In-Session Predictors of Treatment Outcome for Intimate Partner Violence: An Examination of Fathers for Change and Treatment as Usual

Abstract

Batterer Intervention Programs (BIPs) are the current standard of care for men who use intimate partner violence (IPV). BIPs focus on reshaping men’s maladaptive beliefs regarding IPV, women, and gender stereotypes. However, group-based BIPs have mixed empirical support. Fathers for Change (F4C) is a father-focused intervention with promising initial outcomes that takes an individualized and family-centered approach, emphasizing reflective functioning and emotion regulation. The current study leveraged data from a small randomized controlled trial (RCT) that reported greater reductions in self-reported IPV in men who received individually delivered BIP (BIP-I) and F4C compared to a traditional group-based BIP. An observational coding system was applied to video-recorded therapy sessions to examine in-session variables hypothesized to predict father self-reported IPV and emotion dysregulation. The sample included 50 treatment-seeking fathers with a recent history of IPV and child protective services involvement (BIP-I: n = 24; F4C: n = 26). Multigroup SEM analyses revealed different predictors of IPV behaviors in BIP-I and F4C. In BIP-I, none of the variables predicted post-treatment IPV, whereas in F4C, less avoidance and more cognitive emotional processing during sessions predicted greater reductions in IPV. Unexpectedly, more rigidity was associated with more improvement in IPV behavior, whereas higher self-efficacy was associated with less improvement. No significant predictors of emotion regulation emerged in either treatment. These findings highlight the importance of examining in-session variables in IPV treatments for fathers, with implications for refining interventions and guiding future research.

Identifying In-Session Predictors of Treatment Outcome for Intimate Partner Violence (IPV): An Examination of Fathers for Change and Treatment as Usual

Intimate partner violence (IPV), which includes physical, psychological, and sexual violence, is a public health crisis that can have adverse downstream effects for families, particularly affecting the well-being of women and children (Jewkes et al., 2017; Oram et al., 2022; Sorenson et al., 2021). At least one in four women in the United States will experience IPV in their lifetime (Smith et al., 2018; World Health Organization, 2021). In women, IPV is associated with numerous emotional and psychiatric issues including depression, posttraumatic stress disorder (PTSD), increased suicidality, and associated functional impairments (Dardis et al., 2018; Dillon et al., 2013; Iverson et al., 2013). Further, children’s exposure to IPV, most often perpetrated by fathers, poses a unique threat to healthy child development (Grasso et al., 2016). Exposure to IPV early in childhood can impact cognitive development (Stover, 2023) and places adolescents at increased risk for substance misuse (Halpern et al., 2018), antisocial behaviors (Braga et al., 2017; Wilson & Widom, 2009), and sexual risk-taking (Kugler et al., 2019; Thibodeau et al., 2017). Further, multiple large-scale studies show that IPV often overlaps with direct child maltreatment, with emotional or physical abuse of children occurring in more than half of families affected by IPV (Brown et al., 2021; Hamby et al., 2010; Jouriles et al., 2018). IPV poses a serious risk to both the child and non-offending parent. Although IPV is often bidirectional (Machado et al., 2024), research consistently shows that men are more likely to inflict severe physical harm, engage in coercive control, and perpetuate sexual violence (Dunn et al., 2022; Ruiz et al., 2023). Male-perpetrated IPV is also more likely to result in injury and long-term psychological consequences for partners and children (Cunningham et al., 2023; Grasso et al., 2016).

First-line treatments provide critical care to IPV-impacted partners and children. However, fathers who use IPV often remain involved in the non-offending parent and child’s lives and share parenting time. This increases the risk of additional incidents of IPV, which can perpetuate mental health risk. This highlights the importance of intervening directly with the IPV-implicated parent. Targeting men’s use of IPV—particularly in the context of fatherhood—represents a high-impact intervention point for improving outcomes for both partners and children.

Current Standard Treatment for IPV

Batterer Intervention Programs (BIPs) are the most commonly court-mandated IPV treatments across many states and countries, and they continue to serve as the foundation for coordinated community responses (Butters et al., 2021; Cheng et al., 2021, Goodmark, 2018). The Duluth Model BIP (Pence & Paymar, 1993) is the most widely implemented and is typically delivered in a group format with 10 to 12 participants per group (Battered Women’s Justice Project, 2018). Sessions include videos and vignettes illustrating maladaptive beliefs and behaviors that are identified and challenged by group facilitators, who model alternative responses (Pence & Paymar, 1993). The goal is to reshape participants’ beliefs about IPV, women, and gender stereotypes by encouraging them to understand the historical and cultural origins of these beliefs. This intervention is designed to be educational rather than confrontational, as the latter can increase hostility and the risk of early dropout (Cunha & Caridade, 2023). Notably, BIPs do not teach specific skills, such as emotion regulation, reflective analysis, or communication.

To date, many BIPs face challenges related to high attrition rates (often ranging from 30%–60%), and there is mixed empirical support for their effectivenes in reducing violence (Cheng et al., 2021; Eckhardt et al., 2013). However for those who do complete the program, recidivism rates (20%–30%) are low (Houston et al., 2004; Stover et al., 2009). Some improvements in dropout rates for modified versions of BIP have been reported, and the Duluth BIP has been associated with lower dropout rates than other BIPs (Jewell & Wormith, 2010; Zarling et al., 2019). A meta-analysis by Santirso et al. (2020) also found that interventions that include motivational strategies were significantly more effective at increasing treatment engagement and reducing dropout than those that did not. In addition, some participants may benefit more from individualized treatments, which can foster greater openness, allow for more tailored intervention, and potentially better outcomes (Portnoy & Murphy, 2020). Consistent with this, a recent pilot study comparing two individualized treatments—the Duluth Model adapted for individual therapy (BIP-I) and a father-focused intervention called Fathers for Change (F4C)—found that these two individualized approaches yielded greater reductions in IPV than the traditional group-based Duluth Model and lower rates of dropout (approximately 30% for F4C and BIP-I versus 41% for BIP-group; Stover et al., 2025).

Fathers for Change (F4C) Treatment for IPV

F4C is a fatherhood-focused clinical intervention that addresses both IPV and child maltreatment. F4C consists of 18, 60-minute weekly sessions that combine individual therapy sessions followed by family intervention, which allows for optional joint coparent and child participation. F4C supports fathers in identifying, understanding, and managing emotions to reduce aggression, which can improve interactions with their coparent and children and their mental health. This program utilizes a family systems framework and examines multigenerational patterns and experiences to reveal processes and dynamics that might perpetrate family violence (Stover, 2023). The inclusion of an optional family or coparenting component also responds to calls from survivors for interventions that engage the whole family in healing processes, when safety allows. Research by Stith and colleagues has shown that, when appropriately screened and structured, family- or couple-based interventions can reduce IPV and improve relationship functioning without increasing risk to survivors (Stith et al., 2004; Stith & McCollum, 2011). In this way, F4C aims to extend individual behavior change into safer and more supportive family relationships and improve mental health outcomes across the family system.

