INTRODUCTION

Schema therapy (ST) is a relatively new, integrative approach that expands traditional cognitive-behavioral treatments and concepts (^{Salicru, 2023}; ^{Young et al., 2009}). It integrates elements from different approaches, such as Gestalt therapy, attachment theory, and psychodynamic understanding (^{Koppers et al., 2021}), and uses mental imagery techniques that stimulate a greater connection with emotions (^{Fiuza & de Godoy, 2021}; ^{Keulen-de Vos et al., 2017}). ST has developed over the last two decades as a medium to long-term treatment (^{Oleśniewicz, 2021}), with a high level of evidence of effectiveness for personality disorder cases (^{Zhang et al., 2023}). It greatly emphasizes the developmental aspects of patients’ symptoms and problems (^{Fassbinder et al., 2019}; ^{Oettingen & Rajtar-Zembaty, 2022}).

Among the main concepts of the ST are early maladaptive schemas (EMSs) and schema modes (^{Young et al., 2009}). EMSs are like cognitive structures formed by memories, bodily sensations, and emotions related to the view of the self and the relationship with others (^{Bach & Farrel, 2018}; ^{Dunne et al., 2018}; ^{Keulen-de Vos et al., 2017}; ^{Westphalen & Méa, 2018}). They develop from the interaction of the genetic basis of temperament with the basic needs to be met by caregivers in childhood, also associated with the experiences of repeated and/or traumatic experiences (^{Bach & Farrel, 2018}; ^{Baldissera et al., 2021}).

The schema modes encompass several EMSs and active coping modes simultaneously (^{Clercx et al., 2021}; ^{Dunne et al., 2019}; ^{Matos et al., 2018}) and can be differentiated from EMSs by the issue of stability and constancy (^{Matos et al., 2018}). EMSs develop early, last throughout life, and tend to recur in different situations (^{Flanagan et al., 2023}; ^{Soygüt et al., 2023}). On the other hand, schema modes represent an individual’s dissociated “state” at a given moment and can be both adaptive and maladaptive (^{Matos et al., 2018}).

Schema modes generate oscillations in individuals’ emotional and cognitive states and their behavioral tendencies (^{Bach & Farrel, 2018}; ^{Damasceno et al., 2019}). Ten basic schema modes are divided into four broad categories: Innate Child Modes, Maladaptive Parent Modes, Maladaptive Coping Modes, and the Healthy Adult Mode (^{Bachrach & Arntz, 2021}; ^{Young et al., 2009}). The ten schematic modes are briefly described in Table 1.

The schema mode inventory (SMI) is used to measure the activation of these schema modes (^{Damasceno et al., 2019}). It has two versions: the original version, with 270 items, and the short version (SMI-R), derived from the original version, with 118 items. The latter is subdivided into 14 subscales, and its items are evaluated through a Likert scale of six points, ranging from “never or almost never” to “always” (^{Barazandeh et al., 2018}; ^{Lobbestael et al., 2010}). The reduced version of the SMI (SMI-R) was proposed in 2010 by Lobbestael et al. (^{Matos et al., 2020}). In its formulation, four modes were added to the ten initially proposed by Young et al. in 2009: Undisciplined Child, Enraged Child, Detached Self-Soother, and Bully-and-Attack (^{Matos et al., 2020}), as described in Table 1.

Through psychoeducation, working with schema modes helps us understand the origin of the modes, their functioning and their triggering situations (^{Dunne et al., 2018}; ^{Maffini et al., 2020}). The objective of the ST is to reduce the intensity and frequency of oscillation of schema modes, allowing the most adaptive ones to predominate in the patient’s life (^{Soygüt et al., 2023}).

Interventions that advocate working with schema modes have been recommended for more challenging personality disorders, such as antisocial, narcissistic, borderline, or paranoid cases (^{Oleśniewicz, 2021}). These clinical conditions present high recurrence levels, which justifies its indication (^{Bernstein et al., 2023}).

