Profile of partner aggressors as a function of risk of recidivism

Abstract

Partner aggressors present psychopathological, criminal, and sociodemographic characteristics that have been used for classification in typologies. The goal of the present work was to identify profile of aggressors as a function of the risk of recidivism, and assess whether there correspondence with type of offenders proposed by Holtzworth–Munroe and Stuart. The sample was made up of 90 men condemned for partner violence, of whom 50 were serving a prison sentence, and 40 mandatory community intervention/programs. The risk of recidivism was assessed with the SARA - Spousal Assault Risk Assessment Guide, completed with information obtained from prison records, clinical interviews for the assessment of personality disorders, and self-reports. The results reveal three profile of aggressors according to their risk of recidivism, related to the above-mentioned classification: high-risk aggressors coincide with the Dysphoric/Borderline (DB) type, medium-risk aggressors with the low-level antisocial type (LLA), and the low-risk group with the type of aggressors family only (FO). The implications are discussed in the penitentiary intervention of these results.

Resumen

Los agresores contra la pareja presentan características psicopatológicas, delictivas y sociodemográficas que han sido utilizadas para su clasificación en tipologías. El objetivo del presente trabajo ha sido identificar el perfil de estos agresores según el riesgo de reincidencia, y valorar si existe coincidencia con la tipología de agresores propuesta por Holtzworth–Munroe y Stuart. La muestra fue de 90 hombres condenados por violencia contra la pareja, de los que 50 cumplían pena de prisión y 40 una medida alternativa: intervención/programa. El riesgo de reincidencia se evaluó con la guía SARA-Spousal Assault Risk Assessment Guide, cumplimentada con la información obtenida de los expedientes penitenciarios, entrevistas clínicas para la evaluación de trastornos de personalidad y autoinformes. Los resultados muestran tres perfiles de agresores según su riesgo de reincidencia que se relacionan con la mencionada clasificación, ya que los agresores de alto riesgo coinciden con el tipo disfóricos/borderline (DB), los agresores de riesgo medio con el tipo antisocial de bajo nivel (LLA) y el grupo de menor riesgo con el tipo de agresores limitados al ámbito familiar (FO). Se discuten las implicaciones en la intervención penitenciaria de estos resultados.

Intimate partner violence (IPV) is one of the most common types of aggression suffered by women, and worldwide, 38% of homicides of women and 42% of physical and/or sexual aggressions were perpetrated by their partners or ex-partners (World Health Organization, 2013). The figures for this type of violence in Europe are also alarming, as 22% of the women have suffered physical and/or sexual violence, 43% psychological violence, and 55% sexual harassment (European Union Agency of Fundamental Rights, 2014). Knowing the risk factors present in each aggressor facilitates decision making by all the professionals working to prevent new violent acts, for example, members of the police force, judges, forensic teams, penitentiary professionals, and therapists, among others, who must adapt their interventions to the profile of the aggressor (Andrés-Pueyo & Echeburúa, 2010). To facilitate this decision making, various guidelines have been developed to rate the risk, depending on the type of professional who uses it, or the type of violence they need to appraise (Belfrage et al., 2012). Accordingly, one of the most widely used guides worldwide is the Spousal Assault Risk Assessment Guide, or SARA (Kropp, Hart, Webster y Eaves, 1999), considered one of the most complete tools due to the number of items it appraises, the clinical-actuarial method employed, and the contexts in which it can be applied (Storey, Kropp, Hart, Belfrage, & Strand, 2014).

