Asking the Right Questions? A Critical Overview of Longitudinal Survey Data on Intimate Partner Violence and Abuse Among Adults and Young People in the UK

We undertake a critical analysis of UK longitudinal and repeated cross-sectional population surveys which ask about experiences of intimate partner violence and abuse (IPVA). Seven relevant UK representative population-based surveys which ask about IPVA among adults and/or young people (16–17 years old) were identified. We critically engage with the questionnaires to analyse the strengths and limitations of existing UK data on IPVA. Several limitations in UK surveys are identified. Many questions still show a bias, partly historical, towards collecting more data about physical abuse. Few surveys ask about financial abuse, abuse post-separation or through child contact, or through technologies, though improvements are under way. Surveys still seek to count incidents of abuse, instead of enquiring about the impact of abusive behaviours on victims. Ethnicity and other demographic variables are not always adequately captured (or accessible to data users), making it difficult to explore aspects of inequality. Potentially useful comparisons within the UK are difficult to undertake given the increasingly divergent questionnaires used in different UK nations. We discuss how future iterations of existing surveys or new surveys can improve with regards to how questions about IPVA are asked. Given that surveys across geographical contexts often suffer similar weaknesses, our findings will be relevant for IPVA survey methodology beyond the UK context.


Introduction
Longitudinal population surveys in the UK have sought to collect data on intimate partner violence and abuse (IPVA). These include cross-sectional crime surveys as well as longitudinal cohort studies of children and families. Together, these surveys provide us with the means to address a range of research questions about IPVA, including prevalence rates, risk factors, impact and changes over time (Campbell & Rice, 2017;Herbert et al., 2021;Murray et al., 2016; Office for National Statistics, 2019b; Skafida et al., 2021;Yakubovich et al., 2020).
Since domestic abuse is significantly under-reported in police records (Campbell & Rice, 2017; Office for National Statistics, 2020), research using representative community and national surveys is important in determining the extent and nature of IPVA in the general population. Though such surveys are fraught with methodological challenges they do provide higher prevalence rates than police reports (Campbell & Rice, 2017;Office for National Statistics, 2020). Population surveys also allow for exploration of social stratification and inequalities in IPVA experiences, a challenging area to examine using convenience samples.
Longitudinal surveys of IPVA offer additional advantages and can help us better understand trajectories of abuse experiences, critical stages or events for abuse onset, long term impacts of abuse (Devries et al., 2013), and can enable causal inference (Herbert et al., 2022). As the legal landscape around IPVA is constantly shifting, longitudinal surveys provide the potential to monitor if and how changes in IPVA prevalence and experiences vary or not over time. Across the four legislatures in the UK, new legislation has recently been introduced which extends the criminalisation of domestic abuse to include coercive control. This is reflected in the new Domestic Abuse Act 2021 (England and Wales), the Domestic Abuse and Civil Proceedings Act (Northern Ireland) 2021, the Domestic Abuse (Scotland) Act 2018, and the Violence against Women, Domestic Abuse and Sexual Violence (Wales) Act 2015.
Study Aims A comprehensive overview of all recent and relevant UK longitudinal and repeated cross-sectional population surveys asking about IPVA, and a critical reflection regarding their strengths and weaknesses, and their similarities and differences, has not yet been undertaken in the current UK survey landscape. It is this gap we wish to address. There have been important contributions from scholars which we review further below.
Our first aim is to provide a comprehensive overview of the existing relevant data regarding IPVA among adults and young people (aged [16][17] which are being collected in recent large scale longitudinal UK surveys. We use the term longitudinal to cover both repeated cross-sectional surveys as well as cohort studies. Our second aim is to understand how these questionnaires represent the experiences of IPVA among adults and young people, given the nature and wording of the questions being asked, the context within which such questions are being asked, such as other variables of interest, different survey structures and sampling frames utilised. Our third aim is to examine the strengths and limitations of the different questionnaires. We discuss whether current surveys are asking the right questions and whether there are evidence gaps in measuring IPVA. We do not assess the development methods and psychometric properties of the surveys and we have not included questions about child abuse. This paper seeks to present a frame of reference for how existing data collected on IPVA can be improved. We reflect on what areas are being ignored in existing surveys and on how questionnaires can be improved to collect more meaningful information. We aim to make a significant contribution to how IPVA is measured both in the UK and internationally.

Key debates around IPVA measurement and conceptualisation
Methodologically, measuring IPVA poses several challenges. For example, IPVA victims often fail to recognise their experiences as abuse, or may not recall details of abusive behaviour often spanning long periods of exposure. Victims can also hesitate to disclose such experiences to researchers. Additionally, it is difficult to create survey appropriate questions which accurately capture the lived experiences of abuse and the complex dynamics of abusive relationships. Some of the key literature on IPVA measurement is summarised below.

Gendered dimension of IPVA
Scholars have debated whether gendered dimensions of violence are appropriately captured or potentially overlooked in population surveys (Myhill, 2015(Myhill, , 2017. Towers (2017, 2018) note that surveys which focus on the victim rather than the crime as the unit of measurement obscure the gender asymmetry of domestic abuse, and that counting frequency of physical assaults is crucial in revealing such gender disparities. In line with Myhill (2015Myhill ( , 2017 we argue that surveys should measure non-violent coercion previously not captured in traditional crime codes, since failure to do so means that the highly gendered nature of IPVA remains hidden. For a gendered understanding of IPVA, contextual information about the dynamics of abusive relationships is important in differentiating between IPVA perpetration and 'violent resistance' (Boxall et al., 2020). Finally, gender-blind reporting of IPVA often arises from how data are analysed and presented, rather than from what data is collected (Mac-Queen, 2016;.

