The use of virtual reality to reduce mental health stigma among healthcare and non-healthcare students: a systematic review

ABSTRACT

Healthcare professionals, including students, may express stigmatising attitudes towards mental illness. Virtual reality is thought to provide a novel insight into the experiences of individuals with mental health conditions and to reduce stigma. This study aims to systematically review the evidence concerning the use of virtual reality as an educational tool to reduce mental health stigma in healthcare and non-healthcare students. Literature searches were conducted across four electronic databases. Studies were eligible if they targeted healthcare or non-healthcare students, used any form of virtual reality, focused on experiences of mental health conditions, and measured changes in stigma-related outcomes. Fifteen studies, of which eight on healthcare students, were included and synthesised narratively. Both immersive and non-immersive virtual reality technologies were used, and most focused on simulations of mental health symptoms. Different outcomes were measured, including stereotypes, prejudice, and discrimination, and all studies relied on self-report instruments. There is support for using virtual reality to reduce mental health stigma among healthcare students, but not among non-healthcare students. While non-immersive technologies might be as effective as immersive ones, a focus on psychopathology and a lack of educational information appear to increase stigma. Stereotypes and discriminatory intentions were the outcomes most susceptible to change.

KEYWORDS:

• Mental health
• stigma
• healthcare education
• virtual reality

1. Background

Mental health conditions (MHCs) are highly prevalent in Western societies (Eaton et al. 2008; Wittchen et al. 2011). It is estimated, for instance, that 17% of the adult population in England meet the diagnostic criteria for a common MHC such as anxiety, depression, or a combination of both (McManus et al. 2016). In a similar trend, data from the US show that about 20% of the adult population suffers from a diagnosable MHC each year (Substance Abuse and Mental Health Services Administration 2021). As expected, and supported by recent systematic review and meta-analytic evidence, these figures increased at the beginning of the COVID-19 pandemic but over time returned to pre-pandemic levels (Ahmed et al. 2023; Robinson et al. 2022). Critically, MHCs are associated with increased human, social, and economic costs, which are thought to exceed those of physical health conditions such as musculoskeletal disorders, infectious diseases, and cardiovascular illnesses (Vigo, Thornicroft, and Atun 2016). However, among those with MHCs, only a small proportion seek and receive adequate treatment (Thornicroft 2007), amplifying both the individual and societal burden. While reasons for this may be multifactorial, mental health stigma has been identified as a common obstacle to accessing healthcare (Clement et al. 2015; Vistorte et al. 2018).

There is a multitude of studies suggesting that mental health stigma operates across all domains of life as people with MHCs face challenges in securing and maintaining work (Brouwers et al. 2016; Sharac et al. 2010), housing (Browne and Courtney 2007; Peterson et al. 2006), and social relationships (Boardman 2011; Jorm and Oh 2009). Furthermore, people with MHCs experience a lower standard of care, not only for their mental health but also for their physical health (Henderson et al. 2014; Mitchell, Malone, and Doebbeling 2009; Thornicroft, Rose, and Kassam 2007). For example, healthcare professionals tend to express more negative attitudes towards patients with MHCs compared to patients with physical health conditions (Minas et al. 2011). Additionally, healthcare professionals who hold stigmatising attitudes towards patients with MHCs are more likely to believe that these would not adhere to treatment, and in consequence, are less likely to refill their prescriptions or refer them to specialist care (Corrigan et al. 2014). More than that, for patients with MHCs, many healthcare professionals tend to misjudge the presentation of physical symptoms as simple manifestations of patients’ mental health rather than indicators of a separate health problem (Mason and Scior 2004; Shefer et al. 2014), leading to the phenomenon of ‘diagnostic overshadowing’.

In turn, healthcare professionals’ stigma, together with that of society at large, could contribute to the internalisation of negative attitudes and beliefs among people with MHCs (Watson et al. 2006). As a result, those who experience self-stigma are more likely to delay treatment seeking (Clement et al. 2015; Schnyder et al. 2017) and less likely to adhere to medication plans (Kamaradova et al. 2016; Uhlmann et al. 2014), while becoming more preoccupied with their condition (Lin et al. 2016; Yen et al. 2009). These findings might explain, at least to some extent, the worryingly high rates of physical morbidity and mortality seen across MHCs (Thornicroft 2011).

In light of this overwhelming evidence, mental health stigma has become an urgent public health concern, and major institutions around the world are now calling for orchestrated efforts to reduce stigma and discrimination, including that of healthcare professionals (e.g. Thornicroft et al. 2022; WHO 2022). Contact with people with MHCs and education about these conditions are two mechanisms regarded to diminish stigma (Corrigan and Penn 1999), and, as a result, have been frequently used as active ingredients in anti-stigma interventions (e.g, Lien et al. 2021; Martínez-Martínez et al. 2019; Pursehouse 2023). Education-based interventions show positive results in increasing knowledge about MHCs (Yamaguchi, Mino, and Uddin 2011), whereas contact-based interventions are effective in reducing stigmatising attitudes, affective bias, and negative behavioural intentions towards people with MHCs (Maunder and White 2019). However, systematic review and/or meta-analytic evidence indicates that these effects are modest and, in fact, might not be sustainable in the long run (McCullock and Scrivano 2023; Morgan et al. 2018; Yamaguchi et al. 2013). Furthermore, contact with people with MHCs is not always possible, and, paradoxically, in some circumstances, increased contact can heighten the desire for social distance (Jorm and Oh 2009; Rubenking and Bracken 2015). These observations highlight the current limitations of mainstream interventions and tools, and, inevitably, make a case for devising better approaches to tackle mental health stigma.

A novel opportunity to target mental health stigma is offered by virtual reality (VR), a technology that allows one’s cognition to immerse into a digitally generated environment and which is thought to produce similar responses as in real-world settings (Christofi and Michael-Grigoriou 2017; Ma and Zheng 2011; Sanchez-Vives and Slater 2005). Existing scholarship has already established the effectiveness of VR in reducing public stigma towards MHCs (Rodríguez-Rivas et al. 2022), advocating for its potential use in educational, organisational, and healthcare contexts. Furthermore, within healthcare, VR-based technologies are becoming more and more used in the assessment, diagnosis, and treatment of MHCs such as anxiety and depression (Baghaei et al. 2021; Chitale et al. 2022), disordered eating (Clus et al. 2018), psychosis (Schroeder et al. 2022), and post-traumatic stress (Kothgassner et al. 2019), but also more widely in the education of healthcare professionals (Kyaw et al. 2019; Mäkinen et al. 2022; Ruthenbeck and Reynolds 2015).