The characteristics considered as IPV risk factors associated with the aggressor have allowed the establishment of diverse typologies, underscoring the one by Holtzworth-Munroe and Stuart (1994). This classification is based on psychopathological and criminal characteristics, and type of violence employed, as a function of its extension and severity, defining three types of aggressors: (a) Family only (FO): aggressors of low-intensity and severity, presenting scarce psychopathology, without a personality disorder, although with passive-aggressive and obsessive traits, abusive alcohol consumption, low level of depression, and moderate level of anger, and no criminal antecedents; (b) Dysphoric/Borderline aggressors (DB): they employ violence more frequently and intensely than the former group, both against family members and unfamiliar, they have penal antecedents, present higher incidence of borderline personality disorder with emotional instability, they are dependent, jealous and controlling, generating cyclical violence with phases of aggression and repentance, also presenting moderate alcohol consumption and high levels of depression and anger; and (c) Generally violent/antisocial (GVA): this is the type of aggressor who generates more severe violence, with the greatest number of penal and criminal antecedents, a typical anti-social profile with an instrumental use of violence, scarce empathy, low levels of depression and moderate levels of anger, and abusive alcohol consumption. In general, this typology has been replicated in different studies (Amor, Echeburúa, &Loinaz, 2009), and in others like that carried out by Holtzworth-Munroe, Meehan, Herron, Rehman, and Stuart (2000), who expanded it with a fourth subtype, called low-level antisocial (LLA), with characteristics of the FO and GVA type; that is, this type presents antisocial characteristics, although they do not perpetrate such intense and generalized violence as the antisocial subtype.

Diverse longitudinal studies assess the recidivism of previously differentiated aggressors according to different typologies. One of them identified the GVA aggressor as having the highest recidivism, followed by the DB type, and lastly, the FO type of aggressor (Thijssen & de Ruiter, 2011). Other studies identified two types of aggressors, Antisocial/pathological and Non-pathological (Loinaz, 2014), or three types Non-pathological, Antisocial/violent and Disturbed batterers (Cunha & Abrunhosa, 2013).

The present study has the goal of identifying the different profiles of aggressors based on the risk of recidivism, using the factors from the SARA guide, and to study their relationship with the typology of aggressors proposed by Holtzworth-Munroe and Stuart (1994), confirming the above-mentioned studies.

Method

Participants

The sample is made up of 90 men condemned for IPV, serving two different types of sentence: 50 are sentenced to prison in the Penitentiary Center of Alicante-II (Spain), and 40 comply with a Mandatory Community Intervention/Program in the Management Service of Penalties and Alternative Measures of Murcia (Spain). The main sociodemographic characteristics of the sample are: mean age 35.6 years (SD = 8.62), 46.7% (n = 42) are Spanish and 20% (n = 18) are Latin American, 70% (n = 66) have only primary studies, complete or incomplete, and 54.4% (n = 49) state that their economic situation is insufficient.

Instruments

Spousal Assault Risk Assessment Guide, SARA (Kropp et al., 1999). We used the SARA guide, in its adaptation for Spanish population carried out by Andrés-Pueyo and López (2005). This is a semi-structured instrument that estimates the risk of recidivism for IPV by means of the study of 20 risk factors, grouped into four blocks: (I) Criminal history, (II) Psychosocial adjustment, (III) Spousal assault history, and lastly, (IV) Index offense factors. These 20 risk factors are described in Table 1. Each factor is rated using the information collected from the judicial, penitentiary, and social file of each participant, together with the diagnostic interview conducted by specialized professionals. Each factor can receive a score of 0, 1, or 2, depending on its absence, partial/discontinuous presence, or permanent presence, respectively. The assessment was conducted jointly by two specialized psychologists, obtaining an inter-evaluator kappa value of .82.

Table 1.

Risk factors in the Spouse Assault Risk Assessment (SARA) and frequency of presence ratings (N = 90).