Early trials suggest that F4C is a promising treatment for fathers who use IPV and their families. In an initial open trial, F4C demonstrated significant pre-post treatment reductions in IPV behaviors (d= .45) and children’s exposure to parental conflict (d= .43) (Stover et al., 2022). F4C has also shown improved attrition rates (20–33%) in this difficult-to-engage population, as well as significant reductions in IPV and improvements in father-child interactions (Stover, 2013, 2015; Stover et al., 2017, 2018). In a recent RCT comparing F4C to an individualized version of BIP and traditional group BIP (Stover et al., 2025), only the individualized treatments produced significant change in IPV behaviors (F4C: d = .70; BIP-I: d = 1.00), and only F4C produced significant change in emotion regulation (d = .55). The current study seeks to examine how these two individualized treatments (F4C and BIP-I) might facilitate improvements in IPV-related outcomes.

Processes of Change in Treatments for Men who Use IPV

While substantial research has examined demographic and stable individual characteristics associated with IPV treatment outcomes, relatively few studies have investigated in-session processes that may explain how change occurs during intervention—particularly within father-focused or individually delivered models. Understanding these dynamic processes of change is critical for improving intervention precision and tailoring support to participant needs. Existing research has identified some consistent predictors of treatment completion, including older age, White racial identity, employment status, and absence of a childhood abuse history (Lauch et al., 2017). A recent systematic review of longitudinal studies also identified several psychological factors associated with IPV recidivism, including personality traits, anger, trauma history, PTSD, and neurocognitive deficits, underscoring the importance of these individual differences in shaping risk for reoffending (McDonagh et al., 2025). However, these predictors are typically assessed at baseline and reflect relatively static characteristics. Identifying how change unfolds during treatment—beyond static predictors—remains an important and underexplored area, especially given the broader impact of these interventions on partners, children, and family systems.

Processes of Change in BIPs

There is some evidence that reductions in rigid beliefs might be associated with treatment outcomes in BIPs (Herman et al., 2014; Zosky, 2016). Zosky (2016) examined how changes in hostile beliefs about women correlated with overall reduction in aggression at the end of BIP intervention. Hostility was significantly and moderately correlated with aggression at the end of treatment. Herman and colleagues (2014) used the Relationship Beliefs Scale and reported that prior to the BIP intervention, 63% of participants agreed that “women provoke IPV by using bad judgement,” and 45% believed that “men have the natural right to be in charge.” There were significant differences between these percentages before and after completion of the program. Although the BIPs are not consistently associated with change in levels of violence, men might be experiencing some change in previously strongly held and rigid maladaptive thoughts about women in general and about their relationship with their partners.

Processes of Change in F4C

BIPs focus on psychoeducation and do not focus explicitly on skills training. One way that F4C is hypothesized to help reduce violent behavior is by enhancing fathers’ emotion regulation (ER) skills which can both break entrenched patterns of behavior and enhance fathers’ sense that they can manage distress and life difficulties and serve as a direct target of intervention outcomes (Stover et al., 2025). This focus is grounded in theoretical models linking poor ER to higher levels of aggression (Luyten et al., 2020) and violent behavior (Luyten et al., 2020; Stover & Spink, 2012), particularly among individuals who misperceive threat – such as during episodes of situational IPV – and respond with reflexive, emotionally dysregulated behavior (Bateman & Fonagy, 2016; Finkel et al., 2009; Fonagy, 2003; Fonagy et al., 1991). Within the context of IPV, deficits in ER and reflective functioning (RF)—the capacity to understand one’s own and others’ mental states—are considered transdiagnostic risk factors that may underlie reactive aggression and coercive parenting (Gumley, 2011; Thompson, 2019). F4C therefore aims to teach skills that can improve fathers’ self-management and prevent impulsive reactions, which then can enhance their sense of control and ability to regulate their emotions (self-efficacy) (Stover, 2023). Importantly, improvements in emotional regulation are not only theorized to drive reductions in IPV, but are also conceptualized as important standalone treatment outcomes given their strong association with long-term relational and psychological health. This focus on emotional regulation aligns with findings from related interventions. For example, one study investigating a group-based acceptance and commitment therapy (ACT) for partner aggression found that decreases in emotion dysregulation partially accounted for reductions in physical aggression over time (Zarling et al., 2015).

Avoidance refers to difficulty engaging with or remaining present with aversive emotions, thoughts, memories, or somatic sensations (Hayes et al., 1996). It includes behaviors such as withdrawing, shutting down, not talking, or emotionally blunting. This type of avoidance is a maladaptive emotion regulation strategy commonly used by men who engage in IPV. Rather than confronting distressing relational issues, individuals tend to minimize, ignore, and avoid communication related to partner and family conflict and violence, which may reduce short-term discomfort but ultimately reinforces aggression, impedes accountability, and disrupts relationship repair (Stover, 2023). Various studies have identified a relationship between avoidance and aggression in this population. For example, a study of veteran men and their partners suggested that experiential avoidance was associated with an increase in physical IPV perpetration (Reddy et al., 2011). This lack of communication and avoidance of the problem not only disrupts the relationship between co-parents, but also contributes to children’s dysregulation (Stover, 2023). Using data from an RCT assessing the efficacy of group-based ACT for IPV, Zarling et al. (2015) found that participants in the ACT group showed significantly greater decreases in psychological and physical aggression from pre- to posttreatment and follow-up than the comparison group which provided general support and ongoing discussion about IPV behaviors. Moreover, 6-month treatment outcomes were partially mediated by lower levels of avoidance at posttreatment. The authors highlight the potential benefit of incorporating strategies that specifically target avoidance. F4C teaches healthy communication skills and also differentiates avoidance of conflict from taking a healthy break and allowing space for processing. Therefore, it is hypothesized that F4C also targets avoidance by teaching effective communication and emotion regulation skills and related strategies (Stover, 2023).

