used
standard methods to derive comparative prevalence figures.
The World-Safe initiative represents a successful model that
has been used in five countries (Brazil, Chile, Egypt,
Philippines, and India) to study intimate partner violence
against women and children (Sadowski et al. 2004). The
WHO multi-country study uses another model, which has
been applied in 10 different countries. While confirming that
physical and sexual partner violence against women is
widespread, the variation in prevalence within and between
study settings emphasizes that this violence is not inevitable,
and needs to be addressed.
Over the last 10 years, a number of prevalence surveys on
intimate partner violence has been published from around
the world. However, despite a number of initiatives, such as
the European Network on Conflict, Gender, and Violence, the
launching of a European Society of Criminology and efforts to
develop an international survey on violence against women
(Hagemann-White 2001), information from these studies has
not been systematically collated and analyzed. The aim of
this systematic review is to systematically summarize the
worldwide evidence on the prevalence of domestic violence
against women, to evaluate the quality of studies, and to try
to account for variation in prevalence rates between studies.
Literature Searches
Parallel literature searches of 6 databases (Medline, Embase,
Cinahl, ASSIA, ISI, and International Bibliography of the
Social sciences) were undertaken for the period1995–2006.
The reference lists from retrieved studies and specilaized
interdisciplinary journals in violence (Violence Against
Women, Journal of Interpersonal Violence) were hand
searched to look for further studies that might not have
been retrieved by the database searches. Authors of
unpublished studies, e.g., PhD theses, were contacted to
obtain copies of their studies. We contacted experts in the
field before and during the process to obtain feedback and
advice with regard to methodology and analysis. All
citations were exported into Reference Manager software
(version 11). Searches included MeSH and text words terms,
with combinations AND OR Boolean operator (Box 1).
Box 1: words used in the search
1. Domestic violence. 13. Frequency.
2. Spouse abuse. 14. Prevalenc$.tw.
3. Battered women. 15. Incidenc$.tw.
4. Partner abuse. 16. Propotion$.tw.
5. Domestic 17. Frequenc$.tw.
6. Spouse 18. 10 or 11 or 12 or 13 or 14
or 15 or 16 or17.
7. Battered 19. Women.
8. Partner 20.
9. 1 or 2 or 3 or 4 or 5 or 6
or 7 or 8.
21. 19 or 20.
10. Prevalence. 22. 9 and 18 and 21.
11. Incidence. 23. Limit 22 to “all adult
(19 plus years)”
12. Proportion. 24. Limit 23 to female.
We included studies on the prevalence of domestic
violence against women conducted between 1995 and
2006, published in English and including women aged
between 18 and 65 years. We excluded studies on women
with special disabilities or certain complicated diseases e.g.,
HIV, women in places of refuge, case reports, reviews, and
non-English studies. We also excluded studies conducted
on women aged >65 years and on violence against pregnant
370 J Fam Viol (2010) 25:369–382
women, where a large number of studies was found, which
possibly merit a separate review.
Our searches identified 1,653 primary studies, which
were reduced to 356 after screening the titles and abstracts
to assess whether the contents were likely to be within the
scope of the review. We also checked for duplicates
between databases, accounting for 180 (10.9%) of the total
studies. A further 176 studies were excluded because they
were largely naratives about domestic violence cases,
studies of risk factors rather than prevalence or were
predominately review articles. A final total of 134 studies
was selected for further analysis (see Fig. 1).
Quality Assessment
These studies was assessed using structured guidelines
(Loney et al. 2000), and were scored on eight quality
criteria as follows: (1) specification of the target population,
(2) use of an adequate sampling method (e.g., random,
cluster), (3) adequate sample size (>300 subjects), (4)
adequate response rate (>66%), (5) valid, repeatable case
definition, (6) measurment with valid instrument, (7)
reporting of confidence intervals or standard errors, and
(8) attempts to reduce observer bias. We recorded the date of
the study, the prevalence (and/or incidence) estimates of
domestic violence (including life-time and/or current estimates),
and the type of violence reported. These variables
were coded from each study as categorical or continuous.
After quality assessment was completed, studies were
stratified according to the total score from 1–8.
Data Synthesis
The study data were coded and analyzed using SPSS
Version 11. Meta-analyses were conducted in STATA version
10. Continuous & categorical variables were expressed as
frequencies and percentages, and are summarized statistically
in tables and are presented in graphic form. Prevalence
estimates in the figures represent the simple weighted mean
prevalence for all the studies done in each continent.
A number of the studies we have included are described
in more than one publication. In some cases, additional
analysis conducted after completion of a study was reported
in additional publications. In these cases, we used both
reports to inform the data extraction. Conflict in quality
scoring of the included studies was resolved by consensus
between the authors (SA & RJ)
Forest plots were produced to give a graphical representation
of the studies and to convey the extent of heterogeneity
between prevalence estimates. Heterogeneity between prevalence
estimates was tested using a chi-squared test. Sensitivity
analyses were used to determine whether any heterogeneity
found could be due to differing study methodologies, study
quality or geographical differences.