As students and trainees in healthcare disciplines may express more negative attitudes towards people with MHCs than qualified healthcare professionals (Arvaniti et al. 2009; Oliveira et al. 2020; Thornicroft, Rose, and Mehta 2010), increasing attention has been paid to the role of VR as an educational tool to reduce healthcare students’ mental health stigma (e.g. Lem et al. 2022; Marques et al. 2022; Zare-Bidaki et al. 2022). However, individual studies paint a rather fragmented view, and much variability exists in relation to the characteristics of these studies, the design of VR interventions, and the measures used to assess efficacy, which limits one’s ability to make any firm recommendations for healthcare educators and policymakers, as Hand, Olaiya, and Elmasry (2020) also noted. This gap justifies the need to gain a better, systematic understanding of the current evidence for the use of VR in healthcare students to reduce stigma against people with MHCs and, eventually, optimise the quality of care they receive. Furthermore, compared to healthcare students, non-healthcare students tend to show similar levels of stigma towards people with MHCs (Ruiz et al. 2022; Totic et al. 2012). However, to our knowledge, no systematic review has attempted to compare the use of VR to reduce mental health stigma in healthcare and non-healthcare students.

This study aims to systematically review the available evidence pertaining to the use of VR as an educational tool to reduce mental health stigma among healthcare and non-healthcare students. More specifically, this study will seek to clarify whether VR interventions have a similar effect on healthcare and non-healthcare students, what type of VR interventions have been used and how they were designed and/or delivered, and how the efficacy of VR interventions was measured. To guide the development of eligibility criteria, the theoretical model put forward by Corrigan (2000) has been used to delineate stereotypes (i.e. generalised beliefs about MHCs/people with MHCs), prejudice (i.e. negative affective reactions towards MHCs/people with MHCs), and discrimination (i.e. unjust, hostile behaviours against people with MHCs) as core elements of mental health stigma.

2. Methods

2.1. Registration

This systematic review was pre-registered on PROSPERO in October 2022 (CRD42022368202) and its reporting was guided by the PRISMA statement (Page et al. 2021).

2.2. Eligibility criteria

A set of inclusion and exclusion criteria was developed as shown in Table 1. The population of interest was any group of university-level students studying healthcare, including but not limited to medicine, nursing, midwifery, paramedic science, nutrition, psychology, social work, and non-healthcare subjects. To broaden the scope of the review, studies based on elements or extensions of VR such as augmented reality (AR) and mixed reality (MR) were also given consideration. As such, using a broad definition, we accepted any form of VR intervention, including those using elements or extensions of VR, that focused on simulating the experiences of individuals with mental health conditions. In terms of outcomes of interest, studies had to evaluate changes in any mental health stigma-related outcome, including cognitive (e.g. stereotypes), affective (e.g. prejudice) or behavioural (e.g. discrimination) measures. Studies with or without a comparison group were accepted, and these must have included experimental or quasi-experimental designs. Additionally, to be considered eligible, studies must have been written in English, peer-reviewed, and published in academic journals no earlier than 2000.

Table 1. Eligibility criteria.

Criteria	Inclusion	Exclusion
Population	Any group of university-level students studying healthcare (including psychology and social work) or non-healthcare subjects.	General, non-student populations; Clinical populations.
Intervention	Any form of VR intervention, including those using elements or extensions of VR (e.g. AR, MR), focused on simulating the experiences of individuals with mental health conditions.	VR interventions focused on the assessment, diagnosis, or treatment of mental health conditions; VR interventions focused on simulating experiences of individuals with neurodevelopmental or neurodegenerative conditions (e.g. autism, ADHD, dementia, etc.)
Comparisons	Studies with or without a comparison group are accepted.	Not applicable.
Outcome	Changes in any mental health stigma-related outcome, including cognitive (e.g. stereotypes), affective (e.g. prejudice) or behavioural (e.g. discrimination) measures.	Studies that do not explicitly measure mental health stigma, e.g. studies targeting only knowledge, empathy or related variables, clinical skills, etc.
Study design	Studies including experimental or quasi-experimental designs.	Observational studies; Descriptive studies; Qualitative-only studies; Systematic reviews and meta-analyses.

2.3. Information sources and search strategy

Literature searches were conducted between November and December 2022 across four electronic databases – PubMed, APA PsycINFO, Scopus, and Web of Science. Search terms included virtual reality (“Virtual Reality” OR “VR” OR “Virtual Environment” OR “VR Simulation” OR “VR Character” OR “Augmented Reality” OR “AR” OR “Mixed Reality” OR “MR” OR “Extended Reality” OR “XR”) AND stigma (“Stigma” OR “Stigmat*” OR “Stereotyp*” OR “Prejudice” OR “Discriminat*”) AND mental health (“Mental” OR “Psychological” OR “Psychiatric”). Reference lists of key articles (e.g. past reviews) were also searched to find additional studies. Search results were then transferred into an open-source reference management software for screening and review purposes.

2.4. Study selection

Once search results were collated and duplicates removed, titles and abstracts were screened. The full text of potentially eligible studies was then reviewed against the pre-defined eligibility criteria and the final pool of studies to be included in the systematic review was determined accordingly. A random sample of 20% of studies was cross-checked by a second researcher at both title and abstract, and full-text stages.

2.5. Data extraction. Quality appraisal

Study characteristics (e.g. sample and setting, research design), VR design (e.g. type, equipment, content, duration), and indicators of efficacy (e.g. outcomes, instruments, psychometric properties, findings, and effect sizes, if reported) were extracted from each study, then populated into a standardised table.