Measurement of non-physical IPVA
Earlier evaluations of IPVA survey questions called for questionnaires which looked beyond just counting incidents of physical violence (Lindhorst & Tajima, 2008;Waltermaurer, 2005), something surveys are increasingly trying to address. Scholars have asked for questionnaires which are better able to capture daily micro-aggressions and relationship power dynamics as well as the perceived emotional impact reported by victims (Lindhorst & Tajima, 2008). Bender's (2017) review of different US national surveys identifies weaknesses which include: narrow definitions of rape; IPV questions framed as criminal acts which may discourage disclosure; questions which do not contextualise violence (i.e. differentiating between self-defence versus intent to injure); and interviewers not specifically trained to ask about IPV. Higher rates of sensitive behaviours were reported in selfadministered surveys compared to surveys that were administered by interviewers (Mirrlees-Black, 1999). To date no review has determined which longitudinal UK surveys have included questions on controlling behaviours and what aspects of coercive control are currently being measured.
In the academic community, there have been differing theoretical perspectives on non-physical forms of IPVA and how these can be measured. For example, Towers (2017, 2018) argue against the concept of coercive control and in favour of what they label domestic violence crime as a lens through which to develop IPVA measures. According to their approach, existing crime codes and legislation allow us to capture domestic abuse and IPVA, and counting discrete incidents of physical violence is of paramount importance when exploring the gendered nature of IPVA. While Walby and Towers propose that crime codes are adequate tools to capture harm, Myhill and Kelly (2021) disagree and argue in defence of coercive control as a conceptual framework. Like Myhill and Kelly, we contend that coercive control as a framework is better able to account for the various ways in which perpetrators can abuse their victims in ways which reflect more accurately the lived experiences of victims of abuse. However, there is little consistency in how coercive control has been conceptualised and measured thus far, and there are challenges in developing measures to appropriately capture dimensions of coercive control which include notions such as 'intentionality', especially in a survey setting (Hamberger et al., 2017).

From incidents to patterns of behaviour.
Early iterations of IPVA modules in relevant surveys (e.g. the crime surveys, and surveys which borrowed their questionnaires) predominantly focused on enumerating incidents of abuse; mostly of physical violence. Though public discourse and legislation around IPVA has evolved (Myhill, 2017) even recent surveys quantify abuse experience by enquiring either about the number or frequency of abuse incidents (Maguire & McVie, 2017). We recognise that data on incidence is of value and can be part of understanding broader patterns of abuse, but there are limitations to a reliance on this approach in isolation. Counting incidents, in isolation from other factors, risks not recognising the nature of abusive relationships and the ongoing impact of these lived experiences for victims (Crossman et al., 2016;Kelly, 1999;Myhill & Hohl, 2019). Yet, if we want to quantitatively measure patterns of abusive behaviour it may prove difficult to define a 'pattern' without any discussion of incidents or frequency (Walby & Towers, 2018). We argue that questionnaires should try to better capture the fact that living in fear of abuse -even in the absence of measurable incidents-is often an inherent part of being a victim of abuse, as is altering one's behaviour to avoid triggering the perpetrator.

Measuring Victim Impacts
While  argue that crime surveys where IPVA questions align with crime codes are adequate tools for capturing IPVA exposure, other scholars have called for measures which better capture IPVA impact and harm not adequately captured by traditional crime categories (Myhill & Kelly, 2021;Stark & Hester, 2019). Though UK crime surveys have improved and are improving their IPVA questionnaires in recent years, they are still limited by their exclusive focus on criminal acts as legally defined at each relevant time point. While this is perhaps expected given the aims of these surveys, it is worth questioning whether crime surveys should also measure the impact and context around IPVA related crimes.
Research with survivors of abuse has noted that victims often develop coping strategies which can involve extensive modification of their behaviour or lifestyle to try and avoid abuse (Kelly, 1999;Pain & Scottish Women's Aid, 2012), leading victims to 'doublethink' their every move (Pain & Scottish Women's Aid, 2012, p. 15). We therefore agree with Myhill (2015) that future survey questions could seek to address the consequences of ongoing abuse such as "generalized fear; degradation; objectification; loss of confidence, self-esteem, and the will and ability to resist; self-blaming; and the distortion of a victims' subjective reality" as well as questions about financial dependency and social isolation (Myhill, 2015, p. 370).
For survey measures to reflect an experiential understanding of IPVA, questions need to address how victims may change their behaviour due to the perpetrator's actions or threats; the extent to which respondents fear their perpetrator (even in the absence of 'incidents'); and the measures victims take to avoid a perpetrator, including post-separation. Research by Troisi (2018) shows how a screening tool for IPVA trauma could focus on questions about 'fear, a state of alarm elicited by the avoidance of the danger; terror, a paralyzing state that hinders an active process of reaction; shame as a strong exposure to the other that disarms the individual and the guilt as a defensive dimension aiming at the restoring of the link with the abusive partner' (2018, p. 1). However, the practical implications of measuring victim impacts are numerous and complex, and include the difficulty in distinguishing between measurement of generalised versus specific fear (Hardyns & Pauwels, 2010) as well as difficulties in capturing lived experiences of abusive relationships through closed survey questions.

Children's exposure to IPVA
Some of the large UK child cohort studies ask study parents to report on their IPVA experiences, and this enables us to estimate how many children have experienced domestic abuse. For example, a Scottish birth cohort estimates that 14% of mothers had experienced IPVA in the 6-year period since the study child was born and mothers from more disadvantaged backgrounds were far more likely to report IPVA (Skafida et al., 2021). The same data shows that children living in homes where domestic abuse is reported are more likely to themselves experience parental aggression (Skafida et al., 2022b) and to experience detrimental impacts on their social and emotional wellbeing (Skafida & Devaney 2023). The ALSPAC birth cohort was one of the first globally to measure exposure to IPVA antenatally, (albeit with only two questions) showing its deleterious effect on subsequent maternal mental health and child behaviour (Flach et al., 2011). Measurement of IPVA in the 1 st generation of the cohort at 21 years showed a high incidence in young adults (Yakubovich et al., 2019), and has been the basis of further research on the relationship between IPVA and depression (Herbert et al., 2022).

