Abstract
Understanding the population distribution and dynamics of intimate partner violence (IPV) perpetration is necessary for designing effective prevention strategies and informing public health responses. While general population health surveys are widely used to estimate the prevalence and burden of IPV victimisation, the extent to which they capture IPV perpetration is under-explored. This review systematically identified and assessed general population health surveys measuring IPV perpetration, focusing on their methodological and measurement approaches. Searches of four electronic databases identified 39 papers reporting on 27 survey data sources. Five broad categories of measurement tools are described: survey data sources using or adapting the UN Multi-Country Study; the 2005 WHO research tool for violence against women and girls; the Conflicts Tactics Scale (CTS) and its subsequent versions, and two categories of ad hoc measures (with either more than two items, or one or two items). Substantial variation was found in the measurement of forms of IPV perpetration measured, particularly in the inclusion of non-physical forms, the use of frequency and reference periods, and the assessment of harm. Most surveys did not collect information on the sex or number of victims, escalation, or repeat perpetration, and few described item development or validation. Improved and standardised measurement of IPV perpetration in general population health surveys is essential to strengthen the evidence base and inform effective prevention strategies, service responses, and public policy.
Introduction
Preventing the perpetration of intimate partner violence (IPV) is a priority for policymakers and researchers around the world. Broadly defined as deliberate or intended physical, sexual, emotional, psychological, and/or financial harm toward a partner or ex-partner, IPV is highly prevalent (World Health Organization, 2023) and associated with mental and physical morbidity and with death from homicide, suicide, and internal medical causes (Stubbs & Szoeke, 2022; White et al., 2024) . It is widely held that research on IPV should attend not only to victims of IPV, but also to those who perpetrate IPV and to other parties who may be affected by IPV including children in the family/household, and the wider population (Oram et al., 2022). Despite decades of research on interventions, theories, and evidence concerning IPV perpetrators (Devaney & Lazenbatt, 2016, policies and programmes to address IPV perpetrators remain underdeveloped. Internationally, there are calls for coordinated policy responses to IPV perpetration amidst widespread concerns about inadequate criminal justice systems responses (Hester et al., 2020). Perpetration of IPV is relevant to public health and health systems, both as a distal driver of morbidity and as an indicator of poor mental health, substance use difficulties, and exposure to interpersonal trauma (Grace & Anderson, 2018; Jewkes et al., 2012). Research studies on IPV perpetrators have clarified the factors which shape IPV perpetration risk, including the role of alcohol and substance misuse, childhood trauma exposure, and mental ill health (reviewed in e.g., [Kim, 2023; Spencer & Stith, 2020]). This work has contributed important evidence on the impact of key factors, including previous use of weapons, educational attainment, and recent separation or threat of separation, in increasing risk of IPV perpetration. However, many of these studies have selected research participants based on receiving behavioural interventions, or criminal justice contact.
General population health surveys employ random sampling of residents within defined geographic catchments to arrive at representative estimates for disease and risk factor burden (Porta et al., 2014). They are sources of evidence on the population distribution of health states and associated risk factors, informing the development of national and local health policy (Corsi et al., 2012; Hatch et al., 2012; Jenkins et al., 1997; Kilpeläinen et al., 2019) to address a range of challenges including mental ill health (Jenkins, 2001) and childhood disease (Rodríguez et al., 2015). In particular, general population health surveys have been critical to the characterisation of the public health burden of IPV suffered by victims (Leung et al., 2019). General population health surveys are an important resource to improve understanding of IPV including the role of mental health and substance misuse, allowing triangulation with data from statutory agencies (Bunce et al., 2023).
The use of general population health surveys to answer research questions on the perpetration of IPV is more limited. The extent and quality of data collection on IPV perpetration within general population health surveys has not been comprehensively reported. In a narrative review of IPV measurement in surveys, Campbell et al. (2006) highlighted the measurement of IPV perpetration in the National Survey on Drug Use and Health (NSDUH), but did not describe or evaluate the measurement of IPV perpetration in depth, and examined only surveys from the United States. More recently, Skafida et al. (2023) considered IPV measurement in cohort and cross-sectional data for any purpose (including health data as well as criminological studies) but did so for England only.
In this paper, we systematically review the literature to consolidate existing sources of general population health survey data collected on IPV perpetration and to describe and evaluate the measurement tools used by these surveys. We then use this evidence to consider the current capability of these data to inform policy and practice, and to make recommendations for future collection of general population data on IPV perpetration.
