Virtual Reality Aggression Prevention Training for People with Mild Intellectual Disabilities: A Feasibility Study

ABSTRACT

Introduction

Problems with adequately regulating aggression are comon in people with mild to borderline intellectual disabilities (MBID). Learning through experience has been shown to be effective in MBID. Therefore, we adapted an aggression prevention training that uses virtual reality (VR), which was initially made for a forensic psychiatric population. We aimed to assess the feasibility of this VR aggression training for MBID.

Methods

Nine people with MBID attended the 12-session VR aggression training. Both quantitative (on session satisfaction, well-being and observed aggression), and qualitative measures (feasibility workbook evaluations and focus groups with participants and therapists) were obtained.

Results

Six participants completed all sessions, and three participants dropped out because of a lack of motivation. Session satisfaction scores were high. Thematic analyses identified that participants learned coping strategies, and gained insights into emotions and triggers. Suggestions for improvement of the training included more personalization, practicing longer in VR, and more involvement of the social network in training sessions. Aggression observation measures were not feasible and no reliable results could be obtained.

Conclusion

Although the dropout was substantial, the training seems feasible and acceptable for participants and therapists. However, improvements should be made to the intervention to increase efficacy, and enable better fits with participants’ specific needs.

KEYWORDS:

• Virtual reality
• aggression
• anger
• intellectual disability
• behavioral training

INTRODUCTION

Problems in adequate aggression regulation are common in people with mild intellectual disabilities to borderline intellectual functioning (MBID). Twelve-month prevalence varies from 2% to 28% for physical aggression, and 6% to 38% for verbal aggression in adult service users with an intellectual disability (Cooper et al., 2009; Crocker et al., 2006). Recurring aggressive behavior is stressful and disabling for the individual displaying the behavior (Cooper et al., 2009; Heyvaert et al., 2010). Furthermore, it can have a strong effect on their environments, such as family, friends, and caregivers (Bruinsma et al., 2020).

Psychological interventions for aggressive behavior of people with intellectual disabilities are both recommended by guidelines and preferred by healthcare professionals (Bruinsma et al., 2020; Unwin & Deb, 2008). A recent systematic review on psychological interventions for aggression identified 15 studies (Prior et al., 2023). They showed that there is moderate evidence for the efficacy of cognitive behavioral approaches for individuals with intellectual disabilities. Cognitive behavioral therapies (CBT) such as anger management and positive behavior support seemed most effective.

Current psychological interventions for aggression using cognitive behavioral techniques have some limitations for people with MBID. Therapies can be hard to tailor to the specific needs of an individual and repeated practice of social interactions in real life is not always possible or safe. However, repeated practice is important as it takes people with MBID relatively long to learn and generalize new behavior and skills to daily life (Kleinert et al., 2009). Further, practicing face-to-face anger-provoking interactions could negatively influence the client-therapist relationship. Finally, many interventions require the ability to reflect on emotions, thoughts, and behavior, and are often predominantly verbal, whereas people with MBID often experience difficulties processing verbal information (De La Iglesia et al., 2005).

These limitations may be overcome by using immersive virtual reality (VR) (Geraets et al., 2021). With VR, people are immersed in computer-generated simulations of real-life situations, such as talking to a cashier in a supermarket, by using a head-mounted display. VR simulations have been shown to induce presence, which is the feeling of actually being in the virtual environment, causing people to react realistically in VR (Skarbez et al., 2017; Slater, 2009). Furthermore, VR has shown to trigger psychological and physical reactions similar to the reactions in real life (Martens et al., 2019). VR can offer a safe, controllable environment that can be adjusted to the level of the patient; allowing repeated, interactive and personalized practicing. Thus, VR interventions can be tailored to the specific needs of the individual. Further, by using VR the focus is more strongly on visual information processing and practicing instead of verbal information processing. Similarly, VR has been used safely in this target group for learning general life skills (Cheung et al., 2022; Nabors et al., 2020).

Recently, the first VR aggression prevention training has been developed and tested in forensic psychiatric inpatients (Klein Tuente et al., 2018; Tuente et al., 2020). Self-reported hostility, anger, and impulsiveness improved after VRAPT, but no decrease in staff observed or self-reported aggression was found in this RCT. Interviews did reveal that participants were positive about VRAPT and able to recall what they had learned (e.g., recognizing physical signs of tension and insights into triggers). The lack of effects on aggression may be explained by the target population as it concerned people with severe and long-lasting mental illness, and for some, the training and the theoretical part of the training might have been too difficult.

Although the effects in forensic inpatients were inconsistent, this innovative and interactive VR training may be of added value for people with MBID due to the above-mentioned advantages of VR. Therefore, we further developed this VR aggression prevention training to fit the cognitive abilities of people with MBID, amongst others by simplifying the theoretical framework and the structure of the sessions and involving the social network in training sessions. In this study, we aimed to explore the feasibility of VR aggression prevention training in people with MBID and aggressive behavior by examining user experiences, acceptability, utility, and preliminary efficacy of the training with a mixed methods approach using both qualitative and quantitative data.

