Screening for Intimate Partner Violence in Family Mediation: An Examination of Multiple Methodological Approaches Using Item Response Theory

Abstract

Many divorcing/separating parties seeking mediation to resolve family-related issues report intimate partner violence (IPV) victimization from the other party in the case. It is imperative that mediation staff screen parties for IPV so they can make informed decisions regarding how to proceed with mediation. Existing IPV screens for mediation have significant limitations. We examined three methodological approaches using item response theory that address these limitations by increasing the efficiency and clinical utility of an existing standardized IPV screen for mediation, the Mediator’s Assessment of Safety Issues and Concerns (N = 904 mediating parties). We identified three subsets of items, with initial evidence for their validity, focused on helping mediation staff identify high levels of IPV or parties at risk for potentially negative mediation outcomes or needing specialized safety accommodations in mediation. Clinical recommendations are provided indicating which approach is most promising to be used in mediation settings. Overall, findings help advance understanding of how item response theory methodology can enhance the precision of IPV screening in mediation.

Keywords

intimate partner violence screening family mediation divorce item response theory

A substantial number of parties seeking mediation services to resolve family-related issues pertaining to a separation or divorce report a history of intimate partner violence¹ (IPV) victimization from the other party in the case. Studies suggest that up to 50% of such parties report physical violence victimization and up to 98% report psychological abuse (Ballard et al., 2011; Beck et al., 2006; Rossi et al., 2015). Family mediation is a commonly used alternative dispute resolution process in which separating or divorcing parties negotiate issues such as child custody, child support, and division of property, with the assistance of a mediator, a neutral third party. The question of whether family mediation is appropriate and safe for parties reporting a history of IPV is a significant point of contention in the field (Rossi et al., 2017).

There is concern that joint mediation (where both parties sit in the same room to negotiate) may increase the risk of physical harm toward the victim due to both parties being present in the same location or if the mediation process angers the perpetrator (Milne, 2004; Rossi et al., 2017). Joint mediation or videoconferencing mediation (where both parties communicate directly via video) may also provide perpetrators with the opportunity to use coercion tactics to push for family arrangements that do not adequately protect the victims’ needs and interests. Mediation agreements that do not decrease opportunities for continued IPV victimization postseparation may put victims and their children at increased risk of danger (Rossi et al., 2017).

However, when mediation is effective, it gives parties the ability to self-determine the arrangements that are most suitable for their family, rather than a judge making such decisions (Folberg et al., 2004). This feature may be useful for cases with a history of IPV, as victims would have the opportunity to be openly present and discuss options regarding the parenting plans that are in the best interests of the family (Emery, 2011). Shuttle and videoconferencing mediation, modified forms of mediation designed for parties reporting IPV, have been shown to offer several benefits over traditional court-based litigation for parties reporting IPV, including increased feelings of party safety and satisfaction, and faster resolution of case issues (Holtzworth-Munroe et al., 2021).

Despite ongoing controversy, extant research suggests that in order to make decisions about the appropriateness of mediation, it is first imperative to screen for IPV in the mediation setting (Ellis, 2008; Holtzworth-Munroe, 2011; Holtzworth-Munroe et al., 2010; Ver Steegh & Dalton, 2008). Mediation staff must be able to identify cases reporting a history of IPV and assess for information that may potentially affect the mediation process, such as the frequency, level, and type of violence and abuse. However, there are few standardized IPV assessment tools designed for the family mediation context. Existing tools are often lengthy, complex, or offer little guidance on how to use IPV screening results to make decisions about the appropriateness of mediation services for parties reporting IPV (Holtzworth-Munroe et al., 2010).

Advanced psychometric techniques, such as item response theory (IRT), can assist in developing an IPV screening tool that addresses these challenges while also increasing the precision with which we measure IPV in mediation. Thus, the current study is an exploration of possible applications of IRT for increasing the efficiency and clinical utility of an existing standardized IPV screening tool for mediation, the Mediator’s Assessment of Safety Issues and Concerns (MASIC; Holtzworth-Munroe et al., 2010; Pokman et al., 2014). We examine multiple IRT approaches, present various iterations of the MASIC resulting from such approaches, and discuss the clinical implications of each approach.

IPV Screening in Mediation

An IPV screen developed specifically for the mediation setting is imperative, as there are aspects of IPV that are particularly relevant to mediation. For example, party reports of coercive controlling behaviors can be especially concerning for mediation staff. Stalking is another subtype of IPV that needs to be assessed among separating parties, as stalking can escalate postseparation (Mechanic, 2004).

Given this need for a specialized IPV screen, a few IPV measures have been developed specifically for the mediation setting, though most have demonstrated significant limitations. A pilot study (Holtzworth-Munroe et al., 2010) compared two of these screens, the Domestic Violence Evaluation (Ellis & Stuckless, 2006) and the Relationship Behavior Rating Scale–Revised (Beck et al., 2013). This pilot study found that mediation staff expressed concerns regarding the administration of each screen, including lengthy administration, costly copyright protections, a need for intensive staff training, complex scoring, and the inclusion of broad, nonspecific questions about IPV (e.g., Have you been “physically assaulted”?) or questions that targeted differing timeframes on different items (e.g., ever in the relationship versus only in the past year). The MASIC (Holtzworth-Munroe et al., 2010) was developed in an attempt to address many of these concerns. The MASIC consists of behaviorally specific questions (e.g., “Yell or scream at you?” “Hit or punch you?”), rather than broad, less specific questions. It inquires about six types of violence and abuse, including psychological abuse, coercive controlling behaviors, physical violence, sexual violence, threats of severe violence, and stalking, as demonstrated in a confirmatory factor analysis (Pokman et al., 2014). In addition, it inquires about other IPV-related behaviors and consequences (e.g., injury, fear, recent changes in frequency or severity). Items ask about the occurrence of behaviors across two timeframes, ever in the relationship and within the past year. In response to staff’s concerns regarding intensive training and complex scoring, the MASIC was designed to not require extensive training for proper administration and scoring. To address concerns about cost, the MASIC is in the public domain and available for use at no cost. There is also initial evidence for the reliability and validity of the MASIC (Pokman et al., 2014; Rossi et al., 2015a).

Despite preliminary evidence for the reliability and validity of the MASIC and this tool’s design to overcome many of the concerns expressed by mediation staff, two significant limitations remain. Since the MASIC’s introduction (Holtzworth-Munroe et al., 2010), various mediation programs have experimented with the MASIC in their setting. Although staff appreciate the MASIC’s comprehensiveness, some have found that its administration is still too lengthy, making it impractical for their mediation settings. These mediation staff have indicated that the length of the MASIC, relative to less formal IPV screening (e.g., clinician using clinical judgement to determine who is experiencing IPV or using available official records, such as civil protective orders), results in significantly longer intake appointments and, therefore, reduced efficiency in process and greater burden placed on parties to take time off work and find childcare. Unfortunately, however, two randomized controlled trials in mediation settings have demonstrated the limited utility of these informal methods to detect IPV, relative to lengthier, behaviorally specific measures such as the MASIC (Ballard et al., 2011; Rossi et al., 2015a). Despite such evidence, conversations with court personnel and practitioners in community agencies make it obvious that there is a critical need for efficient administration of an IPV measure in the mediation setting given the time constraints and large service demands faced by mediation staff.

The second limitation is that although the MASIC has simplified scoring relative to some other IPV measures for mediation, it provides little guidance on how to interpret IPV information or how to use such information to make empirically informed decisions about mediation. For example, using the MASIC, mediation staff might learn of psychological abuse but have little instruction on how to proceed with mediation in a way that increases parties’ chances of experiencing positive outcomes. Such guidelines are lacking given the absence of research examining which levels or types of IPV lead to unsafe or possibly concerning outcomes for mediation parties. This study used IRT methodology to address these two limitations identified in the MASIC and sought to develop a more efficient and clinically useful MASIC.

IRT and the MASIC

IRT is a statistical method that allows for the empirical evaluation of each item within a questionnaire by estimating the item’s difficulty and discrimination. The difficulty parameter provides information about the level of the latent trait (or theta; i.e., in MASIC, the latent trait or construct of interest is IPV victimization) at which 50% of respondents endorse the item. For example, an individual who endorses having been kicked or stomped on by the other party will likely report higher levels of IPV victimization than an individual who endorses having been yelled or screamed at by the other party. Therefore, the item “kick or stomp” will demonstrate a higher difficulty estimate than the item “yell or scream” for IPV victimization. In contrast, the discrimination parameter indicates the degree to which an item distinguishes between individuals at higher and lower levels of the latent trait. Items that are endorsed by individuals reporting higher and lower levels of IPV victimization will demonstrate poorer discrimination while items endorsed by individuals reporting only lower or higher levels of IPV victimization will demonstrate greater discrimination.