Methods
Study Design
We carried out a systematic review of peer reviewed research studies/articles. The review was conducted and reported in accordance with the PRISMA 2020 guidelines. The review protocol was not prospectively registered. The review question, eligibility criteria, and analytic approach were specified a priori and adhered to throughout the review process.
Inclusion and Exclusion Criteria
We included peer-reviewed studies/articles which measured IPV perpetration using adult general population health survey data. We defined a general population health survey (or survey data source) as a survey which used a specified sampling frame for random (or probability) sampling and reported collection of data on information on one or more health outcome, with the aim of generating representative estimates for the prevalence of health-related states, or risk factor associations (Porta et al., 2014). We included surveys where the target population was restricted based on age, gender, and ethnicity. We placed no limits on geographic location but only included studies/articles reported in English. We excluded studies/articles which, for example, examined only samples of IPV perpetrators, only sampled currently coupled individuals, or only sampled military populations. We also excluded cohorts of individuals of fixed age (i.e., birth cohorts and school cohorts).
Search Strategy
We searched the literature for the following search terms: “partner violence” or “couple violence” or “spousal violence” or “domestic violence” or “domestic abuse” and perpetration or offending or batterer or aggress*; AND survey or cross-section*; AND representative or representative*. Search terms were applied to the following databases on December 14, 2025: Embase, Ovid MEDLINE/PubMed, Global Health, and APA PsycINFO.
Screening and Extraction of Measurement Characteristics
Screening and data extraction were carried out by VB. Titles and abstracts were screened against inclusion criteria, and potentially eligible articles/studies retrieved for full-text screening. Data were extracted from both eligible articles/studies and web-based information on the underlying survey data sources (where publicly available), based on a pre-specified extraction form. We extracted details of the items used by each survey data source to measure IPV perpetration, the study setting and sampling procedure, forms of IPV harmful behaviour captured, health outcomes collected/analysed, and prevalence estimates (disaggregated by sex where possible) for IPV perpetration. Where multiple articles/studies reported the same survey data source, discrepancies between studies (e.g. in the number of forms of IPV harmful behaviour captured) were described.
We also extracted information to support the description and evaluation of the measurement of IPV perpetration in the included articles/studies and associated survey data sources, including findings on the prevalence of IPV perpetration disaggregated by gender where possible. As there is no current consensus on the adequate measurement of IPV perpetration in epidemiological research, we defined a set of characteristics based on the work of Heise and Hossain (2017), Bender (2017), and Walby et al. (2017):
(1) IPV perpetration items align with a stated definition of IPV or IPV perpetration.
(2) Measurements of perpetrated IPV are not limited to physical IPV.
(3) Information is gathered on the sex of the victims of IPV perpetrated by the respondent.
(4) Information is gathered on the nature of intimate relationship in which IPV was perpetrated by the respondent (for example, whether it was in a current or previous relationship).
(5) The frequency and/or repetition of IPV perpetration (including the number of separate victims of IPV) are measured, and reference periods for IPV perpetration provided.
(6) The harmful impact of perpetrated IPV (in addition to IPV behaviours) is measured.
(7) There is description of item development and testing.
(8) There is reported involvement of victim/survivors and/or perpetrators of IPV in item development.
Results
Searches identified 3,283 records (2,283 from four databases and 1,000 from Google Scholar). After title and abstract screening, 231 full-text reports were sought and retrieved. Following full-text assessment, 39 reports, analysing 27 eligible survey data sources, were included in the review; the most common reasons for exclusion were non-representative sampling, restriction to currently partnered samples, and studies focusing on IPV victimisation rather than perpetration (a PRISMA flow diagram is provided as supplementary information). Survey data sources are reported in Table 1, grouped by measurement tool/approach. All survey data sources used multi-stage stage area-based random (or probability) sampling procedures. Fourteen survey data sources collected information on perpetration of physical and sexual IPV, with eight of these survey data sources further collecting information on emotional/psychological IPV. Among these eight survey data sources which collected information on emotional/psychological IPV, three collected data on economic/financial IPV perpetration. One survey data source collected information on perpetration of cyber IPV as well as physical, sexual, and emotional/psychological IPV. We identified no survey data sources collecting general population data on perpetration of stalking and harassment in IPV. No survey data sources assessed the impact on child victims of IPV perpetrated by respondents. Thirteen survey data sources included information on IPV perpetration in men only, with three sources reporting information on IPV perpetration by women only. Nine survey data sources contained information on IPV perpetration in both men and women.