SARA	Frequency/%)
	No/Absent (0)	Possibly/Partially present (1)	Yes/Present (2)
I. Criminal history
1. Past assault of family members	87.8	0	12.2
2. Past assault of strangers or acquaintances	45.6	8.8	45.6
3. Past violation of conditional release	84.5	1.1	14.4
II. Psychosocial adjustment
4. Recent partner relationship problems	1.1	0	98.9
5. Recent employment problems	41.1	10	48.9
6. Victim and/or witness of childhood violence	51.1	12.2	36.7
7. Recent Drug Consumption/Abuse	38.8	5.6	55.6
8. Suicidal and/or homicidal ideas/attempts	66.7	0	33.3
9. Psychotic and/or manic symptoms	72.2	4.4	23.4
10.- Personality disorder with anger, impulsivity, or behavioral instability	33.3		66.7
III. Spousal assault history
11. Past physical assault	54.4	1.1	44.4
12. Past sexual assault and/or jealousy	37.8	30	32.2
13. Use of weapons and/or threats of death	47.8	30	22.2
14. Increase/severity of aggressions	4.4	48.9	46.7
15. Noncompliance of restraining orders	28.9	28.9	42.2
16. Minimization or denial of IPV	12.2	48.9	38.9
17. Attitudes justifying IPV	32.2	41.1	26.7
IV. Index offense factors
18. Severe and/or sexual assault	95.6	0	4.4
19. Use of weapons and/or threats of death	48.9	30	21.1
20. Noncompliance restraining orders	27.8	5.6	66.7

Note. IPV: Intimate Partner Violence.

The information required to complete the SARA guide and classify the aggressors according to Holtzworth-Munroe and Stuart's (1994) typology was obtained from the review of expert technical case files (penal, penitentiary, and social), structured interviews and self-reports, described below:

• a)

  Review of the expert technical case files (criminal, penitentiary, and social). These documents contain the information collected by professionals such as police officers, psychologists, social workers, health professionals, judicial agents, penitentiary officials, etc. These three case files contain the historical data of the different sentences that the prisoner had served, and his evolution in prison.

• b)

  Structured Clinical Interview for DSM-IV Axis-II Personality Disorders (SCID-II; in the authors’ Spanish version (First, Gibbon, Spitzer, Williams, & Smith, 1999), to assess the presence of antisocial, borderline, and passive-aggressive personality disorders. The reliability studies reveal a kappa index between .78 and .91 (Lobbestael, Leurgans, & Amtz, 2010) and, in our case, an inter-interviewer kappa of .81, and an inter-encoder kappa of .73.

• c)

  Self-reports:

  • •

    Dominating and Jealous Tactics Scale of Kasian and Painter (1992), in its version adapted for Spanish population of González (2008). This scale describes 7 tactics of dominance or control that is exerted on the partner, and 4 tactics of jealous feelings and behaviors in the relationship. According to the author of the adaptation employed, this scale presents Cronbach alphas of .67 and .73 for dominance and jealousy, respectively. In the present study we have obtained a Cronbach's alpha of .71.

  • •

    Inventario de Pensamientos Distorsionados sobre la Mujer y la Violencia (IPDMV [Inventory of Distorted Thoughts about Women and Violence]; Echeburúa & Fernández-Montalvo, 1998). We used the version by Ferrer, Bosch, Ramis, Torres, and Navarro (2006), presenting four factors, two related to cognitive distortions about women and two concerning violence, with internal consistency (Cronbach alpha) of .84.

  • •

    Conflict Tactics Scales-2 (CTS-2) of Straus, Hamby, Boney-McCoy, and Sugarman (1996), in the Spanish version of Loinaz (2009). This scale yields five factors: Negotiation, Psychological Violence, Physical Violence, Sexual Violence, and Injuries. In the present study, we only used the subscale of Psychological Violence, made up of 8 items, which provides two measures, Minor and Severe Psychological Violence. The other subscales are discarded to obtain the expert technical files. It presents adequate internal consistency with a Cronbach alpha of .82 (Loinaz, Echeburúa, Ortiz-Tallo, & Amor, 2012).

  • •

    Revised NEO Personality Inventory, NEO–PI–R (Costa & McCrae, 1992) in the Spanish adaptation (Inventario de Personalidad Revisado, NEO PI-R; Arribas, 1999). This is a non-pathological Personality Questionnaire, rated on five domains: Neuroticism, Extroversion, Openness to Experience, Agreeableness, and Responsibility. Each domain is made up of six facets; in this investigation, we used two of them, Hostility and Depression, included within the Neuroticism domain. In the adaptation used, the inventory presents an internal consistency ranging between .82 (Openness) and .90 (Neuroticism).