The Standard Quality Assessment Criteria for Evaluating Primary Research Papers (Kmet, Cook, and Lee 2004) was applied to appraise the quality of studies included in the review. Specifically, studies were assessed on 14 criteria, including research question, design, participant characteristics, outcome measures, analysis, and results reporting, and, where applicable, scored with ‘–’ (the study failed to meet a given criterion), ‘+’ (the study partially met a given criterion) or ‘++’ (the study fully met a given criterion). A global score for each study was also computed to determine the overall quality of research evidence, with scores closer to 1 indicating a better quality. A second researcher independently rated all eligible studies and any disagreements concerning quality ratings were resolved by consensus.

2.6. Synthesis strategy

Data were synthesised using a narrative approach and in line with best practice recommendations on the conduct of narrative synthesis in systematic reviews (Popay et al. 2006). This was organised around three main themes (1) study characteristics, (2) VR design, and (3) indicators of efficacy.

3. Results

A total of 4271 records were identified. Of these, 997 were removed before screening. 3274 unique records were screened based on their titles and abstract content. After records deemed irrelevant were excluded, the full text of 34 records was assessed for eligibility. In total, 15 records met the eligibility criteria and were included in the systematic review. Figure 1 illustrates the selection process.

Figure 1. PRISMA flow diagram illustrating the study selection process.

The publication years of the 34 records that were screened at full-text and the 15 records included in the review were also analysed to examine the trend of publications over time. The graph in Figure 2 highlights an increase in the number of publications from 2019 onwards, suggesting a growing research focus and interest in VR and mental health stigma.

Figure 2. Graph showing trends by publication year for records screened at full-text (n = 34) and records included in the review (n = 15)

3.1. Quality appraisal

All 15 studies included in the review were rated using the Standard Quality Assessment Criteria for Evaluating Primary Research Papers (Kmet, Cook, and Lee 2004). The results of the quality appraisal can be found in Table 2. One study demonstrated low methodological quality (global score = 0.09) and, consequently, was given less evidentiary weight. Excluding the low-quality study, global scores ranged between 0.57 and 0.89 (M = 0.71, SD = 0.09). Overall, studies in this review demonstrated moderate to high levels of quality.

Table 2. Quality appraisal results.

Study	Criteria														Global score
	1	2	3	4	5	6	7	8	9	10	11	12	13	14
Brown (2010)	++	++	+	++	+	–	–	++	++	++	++	++	++	++	0.79
Brown (2020)	+	+	+	++	+	–	–	++	++	++	++	–	++	++	0.64
Brown et al. (2010)	++	++	+	+	+	–	–	++	++	++	++	–	++	++	0.68
Cangas et al. (2019)	++	+	+	+	++	–	–	++	++	++	++	–	++	++	0.68
Dearing and Steadman (2008)	++	+	+	++	+	–	–	+	++	++	–	++	+	+	0.57
Ferchaud et al. (2020)	++	++	+	+	+	–	–	++	++	++	++	++	++	++	0.75
Kalyanaraman et al. (2010)	++	++	++	+	+	–	–	+	++	++	++	+	++	++	0.71
Lam et al. (2020)	++	–	++	++	N/A	N/A	N/A	++	++	++	++	++	++	+	0.86
Lem et al. (2022)	+	++	++	++	++	–	–	++	+	+	+	+	++	++	0.68
Marques et al. (2022)	++	++	+	+	+	–	–	++	++	++	+	+	++	+	0.64
Mullor et al. (2019)	++	++	+	+	+	–	–	+	++	++	++	+	++	++	0.68
Silva et al. (2017)	+	–	–	–	N/A	N/A	N/A	+	–	–	–	–	–	–	0.09
Stelzmann, Toth, and Schieferdecker (2021)	+	++	+	+	+	–	–	+	++	++	–	++	++	++	0.61
Yuen and Mak (2021)	+	++	++	++	++	–	++	++	++	++	++	++	++	++	0.89
Zare-Bidaki et al. (2022)	++	+	++	++	+	–	–	+	++	++	++	+	++	++	0.71

Notes: 1 = Question described?, 2 = Appropriate study design?, 3 = Appropriate participant selection?, 4 = Participant characteristics described?, 5 = Random allocation?, 6 = Researchers blinded?, 7 = Participants blinded?, 8 = Outcome measures well defined?, 9 = Sample size appropriate?, 10 = Analytic methods well described/justified?, 11 = Estimate of variance reported?, 12 = Controlled for confounding?, 13 = Results reported in detail?, 14 = Conclusion supported by results?. ‘++’ = Yes, ‘+’ = Partial, ‘–’ = No, ‘N/A’ = Not applicable.

3.2. Study characteristics

The characteristics of each included study were summarised in Table 3. Eight studies were based on samples of healthcare students such as medicine, nursing, and psychology (Dearing and Steadman 2008; Kalyanaraman et al. 2010; Lam et al. 2020; Lem et al. 2022; Marques et al. 2022; Mullor et al. 2019; Silva et al. 2017; Zare-Bidaki et al. 2022), two studies were based on mixed samples of healthcare and non-healthcare students (Cangas et al. 2019; Yuen and Mak 2021), and five studies were based on samples of non-healthcare students (Brown 2010, 2020; Brown et al. 2010; Ferchaud et al. 2020; Stelzmann, Toth, and Schieferdecker 2021).

Table 3. Summary of study and VR intervention characteristics.