METHODS

Design and Participants

The current study is a feasibility study in which we aimed to include 15 participants. The study was performed at Cosis, a healthcare institution in the Netherlands for people with intellectual disabilities as well as problems in adaptive functioning. Many of the clients at Cosis also have a psychiatric disorder. Inclusion criteria were residing or in therapy at Cosis, referred to aggression training by their therapist or primary mentor, sufficient command and understanding of the Dutch language, IQ between 50–85 (based on the WAIS, WISC, or indication of the clinician in case a structured assessment could not be performed), and age between 16–65 years. Exclusion criteria were a (history of) epilepsy and substance dependence as it can impact treatment effects and adherence. This study was approved by the Medical Ethical Committee of the University Medical Center Groningen (number 2019/516). Participants received €30 compensation after completing the training.

Potential participants were informed about the study by their mentor and/or therapist. If they would like to participate, they were further informed by a member of the research team. Written informed consent was obtained. In case a client was aged between 16–18 or incapacitated, their parents and/or their legal representative were contacted and involved in the information-giving and informed consent process.

VR System

During the training, the Social Worlds VR software of CleVR BV (Delft, The Netherlands) was used, in which multiple virtual environments are available (e.g., a street, supermarket, bus, and home environment). Participants viewed the virtual environments by wearing an Oculus Rift 2 head-mounted display (Oculus VR, California, U.S.) and headphones, and moved around by using a Microsoft Xbox One controller. Therapists could interact with the participant by using a microphone with a speech scrambler, and they could adapt the virtual environments by using a tablet interface, see Figure 1 for the VR setup. At all times the therapist is in control of the virtual environment and can immediately change and/or stop the VR if necessary.

Figure 1. a) VR emotion recognition task and b) the VR set-up.

Intervention

The intervention is a practical, behavioral training, consisting of 12 individual sessions. The training was based on the VR aggression prevention training (VRAPT) protocol for forensic psychiatric inpatients with aggressive behavior problems (Klein Tuente et al., 2018). To make the VR training suitable for people with MBID, we adjusted the protocol and workbooks with two therapists who specialized in MBID in the lead. The following changes were made: 1) the training was shortened from 16 to 12 sessions, to increase the feasibility and make the training more manageable for participants. 2) The original VRAPT is based on the Social Information Processing model (SIP), which describes how people interpret (early information processing) and respond to social situations (late information processing) and how this can result in aggressive behavior (Klein Tuente et al., 2018). In the current training, we used a highly simplified and practical model to explain social information processing by using the action-oriented “Stop-Relax-Think-Act” (Nederlands Jeugd Instituut, 2020), as the SIP model was found to be too complex to work with for this target group based on the experience of the study of Klein Tuente and colleagues. As a first step, participants are taught how to stop or get out of unpleasant or stressful situations (S = “stop”), then they should create a brief moment of relaxation (R = “relax”), after which they can think about how to respond (T = “think”). Finally, after relaxing and thinking, they can act (A = “act”). Consciously competent use of the Stop-Relax-Think-Act strategy is a goal for participants and can be repeatedly practiced in VR environments. 3) Training sessions were highly structured and had a more simple structure, as this is a factor that is known to be advantageous for people with MBID. Also the training always ended with a relaxation exercise to make sure participants left in a positive mental state. This is especially important due to emotion regulation problems participants might have. 4) Mentors were involved during the training and participated in two to four sessions. Moreover, friends or family could also be part of a session. Involvement could amongst others include experiencing the VR, practicing in VR with the participant (both in the role of therapist and/or participant), discussing situations in which the participant gets aggressive, and providing input/examples on situations where aggression occurred if the participant was unable to do this. 5) Interactive scenarios focussed on recent personal experiences of the participants, making the intervention personalized and easier to empathize with. During the training, the participants focussed on one or two different personal scenarios. 6) The language of the training materials was adapted to fit MBID, less theoretical terms and less language in general was used in the training in general, including less focus on verbal reflections. 7) Workbooks were simplified and visual aids were added e.g., for the Stop-Relax-Think-Act strategy.

Therapists received two days of training in the protocol and one day of practical training in the VR software. Three therapists were trained, and group intervision meetings were held regularly. Sessions took approximately 50–60 minutes and were planned biweekly. Two workbooks (i.e., a therapist and participant workbook) were used throughout the training. All sessions were structured similarly. The sessions started with a short introduction. Then practicing in VR took place and the VR exercises were discussed. Finally, sessions ended with a relaxation exercise to unwind. An overview of the sessions is provided in Table 1.

Table 1. Session overview.

