Domestic violence and abuse (DVA) is known to have long-term and wide-ranging detrimental impacts on the health and well-being of children and families [1-7]. The UK Domestic Abuse Act 2021 [8] defines “abusive behaviour” as occurring between “personally connected” parties aged ≥16 years and includes “physical or sexual abuse; violent or threatening behaviour; controlling or coercive behaviour; economic abuse; and psychological, emotional, or other abuse.”

Annual DVA statistics are published for Wales and England by the Office for National Statistics (ONS), using information from the Crime Survey for England and Wales (CSEW). The latest available DVA data from the CSEW highlighted that women are much more likely to be exposed to DVA than men, with an estimated 7.3% of women (1.6 million) and 3.6% of men (757,000) aged 16 to 74 years experiencing DVA in the year ending March 2020 [9]. Research studies report varying estimates of the prevalence of DVA toward women in the United Kingdom; in 1 systematic review, the lifetime prevalence of partner violence against women in the general UK population ranged from 13% to 31%, whereas the 1-year prevalence ranged from 4.2% to 6% [10]. The 1-year prevalence of DVA for women attending a general practitioner (GP) in the United Kingdom was estimated to be 17% in a cross-sectional survey study [11] and 0.15% in a recent data linkage study that used Read (Read refers to originator Dr James Read) codes (hierarchical nomenclature used to record clinical information) to search for records of DVA within GP administrative data’ [12]. Another recent primary care data linkage study estimated that the prevalence of women in the United Kingdom aged ≥18 years recorded as having lifetime experience of DVA in 2017 was 0.37% [13].

Hard-to-reach populations are often excluded from DVA research and statistics [14,15]; for example, the CSEW excludes those living in nonpermanent households and therefore risks excluding individuals considered vulnerable in women’s refuges as well as homeless populations [14]. One such group that is considered vulnerable and often hard to reach comprises individuals involved in public law family court proceedings. When a child is identified as having experienced, or is at risk of experiencing, significant harm at the hands of a parent or caregiver, the local authority may issue care or supervision proceedings under section 31 of the Children Act (1989; s.31). Since the advent of the Adoption and Children Act 2002, harm to children now includes being exposed to, or witnessing, DVA [16]. This change to the definition of significant harm reflected growing awareness of the negative impact of DVA on children and families. Nevertheless, this growing recognition has not been matched by sufficient analysis of DVA for families involved with the family courts.

There is an important evidence base on the heightened health vulnerabilities of parents and children involved in public family law proceedings, including high rates of mental health need and substance use [17-25]. However, published research on DVA using large-scale data is more limited, despite DVA being classed as a major public health concern [26]. Drawing on the available research, 2 studies [27,28] conducted in England suggest that the prevalence of DVA in women involved in s.31 care proceedings is much higher than that in women in the general population. In a study of a sample of 386 s.31 applications made by 15 English local authorities, a little more than half of the mothers (51.1%) had their concerns about DVA recorded by children’s services [27]. Similarly, a case file analysis of a random sample of 354 mothers involved in recurrent care proceedings in England from 50 local authorities revealed that 65% of the mothers had been exposed to DVA by the time of the initial proceedings [28]. Local authorities have highlighted that DVA is one of the most important drivers of the increased demand for children’s social care services in recent years [29]. Annual statistics on children “in need” assessments in England have consistently demonstrated that DVA within the household is the most common concern identified at the end of assessments, and it remained the most common factor in 2020-2021 [30,31]. Although comparable data from Wales are unavailable, a survey of adverse childhood experiences conducted in 2015 reported that 16% of the Welsh adult population were exposed to DVA during childhood [32].

Regarding studies using population-level data to understand DVA in the broader child welfare population, a small number of studies are notable. A recent data linkage study reported that children of mothers with experience of DVA in Manitoba, Canada, were significantly more likely to be receiving services from child welfare organizations than children of mothers who reported no prior experience of DVA [15]. Another longitudinal ecological study found that increases in the rate of infant entry into care over time in England were associated with an increased prevalence of maternal history of adversity-related hospital admissions, including hospital admissions related to exposure to DVA [33]. A further study examined patterns of co-occurrence and severity of DVA, substance misuse, and depression in female caregivers involved with child protective services in the United States; one-third of the sample reported high rates of all factors, including both minor and severe DVA, whereas only 9% reported none [34]. Among women with experience of DVA, mothers whose children were taken into care had more children, were more likely to have self-reported addictions, and had significantly less education than mothers without child protective services involvement. In addition, those with exposure to DVA and with children in care reported significantly more physical abuse from their partners, more psychological distress, and lower quality of life than mothers whose children were not in care; however, there were no statistically significant associations with clinical depression or posttraumatic stress disorder [35].

