Maintaining Relationship Safety while Promoting Relationship Health: Family Service Referrals among Veterans Screened for Intimate Partner Violence

Abstract

Intimate partner violence (IPV) poses a tremendous public health burden across large healthcare systems. While the predominant response to IPV focuses on individual screening and referral, relationship and family services provide an opportunity to prevent low-risk verbal aggression from escalating and to address some forms of bidirectional physical IPV. This study examines mental health and screening variables associated with IPV experience and subsequent referral to family services using a large national dataset of N=256,894 patients screened for IPV in the first year of the Veterans Health Administration’s adoption of a national IPV screen. Of those screened, 7.4% reported any IPV, 1.4% reported severe IPV, and only 0.1% received a referral for couple and family services within the next 90 days. Mental health conditions associated with greater likelihood of IPV (e.g., personality disorders, PTSD, substance use disorders, and depression) were in turn were associated with greater likelihood of referral. Despite the overlap between IPV and mental health conditions that benefit from couple-based interventions, referrals remained infrequent. Referrals were often placed by a different clinic than the service assessing the IPV. Risk behaviors for severe violence (e.g., choking) and suicide ideation were also associated with increased couple and family service referrals, underscoring the need for integrated assessment and risk management. These findings highlight important opportunities for healthcare systems to enhance the alignment of IPV screening and appropriate treatment recommendations by strengthening screening-to-referral pathways for family services and integrating systematic safety assessments.

Intimate partner violence (IPV) – which includes verbal and psychological aggression, physical violence, sexual violence, and stalking – poses a tremendous public health burden (Breiding et al., 2015). Across studies, IPV experience has been linked to a range of mental health symptoms and conditions such as PTSD, alcohol misuse, and depression (Lagdon et al., 2014). In severe cases, IPV can lead to severe injury or even death, making the identification and management of IPV an important goal for maintaining individual health and well-being. At the same time, many couples themselves seek to not just eliminate IPV, but cultivate fulfilling intimate relationships. In many cases, patients actively request the involvement of their partners in their own mental health treatment and seek out available couple and family services – formal services from a mental health professional that include some level of participation by a partner or family member. The competing demands of potential IPV safety risks against patient preferences for partner involvement can be challenging for healthcare systems and providers when selecting the best treatment. The current study examines the intersection between the identification and management of relationship harms and the use of couple and family services designed to support relationship health in the context of Veterans Health Administration (VHA) care.

Potential Role of Couple Therapies across IPV Risk Levels

The dominant medical perspective on IPV focuses on identifying and intervening with patients experiencing IPV (often termed “victim” or “survivor”). This framing is reflected in both the 10th edition of the International Classification of Diseases (ICD-10; World Health Organization, 2016), which includes codes for experiencing IPV but none for using IPV, and in the current U.S. Preventive Services Task Force (USPSTF, Curry et al., 2018) recommendation to screen for IPV experience (but not usage) in women of childbearing age. However, epidemiological work examining dyadic patterns of IPV suggests a large portion of IPV occurs “bi-directionally,” with both partners taking part in using IPV on one another (Johnson, 2008; Stover et al., 2025), suggesting a possible use for couple therapies – psychotherapeutic treatments that involve both members of the couple participating together.

Research creating typologies of IPV most commonly reported form of IPV is bidirectional verbal aggression reflecting recurring arguments while a smaller subset of relationships uses more varied forms of physical IPV, sexual IPV, and psychological coercion (Portnoy et al., 2024; Straus & Douglas, 2004). Bidirectional verbal aggression is addressed by almost all behavioral couple therapies under the framing of developing “communication skills” or “emotion regulation” skills. Meta-analyses suggest that early relationship education-based skills training can reduce the emergence of later verbal aggression (Blanchard et al., 2009) and couple therapies can reduce verbal aggression in distressed couples (Roddy et al., 2020). Even among relationships that use physical violence, studies of IPV typologies suggest that a majority may be characterized as “situational couple violence” -- bi-directional physical IPV that emerges from partners’ emotional dysregulation – that is, a lack of self-control rather than an attempt to control others (Johnson, 2008). In the presence of bi-directional IPV, classic couple therapies maintain their benefits in reducing verbal aggression and increasing relationship satisfaction (Simpson et al., 2008). Furthermore, a new class of couple therapies specifically designed to address physical and/or sexual IPV in couples with situationally violent patterns demonstrate large reductions in physical IPV that are comparable to individual approaches (d=0.84 and d=0.85, respectively; Karakurt et al., 2016, 2019).

Couple therapies may be inappropriate for managing the highest risk IPV. Research suggests that approximately 20% of couples with physical violence would be classified as part of an “intimate terrorism” relationship where one partner predominantly uses physical, sexual, and severe psychological IPV (e.g., isolation, intimidation/threats, and coercion) to control the other partner (Johnson, 2008; Johnson et al., 2014). Clinical researchers and theorists have expressed concerns that couple therapies might reinforce controlling patterns by discouraging threatened partners from leaving the abusive relationship or supporting victim blaming (Stith & Spencer, 2021). There is also concern that couples may experience escalations in conflict during couple therapies that could be dangerous for more severe IPV behaviors (Stith & Spencer, 2021). Therefore, even couple therapy programs specifically developed for IPV exclude couples when there are signs of high injury risk (e.g., strangulation; IPV with weapons; recent injuries) or one-way power are present (Karakurt et al. 2016, Taft et al. 2016). In contrast, many individual IPV treatments show efficacy without excluding for these behaviors and a substantial portion demonstrate effects while working with court-mandated participants who often use controlling or severe assaultive behaviors (Karakurt et al., 2019).