Another way that F4C is hypothesized to facilitate change in IPV (Stover, 2023) is to identify and examine fathers’ maladaptive thinking associated with women, male roles in relationships, and IPV. This focus on rigidly held beliefs and behavioral patterns is thought to increase reflective functioning (the ability to understand their own actions as functions of underlying emotional states and motivations; Fonagy, 2003; Fossati et al., 2009; Taubner et al., 2013). Recent findings on psychological inflexibility—a core ACT construct—further highlight its role as a mechanism of change. A new study examining court-mandated IPV treatment using ACT found significant decreases in both psychological inflexibility and IPV behaviors over time, with medium to large effect sizes across three timepoints (Zarling et al., 2025). Inflexibility and IPV were positively correlated at all time points, and inflexibility significantly predicted IPV at posttreatment. Notably, the construct of psychological inflexibility shares conceptual overlap with the rigidity code used in the present study, which captures entrenched beliefs, resistance, and a lack of openness to new perspectives—further supporting its relevance as a potential mechanism of change in IPV interventions.

The Current Study

The focus of the current study was to identify predictors of treatment outcomes – IPV behaviors and emotion regulation – of F4C and an individually-delivered BIP based on the Duluth Model (hereafter referred to as BIP-I) by coding video recordings of therapy sessions from an RCT that compared these two treatments (Stover et al., 2025). We coded video recordings from both treatments using an observational coding system (CHANGE; Hayes et al., 2007) designed to capture variables that facilitate and inhibit therapeutic change. We coded video recordings from the individual therapy sessions (9 sessions) of F4C that focus on changing beliefs and emotion regulation skills and the corresponding BIP-I sessions that occurred in that time period. The CHANGE variable ratings on avoidance, rigidity, cognitive emotional processing, and self-efficacy were averaged across the available sessions and examined as predictors of self-reported IPV and emotion regulation at posttreatment.

We hypothesized that higher levels of cognitive emotional processing and self-efficacy would predict less IPV and better emotion regulation in the F4C condition. Because BIP-I does not focus on processing or on teaching specific skills and emotion regulation strategies, we did not expect cognitive emotional processing or self-efficacy to have significant effects on the outcome variables. We also predicted that higher average levels of client rigidity and avoidance would predict higher rates of IPV and worse emotion regulation for both treatments.

Method

Data Source

Outcome data and sessions for the current study were from participants in a randomized controlled trial (RCT) of F4C and BIP-I, which has been described in detail elsewhere (ClinicalTrials.gov identifier NCT04165291; see Stover et al., 2025 for trial outcomes, study procedure, and CONSORT diagram). Stover et al. (2025) reported that both treatments produced significant reductions in IPV, but only F4C produced significant changes in emotion regulation. There were no significant differences between conditions on either of the outcome measures at posttreatment. Within-condition effect sizes of changes in IPV were moderate to large in both treatments from pretreatment to posttreatment (F4C: d = .70, BIP-I: d = 1.00). Within-condition effect sizes of changes in emotion regulation were moderate in F4C and small in BIP-I (F4C: d = .50, BIP-I: d = .21). Video recordings from both conditions were archived and then coded for the present study.

Participants

Participants were 60 treatment-seeking fathers. Fathers referred to the Department of Children and Families (DCF) due to IPV and child maltreatment risk were provided information about the clinical trial. Interested individuals participated in a pre-screening process that assessed for incidents of IPV, substance use, and suicidal and homicidal tendencies. Participants were included in the study if they: 1) reported an incident of IPV (physical, psychological, or emotional) within the last 12 months prior to screening (based on court/police records or partner- or self-report), 2) had at least one biological child aged 6 months to 12 years with whom they had at least monthly contact in person or by phone, 3) were able to complete assessments in English, and 4) agreed to have their coparent contacted as a collateral informant. Coparent participation was not required. Participants were excluded from the study if they had an active full/no contact protective order pertaining to their child, physiological addiction to a substance that required detoxification, cognitive impairment that would not allow for understanding of the study intervention, a current untreated psychotic disorder, or were currently suicidal or homicidal.

Of the 60 fathers who participated in the trial, 30 (50%) were randomized to F4C and 30 (50%) to BIP-I. Participants in F4C were an average age of 35.6 years (SD = 7.7), and participants in BIP-I were an average age of 30.6 years (SD = 7.4). In F4C, 50% reported their race as African American or Black, 13% as White, and 3% Native American, 7% mixed race, 13% another race, 13% preferred not to say. Forty percent reported their ethnicity as Hispanic or Latino. In BIP-I, 57% reported their race as African American or Black, 10% as White, and 20% mixed race, 10% another race, 3% preferred not to say. Twenty-seven percent reported their ethnicity as Hispanic or Latino. More than 80% of fathers had at least one prior IPV arrest (F4C: 86.7%; range=0–14; median= 2.00; BIP-I: 83.3%; range=0–5; median=1.00). More than half of the fathers had attended previous IPV programs (F4C: 60%; BIP-I=53%). Additional background characteristics are presented in Table 1.

Table 1.

Demographic and Background Characteristics

	Treatment Condition
Characteristics	F4C (n = 30)	BIP-I (n=30)
Age: mean (SD)	35.6 (7.7)	30.6 (7.4)
% (number) of Participants
Ethnicity
Hispanic/Latino	40% (12)	27% (8)
Not Hispanic/Latino	60% (18)	73% (22)
Race
White	13% (4)	10% (3)
Black/African American	50% (15)	57% (17)
Native American	3% (1)	0% (0)
Mixed Race	7% (2)	20% (6)
Other	13% (4)	10% (3)
Prefer Not to Say	10% (3)	3% (1)
Unknown	3% (1)	0% (0)
Employment
Full-time	50% (15)	53% (16)
Part-time	13% (4)	4% (1)
Unemployed	37% (11)	43% (13)
Marital Status
Never married	70% (21)	70% (21)
Married	17% (5)	10% (3)
Separated	0% (0)	7% (2)
Divorced	13% (4)	10% (3)
Remarried	0% (0)	3% (1)
Widowed	0% (0)	0% (0)
Previous IPV Program	60% (18)	53% (16)
IPV Arrests, mean (SD)	2.3 (2.6)	1.8 (2.0)
DCF Involvement as Child
Yes	40% (12)	13% (4)
No	53% (16)	87% (26)
Unknown	7% (2)	0% (0)

Note. Percentages do not add up to 100% due to rounding.

Procedures

After consenting to the study and completing eligibility assessments, 60 participants were randomly assigned to receive F4C or BIP-I. Participants completed self-report questionnaires to assess the frequency and type of IPV (Abusive Behavior Inventory: ABI; Shepard & Campbell, 1992) and emotion regulation (Difficulties in Emotion Regulation Scale: DERS; Victor & Klonsky, 2016). These questionnaires were administered at pretreatment and again at posttreatment.