Session 1	Introduction session during which information was given about the training and two short, neutral VR exercises were performed to get familiar with the VR system and equipment. During exercise 1 participants could choose and watch their favorite avatar (i.e., virtual character), and during exercise 2 participants walked through an empty virtual shopping street.
Sessions 2–4	Focussed on recognizing and differentiating between emotions. Participants were shown pictures of the basic emotions (scared, angry, happy, sad) in their workbook and discussed and mimicked the emotions. Further, participants practiced with facial emotion recognition of avatars in the VR street. In addition, the Relax step of the Stop-Relax-Think-Act strategy was introduced. Situations in which the participant has difficulties were discussed, to be able to create a personalized VR scenario for the following sessions.
Sessions 5–6	Focussed on experiencing and handling provocation by other people in VR and further developing the personal scenario. In VR, participants experienced provocative situations and were instructed to either react, or remain silent. During these sessions, the other Stop-Relax-Think-Act steps were introduced and practiced.
Sessions 7–12	Focussed on practicing with the personalized provocative interactive social scenarios. The main goal of the interactive scenarios was to let participants practice and learn to cope with their aggression adequately (using the Stop-Relax-Think-Act strategy). During the interactive scenarios, the therapist took the role of the avatar using a microphone with voice distortion for speech, and manually controlled the facial emotional expressions, and body movements of the avatar.

Feasibility Outcomes

Recruitment & retention

Data on the duration of recruitment as well as measurement dropouts (retention) were collected.

Adherence

The number of attended sessions, duration of the intervention and reasons for stopping with the intervention were recorded. Furthermore, at the end of each session the therapist completed the following questions: 1) how long did the session take, 2) were there deviations from the protocol, and if yes, what and why, 3) were there any technical issues, 4) remarks and 5) do you have any tips or suggestions for improvement.

Session Satisfaction

Session satisfaction was measured at the end of each session with the four-item Child Session Rating Scale (CSRS) (Duncan et al., 2003). The CSRS has a visual analog scale (VAS) with at each end a description and a frowning or a smiling smiley that corresponds to a scale ranging from 0 to 10. Items focus on the relationship with the therapist (“the therapist listened to me”), goals and topics (“what we did and talked about is important to me”), approach (“I liked what we did today”), and an overall score (“I hope we do the same kind of thing next time”). The CSRS has moderate validity and solid reliability (Duncan et al., 2006). The CSRS was scored at the end of each session.

Acceptability

Structured treatment evaluation forms with mainly open and some multiple choice questions were completed during the final session by the client, mentor, and therapist. The treatment evaluations assessed which parts of the training were useful and which were not, what participants learned, the effect on aggressive behavior, what improvements could be made, the difficulty level, and the extent to which the training fits with the clients and with the current treatment offer at Cosis.

After all participants completed the training, two focus groups were held. This method was chosen as it provides rich qualitative data and multiple perspectives can be obtained. Furthermore, group dynamics can encourage participants in sharing their experiences. One focus group was done with the three therapists and one mentor, and the second focus group with two participants, one therapist (who was also present in the first focus group), and one mentor. The goal was to evaluate the acceptability, feasibility and to identify areas of improvement. To this end, two scripts with respectively 9 (for the focus group with clients) and 10 (focus group with all therapists) questions were constructed. Focus groups as well as the feasibility evaluation forms were transcribed using AmberScript software.

Effect Outcomes

Well-being

Well-being was measured with the four-item Child Outcome Rating Scale (CORS) which also uses a VAS with smileys with scores ranging from 0 (frowning smiley) to 10 (smiling smiley) (Hafkenscheid et al., 2010). It assesses how somebody is feeling about themselves (“how am I doing”), home (“how are things going at home”), socially (“how are things with me at work and with social contacts”), and in general (“how is everything going”). The CORS has satisfactory levels of validity and reliability (Casey et al., 2020). The CORS was scored at the start of each session so that the session and contact with the therapist would not influence the well-being measure.

Aggression

Aggression was measured with the Modified Overt Aggression Scale (MOAS) (Kay et al., 1988). The MOAS is a behavior rating scale designed to measure four types of aggressive behavior: verbal aggression, physical aggression against objects, physical aggression against self, and physical aggression against others. The MOAS was rated by the mentor of the participant. The MOAS is a valid and reliable measure in people with intellectual disabilities (Oliver et al., 2007; Ratey & Gutheil, 1991).

Analyses

Descriptive statistics were calculated for all quantitative data by presenting the mean and standard deviation, or count and percentage. To analyze the preliminary effects of the treatment, the pre-treatment (session 1) and end-of-treatment (session 12) CORS scores were compared with the non-parametric Wilcoxon matched-pairs test. Significance was accepted at .05.

Analysis of qualitative data (focus groups and treatment evaluations) was conducted with the coding software ATLAS.ti 22. Thematic analyses were performed using an inductive, realist approach. First, we transformed the script into codes and subcodes and then the codes were summarized into semantic themes in an iterative process by two authors CG and LR.