Methods
We sought to identify UK longitudinal and repeated crosssectional population surveys with measures on IPVA among adults and young people. We searched the UK Data Service (https:// ukdat aserv ice. ac. uk/) where large UK surveys are deposited using search terms 'domestic abuse', 'domestic violence' and 'intimate partner violence and abuse' and variants of these. Search results returned some of the surveys which we knew a priori existed. Additional surveys were identified via the UK Data Service's 'variable and question bank' search function where we searched using question wording extracts from other already identified questionnaires (UK surveys often borrow modules from other surveys). This included terms such as: 'controlling'; 'threatened'; 'unwanted letter'; 'belittled'; 'used a weapon against you'. Finally, we also searched Google Scholar for literature using combinations of the above search terms along with terms: national; representative; UK surveys; population surveys; prevalence; and variations of these. It is possible -though we believe unlikelythat a relevant survey has been missed. There is no established protocol for reviewing surveys and our methodology resembles that of similar survey reviews in the field (Wood et al., 2017), though future research could explore and implement a systematic protocol to reviewing survey measures. Table 1 provides an overview of the relevant surveys included in this study. We read the questionnaires and extracted all IPVA related questions into Online Appendix A. Details of surveys not included are provided in Online Appendix B. We noted the broader survey methodology, sampling strategy, demographic variables collected, and the positioning of IPVA questions in the larger survey. To engage critically with the questionnaires we use a qualitative methodology for survey quality assessment which can be best described as an 'experts review' (Biemer & Lyberg, 2003;Italian Institute of Statistics, 2017). The process of identifying strengths, weaknesses and limitations of the reviewed surveys, which we subsequently discuss, drew on both deductive and inductive reasoning. Some weaknesses have been previously discussed in relevant literature on IPVA measurement methodology and we were able to compare known pitfalls with the reviewed surveys. Other limitations emerged by reflecting on whether qualitative accounts of lived experiences of IPVA are appropriately captured in the survey questions at hand. Finally, further insights came from the process of comparing and contrasting the seven identified questionnaires, and reflecting on what IPVA data they provide as a whole. Expert reviews typically take place prior to a survey being issued and they aim to aid questionnaire development. In our study we employ the same principles and reflect on how selected questionnaires meet the intended analytic objectives of surveys measuring IPVA.

Overview of Relevant Surveys
In this section we summarise, in survey name alphabetical order, which surveys enquire about IPVA among adults and young people, and we provide details regarding the broader context and aims of the survey and the methodology around collecting the data. The range of survey topics within which IPVA questions feature is fairly broad, and includes surveys focusing on criminal behaviour, as well as surveys of families focusing on children's development. Table 1 provides a brief overview of some general survey characteristics.

Adult Psychiatric Morbidity Survey (APMS)
The APMS, also known as the Mental Health and Wellbeing Survey, is a repeated cross-sectional survey (every 7 years) of psychiatric morbidity in adults living in private households in England (McManus et al., 2020). This survey first ran in 1993, and the latest data is from 2014. IPVA questions are part of a module on factors associated with mental health problems. In 2014, 57% of those approached agreed to take part in the survey (N:7528). Data was collected via Computer Assisted Personal Interviewing (CAPI), and IPVA questions featured in a self-completion module (McManus et al., 2016). Most of the questions in this survey were borrowed verbatim from the 1 3 Crime Survey for England and Wales described further below. IPVA questions are asked in relation to a partner or ex-partner, for abuse taking place since age 16. A consultation on APMS modules which ran in 2021 resulted in a number of recommendation to improve domestic abuse questions which are being redeveloped for the forthcoming iteration of the survey (Gill et al., 2022).

Avon Longitudinal Study of Parents and Children (ALSPAC)
ALSPAC is an ongoing prospective-longitudinal study based in the south west of England. All pregnant women resident in one of three health districts in the former county of Avon in the UK due between April 1, 1991, and December 31, 1992, were eligible to participate (http:// www. brist ol. ac. uk/ alspac/). Initially, 14,541 pregnant women (and their eventual babies) were enrolled. When the oldest children were approximately 7 years of age, an attempt was made to bolster the initial sample with eligible cases who had failed to join the study originally, resulting in an additional 913 children being enrolled. This resulted in a total sample of 15,454 mothers (76% of all eligible) with 14,901 babies alive at age 1. The ALSPAC Ethics and Law Committee and Local Research Ethics Committees provided ethical approval. Information has been regularly collected since enrolment until present. Questionnaire data in ALSPAC was collected by postal questionnaires until 2014 (when the study children were aged 22 years) when data collection moved to online self-completion. IPVA was reported by parents of child participants when the child was 8 months, 2, 3, 4, 5, 6, 8, 9, 11, 12 and 18 years old, and by participants themselves at aged 21. IPVA measures were reported by participants (retrospectively) at 21 and 23 years old. At age 21, women responded to a validated 8-item scale on physical, psychological, and sexual IPV experiences before and/or after age 18 (Table 1, α = 0.95; Yakubovich et al., 2019). The measure was developed by a team of IPV researchers based on questionnaires used with young people (Barter et al., 2009) and a clinical sample in Bristol (Hester et al., 2015) and piloted for acceptability with the ALSPAC participant advisory group. Items were conceptually similar to those from existing IPV scales but with the benefit of not limiting measurement to conflicts or disagreements or overburdening participants with a large inventory of items (Yakubovich et al., 2019). Moreover, unlike most short-form IPV measures, the current measure captured physical, psychological, and sexual IPVA.