Authors, Year	Sample, Setting	Research design	VR Type, Equipment, Intervention content	Duration
Brown (2010)	134 non-HC students, US	Parallel RCS with 2 conditions: intervention (simulation + task) vs control (simulation + no task).	Non-immersive, portable CD player. Audio segment simulating auditory hallucinations characteristic of schizophrenia.	16 min
Brown (2020)	244 non-HC students, US	Parallel RCS with 3 conditions: simulation vs filmed contact vs control	Non-immersive, video screen. Audio-visual segment simulating auditory and visual hallucinations characteristic of schizophrenia.	7 min
Brown et al. (2010)	143 non-HC students, US	Parallel RCS with 3 conditions: simulation vs filmed contact vs control	Non-immersive, Equipment NS. Audio segment simulating auditory hallucinations characteristic of schizophrenia.	16 min
Cangas et al. (2019)	530 HC and non-HC students, Spain	Cluster RCS with 2 conditions: intervention (serious video game) vs control (other video games)	Non-immersive, the Stigma-Stop game, other equipment NS. Serious video game – the player learns about and interacts with characters with depression, schizophrenia, bipolar disorder, and panic disorder across 4 mini-games.	NS
Dearing and Steadman (2008)	94 HC students, US	Cluster RCS with 2 conditions: intervention (orientation + simulation) vs control (orientation alone)	Non-immersive, audiotape, tape player, headphones. Audio segment simulating auditory hallucinations characteristic of schizophrenia.	45 min
Ferchaud et al. (2020)	207 non-HC students, US	Parallel RCS with 2 conditions: intervention (playing a serious video game) vs control (watching video footage of the game)	Non-immersive, the Hellblade: Senua’s Sacrifice game, computers. Serious video game – the player follows the story of the main character who is completing quests while experiencing auditory and visual hallucinations.	45 min
Kalyanaraman et al. (2010)	112 HC students, US	Factorial RCS with 4 conditions: empathy alone vs VR simulation alone vs VR simulation + empathy vs control	Immersive, VR goggles (type NS) and earphones, processing unit, laptop, motion-tracking sensors. Experience portraying a visit to the pharmacy from a first-person perspective while being exposed to simulated visual and auditory hallucinations characteristic of schizophrenia.	4–5 min
Lam et al. (2020)	159 HC students, China	Pre-post uncontrolled study	Immersive, VR cardboard headset. Simulation from a first-person perspective of challenges encountered by people with psychosis, anxiety, and mood disorders, including symptom simulation and experiences of stigma in different contexts. The intervention took place after a lecture on mental illness.	5–7 min
Lem et al. (2022)	16 HC students, Japan	Parallel RCS study with 2 conditions: intervention (VR app) vs control (educational video)	Immersive, the Virtual Reality Anti-Stigma (VRAS) app, Android OS smartphones, head-mounted displays (type NS). Multiple scenes including a third-person perspective of discrimination at work, a second-person perspective of a counselling session, and a first-person experience of discrimination at work. Educational, anti-stigma messages were also displayed.	5 min
Marques et al. (2022)	102 HC students, Portugal	Parallel RCS with 2 conditions: intervention (VR simulation) vs control (2D video)	Immersive, Oculus Quest VR headset. Experience portraying a self-avatar completing a cognitive task on a virtual laptop while being exposed to simulated visual and auditory hallucinations characteristic of schizophrenia.	7 min
Mullor et al. (2019)	81 HC students, Spain	Parallel RCS with 4 conditions: serious video game vs direct contact vs lecture vs control	Non-immersive, the Stigma-Stop game, multimedia screen. Same as Cangas et al. (2019).	1h
Silva et al. (2017)	21 HC students, Brazil	Pre-post uncontrolled study	Immersive, workstation, Panasonic RP-BTGS10 headphones, Sony HMZ-T2 3D Glasses coupled with a Microsoft LifeCam Studio webcam. Augmented experience from a first-person perspective in which virtual elements characteristic of schizophrenia, such as simulated visual and auditory hallucinations were added over a real scene.	3 min
Stelzmann, Toth, and Schieferdecker (2021)	114 non-HC students, Germany	Parallel RCS with 3 conditions: VR video vs 2D video vs control	Immersive, Zeiss VR One headset, Samsung Galaxy S8 smartphone. 360-degree VR video (Insta360 Pro) in an outdoor setting in which an actor portraying a young man with schizophrenia presents his experience. The monologue was informed by real-life histories of individuals with schizophrenia.	NS
Yuen and Mak (2021)	206 HC and non-HC students, China	Parallel RCS with 3 conditions: immersive VR animation vs text vs control	Immersive, Oculus Go headset with 360-degree head-tracking. Animated story from a first-person perspective of an individual with mixed anxiety-depressive disorder who experiences stigma from family and colleagues. Educational messages were also included.	10 min
Zare-Bidaki et al. (2022)	144 HC students, Iran	Parallel RCS with 2 conditions: intervention (education + VR simulation) vs control (education + psychiatric ward visit)	Immersive, VR headset (type NS), Android or iOS smartphone. Simulation from a first-person perspective of an individual who experiences delusions and hallucinations characteristic of schizophrenia. The intervention took place after a lecture on psychosis.	4–5 min

Notes: HC = healthcare, non-HC = non-healthcare, RCS = randomised controlled study, VR = virtual reality, NS = not specified.

Two studies included thirty participants or fewer (Lem et al. 2022; Silva et al. 2017), two studies included more than thirty but fewer than a hundred participants (Dearing and Steadman 2008; Mullor et al. 2019) and eleven studies included a hundred participants or more (Brown 2010, 2020; Brown et al. 2010; Cangas et al. 2019; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Lam et al. 2020; Marques et al. 2022; Stelzmann, Toth, and Schieferdecker 2021; Yuen and Mak 2021; Zare-Bidaki et al. 2022). These were conducted in eight different countries – six studies in the US (Brown 2010, 2020; Brown et al. 2010; Dearing and Steadman 2008; Ferchaud et al. 2020; Kalyanaraman et al. 2010), two in Spain (Cangas et al. 2019; Mullor et al. 2019), two in China (Lam et al. 2020; Yuen and Mak 2021), one in Portugal (Marques et al. 2022), one in Germany (Stelzmann, Toth, and Schieferdecker 2021), one in Brazil (Silva et al. 2017), one in Iran (Zare-Bidaki et al. 2022), and one in Japan (Lem et al. 2022).

All studies were experimental or quasi-experimental. Ten studies used parallel randomised controlled designs (Brown 2010, 2020; Brown et al. 2010; Ferchaud et al. 2020; Lem et al. 2022; Marques et al. 2022; Mullor et al. 2019; Stelzmann, Toth, and Schieferdecker 2021; Yuen and Mak 2021; Zare-Bidaki et al. 2022), two studies used cluster randomised controlled designs (Cangas et al. 2019; Dearing and Steadman 2008), two studies used pre–post uncontrolled designs (Lam et al. 2020; Silva et al. 2017), and one study used a factorial randomised controlled design (Kalyanaraman et al. 2010).