Overview of Data Sources and Measurement of IPV Perpetration.
Note. BNADS = Brazilian National Alcohol and Drugs Surveys; CTS/CTSSF = Conflicts Tactics Scale/Conflicts Tactics Scale Short Form; DHS = Demographic and Health Survey; HANDLS = Healthy Aging in Neighborhoods of Diversity across the Life Span; IPV = Intimate Partner Violence; MCS = Multi-Country Study; NAALS = National Asian American and Latino Survey; NCS-R = National Comorbidity Survey-Replication; NESARC = National Epidemiologic Survey of Alcohol and Related Condition; NSAL = National Survey of American Life; NSDUH = National Survey on Drug Use and Health; PTSD = Post-traumatic Stress Disorder; RHS = Reproductive Health Survey; RRS = Rural Response System; UN-MCS = UN Multi-Country Study; YSS = Youth Sexuality Survey.
Five groups of measurement approaches were identified: (a) measures using or adapted from the United Nations Multi-Country Study (UN-MCS); (b) measures adapted from the WHO Instrument for Research on Violence Against Women and Girls; (c) measures using or adapted from the Conflict Tactics Scales; (d) ad hoc scales with more than 2 items; and (e) ad hoc scales with two items or fewer. We assess each included measurement approach against criteria described above, in Table 2.
Critical Assessment of Measurement Approaches.
Note. Ph = physical IPV perpetration; Sex = sexual IPV perpetration; Em/psy = emotional/psychological IPV perpetration; Ec/fin = economic/financial IPV perpetration
One NESARC publication does not report sexual violence perpetration data (Lee et al., 2023).
HANDLS does not further describe how the eight items measure different forms of IPV harmful behaviour, but states that physical, sexual, and emotional/psychological IPV perpetration were measured.
Ghent survey also collected information on cyber IPV perpetration.
APMS = Adult Psychiatric Morbidity Survey; BDHS = Bangladesh Demographic and Health Survey; BNADS = Brazilian National Alcohol and Drugs Surveys; CPES = Collective Psychiatric Epidemiology Surveys; CTS/CTSSF = Conflicts Tactics Scale/Conflicts Tactics Scale Short Form; HANDLS = Healthy Aging in Neighborhoods of Diversity across the Life Span; JRHS = Jamaica Reproductive Health Survey; MCS = Multi-Country Study; MHFRS = Men’s Health, Fatherhood, and Relationships Study; MHMLS = Men’s Health and Modern Lifestyles Survey; NCSR = National Comorbidity Survey Replication; NESARC = National Epidemiologic Survey of Alcohol and Related Condition; NFHS = National Family Health Survey; NLAAS = National Latino and Asian American Study; RRS = Rural Response System; SASH = South Africa Stress and Health Study; UN-MCS = UN Multi-Country Study; YSS = Youth Sexuality Survey.
Note. CTS = Conflicts Tactics Scale; UN-MCS = UN Multi-Country Study.
Measures Using or Adapted from the UN-MCS tool
Three survey data sources were identified which used the UN-MCS tool to measure IPV perpetration, reporting prevalence estimates for Bangladesh, China, Indonesia, Papua New Guinea, and Sri Lanka (Fonseka et al., 2015; Fulu et al., 2017; Jewkes et al., 2025; Willie et al., 2022), Zimbabwe (Machisa & Shamu, 2018), and Germany (Berthold et al., 2024; Clemens et al., 2023). Health outcomes studied in these data sources included depression, self-rated health, and self-harm/suicide attempt.
The original multi-country study (MCS) was accompanied by online information that framed IPV as a form of gender-based violence, defined as “an umbrella concept that describes any form of violence used to establish, enforce, or perpetuate gender inequalities and keep in place unequal gender-power relations. In other words, it is violence that is used as a policing mechanism of gender norms and relations and is intended to result in the subordination of women.” Papers reporting the original MCS data and the Zimbabwe survey did not provide a definition of IPV, and the report of the Germany survey referred to the WHO (2021) definition, where IPV “refers to behaviour by an intimate partner or ex-partner that causes physical, sexual or psychological harm, including physical aggression, sexual coercion, psychological abuse and controlling behaviours” (World Health Organization, 2021). The original MCS tool measures four forms of IPV harmful behaviour. It includes five questions about psychological/emotional IPV (e.g., “Have you ever threatened to hurt a partner?”); four questions about economic/financial IPV (e.g., “Have you ever prohibited a partner from earning money?”); five questions about physical IPV (e.g., “Have you ever pushed or shoved a partner?”); and two questions about sexual IPV (e.g., “Have you ever forced your current or previous partner to have sex with you when she did not want to?). Studies using this measurement approach varied in the number of forms of IPV harmful behaviour they measured (one did not measure economic/financial IPV perpetration) and in the number of items they used to measure forms of IPV perpetration (between 3 and 6 for physical IPV, between 2 and 4 for sexual IPV perpetration, between 5 and 6 for emotional/psychological IPV perpetration, and 3 and 5 for economic/financial IPV perpetration).