IPV Screening’s Potential to Inform Couple and Family Service Referrals

The above research suggests there is no one type of therapy appropriate for all the various typologies of IPV. But this literature does suggest clinical decisions can be made along a risk continuum. Couple and family services can be offered widely for patients with verbal aggression only, can be offered in select situations when physical or sexual IPV are used, and are unsupported when there are signs of control or injury risk. However, both patient level and provider level factors may lead to either over- or under-referring to family services. At the patient level, patients with more severe relationship problems have stronger preferences for couple-based over individual therapy (Crasta et al., 2022), paradoxically increasing the desire for couple/family services even when they may be inappropriate. At the provider level, there is an active debate about the appropriateness of conjoint treatment for IPV, with differing views based on professional discipline and training (Stith & Spencer, 2022). Providers trained in traditional models (e.g., Duluth model; feminist critiques) may be hesitant to refer for couple and family services after finding any level of IPV.

Routine IPV screening in healthcare settings could reduce these forces, both increasing identification of patients who may benefit from couple therapy and flagging at-risk couples for referral to individual services. One screening tool was specifically designed with this stratification in mind, the Screener for Clinically Significant IPV (Heyman et al., 2021). This tool combines high sensitivity for verbal-only IPV (i.e., potentially appropriate for couple/family treatment) with a high-specificity detection for physical IPV (i.e., possibly inappropriate; further evaluation is needed). Another way to support stratification is through multiple screenings, such as the “Relationship Health and Safety” protocol (RHS) used by the VHA. RHS uses a staged approach to minimize time burden and enable implementation on a large scale across the healthcare system (Iverson et al., 2019). The first stage of this screener uses the Hurt, Insult, Threaten, and Scream scale (HITS; Sherin et al., 1998) supplemented with an additional sexual IPV item. At this stage, sensitivity is maximized to detect participants experiencing any level of psychological, physical, or sexual IPV (i.e., potentially appropriate for couple/family treatment, although further evaluation would be needed depending on the items and the frequency of experiences endorsed). Individuals screening positive on that primary screener then move onto the second stage composed of three items from the Danger Assessment 5-item screener (Messing et al., 2017) that were selected for their ability to predict increased likelihood of injury in civilian samples (i.e., likely inappropriate for couple/family therapy at this time). Despite the potential for routine IPV screening to inform couple and family service referrals, there has been limited exploration of this possibility.

Development of meaningful referral pathways begins by understanding how IPV screening and family services currently interact in healthcare systems. As a case study, we consider VHA’s expansive nationwide rollout of the RHS screening procedure through the “Intimate Partner Violence Assistance Program” directive (IPVAP; VHA, 2019a). The directive followed from facility-specific efforts to implement IPV screening programs and was informed by recognition of barriers (e.g., time/task constraints in primary care, lack of policy, lack of training) and facilitators (e.g., champions, external supports, positive feedback, and national attention) that differentiated early-adopters of routine screening from facilities that did not take up this practice (Iverson et al., 2019). To this end, IPVAP established policy requiring every facility to implement the RHS screening procedure (at minimum for women of childbearing age and by facility discretion in other groups) and established dedicated IPVAP coordinators at every VHA medical center to promote training, develop IPVAP champions within teams, promote IPV awareness, and provide positive feedback around RHS screening rates. Primary care was particularly targeted for implementation given its alignment with the US Preventive Services Task Force’s focus on preventative care settings (Curry et al., 2018), but screening was also implemented in other clinical settings (e.g., mental health and social services, HIV clinics). An important component of RHS is ensuring patients who endorse IPV, and especially the highest risk patients, are offered and connected to appropriate services (e.g., warm handoff to a social worker for IPV-specific counseling; referral to individual therapy to address related mental health concerns). Similarly, following VHA’s fifteen years of investment in couple and family services, its 2019 update to the “Family Services in Mental Health” directive (VHA, 2019b) was the first to explicitly state that all patients with ICD-10 mental health concerns should be asked annually if they would like their family members involved in their mental healthcare care and offered opportunities to receive “marital and family counseling,” either within the facility or through community providers. To support this explicit standard, the directive was accompanied by expanding dissemination and training efforts of evidenced-based couple therapy services for Integrative Behavioral Couple Therapy (Christensen et al., 2020) and Cognitive Behavioral Conjoint Therapy for PTSD (Monson et al., 2012). Thus, beginning in 2019, facilities were expected to ask patients annually about the experience of IPV (at minimum female patients) and their preferences for family involvement in mental healthcare (at minimum for those with mental health diagnoses), but it is not clear at this time how these two expectations interfaced. Similarly, VHA facilities were required to provide a range of offerings for IPV and relationship concerns that account for individual mental health concerns, but it is not clear to what extent these options overlapped.

Current Study

The current study uses the parallel expansions of IPV screening and family services to understand clinical factors that impact the intersections between IPV experience and referrals for family services. Using VHA electronic health record data (EHR), we examine factors that predict (1) disclosures of IPV on the RHS screen and (2) whether a referral for family services is placed in the 90 days following that screening. As predictors, we include mental health conditions that are a specific requirement for family services in VHA and have been previously linked to IPV experience and family service utilization (McKee et al., 2022). We also consider screening circumstances and clinical functioning measures which may be considered by providers when evaluating appropriateness of family services after an IPV screen.

Method

Procedures

This study was approved by the Institutional Review Board at the Central Virginia Veterans Affairs (VA) Health Care System with a waiver of informed consent. Once approved, all study variables were extracted through the VA Informatics and Computing Infrastructure’s Corporate Data Warehouse for all patients who completed the RHS screen from April 2019 through February 2020 – after the national launch of the RHS protocol, but before the COVID-19 pandemic began to disrupt screening and family service utilization (McKee et al., 2023). Extracted data included IPV screen items, demographics, military service characteristics, mental health conditions, clinical severity indicators, and screens.

A total of 439,298 IPV screens were initiated between April 2019 and February 2020. Approximately 39.5% (n = 173,670) screens were incomplete. While reasons for non-completion were not always documented, potential reasons could include but are not limited to declining the screen, or not currently able to complete the screen privately (e.g., partner present for a medical visit). Of the remaining 265,628 completed Stage 1 screens, 64 of these were excluded from the study because they did not have item-level data that would allow us to determine the severity of IPV (see below). Some patients had multiple IPV screens during the study timeframe. In these cases, we retained the earliest positive primary screen with the fewest missing items or the earliest screen overall in cases where the patient never had a positive screen, resulting in a final sample of N=256,658 unique patients (see Table 1).