3.3. VR design

Eight studies were based on immersive VR technologies such as head-mounted displays (Kalyanaraman et al. 2010; Lam et al. 2020; Lem et al. 2022; Marques et al. 2022; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021; Yuen and Mak 2021; Zare-Bidaki et al. 2022), while the other seven were based on non-immersive technologies such as audio systems, game consoles or computer screens (Brown 2010, 2020; Brown et al. 2010; Cangas et al. 2019; Dearing and Steadman 2008; Ferchaud et al. 2020; Mullor et al. 2019). The duration of each VR intervention ranged from three minutes to one hour.

Ten VR interventions involved a first-person perspective (Brown 2010, 2020; Brown et al. 2010; Dearing and Steadman 2008; Kalyanaraman et al. 2010; Lam et al. 2020; Marques et al. 2022; Silva et al. 2017; Yuen and Mak 2021; Zare-Bidaki et al. 2022), three involved a second-person perspective (Cangas et al. 2019; Mullor et al. 2019; Stelzmann, Toth, and Schieferdecker 2021), and two involved a combination of first-, second-, and/or third-person perspectives (Ferchaud et al. 2020; Lem et al. 2022).

In terms of their content, three studies used serious video games (Cangas et al. 2019; Ferchaud et al. 2020; Mullor et al. 2019), eight used audio and/or visual simulations of mental health symptoms (Brown 2010, 2020; Brown et al. 2010; Dearing and Steadman 2008; Kalyanaraman et al. 2010; Marques et al. 2022; Silva et al. 2017; Zare-Bidaki et al. 2022), three used virtual portrayals of stigmatising experiences in different contexts (Lam et al. 2020; Lem et al. 2022; Yuen and Mak 2021), and one used a 360-degree video of an actor portraying a man with schizophrenia and presenting his experience with the condition (Stelzmann, Toth, and Schieferdecker 2021). The interventions in six studies also included educational elements about mental health (Cangas et al. 2019; Lam et al. 2020; Lem et al. 2022; Mullor et al. 2019; Yuen and Mak 2021; Zare-Bidaki et al. 2022).

Ten interventions focused on experiences of psychosis/schizophrenia (Brown 2010, 2020; Brown et al. 2010; Dearing and Steadman 2008; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Marques et al. 2022; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021; Zare-Bidaki et al. 2022), two focused on depression and/or anxiety (Lem et al. 2022; Yuen and Mak 2021), and three focused on more than one MHC (Cangas et al. 2019; Lam et al. 2020; Mullor et al. 2019).

Of the 15 studies included in the review, eight studies reported the involvement of people with lived experience and other mental health experts in the design of the VR interventions (Brown 2010, 2020; Brown et al. 2010; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Lam et al. 2020; Silva et al. 2017; Zare-Bidaki et al. 2022).

3.4. Indicators of efficacy

All studies used self-report instruments to evaluate the efficacy of VR interventions in reducing mental health stigma. Most studies collected data at two time points – before and after the VR intervention. Three studies also collected follow-up data at 1-week (Brown 2020; Brown et al. 2010; Yuen and Mak 2021). Stereotypes about mental health and/or people with MHCs were measured in ten studies (Brown 2010, 2020; Cangas et al. 2019; Dearing and Steadman 2008; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Lam et al. 2020; Marques et al. 2022; Mullor et al. 2019; Silva et al. 2017). Prejudice against people with MHCs was measured in seven studies (Brown 2010, 2020; Brown et al. 2010; Marques et al. 2022; Mullor et al. 2019; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021). Discriminatory behavioural intentions towards people with MHCs were measured in ten studies (Brown 2010, 2020; Brown et al. 2010; Cangas et al. 2019; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Lem et al. 2022; Marques et al. 2022; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021). Three studies used instruments that measured mental health stigma globally (Lem et al. 2022; Yuen and Mak 2021; Zare-Bidaki et al. 2022). In addition, three studies also measured changes in mental health knowledge (Lam et al. 2020; Marques et al. 2022; Zare-Bidaki et al. 2022).

Eleven studies assessed the psychometric properties of their measurement instruments using internal consistency analyses (Brown 2010, 2020; Brown et al. 2010; Cangas et al. 2019; Dearing and Steadman 2008; Ferchaud et al. 2020; Kalyanaraman et al. 2010; Lam et al. 2020; Mullor et al. 2019; Stelzmann, Toth, and Schieferdecker 2021; Yuen and Mak 2021). Most studies reported acceptable, good, or very good levels of internal consistency for stigma-related measures with Cronbach’s α ranging from .72 to .95. One study reported unacceptable levels of internal consistency for two subscales and, therefore, the authors decided to remove these from further analyses (Mullor et al. 2019). One study also reported the test-retest reliability of the instrument used (Dearing and Steadman 2008), while four studies did not report any psychometric properties for their measurement instruments (Lem et al. 2022; Marques et al. 2022; Silva et al. 2017; Zare-Bidaki et al. 2022).

As shown in Table 4, among the studies measuring stereotypes, five reported significant improvements, in that participants perceived people with MHCs more positively following the VR intervention (Cangas et al. 2019; Dearing and Steadman 2008; Kalyanaraman et al. 2010; Marques et al. 2022; Mullor et al. 2019), two reported significantly more negative attitudes (Brown 2010; Silva et al. 2017), and three reported non-significant changes (Brown 2020; Ferchaud et al. 2020; Lam et al. 2020). In terms of prejudice, one study reported significant decreases in negative emotions towards people with MHCs (Mullor et al. 2019), three reported significant increases (Brown et al. 2010; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021), and three reported non-significant changes (Brown 2010, 2020; Marques et al. 2022). Discriminatory behavioural intentions decreased significantly in four studies after the VR intervention (Cangas et al. 2019; Ferchaud et al. 2020; Mullor et al. 2019; Silva et al. 2017), three studies reported significant increases in discriminatory intentions (Brown 2010; Brown et al. 2010; Kalyanaraman et al. 2010), two studies reported non-significant changes (Brown 2020; Lem et al. 2022), and one reported mixed findings (Stelzmann, Toth, and Schieferdecker 2021). Of the studies measuring mental health stigma globally, two reported significant decreases after the VR intervention (Yuen and Mak 2021; Zare-Bidaki et al. 2022) and one reported non-significant decreases (Lem et al. 2022).