The original MCS instrument does not collect data on victim sex; instead, it asks specifically about perpetration of IPV against female partners. Data on the nature of intimate partnerships are not collected, and skip questions limit data collection to perpetration of IPV within sexual relationships. Frequency of perpetrated IPV is measured via Likert responses in the original MCS tool, but not in data reported by Machisa and Shamu (2018) on Zimbabwe or by Clemens et al. (2023) on Germany. Reference periods for measurement of IPV perpetration are 12 months for the original MCS tool. Machisa and Shamu (2018) report prevalence estimates for both lifetime and 12-month perpetration, while reference periods are not made explicit in Clemens et al. (2023). With regard to the collection of data on the impact of IPV perpetration, MCS items for emotional IPV refer to causing fear, but physical and sexual items do not refer to the occurrence of harm (e.g., occasioning medical care or disability).
The development of the MCS questionnaire is described in an appendix to the main report on the study website (Fulu et al., 2013). Questionnaire development focused on sequencing items (going from less sensitive to more sensitive topics), inclusion of sources of support, non-judgmental phrasing, deliberate non-use of the words “violence” and “rape,” and welcoming/reassuring introductory statements. Gibbs et al. (2019) report the test–retest reliability of MCS items for physical IPV perpetration in a sample of 112 South African men. User involvement in the development of MCS items is not reported.
Measures Adapted from the 2005 WHO Instrument for Research on Violence Against Women and Girls
Four survey data sources were identified that measured IPV perpetration by “transposing” items from the 2005 WHO Instrument for Research on Violence Against Women and Girls, which was designed to measure IPV victimisation (Ellsberg & Heise, 2005). These datasets provide estimates of IPV perpetration in Ghana (Chirwa et al., 2018), the district of Dagoretti in Nairobi, Kenya (Ringwald et al., 2022), and two settings in South Africa (Gauteng and North West Province, (Machisa et al., 2016; Treves-Kagan et al., 2021). Health outcomes collected in this group of studies included depression, post-traumatic stress disorder (PTSD), and substance use.
The original WHO instrument is aligned to the 2003 WHO definition of IPV, referring to aggressive or coercive behaviours among marital, dating, or cohabiting partners. Among the four survey data sources employing this measurement approach, reports from two data sources did not state a definition of IPV perpetration. A report from the Ghana Rural Response System (RRS) referred to a United States Centre for Disease Prevention and Control (CDC) definition (Chirwa et al., 2018), and a report from the North West Province survey referred to a definition from a 2013 article with WHO authors (Devries et al., 2013; Treves-Kagan et al., 2021).
Three forms of IPV victimisation are measured by the original WHO instrument, ever and in the past year, using binary “yes” or “no” response options. Transposing the tool in full would allow for four items about emotional IPV (e.g., belittling or humiliating someone in front of other people), six items about physical IPV (e.g., slapping or throwing something that could hurt someone), and three items on sexual IPV (e.g., physically forcing someone to have sexual intercourse when they didn’t want to). The four survey data sources varied in their measurement of these forms of IPV. The Dagoretti IPV/HIV survey asked about the perpetration of all three forms of IPV described above in the past year only and used a binary response option. The Ghana RRS measured economic/financial IPV perpetration with a single item, in addition to these three domains. The Gauteng and NW Province surveys measured lifetime perpetration of physical and sexual IPV only (i.e., omitting measurement of emotional/psychological and economic/financial IPV). Likert systems were used to measure frequency of IPV behaviours by two of the four studies.