Risk factors for DVA among women in the general population are listed in the National Institute for Health and Care Excellence (NICE) public health guidance on identifying, preventing, and reducing DVA [36], and they include age (women aged between 16 and 24 years are at higher risk than older women), long-term illness or disability, poor mental health, being separated from a partner, pregnancy, postnatal depression, and substance misuse. Other risk factors reported in the literature include intentional and nonintentional injury hospitalizations [15], unemployment, and living in a low-income or single-parent household [37].

The availability of routinely collected population-level administrative data for research provides the opportunity to study cohorts of populations that are hard to reach or considered high risk [38]. This study analyzed administrative records held within the Secure Anonymised Information Linkage (SAIL) Databank [39-41] for a cohort of mothers involved in public family law court proceedings and a matched group of mothers in the general population of Wales. The primary aim was to investigate risk factors associated with primary care–recorded DVA for women involved in public family law proceedings. In addition, we aimed to (1) estimate involvement in public family law proceedings as a risk factor for primary care–recorded DVA compared with a comparison group, (2) estimate the population and cohort prevalence of primary care–recorded DVA, and (3) evaluate whether routinely collected data could be used to create and identify additional risk factors associated with DVA.

Hospital data for inpatient and day case activity were extracted from the PEDW; data from any episode and any diagnostic position were included. ED data were extracted from the EDDS for new attendances only. All measures were calculated across a 2-year baseline period before the index date.

Mothers were linked at the individual level to health and demographic data sets using the ALF, and RALFs were linked to LSOAs to obtain area-level measures. Age, number of children, and youngest child’s age were calculated as of the index date and derived from the Cafcass Cymru and WDSD data for the cohort and comparison mothers, respectively. Deprivation quintiles (most to least deprived using the WIMD income domain) and ONS urban/rural classifications (further divided into sparsely populated areas and nonsparsely populated areas) were used for area-level measures.

The DVA measure was represented as a binary variable where 1 indicated that DVA was recorded, based on the presence of any codes indicative of DVA in a GP-validated clinical Read code list [12]. The Read code list is provided in Multimedia Appendix 1 [12].

Risk factors were created and aimed to represent NICE risk factors. Age was categorized as <25 years or ≥25 years. Binary variables were derived for health measures, where 1 indicated the presence of at least 1 record of interest during baseline measurement. An intellectual disability variable was created from Welsh Longitudinal General Practice data using a validated Read code list [53] as a proxy measure for long-term illness. Mental ill-health (including depression, anxiety, attention-deficit/hyperactivity disorder [ADHD], and autism) and substance use variables (suggesting problematic use of illicit drugs and alcohol) were created using a validated code list [54] for any hospital admissions (PEDW) or GP records. A pregnant or recent childbirth measure was created and recorded for mothers with at least 1 hospital admission record during baseline categorized under the International Classification of Diseases, 10th Revision (ICD-10), as “Pregnancy, childbirth and the puerperium” [55]. A proxy measure of house moves was created as a binary variable to represent the separation risk factor [56,57] using the WDSD, where 1 represents any mother with >1 address registration during the baseline period.

Additional risk factors not included in NICE guidelines were added: assault-related emergency department attendances using the EDDS attendance category, more than one child and child aged <5 years 5 using Cafcass Cymru and WDSD data, living in the 2 most deprived quintiles using the WIMD, and ONS rurality measures further classified into sparsely or nonsparsely populated areas.

The initial analyses, referred to as population-risk analyses, included a general population group of mothers categorized into 2 groups: mothers with family court involvement and those without family court involvement. A second set of analyses were restricted to family court mothers, grouped by the presence or absence of primary care–recorded DVA, referred to as cohort-risk analyses. We examined demographic differences between the groups using the Pearson chi-square test for categorical variables and the Welch 2-tailed t test for numeric variables and reported the corresponding effect sizes and P values.