Crime Survey for England and Wales (CSEW)
The CSEW is the current version of what originally started as the British Crime Survey in 1982. The most recent iterations of the survey contain some of the most detailed questions on IPVA in the UK survey landscape. The survey first collected data on IPVA in 1996 and has since 2001 collected IPVA data regularly, seeking to run the relevant module on alternate years. Respondents are interviewed on a rolling basis during the course of a year, and the CSEW reports average crime data over a moving reference period which allows the data to be more comparable to police recorded crime figures (Office for National Statistics, 2021). The Office for National Statistics, the official UK body responsible for producing national statistics on a vast range of topics, regularly reports on CSEW data including but not limited to IPVA. Since 2019, reporting on IPVA has improved and now includes reporting on a range of relevant sub-topics, such as IPVA prevalence, and victim characteristic, detailed data on partner abuse, interactions with the criminal justice system and victim services and more (Office for National Statistics, 2019b).
Though most of the survey is run using CAPI, the IPVA module is a self-complete module which respondents complete using the interviewer's tablet by themselves. Respondents' answers are hidden from the interviewer during and after the self-completion. The self-completion modules feature at the end of the face-to-face interviews (Office for National Statistics, 2021). From 2017 onwards, the selfcompletion module of the CSEW questionnaire is given to all respondents aged 16 to 74 years (the upper age limit was previously 59 year). In 2009, the survey featured a subsample of respondents aged 10-15 years old which enquired about their experiences of crime, including hate crime and cyber security.
Growing Up in Scotland Survey (GUS) GUS is a longitudinal nationally representative prospective study of children and their families in Scotland. The main birth cohort included 5,217 babies in the first survey, born between 06/2004-05/2005 (Bradshaw et al., n.d.). Babies were 10 months old at the first sweep. Interviews were carried out in participants' homes usually with the child's mother, and mothers and partners (if present) were asked a range of questions about themselves and their children's development in oneto-one interviews via CAPI. The survey used a stratified random sample. Attrition rates are relatively low (87% response rate of surveys issued).
Maternal experiences of IPVA were recorded when children approached their 6 th birthday (3646 mother-child pairs were still part of the survey) (Skafida et al., 2021). Though the survey was interviewer led, the IPVA module was a selfcomplete feature. Mothers reported whether they had experienced a range of different types of violence covering coercive control, physical and sexual violence. Questions were based on the Scottish Crime and Justice Survey (described below) at the time of the GUS questionnaire design, but GUS asks about experiences of such abuse in a 6-year period from the birth of the study child to the present day, whereas the Scottish Crime and Justice Survey asks respondents to report on violence occurring in the last 12 months. Table  1 3 E in the Online Appendix A shows the original question wording.

Northern Ireland Safe Community Survey (NISCS)
Devolved nations Scotland and Northern Ireland run their own national crime surveys for their nations. The NISCS, previously known as the Northern Ireland Crime Survey which launched in 1994/95, has been running regularly since 2005. Survey results are published by the Northern Irish Department of Justice. It includes a self-complete module entitled 'Experience of Domestic Violence and Abuse' module which runs biannually. The most recently collected data are from 2015/16 (Campbell & Rice, 2017). The 2020/21 version of the survey would have originally featured the IPVA module again, but due to the Covid pandemic and the survey being carried out by telephone the relevant self-complete module could not be included (Department of Justice et al., 2022).
The latest available domestic abuse data from the NISCS dating back to 2015/16 was based on self-completion modules that featured in face-to-face interviews with respondents aged 16 to 64. Since 2008/09 the module has changed in multiple ways, making comparability difficult in terms of comparing to earlier data. The changes included extending the upper age limit to 64 (previously 59), and including an additional domestic violence question hoping to capture a more psychological aspect of abuse (i.e. a question about being 'repeatedly belittled so that you felt worthless'). Additional questions were also introduced to capture experiences of abuse in the wider family, beyond intimate partner violence.

Millennium Cohort Study (MCS)
MCS is a longitudinal survey of just under 19,000 children born in the UK in 2000-2001. The survey is broad in its focus and aims to capture various aspects of interest with regards to children's development, their family and immediate environment. Children were 9 months old at the first survey sweep, and subsequent sweeps took place at regular intervals tracking children into adulthood. In the surveys where children were 9 months, 3, 5, 7, 11 and 14 years old, a question about domestic violence was included in the questionnaire, and only mothers and fathers with a full-time resident spouse or partner were asked to respond (via a self-complete module) (Bunting & Galloway, 2012). The survey features one single question about IPVA which focused on physical abuse (see Table F in Online Appendix A) (Jofre-Bonet et al., 2016). The question was identical in wording across all sweeps. Cohort children are now adults, and recent data collection with the study-children themselves sought to measure experiences of serious violence among the participants (both as victims and perpetrators) (Smith & Wynne-McHardy, 2019). Children were 14 years old when this data was collected, and specific questions about IPVA were not part of the questionnaire. Scottish Crime and Justice Survey (SCJS) As discussed above, devolved nations Northern Ireland and Scotland have their own crime surveys. The SCJS is a national survey about crime and victimisation. It draws on a representative sample of adults aged 16 and over. Unlike the English/Welsh and Northern Irish counterparts the SCJS does not have an upper age limit. As with the other crime surveys, questions about IPVA are asked through a self-complete questionnaire at the end of the face-to-face survey. Questions cover a range of different types of abuse, including stalking, harassment, physical violence and sexual abuse, as well as threats of violence and questions about coercive control (MacQueen, 2016). The most recent iteration of the survey (2019-20) also contains questions about physical and psychological impacts of abuse and questions about children's experiences and involvement in abuse incidents.