Regarding victim sex and the nature of the relationship, questions in the original WHO Instrument refer specifically to behaviour towards a husband or partner (Ellsberg & Heise, 2005). In adapting these questions, the Ghana RRS, Gauteng Survey, and Dagoretti Survey specified behaviour by men towards female partners but did not further differentiate between types of intimate relationships. Whether the North West Province Survey specified behaviour from men towards women partners is not stated. Data collection on IPV related harmful impact was limited in studies in this group. Items for emotional IPV referred to causing someone to feel bad, belittled, or humiliated, and sexual IPV items referred to behaviour causing humiliation. Neither physical nor sexual IPV items referred to the occurrence of physical harm (e.g., resulting in medical care or disability).
No studies/articles reported work to assess the suitability of the adapted items for measuring IPV perpetration, including wording of items or phrasing of introductory statements. No articles/studies referred to validation work on transposed WHO instrument items, although the WHO instrument has been evaluated in relation to experiences of IPV victimisation (Schraiber et al., 2010). User involvement in the development of items was not reported.
Adaptations of the CTS and the Subsequent CTS Short Form
Ten survey data sources employed items from the CTS short form (CTS-SF) to measure IPV perpetration. These included the National Epidemiologic Survey of Alcohol and Related Condition (NESARC) (Hahn et al., 2015; Lee et al., 2023; Maldonado et al., 2020; Okuda et al., 2015; Roberts et al., 2010), the Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) study (Maldonado et al., 2022), the Bangladesh Demographic and Health Survey (Murshid, 2017), the Brazilian National Alcohol and Drugs Surveys (BNADS) (Ally et al., 2016; Madruga et al., 2017), the Men’s Health, Fatherhood, and Relationships Study: Walsh et al. (2020), the Ghent Survey (Schokkenbroek et al., 2024), and the Jamaica Reproductive Health Survey (Priestley & Lee, 2021). We grouped measurements of IPV perpetration in the Collective Psychiatric Epidemiology Surveys into three survey data sources to align with eligible reports: studies based only on the National Comorbidity Survey Replication (NCS-R) (Afifi et al., 2010; Singh et al., 2014), an analysis of the combined NCS-R and National Asian American and Latino Survey data source (Cho, 2012), and a separate analysis of National Latino and Asian American Study (NAALS) data only (Chang et al., 2009). Collectively, they provide estimates of IPV perpetration for the USA, Bangladesh, Brazil, Jamaica, and Ghent, Belgium. Health outcomes collected in these sources included problem drinking, smoking, suicide attempt, depression, and PTSD.
The original CTS defines IPV as means for resolving conflict within relationships and measure both IPV victimisation and perpetration. Studies/articles employing CTS-derived items typically did not provide a definition of IPV.
The CTS measures physical IPV, with a later version—the revised CTS (CTS2)—additionally measuring sexual IPV and psychological IPV (Straus, 1996). The CTS-SF measures physical, sexual, and psychological IPV (Straus & Douglas, 2004). Survey data sources employing CTS-derived measurements for IPV perpetration varied with regards to the forms of IPV perpetration measured (although all measured physical IPV) and the number of items used to measure these forms of IPV perpetration. This ranged from a single item to measure past year physical IPV perpetration in the Ghent Survey (purposefully hitting, scratching, pushing, biting or in another way physically hurting a partner) to 7 items measuring lifetime IPV perpetration in the 2007 Bangladesh Demographic and Health Survey (DHS) Survey (pushing, shaking, or throwing an object; slapping; pulling hair or twisting an arm; punching or hitting with a fist or something harmful; kicking or dragging; choking or burning; and threatening or attacking with a knife or gun). Four data sources measured sexual IPV perpetration (NESARC, HANDLS, BNADS, the Ghent Survey) using one or two items, and the Ghent survey measured psychological IPV perpetration using seven items. The report of the HANDLS data source reported using eight items to measure IPV perpetration but did not state which forms of perpetration were captured or how many items related to each form. Cyber IPV perpetration was measured by the Ghent survey using an abbreviated three-item scale adapted from the Cyber Dating Abuse Questionnaire (Borrajo et al., 2015).
Studies/articles employing CTS/Conflicts Tactics Scale Short Form (CTSSF) derived items did not collect information on the sex of the victim or the nature of the relationship in which IPV was perpetrated, although the 2007 Bangladesh DHS measured IPV perpetration only in men and questions referred specifically to female partners/wives. Likewise, the Men’s Health, Fatherhood, and Relationships Study studied only men.