We calculated the prevalence of primary care–recorded DVA for the cohort and the general population comparison group within a 2-year period before the index date. Univariable and multivariable logistic regression analyses were used to calculate differences in primary care–recorded DVA between the groups and are presented as unadjusted odds ratios (ORs) and adjusted ORs (AORs) with 95% CIs to test for associations between primary care–recorded DVA and risk factors. All risk factors that were created to represent those from the NICE guidelines were included with no further variable selection in model 1 and model 2 for the population-risk analyses and the cohort-risk analyses, respectively. This was to allow for comparison of results between the population-risk analyses and the cohort-risk analyses. A final adjusted model, model 3, was created with the aim of finding the best-fitting model for the cohort mothers balanced with achieving a parsimonious model, combining NICE risk factors and additional risk factors. Model 3 variable entry was based on the inclusion of model 2 terms with P values <.20, along with inclusion of additional risk factors, which improved the model fit using the area under the curve (AUC).

The SAIL Databank uses an independent governance review panel (IGRP) that assesses each application for approval. For general data processing activities, the SAIL Databank relies on General Data Protection Regulation (GDPR) Articles 6 and 9 (provisions for a task carried out in the public interest). Cafcass lawfully shares information about family court proceedings for approved research within the SAIL Databank under Practice Direction 12G–Communication of Information (paragraph 2.1).

After cohort creation, the total sample of 237,866 consisted of 4695 (1.97%) cohort mothers (ALF match rate: 5204/5988, 86.91%; Figure 1) and 233,171 (98.03%) comparison mothers. The 2-year prevalence of DVA was 5.73% (269/4695) for the cohort and 0.27% (626/233,171) for the general population comparison group.

The demographic characteristics varied considerably between the cohort mothers and the general comparison group with statistically significant differences in age, deprivation, and rurality measures (Table 1). Cohort mothers were approximately 9 years younger, more likely to live in deprived areas (2355/4695, 50.16%, vs 53,849/233,171, 23.09%, in the most deprived quintile), and more likely to live in urban city and town areas than more rural and sparsely populated areas. No within-cohort differences were observed for deprivation, but there were statistically significant differences in terms of rurality measures, with mothers with primary care–recorded DVA more likely to live in sparsely populated areas than those with no primary care record of DVA (50/269, 18.6%, vs 282/4426, 6.37%, respectively). Mothers exposed to DVA were also on average slightly younger (27.8, SD 7.0, years vs 29.0, SD 8.2, years), which was statistically significant but with a small effect size (t₃₁₄=2.7).

Compared with the general population group, the cohort had a higher proportion of mothers aged <25 years (1607/4695, 34.22%, vs 11,230/233,171, 4.82%), mothers with mental health conditions (2638/4695, 56.19%, vs 39,539/233,171, 16.96%), mothers with an intellectual disability (120/4695, 2.56%, vs 197/233,171, 0.08%), mothers with substance use conditions (1007/4695, 21.45%, vs 2430/233,171, 1.04%), mothers who had a recent pregnancy or childbirth (2739/4695, 58.34%, vs 44,720/233,171, 19.18%), and mothers who had multiple house moves (2648/4695, 56.4%, vs 48,579/233,171, 20.83%).

The unadjusted odds of mothers in the cohort having a record of DVA was almost 23 times higher than those of mothers in the general population (OR 22.8, 95% CI 19.5-26.1). All NICE risk factors were found to have statistically significant positive associations with primary care–recorded DVA with large effect sizes (Table 2). Mothers with an intellectual disability (OR 10.6, 95% CI 5.6-18.0) or substance use conditions (OR 10.3, 95% CI 8.4-12.5) were 10 times more likely to have primary care–recorded DVA than those without an intellectual disability or without a substance use condition, respectively.

Adjusting for the differences in risk factors between the groups, the cohort were 8 times more likely to have primary care–recorded DVA (AOR 8.0, 95% CI 6.6-9.7; model 1). All 6 NICE guideline risk factors retained the same direction of association with primary care–recorded DVA, albeit with smaller effect sizes, and 4 (%) remained statistically significant after adjustment: being aged <25 years, having a mental health condition, having multiple house moves, and having substance use conditions. Model fit metrics showed AUC=79% and Hosmer χ²₈=18.4 (P=.02).