Procedure

This is a descriptive, relational, cross-sectional study, which received the corresponding authorizations of the Secretary of State of Penitentiary Institutions and Penitentiary Center (Spain). Inclusion criteria for sample selection were to be serving a sentence for partner violence and voluntary participation in the study, and as exclusion criterion, the incapacity to participate in a diagnostic interview. The assessment was carried out jointly by two psychologists from the penitentiary center, specialized in expertise of aggressors.

This investigation was authorized by the bioethical committee of the University of Murcia (Spain), meeting the ethical criteria of psychology and code of conduct proposed by the American Psychological Association (2002, 2010): beneficence and nonmaleficence; professional responsibility and confidentiality; personal integrity, no deception; justice and equity in the benefits from the contributions; and respect for the person's dignity, not excluding any collective of persons from the benefits. The participants received prior information about the study and provided their informed consent to participate. The results will have implications in preventive measures against IPV, the assessment, risk management and treatment of aggressors, providing benefits for society in general.

Data analysis

The procedure used to classify the participants into homogeneous groups was hierarchical cluster analysis with Ward's method. This method is recommended for qualitative variables (Campbell, Greeson, Bybee, & Raja, 2008), and is one of the most commonly used to obtain typologies of aggressors (Huss & Ralston, 2008). Nonparametric statistics were used to identify significant group differences, as the assumptions of normality and variance homogeneity were not met. Specifically, we used the Kruskal-Wallis test with exact significance and the Mann-Whitney U for paired comparisons, in this case with the Bonferroni correction for the level of significance (α = .05/3 = .017). For categorical variables, we used the chi-squared test, measuring the effect size by means of phi and Cramer's V.

Results

Cluster analysis

Cluster analysis was carried out with the values of the factors the SARA guide, as shown in Table 1. As the best option, three homogeneous groups were obtained, Group 1 with 43 participants (47.8%), Group 2 with 20 (22.2%), and Group 3 with 27 participants (30%). According to the mean value of the total SARA score, Group 1 obtained the highest value of the three (M = 26.56, SD = 3.83), Group 2, an intermediate score (M = 14.05, SD = 3.50), and Group 3, the lowest score (M = 9.59, SD = 3.36). According to these results, the groups could be denominated high risk (HR), medium risk (MR), and low risk (LR).

The three groups presented significant differences in the global SARA score, according to the Kruskal-Wallis test, χ²₍₂₎ = 70.30, p < .001. Multiple comparisons also indicate differences, according to the Mann-Whitney U (HR/MR: z = - 6.32, p < .001; HR/LR: z = - 7.01, p < .001; MR/LR: z = - 3.76; p < .001).

Figure 1 shows the percentage of subjects with a score of YES/Present (2), in each one of the SARA risk factors. As reflected in Figure 1, there are significant group differences in 17 risk factors, and only Factors (4) Recent partner relationship problems, χ²(2, N = 90) = 3.53, ns, ϕ = .19; (16) Minimization or denial of IPV, χ²(2, N = 90) = 1.99, ns, ϕ = .11; and (18) Severe and/or sexual assault, χ²(2, N = 90) = 4.57, ns, ϕ = .22, presented no differences.

Figure 1.

Porcentages with Yes/Present (2) Score.

Note. FR: Factor Risk; HR: High Risk; MR: Medium Risk; LR: Low Risk; *p < .05; **p < .01; ***p < .001.