Treatments

Batterer Intervention Program-Individualized

BIP-I is an adaptation of the Duluth Model Domestic Abuse Intervention (Pence & Paymar, 1993). To isolate treatment-specific effects, Fathers for Change (F4C) was compared to an individually delivered version of the Duluth Model (BIP-I), rather than the traditional group-based format. Because F4C is delivered individually, comparing it to a group-based BIP would have confounded treatment content with delivery format. Using BIP-I allowed for a more precise evaluation of treatment effects while holding delivery format constant. The manual, handouts, videos and materials used in Duluth BIP groups were used in individual sessions. BIP-I consisted of 18, 50-minute individual therapy sessions delivered weekly. Sessions focused on psychoeducation about the impact of violence on victims, power and control tactics, and societal influences that perpetuate men’s violence toward women. Videos and experiential exercises, such as role plays, were included to enhance learning.

Fathers for Change

This novel treatment (Stover, 2023) consists of 18, 50-minute individual therapy sessions delivered weekly. The intervention focuses on 1) motivational enhancement, 2) emotion regulation and skills building, 3) coparenting communication, and 4) restorative parenting. In the first four sessions, the therapist employs motivational enhancement in the context of discussing the father’s view of fatherhood and how he hopes his relationship with his children and coparent will change over the course of treatment. Genograms and discussions of multigenerational transmission of IPV are used to help him recognize ways in which he might not have been prepared for healthy and effective relationships with coparents and children. The subsequent phase of treatment, sessions five through nine, focuses on improving the father’s understanding of his own emotions, triggers, thoughts, and behaviors; perception of his partner’s and children’s behaviors; and his capacity to regulate and manage emotional arousal. CBT skills including identification of triggers, relaxation, and cognitive processing are reviewed in relation to the father’s use of aggression. Following the individual phase, F4C allows for optional joint coparent sessions. Conjoint sessions target communication and problem-solving skills pertinent to coparenting. If it is not possible to hold conjoint sessions with coparents due to safety and appropriateness, fathers work individually with their clinician to improve their coparenting skills with the same session content, but they practice these skills with the clinician in session rather than with their coparent. Father-child sessions focus on accountability and reparations that benefit the child and also build more positive relationships with the child. These sessions focus on the beginning of rebuilding the father-child relationship through co-regulation activities, play, and discussion.

Therapists

The five therapists on the trial were master’s level clinicians. Therapists delivered both BIP-I and F4C. Therapists received detailed training in both interventions and delivered each treatment to an equal number of fathers. Therapists met weekly for BIP-I supervision with a psychologist with expertise in the Duluth Model and other BIPs, and who had completed Duluth Model intervention training and modified the intervention for individual delivery for this trial. Therapists also met weekly for F4C supervision with the treatment developer (Dr. Stover).

Treatment Fidelity

Trained coders rated 10% of sessions for fidelity using F4C/BIP-I fidelity measures developed for the parent study (Stover et al., 2025). Coders were trained to reach interrater reliability with the principle investigator. Supervisors provided ongoing fidelity feedback to clinicians throughout the trial. Fidelity ratings confirmed differentiation between interventions on treatment-specific content (e.g., power and control in BIP-I; parenting, child development, and emotion regulation in F4C) and no significant differences in general therapeutic behaviors (e.g., reflective listening, rapport- and trust-building, mirroring), suggesting consistent implementation across therapists (Stover et al., 2025).

Measures

Measures of treatment outcomes

Abusive Behavior Inventory (ABI).

Intimate partner violence was measured via father self-report with the Abusive Behavior Inventory (Shephard & Campbell, 1992) at pretreatment and again at posttreatment. The ABI is a 30-item questionnaire designed to measure the frequency of abusive behavior in the previous six-month period. The ABI assesses various types of IPV including physical, sexual, and psychological abuse. Respondents rate on a 5-point Likert scale (1= never to 5= very frequently) statements such as “You slapped, hit, or punched them,” “You told them that they were a bad parent,” and “You physically attacked the sexual parts of their body.” Items are summed to create a total violence score, with higher scores indicating more violence. The ABI has demonstrated good construct validity and internal consistency (Heron et al., 2023; Shephard & Campbell, 1992). Cronbach α for the ABI at baseline was .92 in the present study.

Difficulties in Emotion Regulation Scale (DERS).

The Difficulties in Emotion Regulation Scale (Victor & Klonsky, 2016) is an 18-item, self-report measure designed to assess difficulties evaluating, responding to, and regulating one’s emotions. The DERS consists of 6 subscales: clarity (e.g., “I am confused about how I feel.”), goal pursuit (e.g., “When I’m upset, I have difficulty getting work done.”), impulse control (e.g., “When I’m upset, I become out of control.”), access to strategies (e.g., “When I’m upset, I believe that I will remain that way for a long time.”), awareness (e.g., “When I am upset, I acknowledge my emotions.”), and acceptance (e.g., “When I’m upset, I feel guilty for feeling that way.”). Respondents are asked to indicate how often the items apply to them, using a 5-point Likert scale (1 = almost never to 5 = almost always). Some items are reverse scored. Fathers completed the DERS at pretreatment and posttreatment. Items were summed to create a total score, and higher scores reflect worse emotion regulation. The DERS has demonstrated good construct validity, high internal consistency, and good test-retest stability (Kaufman et al., 2016). Cronbach α for the DERS at baseline was .87 in the present study.

CHANGE

The CHANGE (Hayes et al., 2007) is an observational coding system designed to assess the process of change in psychotherapy. It includes variables thought to facilitate (e.g., cognitive-emotional processing, accommodation, decentering) or inhibit (e.g., unproductive processing, rigidity, avoidance) therapeutic change (Hayes at al., 2007). The CHANGE was used to code video-recorded sessions in the F4C and BIP-I arms of the trial. The CHANGE coding system has been used to identify predictors of outcome in narratives written by clients in exposure-based cognitive therapy for depression (Hayes et al., 2005; Hayes, Feldman, Beevers, et al., 2007) and adult PTSD (Alpert et al., 2020; Alpert et al., 2023; Shayani et al., 2024), as well as audio-recorded sessions of treatments for youth PTSD (Alpert et al., 2021; Ready et al., 2015; Yasinski et al., 2016), adult treatment-resistant depression (Abel et al., 2016), personality disorders (Hayes & Yasinski, 2015), and adult PTSD (Shayani et al., 2023). Each CHANGE coding variable is coded on a four-point scale from 0 (absent to very to) to 3 (high). Codes are not mutually exclusive and can co-occur.