Limitations of Current Surveys
Overemphasis on Physical Abuse We note that there is still a bias, in the existing survey milieu, towards data on physical abuse (Devries et al., 2013). This is in large part due to the fact that some of the reviewed questionnaires were developed and issued before public discourse and UK legislation across jurisdictions changed to recognise coercive control as a crime. Earlier versions of different crime surveys contain several questions about physical abuse and fewer questions about emotional abuse, a pattern that is subsequently also reproduced in non-crime surveys such as GUS and APMS which borrowed these modules.
MCS only measures physical violence using one question repeated over six sweeps: 'People often use force in a relationship-grabbing, pushing, shaking, hitting, kicking etc. Has your husband/wife/partner ever used force on you for any reason?' ALSPAC first included physical abuse questions when children were 8 years old, and only four years later (children aged 12 years old) were detailed questions about coercive control introduced. ALSPAC features repeated measures of physical and psychological abuse starting during pregnancy with the study child. Both questions however rely on a single item which requires respondents to identify a specific experience, in this case 'cruelty' (i.e. Your partner was physically cruel to you; Your partner was emotionally cruel to you). Previous commenters (Barter, 2014;Devries et al., 2013) have recognised that these limitations can 'open the possibility of substantial misclassification of total violence exposure' (Devries et al., 2013, p. 4).
Evolving Non-Violent IPVA Measures Except for MCS, all reviewed surveys continue to try to better capture nonphysical aspects of IPVA. Relevant questions seek to identify which respondents have experienced specific types of controlling behaviours. Though the inclusion of survey items measuring non-physical abuse is welcome, such questions are often under-represented compared to questions about physical violence, even in some of the most recent survey sweeps. For example, in the 2019-20 CSEW, there were three questions about coercive control (i.e. questions about being belittled; being frightened or threatened; being stopped from seeing friends and relatives), and eight questions about physical or attempted physical violence (though we note that there are also specific questions about stalking and abuse through technologies which would be categorised as psychological abuse).
The CSEW and APMS are both seeking to develop better measures of controlling or coercive behaviour for future surveys (Gill et al., 2022; Office for National Statistics, 2019a). The SCJS has five questions about controlling behaviours, seven questions about threats of physical abuse, and seven questions about physical and sexual abuse. The SCJS also has a detailed module on stalking and harassment, which asks whether or not partners or ex-partners were the perpetrators. It is worth noting that some elements of coercive control are more likely than others to feature in surveys. For example, in the SJCS, threats of physical violence make up seven of the total 12 items capturing non-violent IPVA. It could be argued that this shows a persistence for questionnaires to still focus on physical IPVA.
Counting Incidents All of the reviewed surveys included questions structured around the aim of counting incidents or 'events'. Questions typically sought to establish the incidence of specific types of IPVA in a certain time frame (e.g. since age 16; in the last 12 months; since the birth of the study child), and in some cases follow-up questions were asked about 'how many times' the abuse took place (SCJS, GUS). It was not always possible to make inferences about 'patterns' of abusive behaviour from questions about incidents, though the APMS did ask about frequency within the last year for each form of abuse. Only one question, versions of which featured in CSEW, SCJS, GUS, NISCS, and APMS, was phrased from the onset to reflect the systemic nature of abuse, and this asked if a perpetrator 'repeatedly put you down so you felt worthless' (SCJS, GUS) or 'repeatedly belittled you' (CSEW).

Measuring Victim Impacts
The surveys we reviewed lacked questions about how victims were affected by IPVA and how the abuse shaped their lives and behaviours. The APMS only asks one question about IPVA impacts, which is about physical injuries. Questions on impacts feature in ALSPAC which asks the survey children at age 21 to record the impact of their IPVA experiences, including asking if their partner's behaviour had made them feel: scared or frightened; upset or unhappy; sad; anxious; annoyed or angry; and/or depressed. Some promising developments can be found in the latest SCJS questionnaire, which asks IPVA victims to report how they were physically and psychologically affected. Potential responses for physical impacts following the 'most recent' incident include: minor and/or severe bruising; minor and/or severe cuts; severe concussion or loss of consciousness; internal injuries; pregnancy. Psychological impacts include: Difficulty sleeping / nightmares; Depression; Low self-esteem; Fear; Anxiety / panic attacks; Isolation from family or friends; Isolation from children in your household. For a complete list of all effects see Table  D in Online Appendix A. While the question about physical impacts is asked following any reported incident of abuse in the last year, the question about psychological impact is only asked once in relation to the 'most recent' incident. It is unclear why this would be the case and presumes impact relates to a single incident. In the SCJS, and other surveys, it would be preferable for psychological impact questions to be asked in a way which recognises that the impact of the abuse is often a result of a continuous and accumulative process of harm (Stark, 2007) rather than a single identifiable event.
Sexual Violence Sexual violence, as a form of IPVA is often overlooked in research (Bagwell-Gray et al., 2015). All surveys we reviewed, except for MCS, had at least one question about sexual violence. However, GUS, NIJSC only ask one and two questions respectively about being forced to have sexual intercourse or another sexual activity. However, ALSPAC child respondents at age 21 years were asked four questions about sexual IPVA which included being pressured or physically forced into a range of sexual activities. The most recent versions of the CSEW and the SCJS have the most extensive questionnaires on sexual violence (some of which also feature in the APMS), including initial screening questions about sexual violence and attempted sexual violence, and survey routing to a follow-up module with a comprehensive list of items about different types of sexual violence experienced, the context surrounding these experiences, and for the SCJS survey only, also about physical impacts on the victim and whether the police was notified (see Table D, Online Appendix A).
Economic Violence There are several dimensions of IPVA which are largely overlooked or not adequately explored in the existing surveys. For example, there is a case for treating economic violence as a separate category of IPVA (Bender, 2017). This includes excluding victims from financial decision making, or limiting their access to funds (Fawole, 2008). In the reviewed surveys, there is usually an item, albeit only one, which attempts to capture this dimension. The SCJS and GUS ask whether a partner 'stopped you having a fair share of the household money or taken money from you'. The NISCS, CSEW and APMS ask a similar question about 'having a fair share of the household money'.