Understanding the schema modes can be relevant for understanding criminal and violent behaviors (^{Bernstein & Navot, 2023}) since, according to the ST, acts of violence are defined as emotional states of anger and hypercompensatory strategies involving threat, intimidation, and aggression (^{Vos et al., 2023}). In these cases, patients easily understand the work from the perspective of schema modes. They use them intuitively as a self-intervention, which allows the therapist to monitor their patients better and choose the most appropriate therapeutic tools (^{Oleśniewicz, 2021}).

Challenging or criminal behaviors originate from a sequence of developments of maladaptive schema modes (^{Bernstein et al., 2023}; ^{de Klerk et al., 2022}). Such a sequence typically begins with schema modes related to vulnerability, abandonment, and loneliness (^{Vos et al., 2023}), intensifies from modes involving impulsivity, anger, and substance abuse, culminating in modes related to reactive or predatory aggression (^{Bernstein & Navot, 2023}).

ST was adapted for the forensic population (^{Bernstein & Navot, 2023}) and in 2007, researchers identified five typical hypercompensating schema modes in the forensic population: Self-Aggrandizer, Bully-and-Attack, Predator, Conning Manipulator, and Paranoid Overcontroller, whose descriptions are shown in Table 2 (^{Bernstein et al., 2007}; ^{Soygüt et al., 2023}). “Forensic schema modes” increase the likelihood of offending and aggressive behaviors, as they typically represent maladaptive attempts to deal with painful and unwanted feelings (^{Oettingen & Rajtar-Zembaty, 2022}).

Although there have been studies on ST for many years, the number of publications achieved in this systematic review highlights the precariousness of the number of empirical studies on schema modes. Considering that “forensic schema modes” also appear in non-forensic environments, the existing bibliography on the use of SMI in cases of aggressive behavior is relevant not only for research purposes but also to assist clinical therapists who may face major problems because they do not know how to recognize or intervene with these modes (^{Bernstein & Navot, 2023}). Such ignorance can also create obstacles to the therapeutic relationship, considered a primary agent of change in ST (^{Salicru, 2023}).

This systematic review aims to encompass studies in which behaviors described as aggressive or violent have a motivation and/or repercussions related to the definition of violence decreed by the World Health Organization (^{Dahlberg & Krug, 2006}):

The intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community that either results in or has a high likelihood of resulting in injury, death, psychological harm, mal-development or deprivation.

The definition used by the WHO associates intentionality with the completion of the act, regardless of the result produced (p. 279).

Therefore, the correlation between the themes “schema modes” and “aggressiveness/violence” constitutes a topic of great relevance since individuals with aggressive behavior represent a risk to themselves and/or society. We will seek to identify how the data obtained from the SMI proved helpful for understanding aggressive behavior or in interventions with violent individuals and its contributions to understanding and treatment of this population, usually segregated and stigmatized as incurable.

METHOD

The design of this study involves a systematic review of the literature. The guiding question was created based on the PICOS strategy, an acronym considered very useful for formulating questions (^{Fuchs & Paim, 2010}), as described in Table 3. Therefore, the following research question arises: “What are the benefits, proven in empirical studies, of using SMI in working with adults with aggressive behavior?”

After determining this guiding question, we selected descriptors previously validated on the platform Health Sciences Subject Descriptors (Descritores de Assunto em Ciências da Saúde - DECS) and keywords related to the components of the PICOS strategy to be used in search strategies. For this purpose, works from the PePSIC, Lilacs, PubMed, SciELO, PsycINFO and SCOPUS databases were considered. No grey literature was included.

The descriptors “terapia do esquema”, “violência”, “agressão” and “raiva”, and their English versions “schema therapy”, “violence”, “aggression” e “anger” were used. The follow ing keyword was also used in Portuguese and English: “modos esquemáticos” and “schema modes”. All the terms mentioned above were connected by the Boolean operators “OR” and “AND”, and the final search strategies are shown in Table 4.

We selected studies published from 2010 onwards because this was the year the SMI was prepared in Portuguese and English. The searches and selection were carried out by two reviewers (JP and FS) independently, and if they disagreed, an independent reviewer (KS) was consulted to discuss consensus.

One hundred two studies were selected in this preliminary phase. After eliminating 39 duplicates, 63 studies were chosen for the abstract reading phase. The inclusion criteria were being an empirical and quantitative study and addressing the ST theme or schema modes. Forty-four articles were then selected for a second screening.