Table 1.

Demographic Characteristics, Mental Health, and Screening Results of Sample (N=256.894)

Demographic Characteristics	N	%	Clinical Characteristics	N	%
Age 65+	117,712	45.82	Past year IPV behaviors (Stage 1)
Sex			Scream	16,913	6.58
Male	194,743	75.81	Insult	12,875	5.01
Female	61,702	24.02	Threaten	2,704	1.05
Marital status			Hurt	2,013	0.78
Married	142,691	55.54	Sexual coercion or assault	1,046	0.41
Separated	8,512	3.31	Risk factors for future injury (Stage 2)
Divorced	54,696	21.29	Increased frequency/severity of IPV	2,129	13.831
Single/never married	36,738	14.30	Past year choking/strangulation	756	4.691
Widow/widower	9,534	3.71	Fear of being killed by partner	674	4.181
Race			Service Category of assessing provider
American Indian or Alaska Native	1,741	0.68	Primary care	129,054	50.24
Asian	1,430	0.56	Mental health	18,923	7.37
Black or African American	45,709	17.79	Other	107,258	41.75
Native Hawaiian or Other Pacific Islander	1,361	0.53	Past-Year Transdiagnostic Screens
White	189,500	73.77	Suicide Plan, Intent or, Behavior	4,415	1.72
More than 1 race reported	2,112	0.82	Heavy Drinking	248	0.10
Hispanic/Latino	9,615	3.74	Mental health conditions
Period of Service			Personality disorders	4,267	1.66
Post 9/11 conflict era	38,787	15.10	Bipolar depression	9,456	3.68
Pre 9/11 conflict era	117,489	45.73	Major depression	60,287	23.47
Non-conflict era	98,049	38.17	Pervasive dysphoric disorders	4,976	1.94
Rurality			Other mood disorder	19,488	7.59
Urban	156,820	61.04	Posttraumatic stress disorders	68,189	26.54
Rural	85,833	33.41	Adjustment disorders	19,620	7.64
Highly rural	13,785	5.37	Other trauma	6,406	2.49
Combat	68,727	26.75	Anxiety disorders	49,753	19.37
Military Sexual Trauma	39,146	15.24	Obsessive compulsive disorder	1,423	0.55
			Substance use disorders	46,306	18.03
			Schizophrenia	5,222	2.03
			History of psychiatric inpatient	21,525	8.38
			Service-Connected Disabilities
			0% SC/Not Service Connected	101,244	39.41
			10–50% Service Connected	61,863	24.08
			60–100% Service Connected	93,787	36.51

Notes. IPV=Intimate partner violence

¹Percentages for Stage 2 responses use the n=16,119 patients that received a Stage 2 screen as the denominator rather than the whole sample.

Measures

RHS screening.

Relationship Health and Safety screens were identified using EHR health factors referring to IPV or domestic/interpersonal violence screens and supplemented through text search for item wordings. Stage 1 of the screen uses an extended HITS tool, which asks how often in the past 12 months a partner or ex-partner has insulted or talked down to you, screamed or cursed at you, threatened you with harm, physically hurt you, and forced or pressured you to have sexual contact against your will or when you were unable to say no Although HITS responses are given in a frequency scale (Never to Frequently) in the first year of the RHS rollout some facilities’ systems saved the results for each item dichotomously without any frequency information (i.e., simply noting whether the behavior associated with that item was present or absent in the past year). The stage 2 screen includes three items from the Danger Assessment-5 (Messing et al., 2017) that predict a high likelihood of injury (i.e., inappropriate for couple therapy), including IPV increasing in frequency or severity in the past 6-month increasing IPV, history of strangulation/choking by the partner/ex-partner, and fear of being killed by one’s partner/ex-partner. Response options for each of these three items are dichotomous (yes/no).

The absence of frequency information for HITS items meant we could not convert the items into scores allowing for potential clinical cut-offs (Iverson et al., 2015). However, we could still look at the items themselves as measures of the types of behaviors used (i.e., Rarely or higher when frequency measures are available; Present for items that only had dichotomous information). Therefore, for the purposes of this study, we developed two outcomes. The first represented the endorsement of “any IPV” behavior, capturing VHA’s broader definition of Stage 1 positivity designed to maximize sensitivity. The second represented a more selective focus on of “severe IPV behaviors” that represent IPV behaviors that are lower in frequency than verbal aggression while leading to increased negative outcomes including severe psychological IPV (“threaten” item), physical IPV (“hurt” item) and sexual IPV (Straus et al., 1996).

Service category of assessing provider was extracted from staff data in the EHR. We used service department information associated with the staff member marked as the primary provider of services during the screening appointment. Service areas were categorized as mental health, primary care/general medical care, or other. Primary care areas included primary care, patient aligned care teams, general medicine, family practice, integrated care clinics; or clinic titles that referenced community-based outpatient clinics without further specification. Mental health services areas included general mental health, social work, or substance use programs. Other could include other specialty medical services and community care coordinators.

Subsequent referrals to couple and family services.

For each patient, we extracted any “Marriage and Family” referrals from the EHR consult system. Veterans were coded as positive if they had at least one referral placed within 90 days following the index screen. As consults are placed by a specific provider, we were also able to extract the referring provider’s staff data to classify their service category (i.e., primary care, behavioral health, or other) using the same categories above.

Demographic controls.

Demographic variables extracted included sex, age, race, ethnicity, marital status, and rurality. Age was calculated based on the date of the IPV screen administration. We extracted the marital status that was current at the time of the screen. Rurality was determined based on the patient’s primary residential ZIP code at the time of the screen, which was matched to the most recent version of the Rural-Urban Commuting Area Codes (Economic Research Service, 2013). Rurality codes were categorized into urban/suburban, rural, and highly rural groups based on methods used by the Office of Rural Health.