The discrimination and difficulty estimates for each item derived from IRT analyses make it possible to determine which items provide more or less information at differing levels of the construct of interest. Ideally, a measurement tool should include items that provide maximum information about the construct of interest at the level that is most useful depending on the measure’s intended purpose. IRT methods can be used to identify the selection of items that are most helpful in assessing for the desired information and those items that are the least helpful or informative. This allows for the development of assessment tools with greater measurement precision and efficiency.

The sets of items identified using IRT differ according to assessment needs. Although the intended purpose of an IPV screening tool for mediation is to provide information about IPV and assist mediators with making decisions regarding mediation services, the specific information that helps achieve this goal is currently unclear in the field. For example, an IPV measure may assist mediators in their decision-making process by helping them identify parties who are reporting the very highest levels of IPV victimization. Alternatively, it is possible that mediators make enhanced decisions using an IPV measure that identifies parties likely to have potentially negative or concerning outcomes in mediation (e.g., not reach agreement²).

Given the lack of clarity in the field, this study had two aims. The first aim was to examine three IRT-informed approaches to developing a shorter and more clinically useful MASIC.

Approach 1 (Selecting for the Highest Levels of IPV)

In the first approach examined, we used IRT to identify a selection of maximally informative items on the MASIC that correspond to IPV victimization levels two standard deviations above the mean. This approach is common within clinical assessments and helps distinguish individuals demonstrating clinically significant symptomatology (e.g., Petersen et al., 2016).

Approach 2 (Selecting for Any Level of Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

In the second approach, we used IRT to identify a selection of maximally informative items on the MASIC corresponding to IPV victimization levels associated with potentially negative or concerning outcomes of mediation (e.g., not reaching agreement in mediation or being screened out of mediation because of IPV) or the need for specialized procedures or accommodations (e.g., escort party to and from building) during the mediation process due to mediator concerns about IPV. This approach allows mediators to identify individuals who are at any level of risk for having a potentially negative or concerning mediation outcome or needing specialized accommodations due to IPV.

Approach 3 (Selecting for Highest Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

In the third approach, IRT was used to identify items on the MASIC corresponding to IPV victimization levels associated with only the highest risk (50% or greater likelihood) for having potentially negative or concerning outcomes in mediation or needing specialized procedures or accommodations during the mediation process due to IPV. This approach allows mediators to identify individuals who have the highest chances of potentially not benefiting from mediation or needing additional accommodations due to IPV. Approach 3 is similar to Approach 2 in that both focus on identifying individuals who may not benefit from mediation or need additional accommodations due to IPV. However, Approach 3 identifies those who have the highest risk whereas Approach 2 identifies those with any risk for having potentially negative or concerning outcomes in mediation or needing specialized accommodations.

The second aim was to test the convergent validity of the three subsets of items resulting from the aforementioned approaches. In all cases, we examined the association between one party’s IPV victimization report and the IPV-related criminal charges and protective orders/no contact orders issued against the other party in the case. To our knowledge, no study to date has incorporated IRT methodology for the purpose of empirically refining an IPV measure so that it is more efficient and assists mediators in making decisions about the suitability of mediation for parties reporting a history of IPV. There is a significant need to advance IPV screening in mediation so that divorcing or separating parties are able to receive safe and appropriate services.

Method

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. As part of a secondary data analysis, the study sample was comprised of 904 participants (768 individuals comprising 384 male–female dyads, and an additional 136 individuals not in a dyad) from four family mediation centers in the United States and Australia. Parties were seeking family mediation to resolve an initial relationship dissolution or to modify the terms of their previous dissolution. All dyads and individuals in the sample were divorcing, divorced, or unmarried parties in different-sex relationships.

Procedures

All study procedures were approved by the (Indiana University) Institutional Review Board. At each clinic, the MASIC was administered prior to starting mediation as part of the intake procedures. It was administered as an interview with each party separately for the majority of participants but for a small subset of cases was self-administered. In all cases, information gathered on the MASIC was not shared with the other party in the case. At one mediation clinic (n = 350 from 175 dyads plus five individuals not in a dyad, for a total of 355 participants), we had data in addition to the MASIC, including criminal and court records, the mediation clinic file for the case, participant research questionnaires completed during the mediation intake inquiring about demographic information and the mediation case, and mediator-completed questionnaires about the mediation process.

Measures

Demographics

Participants at all four family mediation centers provided their age, sex/gender, and employment status.

MASIC

Participants at all four mediation centers completed the MASIC, which assesses party-reported victimization of IPV based on the other party’s behaviors. It has 37 behaviorally specific (yes or no) items assessing for different types of abusive behaviors (i.e., psychological abuse, coercive control, threats of severe violence, physical violence, severe physical violence, sexual violence, and stalking) that had ever occurred in the relationship (e.g., “Has the other party ever hit or punched you?”).

Mediation Outcomes

For 161 cases from one clinic, each case was classified on whether it came to full or partial (n = 122) or no (n = 38) agreement, or whether parties were screened out of mediation due to concerning levels of IPV victimization (n = 1). An additional 19 cases did not start mediation due to other reasons (e.g., mediator conflict of interest) and were excluded from analyses that examined mediation outcomes.

Additionally, at the same clinic, mediators completed questionnaires at the end of mediation regarding their decision to make any accommodations to the mediation process due to believed or suspected IPV in the case. Possible procedures or accommodations due to IPV concerns included shuttle mediation (i.e., parties remain in separate rooms at all times), a mixture of joint and shuttle mediation (i.e., parties both in the same room at some times and in separate rooms at other times), having an advocate, support person, or attorney attend mediation with a party, parties escorted to and from the building, and the party being instructed on how to leave the building without being seen by the other party. Mediators indicated which of these procedures or accommodations they used in the case. At least one of the accommodations or procedures was made for 47 out of 160 cases. We used the aforementioned information to construct a summary binary (yes/no) variable representing whether a case had any potentially negative or concerning outcomes (i.e., not reaching agreement, being screened out of mediation due to IPV) or procedural accommodations in mediation (n = 72 cases out of 161 cases included in analyses).

Court Records

Also for participants at the same clinic, we coded parties’ criminal and civil record information from the state in which the clinic was located (Pokman et al., 2014), including all previous charges potentially associated with IPV (e.g., domestic battery, stalking, criminal confinement) and all previous protective orders/no contact orders that had been petitioned against a party since the age of 18. Charges and orders may or may not have been related to the other party in the case. This information was collected from a court database including 86 of the 92 counties in the state. Therefore, parties may have had charges and orders issued against them in other jurisdictions that we were unable to examine.³

Two coders coded information for approximately 20% of the sample to determine interrater reliability. The average intraclass correlation coefficient was 0.97, indicating excellent reliablity. Any discrepancies in the data were discussed by the coders and resolved through consensus.

When all IPV-related criminal charges and protective/no contact orders were summed, the number of charges and orders ranged from 0 to 17 (M = 0.50, SD = 1.56; females: M = 0.17, SD = 0.64; males: M = 0.82, SD = 2.06); 73 parties (21%; 9.5% of females and 33% of males) had one or more criminal charges, protective orders, or no contact orders issued against them.

Analytical Method

IRT Assumptions

IRT is predicated on the rationale that each item may function differently and offers unique information to the measure. Items need to meet three criteria to provide reliable model estimates: unidimensionality, local independence, and monotonicity. Unidimesionality indicates that items must reflect only one underlying construct or latent trait. The MASIC assesses for multiple subscales rather than just one (Pokman et al., 2014), thus violating this assumption. Therefore, to address the multidimensionality of the data, we conducted multidimensional IRT (MIRT). Furthermore, to account for interdependence resulting from dyadic level data, we included males and females as separate dimenions in the same model.

The second assumption, local independence, states that items are uncorrelated after controlling for the latent trait. This assumption is generally violated when unidimensional IRT models are estimated using data sets that contain more than one dimension (Edwards, 2009). Use of a MIRT model improved the study’s ability to meet this assumption.