In the second screening, studies with a population under 18 years of age were excluded, as well as studies that correlated schema modes and psychopathologies or symptoms not linked to aggressive behaviors. Studies whose method was a case study were also excluded (due to their low level of scientific evidence).

The remaining studies were read in full, and the inclusion criteria established were the application of the schema modes inventory (SMI) and the concept of aggressive behavior consistent with that specified by the authors. From this, texts that fit the research theme were selected, resulting in nine studies. The study selection flow is detailed in Figure 1.

Note: Translated by: José Luís Sousa*, Sónia Gonçalves-Lopes*, Verónica Abreu*, and Verónica Oliveira / *ESS Jean Piaget - Vila Nova de Gaia - Portugal by: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. doi: 10.1136/bmj.n71

Figure 1 PRISMA 2020 flow diagram for systematic reviews. 

The risk of bias in the randomized clinical trial (RCT) included in the final selection was carried out through the Joanna Briggs Institute Critical Appraisal Checklist, and the risk of bias in cross-sectional studies was assessed using the instrument ROBUST - Risk of bias used for surveys tool, an accessible and flexible instrument designed to evaluate the risk of bias in survey studies (^{Nudelman & Otto; 2020}). Analyses were performed independently by two reviewers (JP and FS), and a third reviewer was consulted (KS) in cases of disagreement.

RESULTS

The results prove the need for more studies on this topic and uniformity in descriptors and keywords related to the subject. Upon consultation of the descriptors by DECS, we found no descriptors related to schema modes. The only descriptor linked to the issue is schema therapy (and its Portuguese version), a term that results in an extensive database search.

Furthermore, the searches were first carried out in October 2022 and redone in August 2023. No articles were found in any of the selected databases that fit the theme of this systematic review in Portuguese. Therefore, the term “inventário de modos esquemáticos” [schema modes inventory] was not helpful when used as a keyword for searches.

Aiming to identify the types of national research being carried out with the schema modes inventory, we used the term as the only keyword without refining the searches according to the inclusion criteria for this study in the databases selected for this review. However, this search yielded no results. This fact corroborates the idea that there is no uniformity in descriptors and keywords related to schema modes, as there are bibliographic reviews and other types of studies on schema modes inventory carried out in Brazil.

After that, we searched its translation into English, “schema mode inventory” (SMI), or just “schema modes”, and yet obtained no results from the SciELO database. On the other hand, in this same database, from the descriptor “schema therapy,” we could locate six studies published in the last five years, only one related to schema modes. However, although this is an empirical study using the SMI, it needed to meet the inclusion criteria of this systematic review as it is a case study.

For the term “schema modes” and the descriptors “schema therapy,” “aggression,” “violence,” and “anger,” connected by Boolean operators “OR” and “AND,” the PubMed platform tracked articles that, when refined, resulted in 16 studies published in the last five years, including four empirical studies on the theme of aggression/violence. Using the same search criteria in the PePSIC database, we found two studies, both from literature reviews (therefore, excluded from this systematic review). In the Lilacs database, accessed through the VHL Regional Portal, we could locate six studies, of which only one involved the topic schema modes but did not fit the inclusion criteria of this study.

The databases that allowed access to more studies were SCOPUS and PsycINFO. With the descriptor “schema therapy,” the keywords “schema modes,” “aggression,” and “violence” connected by Boolean operators “OR” and “AND,” we tracked 46 articles. Among the 15 articles located in Scopus, four met the inclusion criteria proposed here, while among the 31 articles found in PsycINFO, five were also in accordance with the proposal of this systematic review. This information is described in the flowchart shown in Figure 1.

An assessment of the level of evidence of the studies was carried out, considering the following factors: sample characteristics, type of sample recruitment, exclusion criteria, final sample size, explanation of the sociodemographic data of the sample, reliability measures of the instruments, location of the study, and method of data analysis. All these aspects were evaluated by two judges separately. The proportion of agreement between the articles’ clarity criteria was 0.89 (or 89.0%), and the kappa agreement index was equal to 0.420 (95%CI: 0.014, 0.825), with moderate strength. These values are detailed according to aspects covered and presented in Table 5. Similarity was observed in the agreements for most of the evaluation items of each article.