Some aspects of veterans’ military service are also noted in the EHR including the years they served, and certain military experiences related to eligibility for specific services. Period of military service was defined as the period associated with the veterans’ first entry date. We also evaluated indicators of combat exposure such as documentation of eligibility for combat-related services at any time in the veteran’s medical record. Finally, we examined experience of military sexual trauma (MST), which can be noted either through documentation and health factors or through positive MST screens.

Transdiagnostic screens.

We also extracted two other screens that represent cross-cutting clinical concerns that are not specific to a single diagnosis and are reviewed at annually in VHA care. The first is the Columbia-Suicide Severity Rating Scale (CSSRS; Posner et al., 2011). Patients were classified as positive if they reported a suicide plan or suicidal intent in the month prior to the screen or suicidal behavior (e.g., attempt, interrupted attempt, or preparatory behavior) in the three months prior to the screen. We also screened for past-year unhealthy drinking levels using the Alcohol Use Disorders Identification Test Consumption questions (AUDIT-C; Bush et al., 1998). For both screens, we looked for any screen that was completed in the 12 months before the RHS screen. In the case of multiple screens over this timeframe, we chose the screen closest to the index IPV screen’s administration date.

Mental health conditions.

Information on patients’ mental health conditions was extracted from their list of diagnoses (“problem list”), diagnoses associated with medical appointments, and disability- or service-connected conditions. Patients were considered positive for a mental health condition if the initial diagnosis or documentation date preceded the date of the IPV screen and had not been marked as “resolved” prior to the screen. We included mental health diagnoses within the year preceding the IPV screen using ICD-10 diagnoses. We then identified whether patients had prominent diagnoses that have been previously associated with family service utilization (e.g., PTSD, Bipolar Disorder; Personality Disorders; Trauma-Related Disorders; Mckee et al., 2022). Variables were coded as binary predictors (1 = Patient had one or more disorders in this category during the assessment period; 0 = not present).

Clinical severity indicators.

Service-connected conditions are illnesses or injuries that were caused or worsened by military service. VHA medical records of veteran patients include “service-connected ratings” of 0–100% to rate the degree to which the sum total of service-connected conditions (including mental health conditions) is expected to interfere with employment. Veterans without any service-connected conditions or who were 0% service-connected were considered not to have service-connected disabilities. Those with service-connected conditions were categorized according to the total percentage of service connection. Veterans were considered positive for a history of inpatient mental health treatment if they had any inpatient visits in a mental health coded department prior to the screen, including psychiatric hospitalization or long-term residential treatment programs.

Analytical plan

All data structuring and analysis was completed using Microsoft SQL Server Management Studio and SAS Enterprise Guide 8.3 (SAS Institute, Inc.). For demographics and other sample characteristics, we report frequencies using unimputed data. For predictive analyses, we used multiple imputation procedures with 20 imputed datasets to account for missing data. We conducted three analyses predicting endorsement of experience of any IPV (i.e., RHS Stage 1 positivity), specifically severe IPV behaviors (i.e., threats, physical IPV, or sexual IPV), and whether a referral for family services was placed in the VHA consult system within 90 days after a completed IPV screen. For each analysis, we explored each predictor both with single logistic regressions to test unadjusted associations among each predictor and the outcome and then a second time adjusting for demographic variables describing the patient and their military service (age, sex, race, ethnicity, rurality, first era of military service, combat status, and MST status). When examining placement of family consults after a screen, the three RHS predictors (any IPV and specific endorsement of physical/sexual/severe psychological IPV on the Stage 1 screen, and endorsement of injury risk indicators on the Stage 2 screen) were added as a set, so that each also adjusts for one another. For categorical predictors with more than 2 categories, Wald tests were used to determine significance level; those with an associated p-value of < .05 were further investigated by estimating all possible pairwise contrasts among the predictor levels. Predictors were considered significant if they were associated with p <.05 and an odds ratio of >1.43 or <.70 (i.e., at least a small effect size; Chinn, 2000).

Results

Descriptives

Table 1 shows descriptive statistics detailing the demographic, military service, mental health, and screening characteristics of the sample. A total of 19,314 patients (7.5%) screened positive at Stage 1 (i.e., reporting any of the five IPV behaviors at rarely or higher), with 3,656 (19%) endorsing at least one of the severe IPV items (i.e., threaten, physically hurting, or sexual coercion/assault, see Table 1 for item-level data). Of the patients with positive Stage 1 screens, 16,119 (83%) had complete the Stage 2 screen. Out of this group, 2,653 were positive for one or more injury risk factors, with 56% of the Stage 2 positives (n=1,494) also reporting one or more severe IPV behaviors based on the Stage 1 screen. Another 3,195 patients with a positive Stage 1 screen were missing Stage 2 screen data, thus precluding us from using Stage 2 data as an outcome and requiring us to treat this as missing data (i.e., requiring multiple imputation) when Stage 2 positivity was used as a predictor.

Within our final analytic sample (N=256,658), only 264 patients (0.10%) received a referral for family services within 90 days of their screen. Half of these participants received their referral within a month of screening (Median=31 days between screening and referral) and 46 (17%) received the consult on the same day. Only 55 (21%) of these referrals were placed by the same provider who completed the screening with another 41 (16%) placed by a different provider within the same clinic type as the screening provider (e.g., a primary care provider conducting a screening followed by a co-located primary care social worker placing a referral). Most consults (76%; n=200) were placed by providers in mental health clinics and only 17 (6%) were completed by providers in primary care clinics.