RESULTS

Feasibility

Recruitment & retention

Nine participants were included in the study between February 2020 and April 2021. We initially aimed to include 15 participants, however, due to the COVID pandemic we could not reach the aimed sample size within the timeframe of this research. Due to the pandemic, information on recruitment rates for a future study are thus not reliable. Therapists reported obstacles for recruitment to be the location of the VR equipment, to which participants had to travel. All participants were able to complete the CSRS and CORS measuring well-being. Table 2 presents the sample characteristics.

Table 2. Participant characteristics.

	n = 9
Age in years	30.0	(13.8)
Male	6	66.7%
Mild intellectual disability (IQ: 50–69)	6	66.7%
Borderline intellectual functioning (IQ: 70–85)	3	33.3%
Level of education
Primary education	4	44.4%
Pre-vocational education	2	22.2%
Secondary vocational education	2	22.2%
Secondary special needs education	1	11.1%
DSM-5 comorbid disorders
ADHD	3	33.3%
Autism spectrum disorder	2	22.2%
PTSD	1	11.1%
Bipolar disorder	1	11.1%

Note: Data are presented with M (SD) or n and percentage.

Adherence

Six out of nine participants completed all 12 sessions. One participant quit after the first session, because of a lack of motivation and feeling overwhelmed due to a previous negative experience with VR. This was reflected in the session rating scale with low scores on importance of the session content (4.9), pleasantness (5.5), and wish to do something similar again (4.7), the participant did feel listened to (8.4). Two participants stopped after session 9, reasons for stopping were lack of motivation amongst others because of stagnation of improvements. However this was not seen in the session rating scale, as both participants scored all CSRS items in the two sessions before stopping between 8.8 and 10.

Sessions took on average 52 minutes (SD = 9 minutes). For participants who completed all 12 sessions, the training took on average 23 weeks (SD = 6, range = 16–32 weeks). The training took longer because of prolonged intervals between sessions due to the COVID pandemic. Furthermore, it was not feasible for several participants and clinicians to engage in more than one session weekly. Technical issues were reported in 21% (19/91) of the sessions, of which 12 concerned synchronizing of the therapist’s speech with the avatar and headphone failures. In 17 out of the 19 technical issues, the session could be continued as usual, because either the problem was quickly solved (by restarting or calling the helpdesk) or it was a minor issue.

Session satisfaction

General session satisfaction scores as measured with the CSRS were high and increased slightly throughout the training, see Figure 2. The average score on the CSRS over the four items was 9.5 (SD = 0.8). Evaluations showed that the participants scored highest on the question “How much did you benefit from the training” with an average of 9.2 (SD = 1.0, n = 5), then mentors (M = 7.4, SD = 1.1, n = 4) and therapist rated the benefit lowest (M = 6.3, SD = 1.7, n = 4).

Figure 2. Mean session satisfaction ratings.

Acceptability

Table 3 shows the results of the thematic analysis on strengths and points of improvement, presenting the themes, subthemes as well as illustrative quotes. Overarching themes for strengths were learning outcomes, VR-specific advantages, and engagement of the social network. Points of improvement focused on more personalization and specific content-related improvements such as practicing more in VR, more participation of the social network, and removing physical arousal exercises that were not helpful.

Table 3. Thematic analysis with illustrative quotes on experiences with the VR aggression prevention training of focus group participators and session forms. In total, information was obtained from all three therapists, five participants, and five mentors.