Response options in the CTS, CTS2, and CTS2-SF allow participants to report that they have never used a particular behaviour, that they have used a behaviour but not in the past year, or that they used a behaviour in the past year once, twice, 3 to 5 times, 6 to 10 times, 11 to 20 times, or more than 20 times. Data sources using this measurement approach typically measured IPV perpetration with reference to the previous 12 months (6 data sources), or based only on the current/most recent relationship (2 data sources), with one survey collecting information with a lifetime reference period (and on number of relationships). Likert methods for collecting frequency were used by 7 data sources (with some inconsistency in reporting among included studies/articles, see Table 2), with frequency recording not measured or described for 4 data sources. Items to measure physical IPV numbered between 1 and 8, and for sexual IPV, 1 or 2. One survey measuring psychological IPV collected information on this with seven. The report from the HANDLS data source described eight items for data collection on IPV perpetration but did not specify how they broke down in terms of forms of harmful behaviour in IPV. The Ghent Survey also measured psychological IPV perpetration and cyber IPV perpetration, using an abbreviated version of the Multidimensional Measure of Emotional Abuse (Murphy & Hoover, 1999) and an abbreviated three-item scale adapted from the Cyber Dating Abuse Questionnaire (Borrajo et al., 2015), respectively.
Few studies collected data on the number of separate victims of IPV; the NCS-R did so by asking about the number of relationships in which each behaviour occurred. With regard to IPV-related harms, the NESARC assessed harmful impact of perpetrated IPV through a CTS item referring to the injured partner requiring medical care. Other studies/articles in this group did not measure harmful impact of perpetrated IPV.
The derivation of the short form CTS from the full CTS is reported in Straus and Douglas (2004). The development of the measure for psychological IPV in the Ghent survey is described in Murphy and Hooper (2005), and the measurement of cyber IPV perpetration described in Borrajo et al. (2015). Test–retest reliability and internal validity of the CTS2 and CTS-SF are reported (Chapman & Gillespie, 2019; Straus & Douglas, 2004; Vega & O’Leary, 2007). Included articles/studies described above did not describe testing carried out on the items selected by each particular article/study (e.g. validation for that specific geographic population). No involvement of victim/survivors was reported.
Ad hoc Measurement (More Than Two Items)
Four included survey data sources measured IPV perpetration using items not directly derived from a specific cited measurement tool. Included in this group were the Men’s Health and Modern Lifestyles Survey 2009 (Roberts et al., 2016), the Youth Sexuality Survey (YSS) (Zhang et al., 2019), the Adult Psychiatric Morbidity Survey 2014 (Bhavsar et al., 2023), and the South Sudan Survey (Khalaf et al., 2022).
One article in this group did not provide a definition of IPV, and three referred to a WHO definition of IPV. The Men’s Health and Modern Lifestyles Survey asked questions only about physical IPV perpetration. The YSS asked questions about sexual and psychological IPV perpetration, as well as physical IPV. Adult Psychiatric Morbidity Survey (APMS) 2014 measured perpetration of physical IPV, sexual IPV, and psychological IPV (threats). The South Sudan Survey measured physical and sexual IPV perpetration. No studies/articles reporting survey data sources in this group gathered information on the sex of victims, or other information on the nature of the relationship. No reports of data sources in this group measured harmful impact of perpetrated IPV, although APMS 2014 included “behaviour causing fear” in one item. No survey data sources in this group measured frequency of IPV perpetration or the number of victims. No item development process, or evaluation of psychometric properties, was reported by studies/articles in this group. No user involvement was reported.
Ad hoc Measurement Using Two Items or Fewer
Six survey data sources used one or two binary items to measure IPV perpetration. These were the Tanzania DHS (Reese et al., 2021),the Timor Leste (TL) DHS 2016 (Pengpid et al., 2018), National Survey on Drug Use and Health (NSDUH), (Lipsky & Caetano, 2011), the South Africa Stress and Health (SASH) Study: (Gupta et al., 2008; Stein, 2008; Stein et al., 2009), the Adult Psychiatric Morbidity Surveys 2000 and 2007 (González et al., 2016), and the National Family Health Survey (Sabri et al., 2014). Health outcomes studied in the included articles included drug and alcohol misuse, health insurance status, and self-reported diagnoses of serious mental illness.