The third assumption indicates a monotonic relationship between the latent trait and responses to an item. In other words, it is assumed that individuals experiencing higher levels of IPV victimization have a greater probability of endorsing more extreme or severe items than those experiencing lower levels of IPV victimization. We checked for violation of this assumption by assessing for model fit. Poor model fit may suggest inconsistencies in the data.

IRT Models and Parameters (Item Discrimination and Difficulty)

The 37 behaviorally specific items on the MASIC were analyzed using a two-parameter logistic (2-PL) IRT model within Mplus Version 7.31 (Muthén & Muthén, 1998-2017). A 2-PL model was most appropriate for the following reasons: (1) 2-PL models allow for the examination of binary items, that is, items with two response categories (“yes” the behavior occurred or “no” the behavior did not occur); and (2) A 2-PL model was selected over a 3-PL model given that participants are unlikely to use “guessing” as a strategy for responding to items about IPV. Missing data were estimated using Full Information Maximum Likelihood. We tested multiple MIRT models in order to determine the best fitting model. Models differed according to number of IPV subscales. The IPV subscales were determined a priori in the IRT models. (Pokman et al., (2014).

Six subscales: psychological abuse, coercive control, threats of severe violence, physical violence, stalking, and sexual violence.

Four subscales: psychological abuse + coercive control, stalking + threats of severe violence, physcial violence, and sexual violence.

Two subscales: physical violence and nonphysical violence.

We first examined a model that accounted for six IPV subscales (a total of 12 dimensions—six subscales for males and the same six subscales for females; Pokman et al., 2014). We next examined a model that accounted for four IPV subscales (a total of eight dimensions), and then a model that accounted for two subscales (a total of four dimensions). The models with six and with four subscales demonstrated difficulties with convergence and stabilization. The model with two IPV subscales terminated normally. Results of this two-subscale model were used for interpretation and in examining the three assessment approaches.

The best fitting model identified (i.e., two subscales) provided information on the difficulty (b₁) and discrimination (a₁) parameters for each of the 37 items on the MASIC. The model estimated four latent variables (or thetas): females’ reported victimization from nonphysical violence and from from physical violence, and males’ reported victimization from nonphysical violence and from physical violence.

Approach 1 (Selecting for the Highest Levels of IPV)

This approach selected items with difficulty estimates two standard deviations above the mean level of each latent variable or theta. Excluded were items with difficulty estimates that fell outside of this range as well as items with poor discrimination (a₁ < 1.2).

Approach 2 (Selecting for Any Level of Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

This approach sought to identify ranges of IPV victimization or thetas that corresponded with increased probability of not reaching agreement, being screened out of mediation due to concerns about IPV, or utilization of specialized accommodations or procedures during the mediation process due to IPV concerns. A summary binary variable represented whether a case had any potentially negative or concerning outcomes or procedural accommodations in mediation (yes or no). This summary variable was correlated with each theta. This allowed identification of a range of thetas or IPV victimization that corresponded to having any level of risk for a potentially negative or concerning outcome or use of specalized accommodations in mediation. Items were excluded that had difficulty estimates falling outside of the identified range of theta as well as items with poor discrimination (a₁ < 1.2).

Approach 3 (Selecting for Highest Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

This approach identified ranges of IPV victimization or thetas that corresponded with a 50% or greater likelihood of having a potentially negative or concerning mediation outcome or using specialized accomodations according to the summary variable described in Approach 2. Excluded were items that fell outside of those ranges as well as items with poor discrimination (a₁ < 1.2).

Validation of the Subset of Items for Each Approach

After identifying the three subsets of items corresponding to each of the three approaches, we tested the convergent validity of these subsets of items. Convergent validity was tested by calculating a MASIC IPV ever variety total score using only the new subset of items; this score represented the total number of violent and abusive behaviors ever having occurred in the relationship based on the new subset of items. Missing values were computed based on averages in each of the original six subscales (Pokman et al., 2014). Next, using a Spearman correlation to account for the interdependence and nonnormal distribution of the data, IPV ever variety scores corresponding to the new subset of items were correlated with a sum of all IPV-related criminal charges and protective orders/no contact orders issued against the other party in the case.

Results

Demographic and IPV Information

Table 1 summarizes the sample’s demographic information. Comparisons of demographic characteristics across the four sites indicated no significant differences on gender or employment status, but participants from the center with complete data, compared with those from at least one other center, were younger, F(3, 805) = 5.18, p = .002, less likely to have a 2-year college degree or higher level of education, χ²(2, 413) = 30.75, p < .01), and significantly more likely to have an annual income below US $40,000, χ²(2, 596) = 34.72, p < .01). There were also significant differences across the four centers on IPV victimization, using IPV ever variety scores, with parties in three of the centers reporting higher levels of IPV victimization than parties in the fourth center, F(3, 831) = 4.22, p = .006 (see Table 1).

Table 1.

Summary of Demographic Information and IPV Victimization Across Mediation Centers.

	Family mediation center
	Center 1 (n = 355)	Center 2 (n = 377)	Center 3 (n = 77)	Center 4 (n = 95)
	M (SD) or %	M (SD) or %	M (SD) or %	M (SD) or %
Full sample (N = 904)
Age, in years	34.43 (9.70)	37.00 (11.24)	38.41 (9.12)	35.62 (7.56)
Sex/gender
Female %	50.7	51.5	55.8	49.5
Employment status
Employed %	68.8	77.3	80.5	69.9
IPV ever variety score	11.30	11.18	7.63	11.09
Partial sample (n = 732)
Annual income (US$)
Less than $40,000	86.9	41.6	—	—
Number of shared children	1.33 (1.00)	1.36 (0.89)	—	—
Cases involving children %	88.6	92.9	—	—
Length of marriage, in years	8.50 (6.45)	8.87 (6.96)	—	—
Partial sample (n = 450)
Education
Two-year college and beyond %	21.1	—	—	50.5
Partial sample (n = 355)
Race
White %	92.3	—	—	—

Note. IPV = intimate partner violence.

Table 2 demonstrates the frequency with which items on the MASIC were endorsed and whether these frequencies differ significantly by gender. Differences in item frequencies among males and females were in the expected direction according to prior literature. For example, female victims were more likely than male victims to report fear of the other party, an IPV-related injury that required medical attention, and that the other party physical abused or threatened to abuse pets (Febres, 2015; Pokman et al., 2014). Regarding court records, males had significantly more IPV-related criminal charges and protective/no contact orders (mean rank = 34.86) compared with females (mean rank = 30.04; Z = −4.97, p < .001).

Table 2.

Percentage of Males and Females Endorsing Each MASIC Item.