The results prove that the theme chosen for this review represents a critical area of psychology that requires additional investigations, especially in Brazil, which is evident because none of the selected empirical studies were done in Brazil, and all articles are in English.

As for bibliometric indicators, the following items were broken down: authors, year of publication, journal of publication with impact factor, country, and keywords that made it possible to locate the article in this systematic review. The main details on the sample content and bibliometric indicators are shown in Tables 6 and 7. As shown in Table 6, we can identify that the constituted study material presents the theme of using the inventory in schematic ways under different approaches. As shown in Table 7, each study was published in a different publication vehicle.

The years of publications confirm that the topic addressed here is recent in scientific circles worldwide. Only two studies have been published, one in 2023 and another in 2022. However, the latter is slightly older, first published in 2020 and republished in 2022.

The sample content of this systematic review is composed of articles published in scientific journals with different impact factors, even in the case of papers that were part of the same study. The study published in the journal with the lowest impact factor (Study 3) is part of another more robust study (Study 1), which, in turn, was published in the journal with the second-highest impact factor in the sample. Study 2 (^{Dunne et al., 2019}) was based on results from Study 4, conducted by the same researcher and other collaborators in 2018.

The countries responsible for most studies found are Australia and the Netherlands. The other two countries responsible for the research that supports this review are Turkey and Denmark, with an investigation being carried out in each country. The latter is the country that publishes the highest impact scientific journal in the sample.

Considering that ST scientific contributions are commonly associated with personality disorders, four of the nine studies in this systematic review also address this topic (Studies 1, 3, 6, and 8). In almost all studies, the sample comprised male offenders. Of the nine studies analyzed, four were conducted in prison units and three in forensic psychiatric hospitals. The studies whose participants did not fit into the forensic population were Gülüm and Soygüt’s (2022) and ^{Bach and Farrel’s (2018)}.

These two studies above involved participants of both genders, not being restricted to males as in other cases. ^{Gülüm and Soygüt (2022)} claim to have tested, through an ANOVA analysis, whether there would be signs of significant differences related to gender in terms of the variables tested. The results proved that gender differences were not important in this study.

In all cases, the SMI was used as a complementary instrument within an evaluation process composed of qualitative and quantitative instruments, depending on each research’s objective. Only two of the eight studies used the SMI as an assessment instrument within a psychotherapy process in ST, with the two studies being correlated, as they are part of the same research.

Bernstein et al. (2021) and ^{Bach and Farrel (2018)} used the SMI-R - Schema Mode Inventory-Revised (^{Lobbestael et al., 2010}) of 118 items. ^{Dunne et al. (2018)} and ^{Dunne et al. (2019)} used the SMI version 1.1, a version with 124 items that, according to the authors, at the time of publication, had not yet had its psychometric properties published but was the most recent version indicated by the Schema Therapy Institute. ^{Lewis et al. (2021)} used this same version of SMI 1.1, and in the bibliographic references, they also do not mention a validation article; they only refer to the materials and inventories of the Schema Therapy Institute 2014 Version 2.0.

^{Vos et al. (2023)} and ^{Clercx et al. (2021)} claim to have used the SMI-R. However, the latter describes the instrument as containing 80 items. Therefore, it is possible to infer that this is, in fact, the SMI-FV, a short and revised version, whose validation is described in the article by Keulen De-Vos et al. (2017), which, in turn, is among those selected for this systematic review. In this version of the SMI, the 80 items are related to the 16 schema modes (FIVE items for each schema mode) and are evaluated based on a Likert scale from 1 (never or almost never) to 6 (always).

Only the study by ^{Gülüm AND Soygüt (2022)} used the version of the SMI adapted for the forensic population, which the authors are still developing and validating. This is the Schema Mode Inventory-Forensic (SMI-F) with additional items specific to this population, totaling 174 items.