For each session, CHANGE variables were coded to assess participants’ verbal and nonverbal expressions in the therapy sessions. The CHANGE variables selected for the current study assess factors relevant to F4C and hypothesized to facilitate or inhibit therapeutic change. The coding categories cognitive emotional processing and self-efficacy capture adaptive responses and beliefs, and avoidance and rigidity capture maladaptive responses. Both rigidity and avoidance can be cognitive, behavioral, and emotional. Table 2 presents more detailed descriptions of the CHANGE variables, examples of in-session content that was coded as high (rating of 3), and ICCs for each code. ICCs (Shrout & Fleiss, 1979) were calculated for each coding pair and averaged across pairs, as recommended by Hallgren (2012). ICCs ranged from .69-.85, which falls within the good (.60) to excellent (>.80) range of agreement (Cicchetti, 1994). We coded early, middle, and late sessions of the fathers’ individual sessions (approximately sessions 4–9). Sessions from that same time period were coded for the BIP-I condition.

Table 2.

Descriptions of CHANGE Coding Categories with Examples of High Levels of Each Variable and Intraclass Correlations (ICCs) of Inter-Rater Agreement

Coding Category	Description	Example	ICC
Avoidance	Difficulty engaging or remaining with aversive emotions, thoughts, memories, somatic sensations, or situations. This includes pulling away, withdrawing, shutting down, or emotional blunting	“I hide my emotions behind my smile.” “I’m not talking about this stuff. It makes me feel weak.”	.82
Rigidity	A lack of receptivity to new information, difficulty switching from one way of thinking to another, or resistance to alternative perspectives and new experiences. This rigidity may be cognitive, behavioral, or emotional.	“There is nothing I can do. This is how the world is. Nothing is ever going to change.”	.85
Cognitive Emotional Processing	Approaching, exploring, and making meaning of a problem. This can include thinking about, questioning a problem, and exposing oneself to new information and is followed by insights or shifts in perspective or meaning.	“There are times when me and my partner are able to talk things out and see different perspectives, even when she might have a different opinion. We are all individuals.”	.69
Self-efficacy	Belief in one’s ability to cope with hassles and difficult life demands. This can also include the ability to manage one’s emotions and adapt after stressful life events.	“I got a car and am able to drive myself places now when I am feeling upset. I feel more in control.” “I’m using my breathing exercises, and they are really helping me calm down.”	.76

A team of 6 undergraduate and graduate students coded video recordings from both treatment conditions. Coders were trained in CHANGE and then coded with experienced coders until they reached good inter-rater agreement (intraclass correlations (ICCs) ≥ .75)). Two coders rated each session, and weekly consensus meetings were held to prevent rater drift and reach consensus on discrepancies of two or more points on the 0–3 scale. Consensus ratings replaced discrepant codes, and then the two coders’ ratings were averaged. Coders were masked to study hypotheses.

Session selection for CHANGE coding

In F4C, we hypothesized that the second phase of treatment, approximately sessions four through nine, would have distinctive implications for the outcome measures. These individually-delivered sessions are centered around enhancing participants’ comprehension of their own emotions and cultivating the skills needed to regulate and manage emotional arousal. When five sessions were available during this phase, we selected sessions from beginning, middle, and end of the sequence. If fewer sessions were available, then beginning and end sessions were selected for coding. In BIP-I, we matched the session selection by coding sessions from the beginning, middle, and end within the same time period. At least one of five potential sessions was coded for each participant. A total of 160 sessions were coded, and participants had an average of 3.3 sessions available for coding (range=1 to 5).

Data Analytic Plan

First, descriptive statistics were conducted in SPSS, including means and standard deviations of the predictor and outcome variables for F4C and BIP-I participants, as well as intercorrelations among predictor and outcome variables.

To examine whether the associations between CHANGE variables and outcomes of interest (ABI, DERS) differed by treatment group, we performed a series of multigroup structural equation modeling (SEM) analyses. The latent change scores of outcomes between pretreatment and posttreatment were separately regressed on the CHANGE variables, resulting in two multigroup SEM models. The CHANGE variables included avoidance, rigidity, cognitive emotional processing, and self-efficacy. Grand-mean centering was applied to the predictors to enhance interpretability.

The multigroup SEM approach was performed in Mplus 8, as Mplus allows for flexible handling of missing data and for modeling latent change scores (Muthén & Muthén, 2012). Latent change scores were used instead of difference scores or residualized change scores because latent variables have the advantage of being modeled as error-free constructs and using all available data (Castro-Schilo & Grimm, 2018). The pretreatment ABI and DERS scores and residual covariance among variables were also included to account for shared variance. The sample size was 26 for F4C and 24 for BIP-I due to treatment dropout or not attending the posttreatment assessment session. To account for missing data patterns, maximum likelihood estimation was used in all models.

For each outcome, we compared an unconstrained model (in which regression paths were freely estimated across groups) to a constrained model (in which regression paths were set equal across groups). Model fit was evaluated using chi-square difference testing and fit indices (RMSEA, CFI, TLI). If the constrained model did not significantly worsen fit, we retained it and interpreted pooled parameter estimates across both groups. If the constrained model fit significantly worse, we retained the unconstrained model and interpreted group-specific estimates. This approach directly tests whether regression paths differ significantly across treatment groups, providing an explicit test of moderation rather than relying on informal comparisons of significance levels of predictors in each of the treatment conditions.

Regression coefficients were tested using the z-distribution, as is standard in Mplus under maximum likelihood estimation.

Results

Descriptive Statistics and Intercorrelations

Means and standard deviations for the demographic variables and the predictor and outcome variables are presented by treatment group in Tables 1 and 3. Intercorrelations among the predictor variables and outcome variables are presented by treatment group in Table 3.

Table 3.