Strengths
Subthemes	Explanation	Quote
Theme 1. Learning outcomes
Using the Stop-Relax-Think-Action strategy	Participators reported the Stop-Relax-Think-Action strategy as the main learning point and mentioned using it during and after the training. The extent to which the strategy was learned differed, some only mastered “Stop” and “Relax.”	“Well, what you saw right away was also a bit of self-correction, the more client W practiced, the more he corrected himself. Like, oh no, I have to do something else, or I have to stop for a while.” (Mentor 1)
(2) Awareness of emotions	Participants reported learning to be more aware of their emotions, both cognitively and by noticing physical signs corresponding with aggression.	“ … he learned what tension feels like for him. He also learned how to talk about this and to set his limits.” (Mentor 3)
(3) Insights into triggers of aggression	Through the VR experiences, participants and therapists gained insights into specific personal triggers of aggression.	“We were involved in a roleplay and then I noticed that at a certain point, that there were certain gestures that really triggered him, certain words that triggered him and because I noticed that in him and could stop him and say: hey, what’s happening now, tell me, at a certain point he learned to recognize that better himself, so he knew much better what made him angry.” (Therapist 1)
(4) Decrease aggression	Participants, mentors, and therapists reported improved aggression regulation and self-correction resulting in reduced aggression.	“I notice that I am much calmer and I also think a lot more before I speak. Because before, I sometimes said things before thinking, so I’m much more thoughtful in my actions now, I notice that, that’s mainly, yes, especially with getting angry indeed, that’s where I actually came for, because I also notice that I am much calmer, I have become less aggressive, also, in general, I am less likely to get angry.” (Participant 1) “But the really big change is that he had several female friends, who actually didn’t even want to hang out with him anymore because he could react quite aggressively. And that’s just gone, those friends no longer say to him we are afraid of you.” (Mentor 1)
(5) Stop-Relax-Think-Action strategy	Participators reported the Stop-Relax-Think-Action strategy and individual tailoring (e.g., implementing “Stop” by saying it or using gestures) of the strategy as a strength.	“Recently I had a problem with the taxi. I held back. I stuck to SODA (Dutch mnemonic for Stop-Relax-Think-Action strategy) and relaxed in between. Breathe in breathe out. Also applied it at work. I know what STOP stands for. You are not rejected.” (Participant 1)
Strengths
Subthemes	Explanation	Quote
(6) Emotion recognition tasks	Emotion recognition tasks helped practice emotion recognition, to get used to VR, and it was mentioned to be fun for some.	“My clients especially found emotion recognition in the shopping streets very nice and pleasant and we used that a lot to get back into the VR world, well, to once again get used to it. Client X asked for going to the shopping street again to recognize emotions, because he liked that so much and he also really liked to notice that he was getting a little better at that because he found that quite difficult. He also said: I find that pretty hard in real life too, so if I can get better at this, maybe I can get a little better at that in the real world too.” (Therapist 1) “Dolls, indicate who is angry” [Answer concerns choosing the avatar that looks angry out of three avatars] [answer on: what did you learn most from] (Participant 3)
Theme 2. VR-specific advantages
(1) Interactive VR scenarios	Interactive VR scenarios triggered emotions and physical arousal enabling on the spot “recognizing physical signs,” “learning what emotions feel like and how one can talk about this,” “naming felt emotions,” “experiencing how reactions influence others,” and practicing new strategies to deal with anger-provoking situations. Scenarios were repeated with different responses, to enhance learning and generalization. The combination of experiencing in VR and reflecting/discussing afterward was mentioned as a strength.	“I had a client who sometimes got very angry around a mentor, and I think we did the same situation 15× with a different response and then see, what does that do to the avatar, in that case the mentor, if you react differently. She got very angry about behavior of her mentor, but did not dare to say anything about it … So, by being able to practice in a safe environment saying to the mentor, I don’t like this or hey, I’d like it if you could just look at me now. … Practicing things like that over and over again and noticing that it does something different for her, that she goes home differently, yields the necessary effect in aggression regulation at home as well, as she went home less angry.” (Therapist 1)
(2) VR is attractive & fun	Participators reported that VR is an exciting and fun method that made participants curious and motivated and made it feel less like therapy than interventions where they only talk.	“They offered to participate because they thought it was so cool. It’s something new … and the attraction is permanent, so those clients come into that room and their head doesn’t go to me first, … but their head goes to that device and they can’t wait for it, to put those glasses on.” (Therapist 3)
Strengths
Subthemes	Explanation	Quote
(3) VR forms a safe environment	VR was mentioned to feel like a safe environment to practice in for participants as well as therapists to engage in aggression-provoking behavior.	“I think it helps a lot that it’s kind of a safe environment where you can practice, where you can also practice different reactions, for example.” (Therapist 1)
Theme 3. Engagement of social network
(1) During the training: Involving mentors & other healthcare workers	Healthcare workers provided input for scenarios, helped with new skills in between sessions, and participated in roleplays. This gave them information on strategies for how they could help participants in provoking situations, and they made others aware of what the participant was working on and how they could help (e.g., help remember the Stop-Relax-Think-Action strategy or how to relax).	“Yes, what I found all in all very positive was actually the visit. We were invited [to the sessions], and I tried a few more times to let a colleague come along as well, then it lives more at the residence as well.” (Mentor 2) “You knew exactly which buttons to press [to trigger him in VR].” (Therapist 3 to Mentor 2)