In the diverse pairwise group comparisons, the greatest number of factors with significant differences occurred in the comparison of Groups HR and LR: in the same 17 above-mentioned factors. In contrast, between HR and MR, the significant differences decreased to 14 factors, to which must be added the 3 factors with no group differences: (1) Past assault of family members, χ²(2, N = 63) = 3.15, ns, ϕ = .22; (10) Personality disorder, χ²(2, N = 63) = 1.46, ns, ϕ = .15; and (12) Past sexual assault and/or jealousy, χ²(2, N = 63) = 3.32, ns, ϕ = .23. There were only 5 risk factors with significant differences between Groups LR and MR: (2) Past assault of strangers or acquaintances, χ²(2, N = 47) = 17.83, p < .001, ϕ = .61; (6) Victim and/or witness of childhood violence, χ²(2, N = 47) = 12.65, p < .01, ϕ = .52; (10) Personality disorder, χ²(2, N = 47) = 47.01, p < .001, ϕ = 1.00; (11) Past physical assault, χ²(2, N = 47) = 7.73, p < .05, ϕ = .41; and (14) Increase/severity of aggressions, χ ²(2, N = 47) = 7.23, p < .05, ϕ = .41.

There were significant group differences as a function of the origin of the participants, Penitentiary Center and Mandatory Community, χ²(2, N = 90) = 52.81, p < .001, ϕ = .76. In the HR group, 95.3% (n = 41) were from prison, and in the other two groups, there was a predominance of aggressors sentenced to mandatory community, in Group MR, 80% (n = 16), and in Group LR, 81.5% (n = 22). In the paired group comparisons, there were significant differences between HR and MR, χ²(1, N = 63) = 37.97, p < .001, ϕ = .78, and between HR and LR, χ²(1, N = 70) = 43.45, p < .001, ϕ = .79. There were no significant differences between Groups MR and LR, χ²(1, N = 47) = .02, ns, ϕ = .02.

Differences in variables with typological value

Table 2 shows the variables with typological value according to the classification of Holtzworth-Munroe and Stuart (1994), with the exception of the variables Drug use and extension of violence (Relatives/Non-relatives), as they are SARA risk factors included in the cluster analysis. There were significant group differences in all of them, except for the variable Abusive alcohol consumption, which was present in almost all the sample.

Table 2.

Defining variables of aggressor typologies.

	High Riskn = 43		Medium Riskn = 20		Low Riskn = 27
	f	%	f	%	f	%	χ^2	Φ
Personality disorder
No Disorder	2	4.7	0	0	27	100	97.68***	.73
Antisocial	11	25.6	14	70	0	0
Borderline	13	30.2	1	5	0	0
Antisocial/borderline	17	39.5	5	25	0	0
Penal antecedents	37	86	16	80	9	33.3	22.98***	.51
Psychopathology
Alcohol abuse	42	97.7	20	100	24	88.9	4.21	.21
NEO–PI-R:
*Hostility (T ≥ 60)	21	48.8	12	60	5	18.5	9.58**	.32
*Depression (T ≥ 60)	29	67.4	10	50	9	33.3	7.86*	.29

Note. NEO–PI-R = Revised NEO Personality Inventory. f = frequency

*p < .05. **p < .01. ***p < .001.

Of these variables, we especially highlight the distribution of the personality disorders assessed (borderline, antisocial, and passive-aggressive), which presented significant differences in all three groups, χ²(2, N = 90) = 97.68, p < .001, ϕ = .73. The pairwise group comparisons also revealed significant differences: HR/MR: χ²(2, N =63) = 12.45, p < .01, ϕ = .45; HR/LR: χ²(2, N = 70) = 62.14, p < .001, ϕ = .94; and MR/LR: χ²(2, N = 43) = 47.01, p < .001, ϕ = 1.00.

As can be seen in the distribution of the disorders, Group HR concentrates most of the participants with a diagnosis of borderline disorder, either as an exclusive diagnosis (30.2%) or in combination with antisocial disorder (39.5%). Antisocial disorder without the association of another disorder is noteworthy in Group MR, with 70% of the participants. Lastly, Group LR includes the participants without a personality disorder (100%). We did not include passive-aggressive disorder in the statistical analyses due to its low incidence in the sample (n = 2).

In the rest variables of this block, in the pairwise group comparisons, there were significant differences between Groups HR and LR in all the variables and, in contrast, there were no differences between Groups HR and MR. There were significant differences between Groups MR and LR in Criminal antecedents, χ²(1, N = 43) = 10.05, p < .01, ϕ = .46, and Hostility, χ ²(1, N = 47) = 8.56, p < .01, ϕ = .43.