Table 4. Summary of study measures, instruments, and main findings.

Authors, Year	Outcome(s)	Instrument(s)	Psychometric properties	Main findings
Brown (2010)	Stereotypes Prejudice Discrimination	Modified AQ with 4 subscales: forcing treatment, fear, negative emotions, help	Good to very good internal consistency (α = .81–.94)	Support for forced treatment sig. ↑ after intervention (d = 0.22). No changes in fear and negative emotions. Helping intentions sig. ↓ after intervention (d = 0.19). No differences between conditions.
Brown (2020)	Stereotypes Prejudice Discrimination	DS, AS, SDS	Acceptable to good consistency for DS (α = .72–.82). Good to very good internal consistency for AS (α = .87–.93). Acceptable to good internal consistency for SDS (α = .78–.87)	Dangerousness sig. ↓ after filmed contact only (d = 0.88). Negative emotions sig. ↓ after filmed contact (d = 1.34) and control (d = 0.46), but not after simulation. Social distance sig. ↓ after filmed contact only (d = 1.28). The effect maintained at 1-week FU.
Brown et al. (2010)	Discrimination Prejudice	SDS, AS	Acceptable internal consistency for SDS (α =  .75). Good internal consistency for AS (α = .81)	Desire for social distance sig. ↑ after simulation (d = −0.90) and sig. ↓ after filmed contact (d = 0.71). Negative emotions sig. ↑ after simulation (d = −0.49) and sig. ↓ after filmed contact (d = 1.07). Social distance and negative emotions sig. ↓ at 1-week FU after simulation. No changes in the control group.
Cangas et al. (2019)	Discrimination Stereotypes	Spanish QSAS with 2 subscales: social distance, stereotypes	Very good overall internal reliability (α = .95)	Social distance (d = 0.66) and stereotypes (d = 0.15) sig. ↓ after the serious video game. No changes in social distance and stereotypes in the control group.
Dearing and Steadman (2008)	Stereotypes	MCRS	Good internal consistency (α = .87) and test-retest reliability (r = .84)	Positive attitudes sig. ↑ after both intervention and control. Larger differences were noticed in the intervention group.
Ferchaud et al. (2020)	Stereotypes Discrimination	Adapted stereotypes scale, Adapted desire for social distance scale	Good internal consistency (αStereotypes = .82, respectively αSocial distance = .89)	Stereotypes ↓ n.s. after playing the serious video game (compared to watching). Desire for social distance sig. ↓ after playing the serious video game.
Kalyanaraman et al. (2010)	Stereotypes Discrimination	Adapted scale measuring attitudes toward people with AIDS, Adapted SDS	Acceptable internal consistency (αAttitudes = .75, respectively αSocial Distance = .78)	Positive attitudes sig. ↑ after VR simulation + empathy (ω^2 = 0.11) compared to empathy alone and control, but not VR simulation alone. Social distance sig. ↑ after VR simulation alone compared to all other 3 conditions (ω^2= 0.10).
Lam et al. (2020)	Stereotypes Knowledge	Chinese CAMI, Chinese MHKQ	Acceptable to good internal consistency for CAMI (α = .72–.81). Questionable to acceptable internal consistency for MHKQ (α = .62–.73)	Positive attitudes ↑ n.s. after the VR simulation with a sig. ↑ among participants who had previous contact with mental illness. Knowledge about mental health sig. ↑ after the VR simulation.
Lem et al. (2022)	Global stigma Discrimination	DSS, ASD	Not reported	Depression stigma ↓ n.s. after both the VR app and educational video. Social distance ↓ n.s. after both the VR app and educational video.
Marques et al. (2022)	Knowledge Stereotypes Prejudice	Portuguese MAKS, APSQ, EFPSSQ	Not reported	Mental health knowledge sig. ↑ after the VR simulation. Positive attitudes sig. ↑ after both the VR simulation (d = 0.48) and control (d = 0.29). Empathic feelings sig. ↑ in the control group (d = 0.28) and n.s. in the VR simulation group.
Mullor et al. (2019)	Stereotypes Prejudice Discrimination	Spanish AQ with 9 subscales: responsibility, pity, anger, fear, help, dangerousness, coercion, segregation, avoidance	Good overall internal consistency (αBefore = .83, αAfter = .82). Responsibility (α = .23–.30) and pity (α = .24–.20) showed unacceptable consistency and were removed from analyses.	Anger, dangerousness, fear, segregation, and avoidance sig. ↓ in all 3 intervention groups but not in the control group. Coercive attitudes sig. ↓ after the serious video game only. Helping intentions sig. ↑ after the serious video game and direct contact only.
Silva et al. (2017)	Stereotypes Prejudice Discrimination	AQ with 9 subscales	Not reported	Overall stigma sig. ↑ after the intervention. Pity, fear, and segregation sig. ↑ after the intervention. Helping intentions sig. ↑ sig. after the intervention.
Stelzmann, Toth, and Schieferdecker (2021)	Prejudice Discrimination	Adapted intergroup anxiety scale, Adapted social proximity scale, Adapted empathy scale, Benevolence subscale of CAMI	Intergroup anxiety and social proximity scales showed good internal consistency (ω = .85, respectively ω = .86). No other psychometric properties were reported.	Anxiety sig. ↑ and social proximity, empathy, and benevolence sig. ↓ after the VR video compared to the 2D video. Social proximity sig. ↓ after the 2D video compared to control. Social proximity sig. ↑ after the VR video compared to control when participants rated the actor more positively.
Yuen and Mak (2021)	Global stigma	PSAS with 2 subscales: public stigma, personal advocacy towards accepting people with mental illness	Very good overall internal consistency (α = .93). Very good internal consistency for the public stigma subscale (α = .92). Good internal consistency for the personal advocacy subscale (α = .87).	Public stigma sig. ↓ after the VR animation and at 1-week FU compared to pre-scores (${\eta }_p^2$= 0.165). Public stigma sig. ↓ after both the VR animation and text compared to control. The effect maintained at 1-week FU (${\eta }_p^2$= 0.05). No differences between VR animation and text.
Zare-Bidaki et al. (2022)	Global stigma Knowledge	Persian MISS, Educational Achievement Test	Not reported	Mental health stigma sig. ↓ after the VR simulation compared to control (effect size = 0.274). Mental health knowledge sig. ↑ after the VR simulation compared to control (effect size = 0.746).