Potentially relevant prevalence
studies identified for retrieval
Papers excluded on the basis of title & abstract
Papers retrieved for more detailed
evaluation (n=356)
Papers excluded with reasons (n=180), duplicates
between databases.
Potentially appropriate prevalence
studies to be included in the
review (n=176)
Prevalence studies excluded from review with reasons (n=134),
studies of risk factors, narratives of domestic violence cases, or
review articles.
Prevalence studies included in the
review with usable information
Fig. 1 Flow chart summarising
literature review
J Fam Viol (2010) 25:369–382 371
Most of the studies (41%) were conducted in North America,
followed by 20% in Europe, 16% in Asia, 11% in Africa,
and 5% in the Middle East (Table 1). Eighty three studies
(56%) were population-based, twenty five (17%) were
conducted in primary care, 12% in emergency care settings
and others in obstetrics and gynaecology, paediatric, psychiatric
and other hospital clinics. The sample size was over 300
in 84% of studies. Approximately 60% used a form of
randomisation in their sampling (Table 2). In 41% of studies
a measurement instrument was developed by the researchers
using focus groups or by reference to other validated
measuring instruments, although a few did not report about
the instrument used. The most commonly used instrument
was the Conflict Tactic Scale (16.9%), followed by the
Abuse Assessment Screen (14%) and the WHO instrument
(13%). The most frequently used method of collecting the
data was face-to-face interviews (55%), followed by selfadministered
questionnaires (30%), and telephone interviews
Only eighteen studies (12%) scored a maximum of 8 on
our quality criteria, with 33 (27%) studies scoring 7, 25
(17%) scoring 6 (Table 3).
The mean lifetime prevalence for physical, sexual and
emotional violence by country is shown in Fig. 2. The
highest levels of physical violence were seen in Japanese
immigrants to North America (about 47%), who also had
high levels of emotional violence (about 78%) along with
respondents studied in South America, Europe, and Asia
The mean lifetime prevalence of physical violence was
found to be highest (30–50%) in studies conducted in
psychiatric and obstetric/gynecology clinics (Fig. 3). The
highest rates of sexual violence were found in studies
conducted in psychiatric, obstetric, and gynecology clinics
(30–35%) and, for emotional violence, the highest rates
Table 1 Summary of frequencies of settings and continents
Frequency %
Geographical setting
60 studies in North America 40.5
29 studies in Europe 19.6
23 studies in Asia 15.5
16 studies in Africa 10.8
8 studies in Middle East 5.4
5 studies in Australia 3.4
4 studies in South America 2.7
Healthcare setting
83 Population studies 56.1
25 studies primary care 16.9
18 studies in emergency care 12.2
8 studies in Obst/Gyn clinic 5.4
5 studies in hospital setting 3.4
3 studies in pediatric clinic 2
2 studies in psychiatric clinic 1.4
2 studies in college students 1.4
One study in surgical clinic 0.7
One study in HMO 0.7
80 population cross-sectional studies 54.1
57 clinical cross-sectional studies 38.5
5 clinical cohort studies 3.4
4 population cohort studies 2.7
Table 2 Summary of frequencies of sampling, methods, and instruments
Frequency %
124 studies >300 sample size 83.8
24 studies <300 sample size 16.2
88 studies used randomization 59.5
54 studies used other methods 36.5
60 studies used their own instrument 40.5
25 studies used CTS 16.9
21 studies used AAS 14.2
19 studies used WHO instrument 12.8
6 studies used PVS 4.1
4 studies used ISA 2.7
3 studies used NorAQ 2
2 studies used women’s health questionnaire 1.4
One study used DVI 0.7
One study used SVAWS 0.7
One study used BRFSS 0.7
One study used WorldSAFE 0.7
Contact with subjects
82 studies used face-to-face interview 55.4
44 studies used self-administered 29.7
19 studies used telephone interview 12.8
Frequency %
18 studies scored 8 12.2
33 studies scored 7 22.3
25 studies scored 6 16.9
34 studies scored 5 23
27 studies scored 4 18.2
8 studies scored 3 5.4
One study scored 2 0.7
2 studies scored 1 1.4
Table 3 Summary of frequencies
of qulaity score
372 J Fam Viol (2010) 25:369–382
were found in accident and emergency and psychiatric
departments (65–87%).
Forest plots of prevalence estimates and their confidence
intervals indicate that there is a large amount of heterogeneity
between studies. Heterogeneity was formally tested
and confirmed by using the chi-squared test. This test
showed strong evidence of heterogeneity (p<0.001). Sensitivity
analyses found that even in studies that: used a
standardized methodology (WHO), scored high in their
quality criteria, were population-based (Fig. 3, 4, 5, and 6),
and in studies that were done in the same continents
(Dickers 2002), heterogeneity was a constant finding.
Pooled estimates across geographical locations and settings
were not calculated due to the extreme heterogeneity and
the difficulty in interpreting them.

Posted in Uncategorized

Leave a Reply