Means and Intercorrelations of Therapy Process Variables (Averaged across Sessions) and Outcome Variables in F4C and BIP-I

			1	2	3	4	5	6	7
F4C	Mean (SD)	Range
1. Avoidance	1.26 (.57)	.00 to 2.13	–	–	–	–	–	–	–
2. Rigidity	1.12 (.68)	.00 to 2.38	.39*	–	–	–	–	–	–
3. Processing	1.52 (.42)	.75 to 2.25	.10	.50**	–	–	–	–	–
4. Self-Efficacy	1.14 (.48)	.00 to 2.25	−.25	.53**	.43*	–	–	–	–
5. ABI pretreatment	38.65 (7.67)	29.00 to 58.00	−.27	.09	.31	.24	–	–	–
6. DERS pretreatment	31.62 (8.33)	19.00 to 56.00	−.25	−.09	.21	.22	.36	–	–
7. ABI posttreatment	34.00 (5.21)	29.00 to 44.00	−.13	−.35	−.16	.10	.44*	.59**	–
8. DERS posttreatment	29.45 (6.19)	19.00 to 44.00	−.25	−.13	−.41	−.09	.11	.40	.24
BIP-I	Mean (SD)	Range
1. Avoidance	1.42 (.60)	.50 to 2.67	–	–	–	–	–	–	–
2. Rigidity	1.69 (.69)	.50 to 2.75	.57**	–	–	–	–	–	–
3. Processing	1.63 (.56)	.50 to 2.83	−.64***	−.70***	–	–	–	–	–
4. Self-Efficacy	.80 (.49)	.00 to 2.33	−.66***	−.60**	.56**	–	–	–	–
5. ABI pretreatment	40.00 (9.06)	30.00 to 64.00	−.22	−.39	.36	.24	–	–	–
6. DERS pretreatment	30.42 (10.59)	18.00 to 58.00	.14	.00	−.26	−.07	.60**	–	–
7. ABI posttreatment	33.00 (6.27)	29.00 to 52.00	−.21	−.24	.26	.03	.71***	.26	–
8. DERS posttreatment	28.05 (8.61)	18.00 to 45.00	.13	.19	−.26	−.08	.65***	.82***	.38

Note. F4C = Fathers for Change; BIP-I = Batterer Intervention Program-Individual; Processing = Cognitive Emotional Processing; ABI =Abusive Behavior Inventory, DERS = Difficulties in Emotion Regulation Scale.

^*p < .05

^**p < .01

^***p < .001

Predictors of IPV

Multigroup SEM analysis was conducted to examine whether the associations between CHANGE variables and IPV were dependent on treatment group. The unconstrained model (paths freely estimated) provided a significantly better fit than the constrained model (paths constrained equal across groups; Δχ²(4) = 12.98, p = .011), indicating moderation by treatment group. The model provided an adequate fit to the data (χ ² = 9.93, df = 8, p = .27, CFI = .93, RMSEA = .098, 90% CI [.00, .27]). Results revealed different predictors of outcome in the two treatment conditions (see Table 4). In BIP-I, none of the predictors significantly predicted change in IPV behaviors. In contrast, all predictors significantly predicted change in IPV behavior in the F4C group. Specifically, in F4C, more avoidance during treatment was associated with higher levels of IPV behavior at posttreatment (B = 6.44, SE = 1.73, z = 3.74, p <.001), and more cognitive emotional processing was associated with less IPV behavior at posttreatment (B = −6.72, SE = 2.16, z = −3.11, p = .002). Unexpectedly, rigidity was associated with more improvement in IPV behavior in F4C (B = −6.88, SE = 1.58, z = −4.35, p < .001) and self-efficacy was associated with less improvement in IPV behavior (B = 6.71, SE = 1.97, z = 3.41, p = .001).

Table 4.

Multigroup Structural Equation Modeling with Latent Change Scores Predicting Abusive Behavior Inventory (ABI) Outcomes in F4C and BIP-I

F4C	B	SE	z	p	95% CI
					Lower	Upper
Avoidance	6.44	1.73	3.74***	<.001	3.06	9.82
Rigidity	−6.88	1.58	−4.35***	<.001	−9.98	−3.78
Processing	−6.72	2.16	−3.11**	.002	−10.95	−2.49
Self-efficacy	6.71	1.97	3.41**	.001	2.85	10.57
BIP-I	B	SE	z	p	95% CI
					Lower	Upper
Avoidance	−1.81	2.39	−0.76	.450	−6.49	2.88
Rigidity	−0.80	2.23	−0.36	.720	−5.17	3.57
Processing	−0.20	2.45	−0.80	.936	−5.00	4.60
Self-efficacy	−3.46	2.83	−1.22	.222	−9.01	2.09

Note. F4C = Fathers for Change; BIP-I = Batterer Intervention Program-Individual; Processing = Cognitive Emotional Processing; CI = confidence interval. z-statistics are reported, consistent with Mplus maximum likelihood estimation.

^*p < .05

^**p < .01

^***p < .001

Predictors of Emotion Regulation

A multigroup SEM analysis was conducted to evaluate whether the relationships between predictors and change in emotion regulation were dependent on treatment group. In contrast to the IPV model, the constrained model did not significantly worsen fit relative to the unconstrained model (Δχ²(4) = 8.98, p = .062). This indicates that regression paths did not significantly vary by group, and thus pooled estimates across groups are reported.

The constrained model provided an adequate fit to the data (χ²(12) = 16.93, p = .15, CFI = .84, RMSEA = .13, 90% CI [.00, .26]). As shown in Table 5 none of the predictors—avoidance, rigidity, cognitive emotional processing, or self-efficacy—significantly predicted change in emotion regulation from pre- to posttreatment (all ps > .09).

Table 5.

Multigroup Structural Equation Modeling with Latent Change Scores Predicting Difficulties in Emotion Regulation Scale (DERS) Outcomes (Pooled Across F4C and BIP-I)

Predictor	B	SE	z	p	95% CI
					Lower	Upper
Avoidance	−3.07	1.83	−1.68	.094	−6.67	0.52
Rigidity	2.44	1.62	1.51	.132	−0.74	5.61
Processing	−1.84	1.99	−0.92	.356	−5.74	2.06
Self-efficacy	−0.74	2.09	−0.35	.725	−4.84	3.36

Note. F4C = Results represent pooled estimates across treatment groups because regression paths did not significantly differ between F4C and BIP-I. Processing = Cognitive Emotional Processing; CI = confidence interval. z-statistics are reported, consistent with Mplus maximum likelihood estimation.

Discussion

The current study examined predictors of treatment outcomes in an individually-delivered BIP and a new treatment for IPV (F4C: Stover, 2023) that harnesses the role of fatherhood to engage men in learning new skills. This study draws on data from a treatment development RCT, that found that both BIP-I and F4C were associated with significant reductions in IPV (Stover et al., 2025). Despite the inconsistent efficacy of the current group-based BIP standard of care, the individual format (BIP-I) showed significant reductions in violence and a large effect size in this trial. A more individualized and tailored approach, rather than a “one-size-fits-all” group model, might be best suited to this population. The findings from the outcome study raise intriguing questions about how BIP-I and F4C might be having their effects, especially given that the two treatments are rooted in different theories and attempt to reduce IPV in different ways. To date, there are few studies investigating how BIPs and other treatments for IPV use may reduce IPV. Findings from the present study suggest that process variables may differ for IPV outcomes but not for emotion regulation outcomes. For IPV, results indicated group differences in the CHANGE process variables. None of the examined variables significantly predicted change in IPV behaviors in BIP-I, but in F4C, more cognitive emotional processing predicted reductions in IPV, and more avoidance predicted less improvement. Unexpectedly, more rigidity was associated with greater improvements in IPV outcomes in F4C, whereas higher self-efficacy predicted less improvement.