Notes: AQ = Stigma Attribution Questionnaire, DS = Dangerousness Scale, AS = Affect Scale, SDS = Social Distance Scale, QSAS = Questionnaire on Students’ Attitudes toward Schizophrenia, MCRS = Medical Condition Regard Scale, CAMI = Community Attitudes toward Mental Illness, MHKQ = Mental Health Knowledge Questionnaire, DSS = Depression Stigma Scale, ASD = Attitudinal Social Distance Scale, MAKS = Mental Health Knowledge Schedule, APSQ = Attitudes toward People with Schizophrenia Questionnaire, EFPSSQ = Empathic Feelings for People Suffering from Schizophrenia Questionnaire, PSAS = Public Stigma and Acceptance Scale, MISS = Mental Illness Stigma Scale by World Psychiatric Association. sig. = significant/significantly, n.s. = non-significant/non-significantly, ↑ = increase/improvement, ↓ = decrease, FU = follow-up.

Effect sizes were estimated using Cohen’s d, partial η², or ω² and the magnitude of intervention effects varied between small and large. Seven studies did not report any effect size (Dearing and Steadman 2008; Ferchaud et al. 2020; Lam et al. 2020; Lem et al. 2022; Mullor et al. 2019; Silva et al. 2017; Stelzmann, Toth, and Schieferdecker 2021) and one study did not specify the type of measure used to estimate the effect size (Zare-Bidaki et al. 2022).

4. Discussion

4.1. Summary of findings

This study aimed to systematically review the evidence concerning the use of VR as an educational tool to reduce mental health stigma in healthcare and non-healthcare students. The 15 studies included had moderate to high levels of quality with one study rated as low. The studies encompassed diverse samples of healthcare and non-healthcare students from various countries, with varying sample sizes and immersive/non-immersive VR technologies utilised. The VR interventions explored a range of perspectives, content types, and targeted MHCs, with some interventions incorporating educational elements.

The literature reviewed here generally suggests that VR interventions have the potential to reduce one or more aspects of mental health stigma among healthcare students. Embodiment, or the sense of perceiving the world from someone else’s bodily perspective, is a central feature of many VR technologies and is also thought to act as a stigma reduction mechanism (Ahn, Le, and Bailenson 2013). This aligns with a recent meta-analysis which found that VR-based interventions are effective at reducing mental health stigma in the general population (Rodríguez-Rivas et al. 2022). However, this finding was not replicated by studies among non-healthcare students, in which VR interventions were more likely to amplify stigma towards MHCs.

When discussing the divergent findings between the two student populations, it could be argued that, as healthcare students have an interest in caring for patients, including patients affected by MHCs, and might already have had some training and/or practical experience in mental health care, they endorse less stigmatising attitudes and behaviours compared to their non-healthcare counterparts, and, as a result, they would be less susceptible to the potential negative effects of a VR intervention on mental health stigma. However, while the former part may be true, the existing body of literature appears to refute the latter as healthcare and non-healthcare students tend to stigmatise people with MHCs to a largely similar extent (Ruiz et al. 2022; Totic et al. 2012). Thus, mental health stigma is more prevalent and stable than popularly believed and, even though differences might exist at the individual level, these do not necessarily translate to population-level outcomes, warranting the need for effective stigma reduction interventions in the education of all students.

A more plausible explanation is offered by the evident differences in the content and design of the various VR interventions. For instance, Brown (2010) showed that a simulation of auditory hallucinations decreased participants’ willingness to help people with MHCs, and, at the same time, increased the endorsement of forced mental health treatment. Likewise, Kalyanaraman et al. (2010) found that participants exposed to simulated audio-visual hallucinations in a VR environment expressed a greater desire for social distance from individuals with schizophrenia. Yet, in the study conducted by Zare-Bidaki et al. (2022), although a VR simulation of hallucinations was also used, participants showed substantially lower levels of mental health stigma after taking part in the intervention. Nevertheless, in contrast with the first two studies, participants in this study were required to attend a lecture on psychosis beforehand, which might have increased their awareness and understanding of MHCs, and, ultimately, contributed to destigmatisation.

Using a different approach but eliciting the same positive results, Yuen and Mak (2021) developed a VR intervention that focused on the daily challenges of people with MHCs. As with prior work (e.g. Mann and Himelein 2008; Norman et al. 2017), this suggests that replacing a symptom-centred view with a person-centred one, in which the lives of these people are put into a wider, more humane context and emphasis is placed on themes of resilience and recovery, might prove more beneficial in reducing stigma. This is consistent with earlier research which found that exposure to negative and violent media portrayals of people with MHCs increases stigmatisation, while positive, sympathetic portrayals result in more favourable attitudes (Rubenking and Bracken 2015). In turn, as Ando et al. (2011) also reasoned, the mere exposure to simulated symptoms of MHCs may be, in fact, counterproductive – especially if no contextual information or educational content is provided. Therefore, it should not be unexpected that students with no or little understanding of MHCs would express more stigmatising attitudes immediately after listening to a distressing voice-hearing simulation.

Another interesting point is that VR interventions based on non-immersive technologies showed results comparable to those based on immersive technologies, suggesting that embodiment may be as effective as other stigma reduction strategies (e.g. virtual contact) as long as they induce the same degree of transportation into the character’s world. Transportation refers to a mental process whereby one becomes cognitively and affectively immersed in a fictional or non-fictional narrative (Green and Brock 2000) and has been associated with changes in stigma-related attitudes (Caputo and Rouner 2011; Ferchaud et al. 2020; Oliver et al. 2012). Therefore, although VR technologies such as head-mounted displays can create a physical sense of immersion (Chessa et al. 2019; Shu et al. 2019), this alone may not be sufficient to enable a greater degree of transportation. What is clear, however, is that interventions tend to be more successful when multiple stigma reduction strategies are employed, echoing findings from other reviews (e.g. Bielenberg et al. 2021; Doley et al. 2017).