Aim 1: Association of Intimate Partner Violence with Clinical Presentation

Past-year screens for transdiagnostic concerns showed moderate associations with IPV (Table 2). Patients with past-year positive suicide screens had moderately increased odds of reporting past-year IPV with OR=3.54 and large increased odds of reporting past-year severe IPV behaviors such as physical, sexual, or severe psychological IPV (OR=6.69). These associations remained after adjusting for patient characteristics (i.e., age, sex, race, ethnicity, marital status, rurality, military service era, combat experience, or sexual trauma), with an adjusted odds ratio (AOR) of 2.82 for any IPV and AOR=4.27 for severe IPV. Small adjusted associations were also observed among patients who disclosed past-year problematic drinking for both any IPV (AOR=1.88) and severe IPV (AOR=2.40). The service category of the screening provider (i.e., primary care, specialty mental health, or other) had large effects on endorsement. In contrast to patients screened by mental health service providers, patients screened by a primary care clinician were less likely to report any IPV (AOR=0.20) or severe IPV (AOR=0.18). Similarly, veterans screened by other types of providers (e.g., specialty medical services; community care coordinators) also had lower odds of reporting any IPV (AOR=0.20) or severe IPV (AOR=0.15) compared to those screened by mental health providers.

Table 2.

Associations between Clinical Characteristics and Intimate Partner Violence (IPV) Screening Results

	Any Past Year IPV Experience			Past Year Experience of Any Severe IPV (Threats/Physical/Sexual)
Variable	% Positive	OR (95% CI)	AOR1 (95% CI)	% Positive	OR (95% CI)	AOR1 (95% CI)
Whole Sample	7.37	---	---	1.42	---	---
Service Category of Assessing Provider2
Mental health	26.36a	--	--	6.58a	--	--
Primary care	5.96b	0.19 (0.18, 0.20)	0.20 (0.20, 0.21)	0.89b	0.13 (0.12, 0.14)	0.18 (0.16, 019)
Other	6.06b	0.19 (0.19, 0.20)	0.20 (0.20, 0.21)	1.11b	0.16 (0.15, 0.17)	0.15 (014, 0.17)
Past-Year Screens of Transdiagnostic Concerns
Suicide Plan, Intent or, Behavior	21.72	3.54 (3.29, 3.81)	2.82 (2.62, 3.04)	8.13	6.69 (5.97, 7.49)	4.27 (3.80, 4.80)
Heavy Drinking	13.36	2.04 (1.95, 2.13)	1.88 (1.80, 1.97)	3.17	2.53 (2.32, 2.76)	2.40 (2.19, 2.63)
Mental health diagnoses (Present vs. Absent)
Personality disorders	17.01	2.58 (2.38, 2.80)	2.09 (1.92, 2.27)	6.87	5.46 (4.83, 6.18)	3.11 (2.74, 3.53)
Bipolar Depression	13.37	1.96 (1.85, 2.09)	1.64 (1.54, 1.74)	4.37	3.44 (3.10, 3.82)	2.16 (1.94, 2.41)
Major Depression	12.05	2.10 (2.04, 2.17)	1.77 (1.71, 1.82)	2.98	3.22 (3.02, 3.44)	2.04 (1.91, 2.19)
Pervasive Dysphoric Disorders	10.83	1.51 (1.38, 1.65)	1.31 (1.20, 1.44)	2.47	1.78 (1.49, 2.14)	1.29 (1.07, 1.55)
Other mood disorder	12.03	1.78 (1.70, 1.86)	1.54 (1.47, 1.61)	2.85	2.21 (2.02, 2.43)	1.59 (1.45, 1.75)
Posttraumatic Stress Disorders	12.09	2.21 (2.14, 2.27)	1.87 (1.81, 1.93)	2.79	3.07 (2.87, 3.28)	2.13 (1.99, 2.29)
Adjustment disorders	12.42	1.85 (1.77, 1.94)	1.51 (1.44, 1.58)	3.20	2.55 (2.34, 2.79)	1.74 (1.59, 1.90)
Other Trauma	15.14	2.26 (2.11, 2.42)	1.74 (1.62, 1.86)	4.36	3.33 (2.94, 3.78)	1.89 (1.66, 2.14)
Anxiety disorders	11.44	1.84 (1.78, 1.90)	1.50 (1.45, 1.55)	2.70	2.45 (2.29, 2.62)	1.57 (1.46, 1.69)
Obsessive Compulsive Disorder	12.30	1.73 (1.48, 2.03)	1.46 (1.25, 1.72)	2.88	2.07 (1.51, 2.83)	1.39 (1.01, 1.90)
Substance use disorders	11.99	1.95 (1.89, 2.01)	1.82 (1.76, 1.88)	2.99	2.82 (2.64, 3.02)	2.45 (2.28, 2.63)
Schizophrenia	8.67	1.17 (1.06, 1.29)	1.15 (1.04, 1.27)	2.80	2.03 (1.72, 2.41)	1.64 (1.38, 1.94)
History of psychiatric inpatient	13.01	1.98 (1.90, 2.07)	1.79 (1.71, 1.87)	4.01	3.48 (3.22, 3.76)	2.56 (2.36, 2.78)
Service connection2
No SC	5.89	---	---	1.07	---	---
10–50% SC	7.17	1.23 (1.18, 1.28)	1.05 (1.00, 1.09)	1.27	1.19 (1.09, 1.31)	0.95 (0.87, 1.04)
60–100% SC	9.51	1.68 (1.62, 1.74)	1.29 (1.25, 1.34)	1.91	1.81 (1.67, 1.95)	1.16 (1.07, 1.26)

Notes. Predictors that are significant (p<.05) and have at least a small effect size (ORs>1.43 & ORs<0.70) bolded for ease of interpretation.

¹Adjusted odds ratios adjust for veteran demographics and military service characteristics including age, sex, marital status, race, ethnicity, period of service, rurality, combat status, and military sexual trauma status. Demographic controls can be found in supplemental Table S1 and S2.

²For predictors with three or more categories (i.e., service category of assessing provider; veteran’s service connection), groups with different lettered subscripts indicate (1) the total effect was significant (p<.05) and (2) the adjusted pairwise comparisons met our effect size threshold (ORs>1.43 & ORs<0.70).