Two survey data sources in this group gave definitions of IPV, and the remaining four did not provide a definition of IPV. Survey data sources in this group measured only physical IPV perpetration. Sex of victim and other information on the nature of intimate relationships in which IPV was perpetrated were not reported by of studies/articles based on these data sources, although DHS introductory text providing context to IPV items referred to heterosexual relationships. Measurement of frequency was mentioned in reports from the Tanzania DHS, the SASH, and the TL DHS, but was not further described. The NSDUH measured frequency using response options for 0, 1, 2, few and many times. Reference periods were 12 months for the Tanzania DHS, the TL DHS, and for NSDUH, and the SASH and NFHS collected information referring to the current or most recent relationship. Information on number of victims was not collected. This group of survey data sources did not include measurement of harmful impact of perpetrated IPV, based on published reports, and there was no reporting of how items were developed or evaluated prior to formal data collection, or of user involvement in the development of items.
Estimated prevalences of IPV perpetration varied across measurement approaches—given substantial heterogeneity in measurement instruments, reference periods, and forms of IPV captured, these differences should not be interpreted as reflecting variation in underlying IPV perpetration risk. Lifetime prevalence ranged from very low prevalences in some national surveys using minimal or ad hoc items (e.g., around 2%—5% in Men’s Health and Modern Lifestyles Survey, NSDUH, and Timor-Leste DHS) to higher prevalences in studies using more extensive instruments or conducted in low resource contexts (e.g., around 50% in the Ghana Rural Response System, 41% lifetime in Zimbabwe, 56–57% for physical and/or sexual IPV in South Sudan, and 71% reporting at least one form of IPV perpetration in HANDLS). In the UN-MCS, lifetime perpetration ranged from 13.5% in Papua New Guinea to 49.4% in Sri Lanka. Surveys using CTS-based measures in high-income countries typically reported 12-month or lifetime prevalence between 5% and 20%, though some sub-analyses showed higher rates for minor or psychological forms of IPV. Across studies that disaggregated by form, psychological or emotional IPV was generally the most prevalent, followed by physical IPV, with sexual IPV least frequently reported.
Discussion
Summary of Findings
We identified 39 studies reporting 27 survey data sources which measured IPV perpetration in general population health surveys. There was substantial heterogeneity in the forms of IPV perpetration measured, the comprehensiveness and conceptual clarity of measurement tools, and the rigour with which these tools were developed and applied. While most surveys captured physical IPV perpetration, fewer included sexual, emotional/psychological, and economic/financial IPV perpetration. No surveys measured perpetration of stalking or harassment, or the impact of perpetrated IPV on children, highlighting critical gaps in the scope of current measurement practices.
Five broad measurement approaches were identified: adaptations of the UN-MCS tool (Fulu et al., 2013), the WHO instrument (Ellsberg & Heise, 2005), the CTS/CTSSF (Straus, 1973; Straus et al., 1996), and ad hoc measurement approaches which we grouped into extensive and minimal approaches. Approaches to measuring IPV perpetration varied widely in how comprehensively they measured different forms of harmful behaviour within IPV, whether they captured frequency or repetition, and whether they collected data on victims’ sex or the nature of the relationship. The absence of data on victims’ sex limits the ability of existing research to examine IPV perpetrated in same-sex relationships. Moreover, no studies collected data on the number of victims, despite evidence suggesting that perpetrators often engage in repeated IPV against multiple partners (Gondolf, 1997, 2004). This lack of data on repetition and relational contexts also limits the ability to disentangle different patterns of abuse, which may have distinct risk factors, consequences, and intervention needs.
Measurement of IPV-related harm was also inconsistent. Where included, harm was typically inferred from specific behaviours (e.g., causing fear, or requiring medical care) rather than being measured directly. This distinction between behaviours and their subjective or objective impact is crucial for understanding severity and consequences of IPV. Accurately capturing severity, frequency, and related harm is also critical to describing the sex differences in IPV perpetration that are indicated by victimisation research. Improving the measurement of harmful impact of IPV perpetration in general population health surveys could be informed by criminological work to develop and evaluate scales for the measurement of harm, including the crime harm index and the crime severity score (Ashby, 2017). There is also a need to develop measures capable of assessing escalation of violence—an important predictor of serious adverse outcomes including homicide.
The variability in approaches identified in this review points to broader challenges and opportunities in developing a stronger evidence base on IPV perpetration. A key issue is the inconsistent use of definitions of IPV; few studies explicitly referenced such definitions, and many employed transposed victimisation items with little or no adaptation. Employing measurement tools anchored in clear, theoretically informed definitions of IPV perpetration could improve the coherence and interpretability of findings and ensure alignment between measurement and research objectives (Walby & Myhill, 2001).