For emotion regulation (DERS), there were no significant differences in the CHANGE process variables in F4C and BIP-I. This pattern of findings indicates that, although process variables may differentially relate to IPV reduction across treatments, their role in predicting emotion regulation outcomes appears to be more similar across treatment conditions. This finding should be interpreted in light of the broader outcome study (Stover et al., 2025), which demonstrated that there were significant pre–post improvements in DERS for fathers in F4C, but not in BIP-I. One possibility is that the specific process variables assessed in the present study (avoidance, rigidity, cognitive-emotional processing, and self-efficacy) do not capture the variables most responsible for change in emotion regulation, such as applying the new emotion regulation skills learned in therapy. Another possibility is that emotion regulation may be more accurately conceptualized as a mechanism of change rather than as an outcome variable. Given that F4C explicitly seeks to strengthen emotion regulation skills as a central component of the intervention, continued investigation of emotion regulation as both a process and an outcome remains important.

Cognitive Emotional Processing and Rigidity

In F4C, more cognitive emotional processing predicted lower levels of IPV at posttreatment, highlighting its importance as a potential therapeutic change mechanism. These findings support the study hypothesis that cognitive emotional processing would enable fathers to engage in deeper self-exploration and reflection, question entrenched patterns, and ultimately demonstrate meaningful shifts in IPV behaviors (Stover, 2023). The CHANGE cognitive-emotional processing category captures different levels of processing, from increased awareness and insight at lower levels to substantial perspective shifts with associated behavioral changes at higher levels (Hayes et al., 2007). F4C can provide fathers with a clearer understanding of their emotional triggers, multigenerational patterns of abuse, and structured processing of their own rigidly held thought processes and behavioral patterns that increase their risk of IPV behaviors. The strong focus on identifying and challenging entrenched patterns in F4C might facilitate the perspective shifts that characterize higher levels of processing. This might be a reason for the unexpected findings that more processing seemed to occur in the context of a focus on rigid patterns (Table 3) and that both more rigidity and processing predicted improvement in IPV at the end of the treatment in F4C (Table 4).

Cognitive emotional processing did not emerge as a significant predictor of IPV behavior in BIP-I, consistent with study hypotheses. BIP-I provides fathers with educational information on power and control in intimate relationships, which is often inconsistent with their maladaptive views of these relationships. BIP-I is thought to increase awareness of unhelpful thoughts and patterns (Herman et al., 2014; Zosky, 2016), but it does not teach specific skills to facilitate perspective shifts, communicate more effectively, or better regulate emotions. In addition, more rigidity in BIP-I was associated with lower processing scores during the sessions (Table 3). In contrast, more of a focus on rigid patterns in F4C was associated with more processing, which might facilitate not only awareness and insight, but also perspective shifts that can promote change in IPV behavior.

We had predicted that rigidity would inhibit therapeutic change in both treatments. In F4C, rigidity was associated with more improvement in IPV behavior at the end of treatment, but not with emotion regulation, which did not differ between the treatment conditions. In BIP-I, more rigidity did not predict IPV outcomes or worse emotion regulation outcomes. The CHANGE rigidity category not only captures a focus on entrenched patterns, but also therapy-interfering processes, such as resistance, degree of conviction in maladaptive beliefs, and a general lack of openness or flexibility. In interventions that emphasize skills for emotion regulation and reflective functioning such as in F4C, the presence of rigidity may present a clear therapeutic target, enabling therapists to directly address and reframe entrenched patterns in ways that facilitate change. It is noteable that rigidity was associated with more avoidance across both treatments, but also with more processing in F4C and less processing in BIP-I. This pattern of findings is consistent with the ideas that rigidity and avoidant coping may represent maladaptive strategies that, depending on the intervention, can be leveraged as points for change (in F4C) or remain unaddressed (in BIP).

Both BIP-I and F4C target maladaptive beliefs related to IPV (Herman et al., 2014; Stover., 2023; Zarling et al., 2025; Zosky, 2016), and future research could more closely examine the types of cognitive change that occur in these treatments and the interventions used to facilitate these changes. Another task is to better understand how F4C and BIP-I therapists differ in whether and how they address rigid patterns associated with IPV.

Avoidance

Avoidance is a common emotion regulation strategy associated with IPV and is therefore an important treatment target (Reddy et al., 2011; Stover, 2023). It is common for men who use IPV to minimize, ignore, or otherwise avoid talking about their violent behavior or to use avoidance in the face of conflict in general (Stover, 2023). Avoidance can be used to reduce exposure to situations that can trigger maladaptive patterns or to reduce a sense of responsibility and feelings of shame and guilt, even if only in the short-term. Avoidance is also associated with a number of skills deficits (Elliott, 2003; Tager et al., 2010) that can perpetuate maladaptive patterns related to IPV.

As predicted, avoidance seemed to interfere with change in IPV behavior in our study, but only in F4C. Our findings extend previous research on a group-based ACT intervention for IPV that reported that more self-reported avoidance at posttreatment partially mediated worse treatment effects on physical and psychological aggression at 6-month follow-up (Zarling et al., 2015). Therapists in F4C encourage fathers to examine their thoughts and feelings in different triggering situations, reflect on why these contribute to emotion dysregulation (e.g. connections with their early life experiences of adversity or trauma), and take responsibility for their actions. Avoidance in F4C might reflect difficulties fathers have engaging with these challenging tasks, which could dilute the intervention and hinder behavioral change. BIP-I involves psychoeducation and support and does not focus directly on emotion regulation, so it is possible that avoidance does not dilute the treatment as much as it does with F4C. This is speculative, but does suggest that it could be useful to examine the content of the avoidance and its relationship with the specific tasks of F4C or BIP-I. These findings also highlight the potential significance of the therapeutic alliance when addressing avoidance and fluctuations of motivation in this difficult-to-engage population (Doran, 2016; Heinonen et al., 2022)

Because the associations between the CHANGE variables and emotion regulation outcomes did not differ by treatment condition, analyses were conducted on the pooled sample. It is surprising that avoidance was a marginal but not significant predictor of worse outcomes in emotion regulation, especially given that avoidance is a type of maladaptive coping strategy. However, avoidance did predict worse IPV outcomes in F4C. Avoidance was also associated with more rigidity in both F4C and BIP-I, and this broader category of rigidity predicted better IPV outcomes in F4C. This further suggests that it might be fruitful to examine the content of avoidance in the two treatments.