Stalking and Harassment
Stalking as a form of intimidation and abuse which can be difficult to measure in surveys (Fox et al., 2011). Among the reviewed surveys, the CSEW and SCJS ask relevant questions in relation to the perpetrator being a partner or ex-partner. These surveys ask two different questions about being followed, (e.g. CSEW asks if 'ever followed you around and watched you on more than one occasion in a manner which caused you fear, alarm or distress') and both surveys also ask about a perpetrator waiting or loitering outside participant homes or workplaces. The SCJS has a separate section about stalking and harassment (including questions about unwanted phone calls and letters) which asks about unwanted attention both from people whom participants know (like partners and ex-partners) as well as strangers. The follow-up question to every reported stalking or harassment incident asks respondents to specify if the perpetrator was, among other options: a partner, expartner, a date, someone from a casual sexual relationship, a family member.
Abuse via Technologies Surveys need to better capture how technologies continue to change the ways in which perpetrators can abuse victims (Barter & Koulu, 2021). The CSEW and SCJS ask about unwanted messages or posts via email, social networks or social media sites. The SCJS asks whether 'a perpetrator has shared intimate images of you without our consent for example by text, on a website, or on a social media site like Facebook or Twitter' and the CSEW asks about a perpetrator sharing personal or threatening information about a victim on the internet. ALSPAC respondents (at age 21) were asked if their partners had 'regularly checked what they were doing and where they were (by phone or text)'. There are analogous questions in GUS regarding bullying by peers, but not in the context of IPVA. An interesting follow-up question in the SCJS harassment and stalking module, where questions about technologies also feature, is whether the police was notified, and if not, why not.
Other Dimensions of IPVA Other dimensions of IPVA rarely feature in survey questionnaires. For example, we find no questions on female genital mutilation (FGM), and there is little robust data on FGM prevalence generally . One of the complications of enquiring about FGM is the possibility of victims themselves not knowing if FGM has happened to them . There is also relatively little data about 'reproductive coercion' (Grace & Anderson, 2018) which can include pregnancy coercion, or controlling the outcome of a pregnancy, or interfering with a woman's birth control decisions. Only the SCJS asks if the 'most recent' incident of abuse or any prior incident led to pregnancy or if the 'most recent' incident left the victim 'feel[ing] forced to terminate a pregnancy'.

Questions about the Perpetrator
MacQueen argues that what is needed is 'a more nuanced understanding of how and why abuse, and violence more broadly, is perpetrated, how it is understood by perpetrators and victims and what its impact is' (MacQueen, 2016, p. 486). Most surveys seek to learn more about the perpetrator via victims themselves, and ask, for example, if the perpetrator is a partner or ex-partner. The CSEW and the NISCS ask whether respondents experienced IPVA from someone other than a partner (e.g. another family member). The SCJS asks similar questions but only in relation to stalking and harassment, and in the sexual abuse module, but the rest of the 'core' IPVA questions only ask about intimate partners and ex-partners. The SCJS also asks if perpetrators were living with respondents at the time of the abuse and whether respondents are still living with perpetrators at the time of the interview. The NISCS asks if the perpetrator consumed alcohol prior to the abuse, and the SCJS asks if perpetrators of sexual abuse were under the influence of alcohol or drugs during the incident, recognising the alcohol and drugs may be a compounding factor, but not a cause of IPVA. Of the surveys reviewed, only APMS and ALSPAC asked about perpetration. In the core module of the APMS there are four perpetrator questions, of which three about physical and sexual violence, and these questions have been criticised for underreporting prevalence (Gill et al., 2022). In ALSPAC when cohort children were 18-21 years old, they were asked four questions about coercive control and psychological abuse, physical violence and sexual abuse.
Better Individual Demographic Variables Socio-economic and neighbourhood variables are collected across all reviewed surveys, and were more granular in the child cohort studies (MCS, GUS, ALSPAC). Crime surveys collected fewer demographic details on respondents, but age, social class, household income, religion, gender, and disability status are collected in some form in all reviewed surveys, though variable quality differs. For example, questions on disability are more detailed in GUS than in SCJS. Research is increasingly showing associations between IPVA risk and neighbourhood and community level factors, suggesting important drivers behind violence which go beyond the individual, relationship or household level (Benson et al., 2003;Lauritsen & Schaum, 2004;Yakubovich et al., 2020).
There are other less frequently explored characteristics, such as sexual orientation, gender identification, religious affiliation and disability; and existing survey data does not always allow for meaningful analysis of how these dimensions affect IPVA exposure and experiences (MacQueen, 2016). The CSEW and SCJS both ask whether respondents experienced a range of crimes (e.g. harassment) which respondents believe were motivated by factors such as: "ethnic origin/race; religion; sectarianism; gender/gender identity or perception of this; disability/condition they have; sexual orientation; age; and pregnancy/maternity or perception of this" (Scottish Government, UK Statistics Authority, et al., 2021).
Nixon and Humphries call for better measurement of how gender, race, class and disability interact and overlap with each other in affecting women's experiences and risk of domestic violence (Nixon & Humphreys, 2010). Ethnicity variables are not always adequately captured in existing surveys, or datasets released to researchers do not always allow for meaningful analysis of how the risk of experiencing IPVA may be heightened for some ethnic minority groups (Skafida et al., 2022b). An added complexity is that since prevalence of abuse in surveys which ask about recent incidents (e.g. the 12 month reference period in the reviewed crime surveys) is relatively low, it is often difficult to undertake meaningful analysis in relation to sub-samples of interest, such as ethnic minorities (MacQueen & Norris, 2016). We believe that depending on the survey context, boosted samples should be considered to address some of these challenges, as has been done for example in the UK Understanding Society survey with ethnic minority boosts (Platt et al., 2021). Commissioned qualitative work for these and other minorities would also be important in addressing evidence gaps.