MASIC item	Males’ reported victimization (n = 444)	Females’ reported victimization (n = 464)	Significant difference by gender?
MASIC item	% Endorsed ever in relationship	% Endorsed ever in relationship	Significant difference by gender?
Nonphysical violence items
Call you names?	75	76	χ²(888) = 0.21, p = .65
Insult you or make you feel bad in front of others?	57	66	χ²(888) = 11.08, p < .01**
Yell or scream at you?	84	81	χ²(886) = 1.80, p = .18
Forbid you to go out without him or her?	24	26	χ²(887) = 0.41, p = .52
Try to control how much money you had or spent?	35	36	χ²(885) = 0.17, p = .68
Try to control your activities in or outside the home?	46	42	χ²(886) = 1.27, p = .26
Try to control your contact with family and friends?	34	36	χ²(887) = 0.34, p = .56
Act extremely jealous, or frequently check up on where you've been or who you’ve been with?	53	47	χ²(888) = 3.77, p = .052
Demand that you obey him or her?	35	31	χ²(898) = 2.07, p = .15
Physically abuse or threaten to abuse pets to scare or hurt you, or when angry at you?	8	14	χ²(885) = 8.71, p < .01**
Punish or deprive the children because he or she was angry at you?	29	29	χ²(863) = 0.01, p = .91
Make threatening gestures or faces at you or shake a fist at you?	42	49	χ²(884) = 5.58, p = .02*
Threaten to take or have the children taken away from you?	57	54	χ²(867) = 0.52, p = .47
Destroy property, for example, hit or kick a wall, door, or furniture or throw, smash, or break an object?	39	51	χ²(888) = 12.38, p < .01*
Drive dangerously to scare you, or when angry at you?	12	26	χ²(886) = 29.69, p < .01**
Throw an object at you to scare or hurt you, or when angry at you?	35	27	χ²(888) = 6.26, p = .01*
Destroy or harm something you care about?	25	27	χ²(886) = 0.23, p < .63
Threaten to hurt someone you care about?	13	19	χ²(886) = 7.63, p < .01**
Threaten to hurt you?	28	35	χ²(886) = 4.80, p = .03*
Threaten to kill himself or herself?	22	28	χ²(890) = 4.40, p = .04*
Threaten to kill you?	14	20	χ²(890) = 6.52, p = .01
Threaten you with a weapon or something like a weapon?	14	13	χ²(887) = 0.00, p = .97
Demand or insist that you engage in sexual activities against your will?	4	15	χ²(890) = 29.27, p < .01**
Follow or spy on you in a way that made you feel frightened or harassed?	20	25	χ²(889) = 2.53, p = .11
Try to contact you against your will or in a way that made you feel frightened or harassed, for example, by unwanted phone calls, mail, text messages, or Facebook contacts?	29	33	χ²(890) = 2.11, p = .15
Stand outside your home, school, workplace, or other places where he/she had no business being, and in a way that made you feel frightened or harassed?	17	20	χ²(889) = 1.81, p = .28
Leave items for you to find in a way that made you feel frightened or harassed?	6	5	χ²(890) = 0.42, p = .52
	Males’ reported victimization	Females’ reported victimization
Physical violence items
Hold you down, pinning you in place?	7	36	χ²(902) = 107.31, p < .01**
Push, shove, shake, or grab you?	42	51	χ²(903) = 7.64, p < .01**
Scratch you, or pull your hair, or twist your arm, or bite you?	27	24	χ²(902) = 0.66, p = .42
Slap you?	34	22	χ²(903) = 17.29, p < .01**
Hit or punch you?	34	28	χ²(901) = 3.55, p = .06
Kick or stomp on you?	13	12	χ²(903) = 0.18, p = .68
Choke or strangle you?	12	24	χ²(903) = 21.56, p < .01**
Burn you with something?	2	1	χ²(902) = 0.78, p = .38
Use a weapon or something like a weapon against you?	13	8	χ²(903) = 6.49, p = .01*
Physically force you to engage in sexual activities against your will?	2	10	χ²(903) = 24.39, p < .01**
	Males’ reported victimization	Females’ reported victimization
Additional items (not in IRT)
As a result of the other party’s behaviors, did you ever feel fearful, scared, or afraid of physical harm to yourself or to others?	26	46	χ²(869) = 38.87, p < .01**
Scratch, small bruise, swelling, or other mild injury?	30	38	χ²(600) = 4.00, p = .046*
Fracture, small burn, cut, large bruise, or other moderate injury?	16	27	χ²(599) = 10.12, p < .01**
Major wound, severe bleeding or burn, being knocked out, or other severe injury?	11	15	χ²(600) = 1.82, p = .18
Blindness, loss of hearing, disfigurement, chronic pain, or other permanent damage?	10	11	χ²(600) = 0.21, p = .65
Did you seek or should you have sought medical attention for any injury caused by the other party?	14	23	χ²(599) = 8.50, p < .01**
As a result of the other party’s behaviors, did you or someone else ever call the police?	30	39	χ²(879) = 8.04, p < .01**

Note. N = 908 participants; item stem “Has the other party ever . . . ”; MASIC = Mediator’s Assessment of Safety Issues and Concerns; IRT = item response theory.

p < .05. **p < .01.

Multidimensional IRT

Based on a four dimensional 2-PL model, Table 3 provides information regarding item discrimination and difficulty for each of the 37 items examined for males, and Table 4 provides the same information for females. We examined the bivariate fit of items, which compares the observed and expected frequencies for pairs of items. Examination of item fit statistics is recommended in IRT even if models demonstrate good overall model fit (Maydeu-Olivares, 2013). A large proportion of item pair comparisons did not produce significant chi-squared tests, suggesting reasonable to good model fit.

Table 3.

Item Discrimination and Difficulty Estimates for Males’ Reports of Their Own Victimization.

MASIC Item	Item discrimination	Item difficulty	Approach 1	Approach 2	Approach 3
Call you names?	1.700	−1.633	Low difficulty	Low difficulty	Low difficulty
Insult you or make you feel bad in front of others?	1.732	−0.306	Low difficulty	^a	Low difficulty
Yell or scream at you?	1.576	−2.386	Low difficulty	Low difficulty	Low difficulty
Forbid you to go out without him or her?	1.472	1.592	Low difficulty	^a	Low difficulty
Try to control how much money you had or spent?	1.466	0.912	Low difficulty	^a	Low difficulty
Try to control your activities in or outside the home?	1.670	0.269	Low difficulty	^a	Low difficulty
Try to control your contact with family and friends?	1.356	0.917	Low difficulty	^a	Low difficulty
Act extremely jealous, or frequently check up on where you’ve been or who you've been with?	1.712	−0.148	Low difficulty	^a	Low difficulty
Demand that you obey her?	1.804	0.992	Low difficulty	^a	Low difficulty
Physically abuse or threaten to abuse pets to scare or hurt you, or when angry at you?	1.199	2.995	Low discrimination	Low discrimination	Low discrimination
Punish or deprive the children because she was angry at you?	1.180	1.151	Low discrimination, Low difficulty	Low discrimination	Low discrimination, Low difficulty
Make threatening gestures or faces at you or shake a fist at you?	2.667	0.774	Low difficulty	^a	Low difficulty
Threaten to take or have the children taken away from you?	1.408	−0.341	Low difficulty	^a	Low difficulty,
Destroy property, for example, hit or kick a wall, door, or furniture or throw, smash, or break an object?	1.998	0.775	Low difficulty	^a	Low difficulty
Drive dangerously to scare you, or when angry at you?	1.594	2.766	^a	^a	^a
Throw an object at you to scare or hurt you, or when angry at you?	2.459	1.264	Low difficulty	^a	Low difficulty
Destroy or harm something you care about?	2.366	1.980	Low difficulty	^a	^a
Threaten to hurt someone you care about?	1.683	2.759	^a	^a	^a
Threaten to hurt you?	3.522	2.323	^a	^a	^a
Threaten to kill herself?	0.990	1.535	Low discrimination, Low difficulty	Low discrimination	Low discrimination, Low difficulty
Threaten to kill you?	2.211	3.063	^a	^a	^a
Threaten you with a weapon or something like a weapon?	2.277	3.141	^a	^a	^a
Demand or insist that you engage in sexual activities against your will?	1.671	4.108	^a	^a	^a
Follow or spy on you in a way that made you feel frightened or harassed?	2.111	2.253	^a	^a	^a
Try to contact you against your will or in a way that made you feel frightened or harassed, for example, by unwanted phone calls, mail, text messages, or Facebook contacts?	1.553	1.326	Low difficulty	^a	Low difficulty
Stand outside your home, school, workplace, or other places where she had no business being, and in a way that made you feel frightened or harassed?	1.963	2.527	^a	^a	^a
Leave items for you to find in a way that made you feel frightened or harassed?	1.566	3.577	^a	^a	^a
Hold you down, pinning you in place?	2.568	4.423	^a	^a	^a
Push, shove, shake, or grab you?	3.018	0.871	Low difficulty	^a	Low difficulty
Scratch you, or pull your hair, or twist your arm, or bite you?	3.284	2.447	^a	^a	^a
Slap you?	2.529	1.344	Low difficulty	^a	Low difficulty
Hit or punch you?	4.691	2.357	^a	^a	^a
Kick or stomp on you?	2.792	3.654	^a	^a	^a
Choke or strangle you?	1.415	2.628	^a	^a	^a
Burn you with something?	2.621	6.213	^a	^a	^a
Use a weapon or something like a weapon against you?	2.763	3.594	^a	^a	^a
Physically force you to engage in sexual activities against your will?	2.237	5.809	^a	^a	^a

Note. MASIC = Mediator’s Assessment of Safety Issues and Concerns.

Item not identified as having low difficulty or low discrimination.

Table 4.

Item Discrimination and Difficulty Estimates for Females’ Reports of Their Own Victimization.