Self-efficacy

Self-efficacy captures the perceived ability to manage and regulate one’s emotions and cope with problems. As predicted, self-efficacy was not associated with change on IPV or emotion regulation outcome for those in the BIP-I condition. BIP-I does not include skills training, and change in emotion regulation skills (DERS) was associated with only small effect sizes in this condition (Stover, 2023; Stover et al., 2022; Stover et al., 2025. Self-efficacy was a significant predictor of IPV in F4C in our study, but this association was in the opposite direction than hypothesized: Higher self-efficacy predicted more IPV at posttreatment.

F4C focuses on increasing fathers’ sense of control over their functioning, but it is possible that the CHANGE self-efficacy code captured two components of control: one adaptive and one not. Adaptive self-efficacy might involve a father feeling in control and using strategies learned in therapy to manage anger when noticing heightened arousal. In contrast, maladaptive self-efficacy might involve a father’s belief that he has control over his partner and children, reflecting a destructive sense of control. For instance, in a study of police officers, which did distinguish the desire to be in control and to have self-control (a component of self-efficacy), the desire to be in control predicted more instances of IPV (Zavala & Melander, 2019). This finding aligns with our unexpected results, suggesting that the desire to be in control might outweigh more adaptive control or self-efficacy. Furthermore, it is intriguing to consider that for those who have experienced IPV (rather than perpetrators), increasing self-efficacy has been recognized as a crucial factor in recovery (Iverson et al., 2021; Iverson et al., 2022). The roles of self-efficacy and one’s sense of control may be more nuanced in those who use IPV, so it could be useful to parse these variables into more and less adaptive components.

It is also possible that some fathers may inflate their feelings of self-efficacy, reporting that they have mastered skills that they have not. Conversely, fathers who are becoming more aware of their emotions and learning new regulation strategies may report lower self-efficacy, recognizing the challenges of change. This kind of self-reflection, encouraged in F4C, may indicate an understanding of the ongoing work needed in treatment. This is especially relevant, given that the coded sessions occurred between sessions 5–9 of an 18-session intervention, when participants were still in the early-to-mid stages of skill acquisition. Thus, future research might also consider when self-efficacy is assessed relative to the timing of skills training in the course of treatment.

Strengths, Limitations, and Future Directions

This study was characterized by a strength often lacking in psychological and/or intervention literature, as the sample included a diverse sample of men from various racial and ethnic backgrounds. Further, this study includes fathers who had recently used IPV rather than a sample who had perpetrated at some time in the past. However, the study included only men that were in current or previous heterosexual relationships with their coparents. Future studies might include and collect data from same-sex coparenting partners, as sexual minorities are especially vulnerable to IPV (Calton et al., 2016; Kim & Schmuhl, 2021). Another strength is that we used data from an RCT that compared F4C to an active comparator, BIP-I, and therapists received careful training and monitoring, including weekly supervision with video session reviews.

The CHANGE observational coding system (Hayes et al., 2007) used in this study provided several advantages. It provides ratings by independent coders on adaptive and maladaptive processes and yields information beyond clients’ reports of their own functioning. The use of archived video recordings enabled us to capture more detailed information and context, including nonverbal cues (e.g., crying, fists clinching) and avoidance in various forms. At the same time, the CHANGE relies on verbal and nonverbal cues that can be detected in session recordings, and it may potentially miss important processes that occur both in session and outside of session that could contribute to treatment outcomes. For example, out-of-session self-efficacy and avoidance could be captured using self-report measures and daily diary measures or ecological momentary assessment. Future studies might incorporate multiple types of measurement to better capture the range of processes that might be related to change in F4C and BIP-I.

While we coded multiple sessions from sessions 5–9 in both treatment arms, participants had different sessions and number of sessions available, so session-by-session associations among processes and outcome variables were not examined, and the temporal sequencing of change in process and outcome variables is not clear. Moreover, in F4C, current analyses only focused on individual-focused sessions. Although this decision was made because this is hypothesized to be where the most change occurs in F4C (Stover, 2023), additional change is also likely to occur in the coparent- and child-focused sessions, which were not coded in this study. The later sessions of BIP-I also include important content, such as economic abuse and coercive behavior. It will be important in future research to explore how these variables facilitate and inhibit change in these additional phases of treatment.

The basic pre- and post-test design of the RCT limits the study of the process of change. The CHANGE variables were coded during sessions 5–9, which is temporally distinct from outcome, but the IPV and emotion regulation outcome variables were only collected at the beginning and end of treatment. Future studies of both treatments should include more frequent assessment of the outcome variables during the course of therapy and also multiple follow-up assessments to evaluate long-term outcomes.

Another limitation of the current study is that the sample size is small and assessment of outcomes was infrequent. While the use of latent change scores was a strength (Castro-Schilo & Grimm, 2018), the study sample size did not allow for more complex analyses of change, including session-by-session longitudinal analyses. In addition, there were only five therapists in this study, which does not allow for explicit examination of therapist-level contributions to treatment outcomes. The findings must be replicated in studies with larger samples of participants and therapists.

Conclusion

Overall, the significant predictors identified in in-session video recordings in BIP-I and F4C highlight the importance of exploring inhibitors and facilitators of change in these two treatments for fathers who use IPV, one of which is a BIP current standard of care, and the other that is a new approach for treating IPV that aims to improve reflective functioning, induce perspective shifts, and improve emotion regulation skills. Our findings suggest that avoidance, rigidity, cognitive-emotional processing, and self-efficacy all significantly predicted change in IPV outcomes in F4C, whereas none of these variables were significant predictors in BIP-I. This study is one of the first to examine processes of change in interventions for IPV. Because there were no predictors of IPV outcomes in BIP-I in this examination of in-session processes, future research is necessary to understand how BIP-I might be reducing IPV behavior. Additionally, no significant predictors of change in DERS were identified in either group, underscoring the need for further investigation into which processes may underlie improvements in emotion regulation, particularly in F4C, which did show significant pre- to post-treatment improvement (Stover et al., 2025). Investigating in-session processes in our larger, ongoing RCT comparing F4C and BIP-I (Stover, NCT06074068) can contribute to the evidence base for both treatments and help guide treatment refinement and improvement. These individually-delivered interventions for IPV can be readily implemented within existing systems of care serving CPS-involved families and can have positive implications of change for the family constellation.