(2) During the training: Involving family & friends	Involving friends and family resulted in more detailed descriptions of situations where aggression occurred. Significant others to whom aggression was directed were also involved sometimes and were helped by practicing in VR how to react. Finally, learned skills were discussed with significant others.	“Client X, he also practiced with his girlfriend and with his mother, which was really important, network. We also phoned the mother a few times at the end of a session, to discuss what did you learn today, because he found that translation difficult.” (Therapist 2)
(3) After the training: support in daily life	Involving others in the training resulted in beneficial effects. Participators reported that the social network became more aware of triggers, and could help participants cope better, make changes in their own behavior/surroundings, and helped participants remember what they had learned.	“What I heard later from mentors at the residential locations, was that … if something happened on location, that they recognized triggers and could keep a closer eye on him, and could stop him much sooner, if he couldn’t do that himself at that moment.” (Therapist 1) “Client Y who always has the same jokes, so others don’t like that … coincidentally last week that happened again, and then I said to Y, you went to that course then and stop, that applies also for others, eh, if they tell you to stop, then you should do that too. And that was very funny, because then he picks it up immediately, that he is like oh yes, I was taught that then.” (Mentor 2)
Strengths
Subthemes	Explanation	Quote
(4) Visiting environment of participant	Visiting the working, daytime, or living environment by the therapist was reported to be informative. Visits provided input for sessions, were experienced as fun by participants, and created awareness of what the participant was working on e.g., healthcare workers at the locations.	“I’ve been on location with clients to see, where do they live, where do they work, who do they deal with? At one point another client came, who was very irritating for my participant … he talked very quickly and mixed up all kinds of things, and he wanted to know everything about you, very curious. … So, you take that with you in the treatment, and then you dare to do what such a client did back, … . that link with the actual situation is absolutely necessary.” (Therapist 3)
Points of improvement
Subtheme	Explanation	Quote
Theme 1. Content-related
More focus & time on practicing in VR	Therapists and mentors reported that more time should be invested in practicing and learning in VR. Herewith the training will become less verbal, which was seen as an advantage. As therapists became more experienced throughout the pilot, more time was invested in practicing.	“The idea was getting stronger that we need to practice more and less cognitively, so I’ve started doing that more and more, so the clients who came last got the most out of it. … that roleplaying of hard situations … . not going in-depth, but more of just practicing.” (Therapist 2)
(2) Physical arousal exercises are not helpful	Exercises were done to experience the physical arousal that accompanies anger and for practicing indicating when to stop. Participators agreed that the physical exercises did not relate enough to feelings of anger and aggression, and that arousal could also be induced in VR.	“Of climbing the stairs, I also think, what connection does it actually have with the training and with being angry?” (Participant 2) “Exercises that required me to do physical activity doesn’t seem like getting angry to me.” (Participant 4)
(3) More participation of family/friends/healthcare workers	Participation of healthcare workers, family, and friends was found to be extremely beneficial by all participators. But participators expected effects to be bigger if the network is included even more often.	“Other people should accompany participants more often during sessions” (Answer on the question what would you like to see changed most) (Mentor 4)
Theme 2. Personalization
7(4) Personalize session number & frequency	Participators reported that the session frequency and number should be fitted to the client. Some participants needed more sessions, and follow-up booster sessions should be offered for repetition and consolidation.	“What I got back from the participants is that they, what they would really like, are a kind of comeback sessions. So, for example, after 3 or 6 months you have another repeat session, to practice, especially because they also notice that it is quite difficult to keep doing it well and that it feels so safe to be able to practice it in such a VR world, occasionally being able to do it again repeat, because the power of repetition is really very important.” (Therapist 1)
Points of improvement
Subtheme	Explanation	Quote
(5) More personalization in VR software	Participators reported that it would be advantageous if the software offered more personalization in terms of surroundings (enabling practice in environments matching everyday life), appearance, and the voices of avatars.	“I’m going to practice with my husband, but my husband has glasses and the avatar has no glasses, that is not right” (Therapist 1 quoting participant 5) “Yes. Because I couldn’t find a voice at all.” [that sounded right] (Participant 2)
(6) More flexibility in the protocol in terms of VR exercises	Emotion recognition exercises were not suitable for all. Offering the training in a more modular form (in which e.g., the emotion recognition module can be shortened or skipped) and practicing with multiple situations can improve the training.	“Instead of increasing triggers, practicing skills in different situations leads to more generalization and prevents the participant from finding it boring.” (Mentor 3)
(7) Change relaxation exercises	The initial relaxation exercises did not fit with the participants’ needs. Personal exercises (with which they had positive experiences) or viewing relaxing nature scenes in VR did fit well.	“No, they (relaxation exercises) were really terrible! No, that did not really help. Those protocol exercises were, no.” (Therapist 1)