Lack of Longitudinal Data
Though we review seven longitudinal and repeated cross-sectional studies in this paper, the type of longitudinal data they offer is limited in two ways. Firstly, there are few high quality variables within surveys providing insight regarding individual IPVA experiences over time and duration of such experiences. Even within cohort studies, several of the questions provide limited temporal insight. Crime survey questions tell us little about the length of time victims spent living with IPVA, or whether the onset of abuse was triggered by key life transitions. SCJS asks respondents to specify if reported incidents took place in the last year or before this, but the rest of the survey (also applies to CSEW and NISCS) focuses on recording incidents within the 12 months prior to interview, and the only added granularity beyond prevalence in the last year is life-time prevalence. Secondly, among the repeated cross-sectional crime surveys, some of the potential which such studies can offer, such as tracking changes in prevalence over time, is limited by the fact that IPVA questions have changed from sweep to sweep over the years -though often this has been for good reason.
ALSPAC has regularly collected data about IPVA over the years, starting with maternal IPVA reports at 18 weeks gestation of the study child, though to IPVA questions answered by the mother and father when study children were 18 and 12 years old respectively. In the most recent sweep, children aged 21 years old are themselves asked about controlling partners, and about their recall of experiences of parental IPVA when they were younger. ALSPAC is potentially the best source to analyse IPVA from a life-course perspective, though it is worth noting that even in this survey, detailed indicators of (mostly) physical violence were only collected once from the mother when children were 8 years old, and four questions about coercive control were asked once when children were 12 years old. In GUS questions about IPVA have only featured once when children were 6 years old. MCS does collect repeated identical measures over six sweeps but it only asks one question about physical abuse.
IPVA onset can vary over the life-course following a curvilinear pattern (MacQueen, 2016). Data on teenagers suggests higher rates of victimisation than the general population (Barter et al., 2009). Other transitions, such as pregnancy (Bailey, 2010), childbirth, onset of menopause, and older age have also been associated with increased risk of experiencing IPVA, yet surveys rarely allows for analysis of these phenomena at a population level.
More Data on Sources of Support Other than in the most recent SCJS, there were few questions about sources of support victims turn to, who-if anyone-they have spoken to about the abuse, and whether they used any of the different support services which they would have access to. The recent SCJS, asks about a range of people or organisations which victims may have discussed domestic abuse issues with, ranging from friends and relatives to health professionals, and different support helplines. The CSEW features some questions about sources of support in a rotating survey module, but they are difficult to interpret as they cannot be analytically associated with a perpetrator or type of abuse experienced. Latzman et al. (2017) found that few measures capture children's direct involvement in IPVA and few assess exposure to coercive control, sexual violence or stalking. In the surveys we review, though GUS and MCS are both child cohort studies, no specific questions address this. ALSPAC does retrospectively ask the study children at age 21 to report on their childhood experiences of parental domestic abuse. The most recent SCJS questionnaire is the most promising source of data on this topic. It asks if children were present, if they saw or heard, if they got involved, if they got hurt or injured as a result of an incident, if they experienced different psychological problems as a result; and if they had called the police. Though the qualitative literature has highlighted the ways in which child contact opens up opportunities for perpetrators to continue to abuse victims (Morrison, 2015), there are no questions which specifically address this.

Questions Tailored to Survey Contexts
Though there is a logic to borrowing validated survey modules across different survey settings our review finds that borrowing is not always the optimal choice. This applies most strongly to cases where crime survey modules developed to reflect crime codes and measure crime incidents have been widely borrowed in completely different survey contexts, such as GUS and APMS. These latter surveys could have developed their own questions, or borrowed non-crime survey questions to, for example, ask about children's IPVA exposure and impacts in the GUS survey, or ask about psychological (instead of only physical) impacts of abuse in the APMS -a survey about psychiatric morbidity.

Within UK Comparability
With devolution, UK nations produced their own versions of what was originally the British Crime Survey. We have discussed in this paper the three different surveys that now run in Scotland, Northern Ireland, and in England and Wales. Although earlier iterations of these surveys had significant overlap in question wording, with the passing of time, each nation has modified their IPVA questionnaires substantially, at times to reflect different changes in legislation between nations. The IPVA module changes have often improved the questionnaires compared to their predecessors. But as changes have not been homogenous across nations, it is increasingly more difficult to compare survey data across the devolved nations. It is important to work towards consistency in IPV measures which would enable comparisons across time and populations Waltermaurer, 2005), though this should be balanced with improving IPV measurements as our understanding of the phenomenon evolves (Waltermaurer, 2005). To return to the point argued in the previous section, this is a scenario where at least some question homogeneity would be beneficial.