MASIC item	Item discrimination	Item difficulty	Approach 1	Approach 2	Approach 3
Call you names?	2.077	−1.899	Low difficulty	Low difficulty	Low difficulty
Insult you or make you feel bad in front of others?	2.196	−1.114	Low difficulty	Low difficulty	Low difficulty
Yell or scream at you?	1.721	−2.101	Low difficulty	Low difficulty	Low difficulty
Forbid you to go out without him?	2.489	2.013	^a	^a	^a
Try to control how much money you had or spent?	1.136	0.754	Low discrimination, Low difficulty	Low discrimination	Low discrimination, Low difficulty
Try to control your activities in or outside the home?	2.146	0.609	Low difficulty	^a	Low difficulty
Try to control your contact with family and friends?	1.824	0.967	Low difficulty	^a	Low difficulty
Act extremely jealous, or frequently check up on where you've been or who you’ve been with?	2.039	0.312	Low difficulty	^a	Low difficulty
Demand that you obey him?	2.398	1.583	Low difficulty	^a	^a
Physically abuse or threaten to abuse pets to scare or hurt you, or when angry at you?	1.309	2.349	^a	^a	^a
Punish or deprive the children because he was angry at you?	1.180	1.176	Low discrimination, Low difficulty	Low discrimination	Low discrimination, Low difficulty
Make threatening gestures or faces at you or shake a fist at you?	3.029	0.216	Low difficulty	^a	Low difficulty
Threaten to take or have the children taken away from you?	1.453	−0.192	Low difficulty	^a	Low difficulty
Destroy property, for example, hit or kick a wall, door, or furniture or throw, smash, or break an object?	1.795	0.009	Low difficulty	^a	Low difficulty
Drive dangerously to scare you, or when angry at you?	1.672	1.545	Low difficulty	^a	^a
Throw an object at you to scare or hurt you, or when angry at you?	2.131	1.749	Low difficulty	^a	^a
Destroy or harm something you care about?	2.427	1.909	Low difficulty	^a	^a
Threaten to hurt someone you care about?	1.699	2.104	^a	^a	^a
Threaten to hurt you?	3.172	1.498	Low difficulty	^a	^a
Threaten to kill himself?	1.130	1.224	Low discrimination, Low difficulty	Low discrimination	Low discrimination, Low difficulty
Threaten to kill you?	2.407	2.518	^a	^a	^a
Threaten you with a weapon or something like a weapon?	2.133	3.03	^a	^a	^a
Demand or insist that you engage in sexual activities against your will?	2.007	2.719	^a	^a	^a
Follow or spy on you in a way that made you feel frightened or harassed?	1.655	1.64	Low difficulty	^a	^a
Try to contact you against your will or in a way that made you feel frightened or harassed, for example, by unwanted phone calls, mail, text messages, or Facebook contacts?	1.256	0.949	Low difficulty	^a	Low difficulty
Stand outside your home, school, workplace, or other places where he had no business being, and in a way that made you feel frightened or harassed?	1.813	2.171	^a	^a	^a
Leave items for you to find in a way that made you feel frightened or harassed?	1.700	3.938	^a	^a	^a
Hold you down, pinning you in place?	3.519	1.554	Low difficulty	^a	^a
Push, shove, shake, or grab you?	3.735	0.087	Low difficulty	^a	Low difficulty
Scratch you, or pull your hair, or twist your arm, or bite you?	3.327	2.714	^a	^a	^a
Slap you?	3.277	2.956	^a	^a	^a
Hit or punch you?	3.272	2.262	^a	^a	^a
Kick or stomp on you?	4.178	5.21	^a	^a	^a
Choke or strangle you?	1.914	1.877	Low difficulty	^a	^a
Burn you with something?	1.84	5.789	^a	^a	^a
Use a weapon or something like a weapon against you?	1.894	3.583	^a	^a	^a
Physically force you to engage in sexual activities against your will?	1.793	3.278	^a	^a	^a

Note. MASIC = Mediator’s Assessment of Safety Issues and Concerns.

Item not identified as having low difficulty or low discrimination.

Selecting a Clinically Useful Subset of Items According to Each Approach

After estimating the final model, we used the model results to select items using several steps described under each approach.

Approach 1 (Selecting for the Highest Levels of IPV)

Step 1: Eliminate items demonstrating low discrimination

We excluded items with discrimination estimates below 1.2. Overall, three items were excluded for males and three items were excluded for females. Refer to Tables 3 and 4 for a list of items excluded due to low discrimination under Approach 1.

Step 2: Identify a clinically useful range of IPV victimization

This approach identified items with difficulty estimates that corresponded to two standard deviations above the mean level of each theta, that is, difficulty estimates equal to or greater than 2.

Step 3: Exclude items with difficulty estimates outside of the ranges of interest

We excluded items with difficulty estimates less than 2. Regarding nonphysical violence, 17 items had difficulty estimates <2 for males and 19 items had difficulty estimates <2 for females. Regarding physical violence items, two items had difficulty estimates <2 for males and three items had difficulty estimates <2 for females. See Tables 3 and 4 for information on item difficulty. Overall, 19 items for males and 22 items for females were excluded.

Final subset of clincally useful MASIC items

In total, considering poor discrimination and low difficulty, 20 items were exluded for males and 22 items were excluded for females. Thus, the final subset consisted of 17 items for males and 15 items for females, with 12 of these items being the same for males and females. See Tables 3 and 4 for the final subset of items under this approach.

Validation of the subset of items

To test the convergent validity of this new subset of items, we examined the association between ever variety scores derived from the new subset of items as reported by one party in the case and the IPV-related criminal charges and protective orders/no contact orders issued against the other party in the case. Using a Spearman correlation to account for the interdependence of the data, results indicated a significant association between reported IPV victimization and all IPV-related criminal charges and protective orders/no contact orders against the reported perpetrating party (r = .22, p < .01). When examining by gender, there was a significant correlation between females’ reports of male-perpetrated IPV and males’ IPV-related criminal charges and protective orders/no contact orders (r = .29, p < .01). The association between males’ reports of female-perpetrated IPV and females’ IPV-related criminal charges and protective orders/no contact orders was not significant (r = .11, p = .162).

Approach 2 (Selecting for Any Level of Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

Step 1: Eliminate items demonstrating low discrimination

Similar to Approach 1, we excluded items with discrimination estimates below 1.2. Three items were excluded for males and three for females. Refer to Tables 3 and 4 for a list of items excluded due to low discrimination under Approach 2.

Step 2: Identify a clinically useful range of IPV victimization

To identify a clinically useful range of IPV victimization under this approach, we examined the relationship between the four latent variables or thetas and the summary variable representing potentially negative or concerning mediation outcomes or use of specialized accommodations. For males, results indicated the need to exclude nonphysical violence items with difficulty estimates below −1.12 and physical violence items with difficulty estimates below -0.83. For females, results indicated the need to exclude nonphysical violence items with difficulty estimate below −0.68 and physical violence items with difficulty estimates below −1.11. Ranges indicated exclude outliers.

Step 3: Exclude items with difficulty estimates outside of the ranges of interest

Regarding nonphysical violence, two items had difficulty estimates below the identified range for males, and three items had difficulty estimates below the identified range for females. No physical violence items had difficulty estimates outside of the identified ranges for males or females. See Tables 3 and 4 for information on item difficulty. Overall, two items for males and three items for females were excluded.

Final subset of clincally useful MASIC items

In total, considering poor discrimination and low difficulty, five items were exluded for males and six items were excluded for females. Thus, the final subset consisted of 32 items for males and 31 items for females, with 30 items the same for males and females.

Validation of the subset of items

To test the convergent validity of this new subset of items, we used the same method as for Approach 1. Using a Spearman correlation to account for the interdependence of the data, results indicated a significant association between reported IPV victimization and all IPV-related criminal charges and protective orders/no contact orders against the reported perpetrating party (r = .20, p < .01). When examining by gender, there was a significant correlation between females’ reports of male-perpetrated IPV and males’ IPV-related criminal charges and protective orders/no contact orders (r = .28, p < .01). The association between males’ reports of female-perpetrated IPV and females’ IPV-related criminal charges and protective orders/no contact orders was not significant (r = .09, p = .126).

Approach 3 (Selecting for Highest Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

Step 1: Eliminate items demonstrating low discrimination

Similar to Approaches 1 and 2, we excluded items with discrimination estimates below 1.2. Three items were excluded for males and three items for females. Refer to Tables 3 and 4 for a list of items excluded due to low discrimination under Approach 3.

Step 2: Identify a clinically useful range of IPV victimization

In this approach, we identified the ranges of thetas that corresponded to a 50% or greater likelihood of having a potentially negative or concerning outcome in mediation or using specialized accommodations. Excluding outliers, results indicated that for males, nonphysical violence items with difficulty estimates greater than 1.6 and physical violence items with difficulty estimates greater than 1.4 were associated with IPV victimization levels related to highest risk for a potentially negative or concerning outcome or using specialized accommodations in mediation. For females, nonphysical violence items with difficulty estimates greater than 1.6 and physical violence items with difficulty estimates greater than 1.4 were associated with these levels.