After adjusting for patient characteristics, personality disorders showed a small association with experience of any IPV (AOR=2.09) and a moderate strength association with severe IPV (AOR=3.11). This pattern of small associations with any IPV and somewhat stronger associations with severe IPV held for most categories of diagnoses including PTSD (AOR_Any=1.87; AOR_Severe =2.13), adjustment disorders (AOR_Any=1.51; AOR_Severe =1.74), other trauma related disorders (AOR_Any=1.74; AOR_Severe =1.89), anxiety disorders (AOR_Any=1.50; AOR_Severe =1.57), bipolar disorders (AOR_Any=1.64; AOR_Severe =2.16), major depression (AOR_Any=1.77; AOR_Severe =2.04), another mood disorder (AOR_Any=1.54; AOR_Severe =1.59), or a substance use disorder (AOR_Any=1.82; AOR_Severe =2.45). A diagnosis of schizophrenia only crossed the effect size threshold when experience of severe IPV behaviors was the outcome, indicating veterans with schizophrenia experience greater threats, physical violence, or sexual violence than veterans generally (AOR=1.64).

Regarding our clinical severity indicators, degree of service connection did not show a meaningful association with IPV experience after adjusting for veteran characteristics. However, veterans with previous psychiatric inpatient stays had greater odds of endorsing IPV experience (AOR_Any=1.79; AOR_Severe =2.56).

Aim 2: Association of Couple and Family Service Referral with Clinical Presentation

As noted above, couple and family service referrals were rarely placed in our sample, with only one in 1,000 participants receiving a family services consultations within 90 days of their RHS screening date. However, the placement of these referrals showed strong raw associations with each level of IPV disclosure on the RHS when entered as a set (Table 3). Encouragingly, even after adjusting for demographic controls and the presence of more severe IPV behaviors, reporting any IPV experience had a moderate association (AOR = 2.39) with placement of a subsequent family services consult. Surprisingly, results on the IPV secondary screen showed a small additive effect (AOR = 1.97) above and beyond reporting IPV more generally. Veterans whose IPV screens were completed by VHA mental health providers were more likely to have a consult placed than screens completed by providers in primary care (AOR=0.30) or another service area (AOR=0.39).

Table 3.

Associations between Clinical Characteristics and Family Service Referrals over the next 90 days

Predictive	% receiving referral	OR (95% CI)	AOR1 (95% CI)
Whole Sample	0.10	---	---
Relationship Screening Results2
Any Intimate Partner Violence (Stage 1)	0.33	4.19 (3.12, 5.63)	2.39 (1.65, 3.47)
Threats/Physical/Sexual Violence (Severe Stage 1)	0.60	6.64 (4.20, 10.49)	1.22 (0.65, 2.28)
Risk for Future IPV-Related Injury (Stage 2)	0.76	8.07 (5.10, 12.76)	1.97 (1.05, 3.68)
Service Category of Assessing Provider3
Mental Health Care	0.31a	---	---
Primary Care	0.07b	0.24 (0.17, 0.33)	0.30 (0.21, 0.42)
Other	0.10b	0.33 (0.24, 0.45)	0.39 (0.28, 0.55)
Past-Year Screens of Transdiagnostic Concerns
Suicide Plan, Intent or, Behavior	0.54	6.01 (3.94, 9.15)	3.90 (2.54, 5.98)
Heavy Drinking	0.17	1.80 (1.25, 2.60)	1.59 (1.09, 2.31)
Mental health conditions (Present vs. Absent)
Personality disorders	0.33	3.44 (2.01, 5.90)	2.28 (1.32, 3.96)
Posttraumatic Stress Disorders	0.21	3.47 (2.72, 4.44)	2.42 (1.86, 3.14)
Adjustment disorders	0.30	3.46 (2.57, 4.64)	2.30 (1.70, 3.10)
Other Trauma	0.37	3.86 (2.52, 5.93)	2.33 (1.51, 3.60)
Obsessive Compulsive Disorder	0.21	2.13 (0.68, 6.65)	1.56 (0.50, 4.89)
Other anxiety disorders	0.23	3.32 (2.60, 4.25)	2.24 (1.74, 2.89)
Bipolar Depression	0.33	3.62 (2.48, 5.27)	2.58 (1.76, 3.79)
Major Depression	0.22	3.24 (2.54, 4.13)	2.19 (1.70, 2.82)
Pervasive Dysphoric Disorder	0.12	1.21 (0.54, 2.72)	0.91 (0.40, 2.05)
Other mood disorder	0.23	2.52 (1.82, 3.49)	1.92 (1.38, 2.66)
Substance use disorders	0.17	2.06 (1.58, 2.68)	1.80 (1.37, 2.36)
Schizophrenia	0.11	1.15 (0.51, 2.58)	1.05 (0.46, 2.36)
History of psychiatric inpatient	0.24	2.72 (2.00, 3.69)	2.15 (1.57, 2.95)
Service connection3
No SC	0.06a	---	---
10–50% SC	0.09a	1.64 (1.12, 2.39)	1.25 (0.86, 1.83)
60–100% SC	0.16b	3.01 (2.21, 4.12)	1.84 (1.33, 2.54)

Notes. Predictors that are significant (p<.05) and have at least a small effect size (ORs>1.43 & ORs<0.70) bolded for ease of interpretation.

¹Adjusted odds ratios adjust for veteran demographics and military service characteristics including age, sex, marital status, race, ethnicity, period of service, rurality, combat status, and military sexual trauma status. Full model including controls can be found in supplemental Table S3.

²The dichotomous IPV screening predictors were entered simultaneously in adjusted analyses, thereby approximating contrast codes.

³For predictors with three or more categories (i.e., service category of assessing provider; veteran’s service connection), groups with different lettered subscripts indicate (1) the whole adjusted effect size was significant and (2) the adjusted pairwise comparisons met our effect size threshold.