The development and validation of IPV perpetration items was rarely reported. While the UN-MCS instrument has documented item development and testing (Fulu et al., 2013; Gibbs et al., 2019), other commonly used instruments were applied without attention to local context or validation for perpetration-specific use. No study reported item development processes for new or adapted measures, nor reported on the involvement of survivors or perpetrations in design. This lack of participatory and context-sensitive development risks introducing measurement error and underreporting, particularly given known challenges of disclosure (Krumpal, 2013).
Under- and over-reporting of IPV perpetration have substantial implications for prevalence estimates and inferences about associations with health. Studies of crime surveys suggest that misreporting can distort theories of IPV and policy responses (Ackerman, 2016). Without better evidence on how IPV perpetration is reported in representative samples, our ability to interpret survey data and use it to inform prevention and intervention remains limited.
Future work should prioritise the development and use of comprehensive, theoretically grounded, and psychometrically robust IPV perpetration measures. These tools must extend beyond physical IPV to encompass emotional/psychological, sexual, and economic forms of abuse, with attention to frequency, pattern, escalation, and context. Items should be developed through transparent processes, include local pilot testing, and involve the perspectives of both survivors and perpetrators of IPV. Incorporating measures of harm and the number of victims could allow better understanding of severity and chronicity, while collecting data on the sex of victims and nature of relationships is essential for inclusive, intersectional analysis. Improving IPV perpetration measurement in general population surveys is not only a methodological priority but a necessary step toward developing research-based policy and prevention strategies grounded in representative data.
Strengths and Limitations
This review provides the first systematic synthesis of how IPV perpetration is measured in general population health surveys across a range of geographic and socioeconomic settings. A key strength is our use of a structured evaluation framework, informed by conceptual and methodological work on violence measurement (Heise & Hossain, 2017; Walby et al., 2017) allowing for a comparative assessment of measurement quality across diverse tools and approaches. By focusing on general population health surveys, the review specifically highlights the extent to which epidemiological studies are equipped to inform policy and prevention efforts.
However, our approach has limitations. While our search terms were designed to capture studies measuring IPV perpetration, eligible studies/articles may have been missed if they did not explicitly refer to IPV in their titles, abstracts, or indexing terms—for example, where IPV was embedded within broader community violence or public health surveys. We excluded studies/articles restricted to currently partnered individuals, which may have led to the omission of studies using robust sampling strategies but narrowly defined target populations. We examined peer-reviewed articles published in English, which may have excluded relevant evidence published in other languages or grey literature sources, including reports that might have provided more detailed accounts of item development or piloting. Our decision to limit the scope of eligible articles to those explicitly addressing public health outcomes in our inclusion criteria may have excluded studies grounded in other disciplinary approaches, such as criminology. We recognise that studies excluded from our review on this basis may also relevant to future improvements in data collection on IPV perpetration in health research, in addition to the health survey data reviewed in this paper. Finally, study screening and data extraction were conducted by a single reviewer, which may have increased risk of selection bias compared with reviews using independent duplicate processes. This may have resulted in the inadvertent exclusion of eligible studies or minor inaccuracies in extracted information. Despite these limitations, this review highlights critical methodological gaps in the measurement of IPV perpetration in epidemiological research and provides a foundation for improving measurement practice in future population-based studies. We did not include the involvement of victim/survivors of IPV into the design of the review. The potential benefit of involving victim/survivors of IPV in developing measurement approaches to IPV perpetration should be evaluated in future research.
Conclusions
Substantial evidence on IPV has been generated by general population health surveys, yet most of this research overlooks IPV perpetration and, therefore, addresses only part of the problem. The limited number of surveys identified which measured IPV perpetration represents a significant gap in the evidence base. Strengthening data collection on IPV perpetration in general population health surveys would enable triangulation with administrative data sources such as police records, health data, and population registers, supporting more comprehensive surveillance and intervention planning. Future general population health surveys should routinely collect data on the sex and number of victims, harmful impacts of perpetrated IPV, and the escalation and repetition of abusive behaviours, as well as explicitly report the definition of IPV perpetration used and the process of item development and validation. Addressing these gaps is essential to generate the high-quality, inclusive, and actionable evidence needed to inform IPV policies and services.
Footnotes
Ethical Considerations
This is a systematic review of existing research and ethical approval and informed consent were not sought.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by an NIHR Advanced Fellowship awarded to Dr Bhavsar (NIHR AF302243).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Data on which this review is based are available from the authors upon request.
Supplemental Material
Supplemental material for this article is available online.