Differences in specific IPV variables

The specific IPV variables considered in this study are shown in Table 3. The Kruskal-Wallis test identified significant differences in the variable Minor psychological aggression, χ²(2) = 8.15, p < .05, and Factor 3, Attitudes favoring IPV, χ²(2) = 8.28, p < .05. The analysis to detect pairwise group differences using the Mann-Whitney U and Bonferroni's correction yielded significant differences in the variable Minor psychological aggression between HR and LR, z = -2.53, p = .011, and between MR and LR, z = -2.42, p = .015, and in Factor 3 between HR and MR, z = -2.81, p = .005.

Table 3.

Specific IPV variables.

	High Risk		Medium Risk		Low Risk
	M	SD	M	SD	M	SD	χ^2(2)^a
CTS-2
Minor psychological aggression	12.58	7.20	12.80	6.72	8.11	6.22	8.15*
Severe psychological aggression	3.09	3.90	2.90	3.14	1.52	2.60	4.88
ATTITUDES
F 1: Acceptance sexist stereotypes	1.46	.45	1.67	.58	1.41	.43	3.36
F 2: Blaming the female victim	2.52	.52	2.50	.56	2.25	.55	5.37
F 3: Acceptance of violence	1.50	.42	1.98	.65	1.65	.47	8.28*
F 4: Minimizing violence	2.33	.63	2.53	.61	2.33	.55	1.51
JEALOUSY	7.86	3.70	7.45	3.10	6.70	3.62	4.88
CONTROL	12.07	4.28	11.50	4.12	9.85	3.07	4.77

Note. IPV: Intimate Partner Violence; CTS-2: Conflict Tactics Scales-2; (a): Kruskal-Wallis test; *p < .05.

Discussion and conclusions

The classification of men sentenced for IPV into homogeneous groups, using the factors from the SARA guide, has allowed us to obtain three groups with different levels of risk of recidivism: high, medium, and low, coinciding with the proposals of the last review of the guide (Kropp & Gibas, 2010). These groups are associated with certain psychopathological and criminal characteristics, revealing a high parallelism with the typology of aggressors proposed by Holtzworth-Munroe and Stuart (1994). For instance, the group with the highest risk (HR) presents characteristics of the Disphoric/Borderline type of aggressor, the MR group is similar to the low-level antisocial type (LLA), and lastly, the group with the lowest risk of recidivism (LR) presents similar characteristics to the type of aggressor family only (FO) (Holtzworth-Munroe et al., 2000). Also agree with the proposal Cunga and Abrunhosa (2013), where the HR group would be the Disturbed batterers type, the MR group would be the antisocial/violent type, and finally the LR Non-pathological type.

According to these characteristics, the HR group is made up of violent men, mainly against unfamiliar, although they may also assault family members. They present penal antecedents, violation of conditional release, and noncompliance of restraining orders, showing that the judicial measures do not achieve the desired behavioral control. The psychosocial adjustment of these aggressors is deficient; they have economic and employment problems, and consume drugs and alcohol. One of their main characteristics is the presence of psychopathological alterations; borderline personality disorder is notable in the majority of the aggressors of this group, as a single diagnosis, or comorbidly with antisocial disorder. They present a long history of IPV, in which violence shows an upward trend in frequency and severity, leading to the use of weapons and death threats, and more than half of these men display chauvinistic attitudes justifying violence. These characteristics coincide with those that define borderline-type aggressors, except for the high use of violence towards unfamiliar and the presence of antisocial disorder comorbidly with borderline disorder. These characteristics justify these men's prolonged history of IPV, in which their high emotional dependence, jealousy, and emotional instability, coupled with their aggressive responses, generate recurrent IPV (Cortés-Ayala et al., 2014).