Step 3: Exclude items with difficulty estimates outside of the ranges of interest

Regarding nonphysical violence, 16 items had difficulty estimates below the identified range for males, and 13 items had difficulty estimates below the identified range for females. Regarding physical violence, two items had difficulty estimates outside of the identified ranges for males and one item had difficulty estimates outside of the identified ranges for females. See Tables 3 and 4 for information on item difficulty. Overall, we excluded 18 items for males and 14 items for females.

Final subset of clincally useful MASIC items

In total, considering poor discrimination and low difficulty, 19 items were exluded for males and 14 for females. Thus, the final subset consisted of 18 items for males and 23 items for females, with 18 of these items the same for males and females.

Validation of the subset of items

To test the convergent validity of this new subset of items we used the same method as for Approaches 1 and 2. Using a Spearman correlation to account for the interdependence of the data, results indicated a significant association between reported IPV victimization and all IPV-related criminal charges and protective orders/no contact orders against the reported perpetrating party (r = .28, p < .01). When examining by gender, there was a significant correlation between females’ reports of male-perpetrated IPV and males’ IPV-related criminal charges and protective orders/no contact orders (r = .30, p < .01). The association between males’ reports of female-perpetrated IPV and females’ IPV-related criminal charges and protective orders/no contact orders was not significant (r = .12, p = .109).

Discussion

This study addressed the need for standardized yet clinically useful and efficient IPV screening tools in family mediation by examining three different IRT approaches to refine the MASIC. The first approach sought to create an IPV screening tool that allows mediators to identify individuals reporting the highest and potentially most clinically significant levels of IPV victimization. The second approach constructed an IPV measure that helps mediators identify individuals at any risk for having a potentially negative or concerning outcome in mediation or needing special accommodations during the mediation process due to IPV. The third approach created a measure that assists with identification of individuals having the highest risk (a 50% or higher chance) for a potentially negative or concerning outcome in mediation or needing additional accommodations due to IPV. In all three approaches, findings revealed several items that may be appropriately excluded from the MASIC. Items excluded differed for males and females. Additionally, there was initial evidence of the convergent validity of each approach.

In all three approaches, items were eliminated based on two criteria. First, a clinically efficient IPV measure should contain items that are able to discriminate well between high and low levels of IPV. Three of the MASIC items we excluded due to low discrimination were the same for males and females across all three approaches. Some of these items inquired about IPV behaviors that can potentially be influenced by external factors. For example, an intimate partner may threaten to kill him or herself as a result of affective instability or borderline personality characteristics (Matusiewicz et al., 2014) regardless of the level of IPV perpetration. Thus, an inconsistent association between the IPV behavior and level of IPV may result in a poorly discriminating item.

Moreover, poorly discriminating items may be unreliable markers of IPV victimization due to a low frequency occurrence. In the present study, the item “Physically abuse or threaten to abuse pets to scare or hurt you or when angry at you?” demonstrated low discrimination ability for male victims only, not female victims. Consistent with previous research indicating that animal abuse is more prevalent among male than female IPV perpetrators (Febres, 2015), in this study, only 8% of male victims endorsed female partners’ perpetrated animal abuse compared with 14% of female victims (p < .01), suggesting that this item may not be as informative in an IPV screen for male victims. However, not all items with a low base rate are poor discriminators. The MASIC item inquiring about being burned by the other party was endorsed by only 2% of the sample, but results demonstrated a more consistent and distinct association between this behavior and high levels of IPV victimization.

The second criterion used for eliminating items from the MASIC was based on item sensitivity to differing ranges of IPV victimization depending on the approach examined. Despite differences in how our ranges of interest were selected within each approach, the final subsets of items derived using the different approaches were relatively similar. For example, all approaches led to the inclusion of various physical violence items and few psychological abuse items. This suggests that generally, regardless of approach, items related to more severe IPV behaviors (e.g., “Hit or punch you?” “Kick or stomp on you?”) are associated with higher overall levels of IPV victimization and having potentially concerning outcomes in mediation or needing special accommodations. Thus, these items are likely more clinically relevant to mediators. Items related to less severe IPV behaviors, such as “Call you names?” and “Yell or scream at you?” have been typically characterized by high endorsement rates of up to 98% among divorcing or separating parties seeking mediation services (Beck et al., 2009; Pokman et al., 2014; Rossi et al., 2015a). In the present study, items inquiring about name-calling and yelling were endorsed by 76% and 83% of the sample, respectively. Given such high endorsement rates, these IPV behaviors tend to be associated with many different levels of IPV victimization among divorcing or separating parties and thus do not necessarily help identify those cases that may need special considerations in mediation due to IPV.

Clinical Implications of the Differing Approaches

The final subsets of items for both males and females across approaches achieved our goals for revising the MASIC in some ways, but not others. Across all approaches, common items inquire about a wide range of IPV behaviors that may be clinically important to mediation staff, such as behaviors related to physical violence, coercive control, threats of physical violence, stalking, sexual violence, and psychological abuse (though no items related to psychological abuse were retained across all approaches for female victims). Additionally, consistent with the goal of developing a more clinically efficient IPV measure for mediation, the final subset of MASIC items derived from each approach was tailored specifically to assist mediation staff with identifying IPV cases that may need particular consideration in the mediation process. Despite these advantages, each approach also demonstrated potential disadvantages.

Approach 1 (Selecting for the Highest Levels of IPV)

This approach was most sensitive to parties reporting the highest levels of IPV victimization. While this is an important group to detect in mediation, it is possible that this approach fails to detect a significant number of parties reporting lower levels of IPV who may also need additional assistance in mediation. Future research is needed to determine the level of IPV at which mediation is unlikely to benefit divorcing or separating parties. Knowing this threshold would allow us to construct an IPV measure that more accurately identifies parties needing assistance in the mediation process.

Approach 2 (Selecting for Any Level of Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

The second approach focused on potentially negative or concerning mediation outcomes, including not reaching agreement and being screened out of mediation due to concerns about IPV, or utilization of specialized accommodations or procedures during the mediation process due to IPV concerns. Though the outcomes examined provide some indication of potential concern in mediation, they do not represent all mediation outcomes possibly indicating concern. Future studies should focus on other negative or concerning outcomes, like continuation of IPV during or following mediation or parties’ inability to abide by the provisions agreed to in mediation. Doing so would allow better identification of divorcing or separating parties least likely to benefit from mediation and/or most likely to need assistance to benefit from mediation.

Furthermore, Approach 2 used not reaching agreement as an outcome by which to select items. Not reaching agreement in mediation is generally considered a negative outcome for the court given the unnecessary use of resources and time. However, some argue that not reaching agreement may be seen as a positive outcome among cases with IPV because it suggests that victims felt empowered and able to voice their interests in mediation rather than being coerced by the perpetrator into making agreements that do not represent their interests (Rossi et al., 2015). Additionally, it is possible that parties may still benefit from the mediation process even when not reaching agreement by having had the opportunity to think through and discuss important issues. It is important that future studies clarify if there are indeed potential benefits to not reaching a mediation agreement and whether reaching or not reaching agreement should be a critical factor in determining mediation appropriateness for parties reporting IPV.

Finally, under this approach, we did not eliminate many items (five for males and six for females). Although one of our main goals was to create a briefer IPV measure suitable for the time constraints that exist in many mediation settings, the final 32 items for males and 31 items for females under Approach 2 may not result in significantly shorter and more efficient IPV screening.

Approach 3 (Selecting for Highest Risk for Potentially Negative or Concerning Mediation Outcomes or Use of Specialized Accommodations)

The third approach, with many of the same disadvantages listed under Approach 2, did allow significant shortening of the MASIC. However, selecting only the items associated with highest risk of potentially negative or concerning mediation outcomes or needing specialized accommodations could result in mediation staff failing to detect parties reporting IPV who still have some likelihood of experiencing potentially negative or concerning outcomes or needing specialized accommodations in mediation. Also, the highest risk items ask about behaviors pertaining to physical violence, threats of severe violence, stalking, sexual violence, and coercive control for both males and females. Yet, there are five highest risk items inquiring about coercive control victimization for females and only one item for males. This may indicate that coercive control victimization is related to higher risk for specialized accommodations or a potentially negative or concerning outcome in mediation for females compared with males.