Within the cohort, comparing differences between mothers with a record of DVA and those without a record of DVA, there was a higher proportion of mothers with DVA with all risk factors, with 1 exception: there was a slightly higher proportion of mothers with no recorded DVA living in the 2 most deprived quintiles (Table 3). Unadjusted ORs indicate significantly higher risk of mental health conditions, substance use conditions, and house moves among mothers with primary care–recorded DVA than among those without primary care–recorded DVA, all with relatively small effect sizes with OR values of <2. Additional risk factors that were significantly associated with primary care–recorded DVA within the cohort included living in a sparsely populated area, assault-related ED attendances, and child-related variables (>1 child and child aged <5 years), with OR values between 1.5 and 3.4. Age indicated a protective effect (OR 0.98, 95% CI 0.97-1.0). Adjusting for all NICE risk factors in the model (model 2), only mental health remained as a significantly associated risk factor: mothers involved in the family courts with primary care–recorded DVA had 75% higher odds of having mental health issues than those without primary care–recorded DVA. House moves and substance use became nonsignificant when adjusting for all other factors. Model 2 fit metrics showed AUC=60% and Hosmer χ²₇=5.6 (P=.69).

The final model 3 included 3 NICE risk factors (mental health, substance use, and house moves), along with 3 additional variables (living in a sparsely populated area, assault-related ED attendances, and having >1 child). The risk factors associated with the largest effect sizes were living in a sparsely populated area (AOR 3.9, 95% CI 2.8-5.5) and at least 1 assault-related ED attendance (AOR 2.2, 95% CI 1.5-3.1) for family court mothers with primary care–recorded DVA compared with mothers without primary care–recorded DVA. Mothers with primary care–recorded DVA were also 82% more likely to have >1 child (AOR 1.82, 95% CI 1.41-2.35), 69% more likely to have a mental health issue (AOR 1.69, 95% CI 1.29-2.23), and 35% more likely to have multiple house moves (AOR 1.35, 95% CI 1.04-1.8). Model 3 metrics showed AUC=66% and Hosmer χ²₇=4.6 (P=.80).

Figure 2 illustrates AORs and 95% CIs of risk factors for all 3 models. The large effect size (AOR 8.00, 95% CI 6.62-9.65) for family court involvement in model 1 is excluded from Figure 2 to aid visualization of smaller effect sizes. Figure 3 shows unadjusted ORs and AORs and 95% CIs for all risk factors tested within the final model 3.

This study found a 2-year primary care–recorded DVA prevalence of 5.73% (269/4695) and 0.27% (626/233,171) for a cohort of public law family court mothers and mothers from the general population of Wales, respectively. We found that mothers involved in the public law family courts were at increased risk of DVA, being 8 times more likely to have primary care–recorded DVA than the general population group. Factors that were significantly associated with primary care–recorded DVA in the general population group included being aged <25 years, having a record of a mental health or substance use condition, and having multiple house moves. Of these factors, only mental health remained significantly associated with primary care–recorded DVA within the cohort of mothers involved in the public law family courts. Additional variables, informed by the literature and derived from linked administrative data, were also significantly associated with primary care–recorded DVA within the cohort of family court mothers; these included living in a sparsely populated area, having an assault-related ED attendance, and having >1 child. However, deprivation, so often associated with risk of poor health, was not found to have a significant association with primary care–recorded DVA.

Our reported 2-year prevalence of primary care–recorded DVA in the general population (626/233,171, 0.27%) is in line with a 1-year prevalence of 0.1% reported in a recent study [12]. For the family court cohort, the prevalence of 5.73% (269/4695) is comparable with a previous data linkage study reporting a 1-year prevalence of DVA of 5% in mothers whose children had entered public care [58]. Furthermore, our finding that these mothers are 8 times more likely to have a record of primary care–recorded DVA than the general population group is similar to that of previous research reporting AOR values of 8.8 for an association between child welfare involvement and DVA screened for at childbirth [15]. Finally, our findings are similar to recent evidence showing that 4% of the women involved in private family law cases in Wales had exposure to DVA noted in their GP records in the year before the proceedings [59].

Our findings are consistent with those of previous studies that have highlighted the co-occurrence of DVA with mental health and substance use problems in the general population [15,60,61]. Studies have also highlighted the high prevalence and co-occurrence of DVA, poor mental health, and substance use among parents whose children have been taken into care [33,34,59,62]. However, in this study, only mental health remained significantly associated with DVA within the cohort of mothers involved with the family courts, once all other risk factors were adjusted for, and substance use was not significantly associated with DVA. Although the relationship between poor mental health and DVA is likely bidirectional, this finding emphasizes that mothers involved with the public law family courts who have experienced DVA are likely to have an additional need for mental health services and ongoing mental health support.