DISCUSSION

Initially, the index of agreement of the assessment of the quality of the selected articles stands out. There was no difficulty in identifying its main characteristics in the studies. Achieving high levels of agreement in the selection of works can be a complicated task, considering that not all articles make clear all the points required for their evaluation. In the literature, there are interpretations for kappa values, and these depend on the prevalence of the effect of interest and the way of classification by different reviewers. In this study, the agreement rate was 89%, considered moderate but satisfactory.

The results revealed several types of limitations in the studies. When the research was carried out in psychiatric or forensic institutions, the institutions previously selected the participants, thus leaving aside the most hostile and unstable individuals. Self-narrative instruments also raise the hypothesis of biased accounts inconsistent with reality, provided as socially desired responses or interposed by cognitive limitations, memory limitations, or feelings of guilt and shame.

It is also necessary to point out that the fact that most studies were carried out only with male individuals prevents generalization of the results. In the study by ^{Lewis et al. (2021)}, the sample was homogeneous in terms of gender and country of origin, while in the study by ^{Bach and Farrel (2018)}, the participants were from a specific region of Denmark characterized by a homogeneous sociodemographic profile.

Furthermore, one of the main limitations in case studies carried out with the forensic population is the principle of voluntariness. Bernstein et al. (2021) point out that previous studies have already raised questions about the effectiveness of coercive forensic treatment, as is the case with the types of treatment usually offered in the institution where their randomized clinical trial took place. However, participation in Bernstein et al.’s (2021) clinical trial was voluntary, and participants had the prerogative to withdraw without any consequences, demonstrating rare withdrawals and very satisfactory results.

In the case of these institutionalized patients, even when participation is voluntary, participants are chosen for convenience, as in the studies by ^{Vos et al. (2023)}, ^{Lewis et al. (2021)}, Bernstein et al. (2021), ^{Clercx et al. (2021)}, and ^{Dunne et al. (2019)}. The studies by ^{Vos et al. (2023)} and ^{Lewis et al. (2021)} were conducted in psychiatric hospitals, where patients with the research profile were invited. Participants who volunteered had to obtain permission from their psychiatrists or head nurses of their units, who checked whether the volunteers were mentally stable and their stress level at the time. In studies by Bernstein et al. (2021) and ^{Clercx et al. (2021)}, participation was also voluntary, and the exclusion criteria were also related to mental instability, chemical dependency, and pedophilia.

In Dunne et al.’s (2019) study, participants were invited through notices and flyers spread throughout the prison. Twentyeight participants were excluded due to illiteracy, while those not proficient in the primary language of the research, English, were also rejected. The same invitation resources were used by ^{Dunne et al. (2018)}; only those under 18 and those who did not know how to speak the primary language of the research (English) were not considered eligible.

In the study by ^{Gülüm & Soygüt (2022)}, the sample was composed of university students who studied the same Psychology subject. Therefore, we can infer that some technical knowledge about the object of study may have interfered with the results. Furthermore, there was no differentiation of diagnostic conditions, and the sociodemographic profile was homogeneous.

^{Clercx et al. (2021)} state that their hypotheses could not be confirmed because the effect of the correlation between emotional states and the risk of violence is small, which would require a much larger sample. Another limitation could be the possibility of this correlation only existing in a later treatment period or shortly after hospital admission. The average length of stay for participants was one and a half years, a period that may have already been used to treat maladaptive emotional states, which may have interfered with the results of this research.

In the study by ^{Lewis et al. (2021)}, the questionnaire items were read aloud so that participants responded directly to the researcher. In the study by ^{Dunne et al. (2019)}, participants answered the self-report instruments in the presence of up to four other participants in the same room, and in the study by ^{Dunne et al. (2018)} with the presence of up to five other participants, under the supervision of the PhD candidate responsible for the research. As it is sensitive content, the different forms of exposure may have impacted the participants’ responses. ^{Dunne et al. (2019)} also point out previous studies that identified a tendency in forensic patients to present themselves disproportionately favourably, which can affect research results.

Keulen De-Vos et al. (2017) corroborate this idea and, to minimize the effect of this bias on the results, they previously carried out a data analysis correlating the results of all inventories used (including the SMI) with the results of the BIRD (Balanced Inventory of Social Desirable Responding), an inventory that assesses social desirability. As a result, they reduced the sample by 10% to reduce the tendency of distorted reports.