Most of the aggressors of the MR group present antisocial disorder with no comorbidity with other disorders, explaining the large number of penal antecedents they present, their use of violence and their social maladjustment, although it is not so generalized as the HR group (Gibbons, Collins, & Reid, 2011). In their history of IPV, they rarely breach restraining orders, possibly because none were imposed because this was their first aggression towards that partner, a fact that does not imply that there were no violent episodes with previous partners. These aggressors resemble the low-level antisocial aggressor type more than the generally violent/antisocial type and, although they coincide in the antisocial disorder, this group's level of aggressiveness and harm to the victim is lower. Typology GVA presents more similarity with the profile of Group HR, with borderline personality/antisocial characteristics (Holtzworth-Munroe & Stuart, 1994; Thijssen & de Ruiter, 2011).

According to the variables considered to have typological value and specific to VCP, there were only significant differences between Groups HR and MR in borderline/antisocial personality disorder and drug use. In contrast, there were differences in 14 of the 20 SARA risk factors. These two groups present similar characteristics to the Antisocial/Pathological group identified in the study of Loinaz (2014), in which they used exclusively variables with typological value for the classification, without specifying the type of personality disorder. The differential diagnosis of personality disorders performed by means of a diagnostic interview has allowed us to differentiate the aggressors with borderline and antisocial disorder, and comorbidity in both of them.

Lastly, the aggressors of the LR group coincide in the characteristics with the FO type of aggressor (Holtzworth-Munroe & Stuart, 1994), or with the Non-pathological type (Loinaz, 2014), are characterized by a lack of IPV antecedents or restraining orders; their social adjustment is adequate and they have no personality disorder, only presenting abusive alcohol consumption. Their violence could be explained as the expression of inadequate resolution strategies for partner conflict.

We note the scarce differences found in the three groups regarding specific IPV variables. We underline the differences in Minor psychological aggression, in Group LR compared to the other two groups, and in Factor 3, with thoughts justifying violence, between Groups HR and MR, with higher scores in Group MR, coinciding with the group with predominantly antisocial disorder (Ruiz-Hernández, García-Jiménez, Llor-Esteban, & Godoy-Fernández, 2015; Ureña, Romera, Casas, Viejo, & Ortega-Ruiz, 2015). These results imply that the main differences between these aggressors are in the way they resolve conflicts, depending on whether or not they have a personality disorder, and what type of disorder, their level of social maladjustment, and their level of stress (García-Jiménez, Godoy-Fernández, Llor-Esteban, & Ruiz-Hernández, 2014; Loinaz, Ortiz-Tallo, & Ferragut, 2012; Loinaz, Ortiz-Tallo, Sánchez, & Ferragut, 2011).

The results obtained identified an important relationship between personality disorder, aggressor characteristics, and risk of recidivism, which, in our view, is essential for decision-making at the police, judiciary, and penitentiary levels. Knowing the psychopathological characteristics, personality disorders, attitudes favoring IPV, conflict resolution strategies, and associated psychopathology is essential for classifying aggressors, appraising their risk of recidivism, and designing the psychosocial intervention, in order to improve adherence to and efficacy of the treatment (Novo, Fariña, Seijo, & Arce, 2012).

The provenance of the sample in the diverse groups presents conclusive differences: the HR group comprises mainly incarcerated aggressors whereas the other two groups (MR and LR) are made up of released aggressors who are sentenced to mandatory community intervention/programs. This relation between type of prison sentence and risk of recidivism indicates appropriate decisions by the judges and, although they do not make their decisions based on risk assessment instruments, they do take into account the risk factors described above, such as criminal antecedents, violence employed, violation of court orders, or the use of weapons.

The main limitation of this study is the sample size. It should be extended in order to verify the stability of the proposed profiles, in which the following novel aspects stand out: the comorbidity of borderline and antisocial disorders, the unclear differentiation of the generally violent profile in contrast to the low-level antisocial profile, and the absence of differences in specific variables, especially in behaviors regarding control and jealousy, and chauvinistic attitudes. These notable aspects should give rise to longitudinal studies in which to apply and assess adherence to and the efficacy of different programs of psychosocial intervention depending on the profiles obtained in recidivism.