Maternal age is commonly reported as having negative associations with DVA, with younger age relating to increased risk [36,37]. In this study, being aged <25 years was associated with DVA in the general population group but not in the cohort of mothers involved in public law family court proceedings, which may be explained by the much higher proportion of mothers aged <25 years in the cohort. Of the additional variables included in this study, perhaps the finding that DVA was associated with living in a sparsely populated area is the most interesting. Previous research links rurality with increased prevalence, severity, and duration of DVA. Possible reasons for this are that perpetrators may commit more chronic and severe DVA, mothers may have less access to social support networks in rural settings, there may be a lack of access to services such as hostels to potentially escape abusive behavior, and differential policing and health care may also have an impact [63,64]. The issue is complex, and there are evidently many plausible reasons to explain such an association. Additional data would be required to allow further investigation into these areas. Interestingly, deprivation, so often associated with risk of poor health, was not found to have a significant association with primary care–recorded DVA. Understanding this result would require further research but could indicate that using area-level measures of deprivation, which lack granularity, is not a suitable approach to examining the relationship between DVA and deprivation and that individual-level deprivation measures may be more robust for identifying risks associated with primary care–recorded DVA. Given underreporting of DVA within primary care settings, it may also be that those who actively engage with their GP and report DVA are not the most deprived.

At the time of writing, reliable ethnicity data were not available in the Cafcass Cymru data; therefore, this aspect was not explored in these analyses. Work is underway to improve the quality of ethnicity recording. However, recent research has highlighted the heterogeneity of the relationship between DVA and mental or physical health based on ethnicity [65], finding that non-Hispanic Black women and non-Hispanic White women who experienced DVA were more likely to experience poor mental health over time, whereas there was no association between DVA and mental health outcomes for Hispanic mothers. In addition, only Black mothers who experienced DVA were more likely to report more physical health problems over time.

This research provides new evidence in terms of DVA prevalence and understanding of risk factors for a group of mothers traditionally hard to reach and considered vulnerable, some of whom may not be represented in official DVA statistics. The findings indicate that risk factors for DVA for mothers involved in public law family court proceedings differ from risk factors for DVA for the general population of women, as noted in the national NICE guidelines. With additional risk factors highlighted, there may be value in creating bespoke guidelines specifically relating to groups considered vulnerable, such as mothers involved in the family courts who were found to be much more likely to experience DVA than mothers in the general population.

The additional information may be useful in providing further information for those working with such individuals considered vulnerable in terms of policy setting or providing frontline assistance; for example, the increased risk of DVA for mothers living in sparsely populated areas could be used as evidence to improve support services in such areas. Increased risks associated with ED assault may be an obvious link but could be used to emphasize or provide additional services or signposting within EDs.

The findings corroborate underreporting of primary care–recorded DVA, an obvious limiting factor to this research. Purposefully improving such data collection would allow more robust analyses to corroborate or improve the evidence base to provide to stakeholders to allow more informed changes to be made to help those in need. Programs aimed at improved identification and reporting of DVA have been developed and run within GP practices [75], and research projects have reviewed and recommended improvements to DVA recording across health settings [70]. Programs such as these will improve the quality of data collection and subsequent use of the data for research.

This study provides information on risk factors associated with DVA for a cohort of mothers involved in public family law proceedings who self-report to a GP compared with a general population comparison group and highlights the increased risk for this cohort of mothers considered vulnerable. Further evidence of the association between having a mental health condition and experiencing DVA should help to corroborate the necessity for continued and further resources targeted at specialist mental health and DVA support services. The additional evidence that this study adds in highlighting risk in sparsely populated areas should further inform service design.

An initial model including the general population of Wales shows common risk factors (NICE) can be reproduced with these data and analytical approach and highlights the increased association of primary care–recorded DVA for mothers involved with the family courts. Analysis of the same risk factors solely within the cohort of family court mothers emphasized key differences in associations with DVA. A final model that included additional risk factors helped to further explain some of the variability for the cohort, which adds more information and understanding of this population.

The results support the case for the use of administrative data to improve understanding of DVA risk factors, especially for groups considered vulnerable, to improve the evidence base and inform policy guidelines. Such groups that are hard to reach and considered vulnerable are more likely to be excluded from more traditionally gathered statistics, whereas, in this situation, their interaction with health services is recorded as standard; the inclusion of such individuals lessens implications of external validity and generalizability [76].

However, caution is urged in the interpretation of the results because the variability in individual situations and the complexity of lived experiences of DVA cannot be fully explained using the available data and modeling approaches used in this study. Improved reporting of DVA in GP administrative data would improve validity in future work. Nonetheless, accepting such limitations, the information provided is valuable and highlights risk factors that may be useful for policy makers and those services working with such mothers considered vulnerable.