Since ^{Lewis et al. (2021)} used an observation instrument, their results may have been impacted by the observers’ (nurses) bias regarding mood, comparison with other patients, and knowledge of the participants throughout the assessment. And because they were participants in psychiatric treatment, their health conditions may also have interfered with the results.

^{Vos et al. (2023)} and ^{Dunne et al. (2018)} point out as a limitation in their studies that they are cross-sectional studies, which limits the possibility of making inferences about the real role of schematic modes in retrospective aggressive behaviors. Differently, ^{Clercx et al. (2021)} assessed follow-up after 6, 12, and 18 months. Follow-up could not be assessed in the clinical trial by Bernstein et al. (2021).

According to ^{Lewis et al. (2021)}, who assessed the follow-up after two weeks, the Angry Child and Undisciplined Child modes were the most identified in aggressive behaviors, corroborating previous studies on this topic. There was no correlation between the schema modes identified in retrospective aggressive behaviors and the follow-up.

Golüm and Soygüt’s (2022) results pointed to the mediating role of anger-related modes, and the Vulnerable Child mode in the relationship between the Punitive Parent and Demanding Parent modes and psychological symptoms. However, the most expressive results are related to the Vulnerable Child mode as a mediator rather than the schema modes more directly related to anger. The authors also point out that anger is more related to the Punitive Parents mode, and this corroborates the theory that patients become their own abusive parents and punish themselves.

According to ^{Dunne et al. (2019)}, the results showed correlations between the cognitive structures related to the aggression of the General Aggression Model (GAM), anger, DSM-5 personality traits, and schema modes. Anger traits are strongly associated with the three schema modes included in the study (Angry Child, Impulsive Child, and Bully-and-Attack), the aggression script is moderately associated with the Angry Child and Bully-and-Attack and little related to the Impulsive Child mode; normative beliefs of consent to aggression were moderately associated with the Angry Child and Bully-andAttack, and there was no correlation with the Impulsive Child mode. Hostility and Risk-taking were positively associated with the three schema modes.

CONCLUSION

This study aimed to list the ways the SMI can contribute to understanding aggressive behavior, a relevant topic to the well-being of society. We consider that this theoretical knowledge can support new empirical research and, in the future, new care protocols and more effective practices than those currently in force.

The research carried out indicates, initially, as we still lack a more incisive volume of scientific articles to support this claim, the viability of using the SMI as a predictor for the risk of violence and propensity for aggressive behavior, and its benefit within a process of long-term psychotherapy, along the lines of the ST, aiming at the rehabilitation of violent offenders with personality disorders. New longitudinal studies are necessary to confirm the relevance of identifying schema modes in predicting the risk of violence.

Therefore, studies that aim to understand the triggers of emotional dysregulation and impulsive behaviors are crucial, since theses aspects are directly related to aggressive behavior. Using the SMI to identify the schema modes that precede or are present in occasions involving violence provides self-knowledge that can be used to regulate emotions and control impulsivity.

This study points out the Impulsive Child, Angry Child, and Bully-and-Attack as the schema modes most related to aggressive behaviors. Recognizing schema modes allows a greater understanding of the circumstances that activate certain emotional states that cause discomfort and how these individuals self-learned to cope with discomfort with aggressive behavior, seeing in it some advantage over emotional discomfort. Therefore, recognizing the schema modes, the role of anger, and whether it is a primary or secondary emotion through the SMI can be a precursor act to discovering new possibilities for interventions in cases of violence.

Aggressive behaviors hide emotional needs. Therefore, without psychological intervention that identifies the emotional need hidden behind angry behavior, professionals who work with violent behaviors may have difficulty understanding and, consequently, dissuading these individuals from not recognizing aggressiveness as the only alternative to let out emotional discomfort.

The limitations of this systematic review are directly linked to the limits of the studies included in it. The fact that only nine studies were identified, most carried out in a forensic context, with male participants and a convenience sample, prevents generalization and greater assertiveness in describing the results. The authors declare that there is no conflict of interest or funding.