Then, to assess the true efficacy of VR interventions, the selection of suitable outcomes is fundamental as these must correspond with the intended mechanisms of change (Coster 2013). In this case, most studies looked at changes in discrete elements of mental health stigma, recognising, therefore, the multifaceted nature of the construct which may be influenced differently by different components of an intervention (Morgan et al. 2018; Ungar, Knaak, and Szeto 2016). Indeed, stereotypes and discriminatory intentions were more likely to change than prejudice as a result of the interventions. As Na et al. (2022) explained, and consistent with their findings, stigma reduction interventions may be more effective at reducing cognitive but not affective elements of mental health stigma. This can be problematic since affective responses are arguably more predictive of actual discriminatory behaviour than intentions are (Corrigan et al. 2003).

4.2. Limitations of the included studies

Although the research reviewed here was rated as having at least moderate quality, it is essential to acknowledge some methodological limitations that could have affected study findings and their interpretations. Firstly, studies in this review relied heavily on self-report instruments to assess changes in mental health stigma, leaving room for social desirability bias. While self-report measures are able to capture explicit aspects of stigma, that is not the case for affective reactions or actual behaviour toward stigmatised groups, which may, in fact, be inconsistent with explicitly expressed attitudes or intentions (Corrigan et al. 2003; Kopera et al. 2015). Secondly, in an attempt to report potential confounders, the vast majority of studies tended to solely focus on demographic characteristics, such as age and gender. However, less consideration was given to more contextually salient factors, including previous contact with people with MHCs, personal history of MHCs, or mental health literacy, which had been found to influence mental health stigma (Eack and Newhill 2008; Moreira, Oura, and Santos 2021; Rüsch et al. 2011). Thirdly, almost none of the studies included in this review reported the blinding status of their participants, which is a prevalent concern among randomised studies of psychological and/or behavioural interventions (Juul et al. 2021).

4.3. Strengths and limitations of the review process

This systematic review presents several strengths. Firstly, the search strategy and eligibility criteria were pre-registered in a publicly available database. This improves the transparency of the review process and allows other researchers to easily replicate the literature search. Secondly, both specialised and general databases were searched, which potentially increased the probability of retrieving relevant studies. Thirdly, to enhance accuracy and objectivity, the screening and quality assessment results were cross-checked by external researchers. Lastly, for consistency, as well as replicability purposes, the structure and content of this systematic review closely followed the well-established PRISMA reporting standards.

Some limitations of the review process should also be noted. Firstly, only English publications were taken into account, potentially missing research written in other languages. Then, as the literature search was limited to peer-reviewed publications, any other type of work was automatically excluded from the review. Publication bias represents a particular issue that readers should be aware of when interpreting the findings of this systematic review. As a third and final limitation, studies varied substantially in the way they measured mental health stigma, but also in the type, content, and length of VR interventions, thus meta-analysis was not a suitable method for synthesising the data quantitively. In consequence, the evidence behind the effectiveness of VR interventions to reduce mental health stigma in healthcare and non-healthcare students should be treated with caution until more research efforts addressing existing limitations are made and a pooled statistic can be calculated.

4.4. Implications. Conclusion

The findings of this systematic review yield a number of preliminary but important implications for healthcare education research and practice. As advancements in VR technologies continue to be made, these become more and more accessible, and, as a result, find extensive applications in the real world. Healthcare is no exception with VR technologies being increasingly used in the education of healthcare professionals (Kyaw et al. 2019; Mäkinen et al. 2022; Ruthenbeck and Reynolds 2015). Beyond the teaching of clinical skills, the possibility offered by VR technologies to easily adjust aspects of reality and induce a sense of presence has started to be explored even more in recent years, which has led to the use of VR as an educational tool to tackle psychosocial aspects of healthcare such as mental health stigma (Riches et al. 2022). In these interventions, the healthcare student has the opportunity to watch, interact with, or even embody characters with MHCs and understand their symptoms, but also to obtain insight into their daily challenges in different scenarios.

As the current body of knowledge appears to suggest, VR interventions can reduce one or more aspects of mental health stigma, supporting its use in the education of healthcare students. Future work is, however, needed to evaluate the comparative effectiveness of immersive and non-immersive VR technologies and to better understand the underlying mechanisms of stigma change. At the same time, even though existing studies demonstrate the short-term efficacy of VR interventions in reducing mental health stigma, more studies should investigate whether these effects are maintained in the long run by collecting longitudinal, follow-up data.

Nonetheless, it is unquestionable that, in order to enhance their success, healthcare educators should not use VR interventions as a one-off activity but rather as part of a wider anti-stigma programme, in which interventions such as education and contact are also being offered. Additionally, the involvement of individuals with lived experience and their families/carers in the design of VR interventions targeted at reducing mental health stigma should be a priority as they can ensure that the virtually generated scenarios are reflective of their actual experiences and are not mere stereotypical or exaggerated portrayals of mental illness.

Furthermore, mental health stigma is believed to be condition-specific (Carrara et al. 2021; Krendl and Freeman 2019), thus future VR interventions may benefit more from focusing on a specified MHC and evaluating stigma changes concerning that condition. On that note, while much work already exists on psychosis/schizophrenia, little attention has been paid to other MHCs such as substance abuse, disordered eating, or bipolar disorder, which are often characterised by high levels of physical comorbidity and poor clinical outcomes (Keaney et al. 2011; Olguin et al. 2017; Young and Grunze 2013).

Lastly, as there was great variation in reporting the design and development of VR interventions, efforts should be made to improve reporting consistency across studies. In that sense, standardised guidelines (e.g. Duncan et al. 2020; Möhler, Köpke, and Meyer 2015) can be used to describe transparently and consistently the key elements and stages in the process of designing VR interventions, maximising the possibility of making fair between-study comparisons and carrying out replications.

All in all, while the current body of evidence has some methodological drawbacks, there should be optimism that, as VR technologies advance, more and better research outputs will be generated, further strengthening the state of knowledge around the novel role of VR in tackling mental health stigma both within and outside of healthcare education.

Acknowledgements

We thank Fatima Ozeto (University of Surrey) and Ciprian Ciobanu (Birkbeck, University of London) for assisting with this systematic review.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data availability is not applicable to this article as no new data were created or analysed in this study.