Veterans were more likely to have a consult placed if they had a past-year positive suicide ideation screen (AOR=3.90) or positive problem drinking screen (AOR=1.59). Although many diagnoses had moderate unadjusted associations with likelihood of a consult to family services (i.e., ORs >2.50), only bipolar depression maintained a moderate association when adjusted for demographic characteristics (AOR=2.58). Other diagnoses maintained small associations after adjustment including personality disorder (AOR=2.28), post-traumatic stress disorder (AOR=2.42), an adjustment disorder (AOR=2.30), another trauma-related disorder (AOR = 2.33), an anxiety disorder (AOR = 2.24), major depressive disorder (AOR = 2.19), another mood disorder (AOR = 1.92), or a substance use disorder (AOR = 1.80). Likelihood of a consult being placed also showed small associations with clinical severity. Veterans with 60–100% service connection were more likely to have a consult placed than veterans with 10–50% service connection (AOR = 1.47) or those without service connection (AOR = 1.84). Similarly, veterans had a slightly higher likelihood of a family consultation being placed if they had a history of psychiatric inpatient hospitalization (AOR = 2.15).

Discussion

For the last decade, VHA has made large investments in both routine IPV screening among the patient population and delivery of couple and family services. The present study examines the parallel expansions of these already impressive programs comprehensively across a healthcare system rather than focusing on a specific subpopulation (e.g., female patients only) or setting (e.g., primary care only), thereby highlighting opportunities that can be leveraged by other organized healthcare systems hoping to promote comprehensive relationship health. The descriptive data highlights a division of labor that likely occurs in many healthcare systems, where a majority of IPV screening occurs in primary care clinics while a majority of couple and family referrals occurs in specialty mental health clinics. This results in patients pursuing family services on their own in separate discussions from the screening where they disclosed their relationship problems, often a month or more after the screening occurred. The predictive data suggests this separation results in two challenges that could be addressed by improving integration between IPV screening and family service referrals at a systems level. The first challenge is that despite the large overlap between the clinical factors associated with the presence of IPV and the placement of family service referrals, only a fractional percentage of individuals reporting these relationship problems receive referrals to further couple and family services. The second challenge is that the providers placing the referrals may not be aware of risk information shared during previous screens and are thus unable to evaluate referral appropriateness or convey IPV risk and safety information to the receiving couple/family therapist. In the remainder of the discussion, we explore avenues to address each challenge in turn and strengthen care across healthcare systems.

Opportunity to Expand Pathways from IPV Screens to Couple and Family Services

Although the primary function of IPV screening is to facilitate service connection for the highest risk patients (Curry et al., 2018), strategies to maximize sensitivity will generally also provide insight into lower levels of aggression and conflict that may benefit from couple-based approaches. Even conservatively assuming all patients with missing Stage 2 screens would be contraindicated for couple therapy, that still leaves over 11,000 veterans (61% of the veterans reporting any IPV) who reported exclusively experiencing verbal aggression while explicitly denying experience of physical IPV, sexual IPV, or risk factors for further injury. It is important to note that a negative screen is not a guarantee of appropriateness for couple therapy. Many patients are likely underreporting IPV, as the positivity rate of 7–8% in both the current sample and other VHA medical record studies (Brignone et al., 2018; Dichter et al., 2018; Iverson et al., 2023) has consistently been lower than the >20% IPV rate given by similar populations of veterans in confidential research surveys (Iverson et al., 2015; Iverson et al., 2024). Additionally, screening tools that inquire only about experiences of IPV, such as the modified HITS, may mischaracterize patients that use more severe IPV than their partners (e.g., if the respondent physically attacks their partner, but the partner only screams back; they would appear as a verbal-only on an experience screen) and may miss other dimensions of power and control (Crasta et al., 2022). However, even recognizing these caveats, it is likely that many more than the 0.33% (i.e., 1/300) of veterans reporting IPV experience that received referrals could have benefitted from a couple and family service referral, which provides an opportunity for further evaluation of IPV, couple/family related needs and dynamics by a specialized therapist.

While reports of IPV experience were associated with greater placement of couple and family service consultations, the adjusted effect sizes are only somewhat larger than similarly adjusted associations between IPV experience and VHA healthcare utilization more generally (Dichter et al., 2018). This suggests that while veterans experiencing IPV may use more health care services overall, couple and family work does not rise in priority relative to other services. Consistent with prior research among female patients in VHA and other healthcare systems (Dichter et al., 2017; Miller et al., 2011), report of any IPV experience among this mixed sex sample was associated with many mental health diagnoses in particular personality diagnoses, PTSD, substance use disorders, and major depression (all AORs >1.75). It may be that providers and clinic managers feel it is important to prioritize addressing these clinical concerns with a single individual therapy referral rather than two simultaneous referrals to avoid overburdening patients. This may explain why family service referrals are higher among providers in dedicated mental health clinics (who would already be doing more comprehensive treatment for the individual mental health issues) than medical providers. By disseminating couple-based interventions for psychopathology (Baucom et al., 2012), treatment systems can reduce the choice between prioritizing mental health or the relationship. VHA sustains delivery of Cognitive Behavioral Conjoint Therapy for PTSD (Monson et al., 2012) and Behavioral Family Therapy for Serious Psychiatric Disorders such as bipolar disorder and schizophrenia (Mueser & Glynn, 1995). Similarly, the United Kingdom National Health Service has been able to disseminate Behavioral Couple Therapy for Depression (Baucom et al., 2018). VHA has also previously disseminated Alcohol Behavioral Couple Therapy (McCrady et al., 2016) and currently disseminates Strength at Home Couples (Taft et al., 2016), both of which bring partners together around substance use and PTSD and their contributions to IPV. While empirical evidence of couple therapies for personality disorders is a growing field (Fitzpatrick et al., 2024), there has been clinical guidance on how to extend mainline behavioral and attachment-based couple therapies for personality disorders (Lebow & Uliaszek, 2010).