Concerns about Survey Structure
The most common way to collect information about IPVA in the three different crime surveys we have reviewed (CSEW, SCJS, NISCS) is via a self-complete module often positioned at the end of a long 'main' face-to-face survey. Thus, response rates for the selfcompletion part tend to be lower than for the main survey. For example, the SCJS technical report noted that 'ran out of time' was the main reason why the self-complete module was not completed, with 34% of the non-response group stating this (Scottish Government, ScotCen Social Research, et al., 2021). Also, in the SCJS, the IPVA module follows the questionnaire about respondents' drug use. The Scottish Women's Aid consultation on SCJS data notes that 'having the gender-based violence questions immediately after questions on behaviour and illicit drug use may be perceived as suggesting a link between them, or potentially, as victimblaming.' (Scottish Women's Aid, 2021, p. 3). Finally, since crime surveys tend to underestimate prevalence of relevant abuse (Brunton-Smith et al., 2022), perhaps IPVA survey modules would be better placed within public health surveys.

Sampling Strategies, Survey Response and Non-Random
Attrition Some population groups are routinely left out of sampling frames for most national surveys. These include populations living in sheltered accommodation, those changing addresses very frequently, those living temporarily with friends or family, homeless people, students, and prison populations (Scottish Women's Aid, 2021;. This means that key parts of the population most at risk of experiencing IPVA are also those who are left out of most prevalence surveys. For example, IPVA is a key cause behind women's homelessness (Bimpson et al., 2021). Frequently changing homes is associated with higher attrition (Watson, 2003), meaning that disadvantaged populations in unstable accommodation are also most likely to be underrepresented in such surveys.
Despite these weaknesses, little work has been done to explore non-random non-response, aside from overall survey attrition, in IPVA questionnaires and what the implications of this are for making population inferences. According to an ONS report (ONS, 2021), analysis of CSEW non-response was commissioned, though little detail is provided regarding findings. Research into item-non response in the GUS IPVA questionnaire notes that non-response in IPVA questions is highly socially stratified, and likely to be masking unreported experiences of abuse (Skafida et al., 2022a). This suggests that the social gradient in experiences of IPVA is likely to be substantially underestimated through population surveys.

Questionnaires Versus Their Reporting
On a final note, we wish to draw attention to the fact that most of the existing questionnaires offer potential for detailed analysis which is usually not fully exploited in the standard reporting of such data. This is especially the case where such reporting is carried out by organisations tasked with producing overall summaries of entire surveys -and not merely IPVA modules (for example reports produced by the Office for National Statistics, Scottish Government, Northern Ireland Department of Justice). This is something that  raise with regards to ONS publications using CSEW data, as does MacQueen (2016) reflecting upon the potential of the SCJS dataset. It is worth considering to what extent the increasingly stringent data access requirements for many of these datasets are dissuading a larger pool of scholars from using the data in more creative ways.

Conclusion
In this paper we review seven UK longitudinal and repeated cross-sectional population surveys which include questions about IPVA. We have critically engaged with the questionnaires and reflected upon the strengths and limitations of the surveys both individually and collectively in terms of the quality of IPVA data that they have to offer. In doing so we have drawn extensively on literature of methodological discussions around measurement of IPVA in surveys. We conclude by outlining a range of recommendations, summarised in Box 1, to aid both in the development of future iterations of the same surveys as well as in the creation of new surveys focusing on IPVA. Though our geographical context is the UK, most of the limitations we identify are of relevance for IPVA population survey methodology beyond the UK context.

Box 1.
Recommendations for future IPVA survey questionnaires and data collection 1. To address a systematic overemphasis on physical IPVA, future questionnaires should give equal attention to physical and nonphysical forms of abuse 2. Future surveys should test ways to measure psychological abuse and coercive control drawing on qualitative evidence of lived experiences of abuse 3. Surveys should balance the measurement of incidents of abuse with the measurement of patterns of abusive behaviours not adequately captured by asking about distinct incidents 4. We recommend the inclusion of more experiential questions which measure the impacts of abuse for victims 5. Development of new modules intending to measure sexual violence should explore the good templates offered by CSEW and SJCS questionnaires 6. Questionnaires should explore economic violence as a separate form of abuse which would require further questionnaire development in this realm 7. Those wishing to measure stalking and harassment should examine CSEW and SCJS questionnaires which also capture these behaviours within and beyond the context of intimate relationships 8. Since abuse is increasingly taking place virtually, questionnaires should evolve to better reflect this 9. Future questionnaire and survey development should consider how to address overlooked aspects of IPVA, such as reproductive coercion and female genital mutilation 10. Future surveys could collect demographic variables on underexplored characteristics (e.g. sexual orientation, gender identification, religion; more granular disability information) 11. Better longitudinal IPVA data could be collected in future surveys, including data on how IPVA experiences change over the life course and during critical transitions 12. More questions are needed about the sources of support which victims use when experiencing IPVA, including whether they confide in anyone, and if they seek support from services 13. Children's experiences of domestic abuse are poorly measured, and future surveys could consider exploring how children are affected by abusive relationships 14. Future crime surveys across UK nations should share some homogenised IPVA measures to enable cross-national comparison within the UK 15. Future surveys could experiment with changing the positioning of IPVA modules, since their current placement typically at the end of longer surveys is detrimental to the response rate 16. Sampling strategies should be reviewed to explore how underrepresented populations at a higher risk of IPVA can be included in surveys used for national prevalence estimates 17. Routine reporting of existing surveys usually fails to fully exploit the available data. Relevant, organisations should consider proactively commissioning more detailed analysis of IPVA modules
Funding The project has been funded by the Nuffield Foundation grant number WEL/43875, but the views expressed are those of the authors and not necessarily the Foundation. GF's salary was supported by the UK Prevention Research Partnership (Violence, Health and Society; MR-VO49879/1), an initiative funded by UK Research and Innovation Councils, the Department of Health and Social Care (England) and the UK devolved administrations, and leading health research charities.

Declarations
Competing Interests None to declare.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.