Evaluation of Effect Outcomes and Preliminary Effects

Well-being

The mean well-being scores over the 12 sessions are plotted in Figure 3. The Wilcoxon matched-pairs test showed that there were improvements in well-being when comparing scores at session 1 with the final session of the participants (either session 9 or 12; n = 8). Median scores improved for me (median_pre = 7.0; median_end = 9.8, Z = −2.5, p = .01), home (median_pre = 6.5; median_end = 9.8, Z = −2.5, p = .01), socially (median_pre = 6.5; median_end = 9.8, Z = 2.3, p = .02) and overall (median_pre = 8.2; median_end = 9.8, Z = −2.5, p = .01).

Figure 3. Mean well-being per session.

Aggression

Limited data were obtained on the MOAS observation aggression measure. Pre-treatment measurements were available for five participants, the mean score for verbal aggression was 2.8 (3.0), for object aggression 1.3 (2.6), aggression toward self 0.7 (1.3), and others 0.4 (1.3). No complete post-treatment data was available; therefore, no analysis could be performed. This had multiple reasons: 1) training trajectories took longer than 16 weeks due to treatment pauses because of the COVID pandemic and because for multiple participants it was not feasible to engage in two sessions per week, 2) missings were caused by changes in mentors and 3) unknown reasons.

DISCUSSION

This was the first study using VR as a tool for aggression training in people with MBID. We aimed to investigate the feasibility of an adapted version of a VR aggression prevention training used in forensic psychiatry and to identify aspects of the training that need improvement. Findings suggest that the training is acceptable and well-tolerated. Six out of nine participants completed the 12-session training, and two completed 9 sessions. Although three participants dropped out of the training, mainly due to motivational issues, session satisfaction ratings were high. Focus groups revealed that interactive practicing in VR was most valuable and helped in learning new strategies to cope with difficult situations and/or gave insights into triggers of aggression that could help to inform the social network of the participant.

The current findings suggest that VR aggression training is promising and might be of added value for people with MBID and aggressive behavior. Even though levels of cognitive functioning were limited and comorbidity was high, the training was acceptable for two third of the participants. However, individual session rating scores were high prior to dropout, indicating that external factors other than the intervention may have played a role for quitting. Such high dropout levels during psychological interventions are common in this target group (Bruinsma et al., 2020), but might be reduced when mentors and family members are engaged more and encourage participants to complete the training. The participants do seem to be a good reflection of the population as high comorbidity is common in people with MBID (Einfeld et al., 2011), which is also related to having higher odds of displaying aggressive behavior problems (Crocker et al., 2014).

In one participant no aggression nor anger could be triggered. The participant reported that it was caused by a lack of realism of the VR surrounding, that the roleplays were not real enough, and that he knew that the therapist aimed to provoke aggression. Thus is seems that he was not immersed sufficiently and not enough presence has been induced. Although increasing realism can enhance presence, also emotional factors (such as feeling anxious, lonely, and memories) can influence the sense of presence (Riches et al., 2019). Lack of motivation or fear of losing control could have played a role as well and might create resistance against immersion. Several ways to increase readiness for therapy in people with MBID have been proposed, such as motivational interviewing, and simplifying or personalizing a therapy (Willner, 2006), which might be necessary for some individuals with MBID to engage in VR therapy.

Participants felt they were listened to during the training, found the content relevant and sessions pleasant. This indicates that the affective attitude was good, which is an important component of acceptability (Sekhon et al., 2017). Mean session satisfaction scores increased slightly over the first three sessions. This might be explained by participants getting used to the VR and small changes of the protocol. For example, if relaxation exercises did not match with the participant these were replaced by other self-chosen exercises to keep participants motivated. As the session satisfaction questionnaire was completed in the company of the therapist, we cannot rule out that some participants may have given socially desirable answers (Langdon et al., 2010; van de Mortel, 2008).

Qualitative data revealed that participants and mentors reported that participants learned to self-regulate better by recognizing triggers and (partly) using the Stop-Relax-Think-Act strategy preventing anger-inducing situations to escalate. Whereas some managed to fully learn the Stop-Relax-Think-Act strategy, others only learned the Stop part, which is still relevant in daily life. Strategy learning occurred largely during interactive scenarios in which person-specific triggers of aggression were created safely, causing real-time emotions and arousal. Consequently, it enabled repeated practicing of specific steps to facilitate learning by practice. Similarly, qualitative studies focussing on CBT in people with intellectual disabilities also reported use of new skills as an important outcome (Evans & Randle-Phillips, 2020). Further, studies on anger management, which is strongly related to aggression treatment, using a CBT approach were found to be effective (Taylor et al., 2016; Vereenooghe & Langdon, 2013). Interestingly, recent research has shown that the experience of the therapist is also a factor influencing effectiveness in anger treatment in this target group (Taylor & Novaco, 2023), which should be taken into account in further research.

Care workers, family, and friends had important functions during the training, as is advised by the NICE guidelines (National Institue for Health and Care Excellence, 2016). They provided input on aggression-inducing situations and made exercises more meaningful and targeted with input from their perspective. People from the network also participated in scenarios, gained insights into triggers, and some learned how to manage aggressive situations better. Finally, the network enabled continued rehearsal and prompting of new learned strategies after the training (Brown et al., 2011), could prevent triggers, and create more awareness of triggers in the participants’ social environment. The relevance hereof cannot be underestimated as many struggle to apply and maintain new skills to everyday life as shown in a review on qualitative studies (Evans & Randle-Phillips, 2020). However, the specific role people from the network can take varies strongly with their capabilities and motivation as well (Willner, 2006). So the involvement of the network including carers should be tailored to the specific individuals (Willner, 2006). Providing multiple options for involvement within a protocol would therefore be advised.