A potential structural reason that referrals are more likely after RHS is conducted in a mental health clinic is that VHA family services are situated within VHA’s general mental health. This means mental health providers will generally have greater awareness of family services available when relationship issues are discussed in the context of the RHS screen. Providers in these clinics also have longer appointments with patients, providing opportunity for both disclosure of IPV and discussion of treatment options. However, since a bulk of IPV screening is conducted by primary care teams across both VHA and other large healthcare systems, these teams should also be equipped to initiate family referrals. Providing automated recommendations within an EHR can ensure that primary care providers are aware of available services and can highlight the potential benefits of family services for managing low-risk IPV cases without adding any burden to a single staff member.

Even more options are available for primary care settings that use an “integrated primary care” model, which incorporates a dedicated behavioral health consultant to support the primary care provider and other team members (Trabold et al., 2023). Within the IPV assessment domain, designating a member of a primary care team to serve as an “internal facilitator” to receive extra training on policies and best practices to disseminate to team members can increase detection of IPV up to 25% (i.e., 8.3% to 10.6% across a wave of facilities; Iverson et al., 2023). Within the treatment domain, patients with lower-level verbal aggression prefer to receive relationship services in primary care (Crasta et al., 2022). The behavioral health consultants can naturally serve as internal facilitators – providing IPV-specific education to the team members who conduct routine screenings and providing more thorough assessment for individuals with vague or idiosyncratic responses – and can also provide brief, focused conjoint therapy to both partners.

Opportunities to Coordinate Safety Among Multiple Levels of Care

As settings establish reliable pathways between initial IPV screening and marriage/family services, it is important to minimize inappropriate referrals to this pipeline. Best practice guidelines for couple therapists (Bograd & Mederos, 1999; Keilholtz & Spencer, 2022) recommend interviewing each partner separately during intake (i.e., allowing each partner to disclose when they are not afraid of their partners’ violent reprisal for discussing IPV and they can discuss whether they feel couples therapy is a safe approach for their situation). In this context, couple therapists can assess IPV frequency, severity, and consequences (e.g., injuries; fear of partner) to identify patterns suggesting intimate terrorism or injury risk as well as assess individual mental health concerns that may require further management (e.g., suicide ideation). While this process is likely to divert couples who would be inappropriate for treatment, it can be resource intensive for therapists and can be disappointing for couples if they need to switch treatment modalities after beginning to build a relationship with the couple therapist.

For providers conducting routine screening, using severe item endorsement based on the screen structure may serve as useful rule out for offering couples therapy referral and instead offering individual therapy. In our VHA-based study, the severe items of the RHS Stage 1 screen (i.e., severe psychological IPV, physical IPV, and sexual IPV) should lead providers to be cautious about offering couple therapies but were instead unassociated with referrals to couple and family services. Similarly, the severe injury risk indicators assessed in the Stage 2 screen (i.e., worsening IPV, fear of being killed, and past choking/strangulation; all of which suggest couple and family therapies may be inappropriate) were instead associated with a greater likelihood of receiving a family consult after adjusting for Stage 1 findings. Such inappropriate referrals can be mitigated if the EHR automatically pulled in recent IPV screening results when providers attempt to place referrals. These automated approaches would be especially important in cases where there are delays between the initial disclosure of IPV and the eventual referral request. As most IPV screening occurs in primary care, greater education on the rationale for stabilizing higher-risk IPV in individual therapy before conjoint therapy as well as training on how to discuss this “safety-first” rationale with patients may be needed to reduce inappropriate referrals.

Another important safety factor is suicide ideation, which showed a stronger association with IPV and family service utilization than any single clinical diagnosis. Specifically, approximately 1 in 5 veterans reporting suicide plan, intent, or behavior reported IPV experience over the same period. Examining likelihood ratios suggest that veterans with suicidal thoughts are more than 5x as likely to report severe IPV and more than 5x as likely to receive a family service referral than the overall screened population. Note that the strength of the association is higher in the present study than has been noted in a previous VHA study examining the link between IPV screening responses the presence of any indication of suicidality in the medical record (Brignone et al., 2018), though this previous work used a broader definition of suicidality and only examined female patients whereas the current study includes male and female patients. Nevertheless, this strong association suggests that referring providers need an integrated and systematic strategy for assessing overlapping IPV and suicide risk, prioritizing between the two concerns, and considering what role (if any) couple and family services can play in reducing risk. Fortunately, a new class of treatments have emerged that can address adult suicide risk at the family level either through dedicated couple therapy programs like Treatment for Relationship & Safety Together (TR&ST; Khalifian et al., 2022) or safety planning treatments such as Couples Crisis Response Planning (May, 2020) and Safe Actions for Families to Encourage Recovery (Goodman et al., 2022).

Limitations and Need for Future Work

This study has several limitations. First, while use of medical record data among all VHA patients increases representativeness in comparison to survey studies that can be declined by participants, it also includes many sources of measurement error in comparison to self-report surveys. Individual diagnoses may be out of date if the chart was not reconciled after a condition was resolved and screening data may represent non-standardized administration. Second, the RHS screening protocol focuses on IPV experience and does not currently address IPV use. However, there have been some piloting efforts related to IPV use screening in VHA (Portnoy et al., 2023). Additionally, while the study highlights the predictors of family service consult placement, we were underpowered to examine consult completion over the same period, leaving the potential possibility that receiving couple/family service providers may be more effective in identifying inappropriate consults and redirecting them to individual services. Finally, the study’s VHA-based findings may not generalize to patients who do not use VHA services.

Conclusion

This study provides valuable insights into the intersection between IPV risk, clinical functioning, and family consult placements in a large integrated healthcare system. Despite the overlap between clinical risk factors for IPV experience and family services referrals, pathways between relationship screening and relationship treatments can be strengthened through increasing referrals for appropriate cases, addressing overlapping concerns, and careful management of safety to concurrently reduce danger to self and others. We hope the current findings can guide the development of future training for individual providers and structural changes for systems.