Across Approaches 1, 2, and 3

As noted, findings demonstrated some differences in item information across males and females, suggesting that the field may need to consider separate IPV screening tools for females and males (O’Hara et al., 2018). However, given the large overlap between the subsets of items for males and females within each approach, practitioners may consider that it is more clinically practical to administer just one combined measure. It is also important for practitioners to consider the clinical implications of administering different sets of items to parties in a case. On the one hand, one party may become concerned or upset that s/he was not asked the same questions as the other party. On the other hand, the mediator can take a transparent approach and tell both parties that they were asked some similar and some different questions with responses kept confidential, potentially providing a victim with some protection if the perpetrator demands to know, from the victim, if IPV was assessed and what answers were provided. It should be highlighted that we were unable to examine same-sex couples in the present study. It is unclear whether the subsets of items identified in this study under each approach would apply to same-sex couples and whether parties within such couples would also require differentiated sets of items. Although research has shown that same-sex couples report similar IPV behaviors and dynamics as different-sex couples, parties in same sex relationships can experience unique stressors related to being a sexual minority (e.g., stigma, discrimination) that exacerbate risk of IPV (Rollè et al., 2018). Thus, IPV screens for same-sex couples undergoing mediation may need to assess for these unique stressors so that parties have sufficient resources during the mediation and separation process. Additional research is needed determining how IPV screens for mediation should be adapted for same-sex couples.

Furthermore, it is important to note that the approaches examined in this study provide empirically based recommendations of IPV items that may be helpful to mediation staff when making decisions about mediation with parties reporting IPV, though, staff members who have more time available for screening may choose to ask additional follow-up IPV-related questions. However, the current study was designed in response to the critical need to develop a shorter IPV screen for family mediation settings. Short IPV screens for mediation are essential because they make conducting standardized and effective IPV screening in mediation practical and feasible, and allow for empirically based decisions that help protect IPV victims and their families. Each approach identified the items that are most helpful to mediation staff in making decisions about the appropriateness of mediation for parties reporting IPV and excluded the items that are less helpful in this process, thus reducing the need to ask supplemental questions.

Clinical Recommendations

Despite the potential advantages and disadvantages of all three approaches examined in this study, we believe Approach 1 is the most promising to be used in mediation settings. Approach 1 successfully resulted in a shorter IPV screen focused on identifying mediating parties reporting the highest and most concerning levels of IPV victimization. Parties reporting such levels of IPV can continue experiencing IPV long after separation even when they are offered specialized forms of mediation (Holtzworth-Munroe et al., in press). Thus, parties identified using Approach 1 likely require special safety considerations in mediation (e.g., staggered arrival and departure times, shuttle mediation where parties do not see or hear one another) and perhaps additional services (e.g., referral to an advocate for consultation and possible presence at the mediation, attorney or support persons’ presence at the mediation, referrals to shelters, law enforcement, and/or health services). These parties can be identified when they endorse at least one item from the subset of items derived from Approach 1. The current authors (Rossi, Holtzworth-Munroe, Applegate, and Beck) have utilized the items in Approach 1 to develop a short, computerized, and adaptive IPV screen for mediation, the MASIC-Short with Danger Assessment (MASIC-S with DA). This screen, described by (Applegate et al., 2020), is publicly available for practitioners once they have completed required training to conduct specific components of the tool (i.e., Danger Assessment; Campbell et al., 2009). Overall, the MASIC-S with DA assesses for level of IPV victimization (based on Approach 1 MASIC items) and risk of lethality from IPV (based on items from the Danger Assessment, a validated tool) as well as provides recommendations for accommodations in mediation to address elevated IPV victimization and risk of lethality from IPV. Currently, we are conducting additional tests of its psychometric properties.

Limitations and Future Directions

Despite the strengths of this study to advance IPV screening within family mediation, there were also several limitations. One was that information about not reaching agreement and use of special accommodations was only available for one center. Thus, the outcome data used to select items in Approaches 2 and 3 may not generalize to other mediation centers. Future studies should incorporate outcome data from several mediation centers to explore possible differences across centers or to identify a final subset of items based on a more representative sample. Despite the potential inability to generalize the outcome data analyzed under Approaches 2 and 3, this study utilized IPV data from four mediation centers across the United States and Australia, increasing diversity and generalizability of the full sample. Although we included data from four centers in the full sample, we were unable to examine racial and ethnic differences across MASIC items as we did not have such information from three of the four centers. We encourage researchers to examine whether IPV screening questions differ for differing racial/ethnic groups as certain IPV behaviors have been associated with specific races and ethnicities. For example, there is evidence that Latina women experience more sexual violence and injury-related physical violence compared with White or African American women (Clark et al., 2016). Thus, items inquiring about these IPV behaviors may be more relevant for screening Latina women than White or African American women. It is possible that our data included mostly White participants and our final subsets of IPV items may not have reflected cultural differences. Though, we believe at least one of our mediation centers, in Washington, D.C., contributed to greater racial and ethnic diversity in the sample, as a later study we conducted there had a large proportion of study participants who self-identified as Black/African American (Holtzworth-Munroe et al., 2021). Additionally, as discussed above, we were unable to examine same-sex couples and suggest that future studies examine possible variation in item information across populations of differing sexual orientations. Future studies should also further explore how IPV screening questions differ by gender in varying populations.

Finally, although the findings provide evidence for the validity of the final subsets of items derived under the three approaches, the identified subsets of items were not readministered as separate IPV measures. It is uncertain whether parties would have responded to the final subsets of items in the same way if the eliminated MASIC items had not been asked. Additionally, protective orders and IPV-related criminal charges issued in only 86 of 92 counties in one center’s state were coded. The sum of IPV-related criminal charges and protective orders used in this study is likely an underestimate of IPV behavior, particularly as a significant number of IPV behaviors are not reported to the police and do not appear on criminal records (Tjaden & Thoennes, 2000). There may also be a discrepancy in the rate at which males and females receive IPV-related charges and protective orders/no contact orders as a result of gender bias by law enforcement (U.S. Department of Justice, 2015). Additionally, males may be less likely than females to contact law enforcement for sustained IPV due to possible stigma or a reduced need to call the police given lack of IPV-related fear or injury (Cook, 2009). All of these explanations may provide some insight as to why the correlation between reported IPV victimization and all IPV-related criminal charges and protective orders/no contact orders against the reported perpetrating party was statistically significant for males but not females. Thus, future studies should further examine the validity of the identified subsets of MASIC items by readministering them to new samples and using more comprehensive measures of IPV.

Conclusion

This study examined the extent to which multiple IRT approaches generated a more reliable, efficient, and clinically useful IPV screening tool for mediation. It suggests that IRT may be successfully used to construct IPV screening measures for mediation that are shorter, have greater clinical utility, and have initial evidence for their validity. Findings advance understanding of how IRT methodology enhances the precision of IPV screening in mediation beyond the current limits and provide a vision for the future of IPV screening. Based on current findings and in weighing the advantages and disadvantages of all three approaches examined, we recommend that practitioners use MASIC items from Approach 1 in their practice. However, we are conducting additional psychometric testing of an IPV screen for mediation with these items. Overall, challenges to be addressed to further advance IPV screening in mediation include determining the level of IPV at which mediation (or specific versions of mediation) no longer is safe, and identifying factors that reliably indicate unsafe, inappropriate use of mediation by parties reporting IPV. Additionally, the field must continue to grapple with the inherent tension that exists between creating a measure that is feasible and practical yet provides sufficient information for practitioners to make clinically effective decisions. Continued advancement in IPV screening methods within the mediation context is critical for promoting the safety and interests of IPV victims and their children.

Footnotes

Acknowledgements

We thank the mediation staff at the Indiana University Maurer School of Law Viola J. Taliaferro Family and Children Mediation Clinic, the Multi-Door Dispute Resolution Division of the Superior Court in Washington, D.C., Relationships Australia, South Australia, and the Anglicare Parramatta Family Relationships Centre for providing us with data or allowing us to collect data used in this study. A special thanks to Jeannie Adams, Director of Multi-Door, for her support and contributions to this research.