Evaluation of Preliminary Intervention Responses

Well-being improved over time, even though at the start of the training it was already relatively high. This forms preliminary indications of a possible treatment effect, but we cannot dismiss the possibility of other unknown factors having influenced well-being. Further research with a control group is needed to investigate the efficacy of the treatment.

Quantitative information on changes in aggression was not available as mentors did not continue to score the MOAS observation scale long enough. Partly because it is highly time-consuming (Matson et al., 2005), mentors were unmotivated and/or forgot to score the MOAS. Further, more weeks of observations than initially planned were needed due to delays because of the COVID pandemic and lower frequencies of sessions. Also, the time that mentors saw participants varied strongly as some participants were more independent (e.g. living independently), and others saw participants daily as they lived in the care facility of Cosis. For this specific target group that partly lives independently, the MOAS does not seem to be a feasible instrument. Import to note, during the training, the MOAS was found to be relevant to provide input for the VR scenarios, as some mentors described incidences that had happened in the past week.

Main Protocol Improvements Based on This Study

Even though the involvement of the social network was a strength, more room is needed in the protocol to include healthcare workers, family, and friends. Practicing with interactive scenarios was found one of the most helpful parts of the training, and more time should be spent on this. Especially in this group, the value of repetition cannot be underestimated. Due to software and hardware advances, cybersickness has become less common with longer VR exposure (Rebenitsch & Owen, 2021). Furthermore, repeated exposure – as was done in the present study – has also been shown to reduce cybersickness and enable longer exposure in VR (Palmisano & Constable, 2022).

Another point of improvement concerned a higher level of flexibility and personalization. This is in accordance with the review of Willner (2006) on factors influencing therapy in people with an intellectual disability, which indicates that changing the therapy on an individual basis is an important strategy for increasing efficacy in this target group. Also, more specifically for aggression, a recent review stresses the importance of tailoring and personalizing treatments to fit individual needs (Royston et al., 2023). In our study, more personalization was found to be relevant on different levels, e.g., on the level of session planning (i.e., number of sessions, frequency, and the possibility for follow-up sessions) and the protocol, as participants had different levels of functioning and goals. More tailor options in the software would also be helpful in terms of the surroundings, looks, and voices of avatars so that past aggression-inducing situations can be mimicked better and to increase realism (Jung & Robert W, 2021). Higher flexibility levels do ask for more input and strain from therapists. However, by acknowledging the heterogeneity between people, it enables better fits between participant and training, which is done in clinical practice on a daily scale.

Limitations

The main limitations were the uncontrolled nature and small sample size of this study. We aimed to include 15 participants, however, due to the COVID pandemic the intervention took on average more than double as long because sessions had to be delayed. Therefore, we could not reach the aimed sample size within the timeframe of this research. Furthermore, only two participants with relatively mild intellectual disabilities and two mentors participated in the focus groups. Possibly the most motivated people participated, which could have influenced our findings. This may limit the generalizability of our findings.

Implementation of the MOAS failed, providing no valid information on changes in aggression which is a strong limitation of the current study (Matson et al., 2005). Also, only simple measures were used on well-being and no extensive questionnaires as these do not require the ability to read and could be completed by everyone, but this limits information regarding effects. Alternative questionnaires measuring anger that have been adapted for intellectual disabilities are the Northgate State‐Trait Anger Expression Inventory and the Northgate Novaco Anger Scale (Willner et al., 2019). No specific self-report aggression questionnaires for people with intellectual disabilities are currently available (Willner et al., 2019). Finally, the use of self-reports forms a limitation as the reliability can be affected by cognitive capacities and limited verbal comprehension (Taylor, 2002).

CONCLUSION

VR aggression prevention training seems a feasible intervention for individuals with MBID, their relatives, mentors and therapists. VR is a tool that seems promising to learn new strategies and inform the social environment about triggers of aggression. However, this study also showed that adaptations of the current treatment protocol are needed; amongst others the amount of time spent on learning by doing/practicing in VR should be increased, the social network should be involved more intensively, and the number and frequency of training sessions should be personalized. Furthermore, choosing feasible and acceptable outcome measures on anger and aggression for this target group was challenging, and a qualitative or mixed methods approach seems to be the most informative.

Acknowledgments

We thank all participants, mentors and therapists (Annet Hageman, Eline Roodakker, Marjet Ekkel and Grieke Olijve) for investing their time and effort, and providing us with valuable feedback. Further, Catheleine van Driel is acknowledged for her contribution to the qualitative analyses.

Disclosure Statement

The authors have no conflict of interest.

Additional information

Funding

This work was supported by Stichting tot Steun VCVGZ under Grant 265).