Author Note

This study is supported in part by the American Psychology-Law Society Minority Affairs Diversity Research Award and by the VA Office of Academic Affiliations and Health Services Research and Development Service Research funds (HSR&D RCS 00-001). The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States Government.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Fernanda S. Rossi

Notes

References

Ballard

R. H.

Holtzworth-Munroe

Applegate

A. G.

Beck

C. J. A.

(2011). Detecting intimate partner violence in family and divorce mediation: A randomized trial of intimate partner violence screening. Psychology, Public Policy, and Law, 17(2), 241-263. https://doi.org/10.1037/a0022616

Beck

C. J. A.

Menke

J. M.

Figueredo

A. J.

(2013). Validation of a measure of intimate partner abuse (Relationship Behavior Rating Scale–Revised) using item response theory analysis. Journal of Divorce & Remarriage, 54(1), 58-77. https://doi.org/10.1080/10502556.2012.743830

Beck

C. J. A.

Walsh

M. E.

Weston

(2009). Analysis of mediation agreements of families reporting specific types of intimate partner abuse. Family Court Review, 47(3), 401-415. https://doi.org/10.1111/j.1744-1617.2009.01264.x

Breiding

M. J.

Basile

K. C.

Smith

S. G.

Black

M. C.

Mahendra

(2015). Intimate partner violence surveillance uniform definitions and recommended data elements. Centers for Disease Control and Prevention National Center for Injury Prevention and control. https://www.cdc.gov/violenceprevention/pdf/intimatepartnerviolence.pdf

Campbell

J. C.

Webster

D. W.

Glass

(2009). The danger assessment: Validation of a lethality risk assessment instrument for intimate partner femicide. Journal of Interpersonal Violence, 24(4), 653-674. https://doi.org/10.1177/0886260508317180

Clark

H. M.

Galano

M. M.

Grogan-Kaylor

A. C.

Montalvo-Liendo

Graham-Bermann

S. A.

(2016). Ethnoracial variation in women’s exposure to intimate partner violence. Journal of Interpersonal Violence, 31(3), 531-552. https://doi.org/10.1177/0886260514555871

Cook

P. W

. (2009). Abused men: The hidden side of domestic violence (2nd ed.). Praeger Press.

Edwards

M. C.

(2009). An introduction to item response theory using the need for cognition scale. Social and Personality Psychology Compass, 3(4), 507-529. https://doi.org/10.1111/j.1751-9004.2009.00194.x

Ellis

(2008). Divorce and the family court: What can be done about domestic violence? Family Court Review, 46(3), 531-536. https://doi.org/10.1111/j.1744-1617.2008.00218.x

10.

Ellis

Stuckless

(2006). Domestic violence, dove, and divorce mediation. Family Court Review, 44(4), 658-671. https://doi.org/10.1111/j.1744-1617.2006.00117.x

11.

Emery

R. E.

(2011). Renegotiating family relationships: Divorce, child custody and mediation (2nd ed.). Guilford Press.

12.

Febres

(2015). The perpetration of adulthood animal abuse and intimate partner violence in men and women arrested for domestic violence [Doctoral dissertation]. University of Tennessee. http://trace.tennessee.edu/utk_graddiss/3578

13.

Folberg

Milne

A. L.

Salem

(2004). Divorce and family mediation: Models, techniques, and applications. Guilford Press.

14.

Holtzworth-Munroe

(2011). Controversies in divorce mediation and intimate partner violence: A focus on the children. Aggression and Violent Behavior, 16(4), 319-324. https://doi.org/10.1016/j.avb.2011.04.009

15.

Holtzworth-Munroe

Applegate

A. G.

Beck

C. J.

Rossi

F. S.

Adams

J. M.

Jiang

L. J.

Tomlinson

C. S.

Hale

D. F.

(in press). Intimate partner violence (IPV) and family dispute resolution: One-year follow up findings from a randomized controlled trial comparing shuttle mediation, videoconferencing mediation, and traditional litigation. Psychology, Public Policy, and Law.

16.

Holtzworth-Munroe

Beck

C. J.

Applegate

A. G.

Adams

J. M.

Rossi

F. S.

Jiang

L. J.

Tomlinson

C. S.

Hale

D. F.

(2021). Intimate partner violence (IPV) and family dispute resolution: A randomized controlled trial comparing shuttle mediation, videoconferencing mediation, and litigation. Psychology, Public Policy, and Law, 27(1), 45-64. https://doi.org/10.1037/law0000278

17.

Holtzworth-Munroe

Beck

C. J. A.

Applegate

A. G.

(2010). The Mediator’s Assessment of Safety Issues and Concerns (MASIC): A screening interview for intimate partner violence and abuse available in the public domain. Family Court Review, 48(4), 646-662. https://doi.org/10.1111/j.1744-1617.2010.001339.x

18.

Matusiewicz

A. K.

Bornovalova

M. A.

Limbrunner

H. M.

(2014). Borderline personality disorder. In Grossman

Walfish

(Eds.), Translating psychological research into practice (pp. 153–158). Springer Publishing Company.

19.

Maydeu-Olivares

(2013). Goodness-of-fit assessment of item response theory models. Measurement: Interdisciplinary Research and Perspectives, 11(3), 71-101. https://doi.org/10.1080/15366367.2013.831680

20.

Mechanic

M. B.

(2004). Stalking vicitmization and intimate partner abuse. In Jackson

T. L.

VandeCreek

(Eds.), Innovations in clinical practice: Focus on violence treatment & prevention (pp. 21–34). Professional Resource Press/Professional Resource Exchange.

21.

Milne

A. L.

(2004). Mediation and domestic abuse. In Folberg

Milne

A. L.

Salem

(Eds.), Divorce and family mediation (pp. 304-335). Guilford Press.

22.

Muthén

L. K.

Muthén

B. O.

(1998-2017). Mplus user’s guide (8th ed.). Muthén & Muthén.

23.

O’Hara

K. L.

Perkins

A. B.

Tehee

Beck

C. J.

(2018). Measurement invariance across sexes in intimate partner abuse research. Psychology of Violence, 8(5), 560-569. https://doi.org/10.1037/vio0000153

24.

Petersen

I. T.

Bates

J. E.

Dodge

K. A.

Lansford

J. E.

Pettit

G. S.

(2016). Identifying an efficient set of items sensitive to clinical-range externalizing problems in children. Psychological Assessment, 28(5), 598-612. https://doi.org/10.1037/pas0000185

25.

Pokman

Rossi

F. S.

Holtzworth-Munroe

Applegate

A. G.

Beck

C. J.

D’Onofrio

B. M.

(2014). Mediator’s Assessment of Safety Issues and Concerns (MASIC): Reliability and validity of a new intimate partner violence screen. Assessment, 21(5), 529-542. https://doi.org/10.1177/1073191114528372

26.

Rollè

Giardina

Caldarera

A. M.

Gerino

Brustia

(2018). When intimate partner violence meets same sex couples: A review of same sex intimate partner violence. Frontiers in Psychology, 9, Article 1506. https://doi.org/10.3389/fpsyg.2018.01506

27.

Rossi

F. S.

Holtzworth-Munroe

Applegate

A. G.

(2015). Does level of intimate partner violence and abuse predict content of family mediation agreements? Family Court Review, 53(1), 134-161. https://doi.org/10.1111/fcre.12135

28.

Rossi

F. S.

Holtzworth-Munroe

Applegate

A. G.

Beck

C. J. A.

Adams

J. M.

Hale

D. F.

(2015a). Detection of intimate partner violence and recommendation for joint family mediation: A randomized controlled trial of two screening measures. Psychology, Public Policy, and Law, 21(3), 239-251. https://doi.org/10.1037/law0000043

29.

Rossi

F. S.

Holtzworth-Munroe

Applegate

A. G.

Beck

C. J. A.

Adams

J. M.

Hale

D. F.

(2017). Shuttle and online mediation: A review of available research and implications for separating couples reporting intimate partner violence or abuse. Family Court Review, 55(3), 390-403. https://doi.org/10.1111/fcre.12293

30.

Tjaden

Thoennes

(2000). Full report of the prevalence, incidence, and consequences of violence against women. U.S. Department of Justice, Office of Justice Programs. National Institute of Justice. https://www.ncjrs.gov/pdffiles1/nij/183781.pdf

31.

U.S. Department of Justice. (2015). Identifying and preventing gender bias in law enforcement response to sexual assault and domestic violence. U.S. Department of Justice, Office of Justice Programs. National Institute of Justice. https://www.justice.gov/crt/file/799316/download

32.

Ver Steegh

Dalton

. (2008). Report from the wingspread conference on domestic violence and family courts. Family Court Review, 46(3), 454-475. https://doi.org/10.1111/j.1744-1617.2008.00214.x