Different from previous attempts, the market success of VR head-mounted displays and 3D controllers that started in the early 2000s has not been stopped short, yet. With pricing and capabilities making VR technology more accessible than ever and sales numbers growing rapidly (Angelov et al., 2020), the need for authoring platforms of VR experiences has grown as well (Ashtari et al., 2020).

The developments towards authoring tools started in the early nineties with technical libraries and toolkits. Scene graph libraries such as OpenGL Performer (Rohlf and Helman, 1994), Open Inventor (Strauss and Carey, 1992), and OpenSceneGraph (Wang and Qian, 2010) were created to realize visual simulations, virtual reality environments and other real-time 3D graphics applications. But oftentimes, most of these libraries focused on performance over ease of programmability. Also, developers often found it difficult to use these general-purpose libraries for specific problems (Bethel et al., 1999). Either these libraries were extended (Hesina et al., 1999), or additional packages and frameworks were developed upon these toolkits that could be used to bridge this gap (Kelso et al., 2002; Pavlik and Vance, 2012).

Further, many platforms supported networking and enabled the development of multi-user virtual environments (Carlsson and Hagsand, 1993; Greenhalgh and Benford, 1995; Allard et al., 2002). Latoschik and Tramberend (2011) created the open-source research platform Simulator X that is targeted at developing real-time interactive systems. Fischbach et al. (2017), for example, demonstrated how this platform can be used to increase the software quality of real-time interactive systems. Efforts like this made multiple developers and applications work in a common environment and communicate with each other without exposing the complex internal architectures. Frameworks like Avocado/Avango (Tramberend, 1999; Kuck et al., 2008), MR Toolkit (Shaw et al., 1993), or VR Juggler (Bierbaum et al., 2001) provided high-level APIs to mask the system level architecture from the author. Also, the generality of the data-flow architecture used in these platforms enabled an easy exchange of data across modules, and their users were empowered to create a wide variety of virtual environments (Figueroa et al., 2002; Allard et al., 2004, 2010, 2005).

These libraries and frameworks focused on software architectures and were thus to be used by programmers. A user without profound knowledge in software development would find it overwhelming to understand their respective workflows. The early developments, starting from the libraries and toolkits, slowly evolved into high fidelity authoring platforms over time. With the current advancements in technology and with the wide acceptance of VR, researchers and developers are innovating more sophisticated and effective solutions that can be used to produce more immersive and realistic content. A-frame ¹ is a relatively new open-source framework from Mozilla to generate WebVR content (Neelakantam and Pant, 2017). Creating a VR experience is cost-effective in A-frame, but the initial learning curve can be an insurmountable obstacle to novice users (Santos and Cardoso, 2019). Many commercial and research solutions are addressing the design challenges faced by educators. CoSpaces ² , InstaVR ³ , or WondaVR ⁴ are examples for commercial platforms primarily targeting the education sector. Among these, we found CoSpaces to be an easy-to-use, beginner-friendly tool, that allows the users to create simple VR environments using a drag and drop facility and allows to make the models interactive by employing a Scratch-like visual programming environment (Andone and Frydenberg, 2019). FlowMatic (Zhang and Oney, 2020) is a VR experience editor that uses a different approach to the design process by introducing a concept called immersive authoring. Most of the authoring platforms use 2D software to design the VR experience which takes off the freedom of using spatial information into consideration. FlowMatic makes use of VR, to create VR experiences taking advantage of the 3D spatial interaction. However, FlowMatic focuses only on making the newly created VR environment interactive using a visual programming tool. The above mentioned tools provide ways to facilitate the creation of VR experiences. However, there are some features that they are missing. Especially in terms of healthcare applications there could be the need for additional features such as, control over set stimuli and thus the experience, the ability to monitor the state of the end-user, or an easy creation process for virtual environments. These are also ideas that are part of our vision for a software authoring platform.

Understanding a user and their perception of technology plays an important role in the user-centered software development process. A user acceptance study is a critical milestone to find out if users are convinced by the implementations, to see if their expectations are met, and to make it clear to developers and designers if the system is usable for the end user. It is generally carried out before the final release of the software system (Leung and Wong, 1997). This seems to be an ideal procedure in a user-centered approach where the design process happens for the user. However, there is a shift towards participatory design approaches where users are a part of the design and development process from the very beginning till the release of the platform (Sanders, 2002).

There is a vast literature that addresses technology acceptance studies. In the area of VR and healthcare, information from various acceptance studies are available. However, these studies tend to focus on specific applications that have already been completed (Snoswell and Snoswell, 2019; da Costa and de Carvalho, 2004; Garcia-Palacios et al., 2007) or VR-acceptance overall (Syed-Abdul et al., 2019). Some studies also set in at an earlier stage and used participatory approaches. In a study by Karaosmanoglu et al. (2021) healthcare professionals provided feedback via interviews before the development of a VR exergame to enhance physical activities by people suffering from dementia. In a study by Hilton et al. (2011) professionals’ feedback was gathered via questionnaires and focus groups to inform the development of a system for stroke rehabilitation. Such studies are very valuable because they ensure that applications meet the needs of users. Besides, they can serve as a guideline for future research in the fields of interest and are a stepping stone for any researchers who further want to explore a similar area with a similar user base. However, their conclusions are to some point limited to specific applications and user groups. Besides, while health professionals are consulted for the development of these applications, the ultimate focus is on their use by patients. To our knowledge, acceptance studies that focus on the needs of professionals in the healthcare domain and their use of VR applications have not been published so far.

One of the most used theoretical frameworks to assess an individual’s acceptance of any technical applications or devices is the Technology Acceptance Model (TAM), widely categorized under the umbrella of Information Systems (Davis, 1989; Lee et al., 2003). It assesses how different features of a system affect the perception and attitude of a user toward using the actual system (Balog and Pribeanu, 2009). The original TAM proposed two major variables to understand user acceptance, namely, Perceived Usefulness and Perceived ease of Use. Perceived Usefulness is defined as to what level a person thinks using a system would improve his or her work performance. Perceived ease of Use, is defined as the degree to which a user thinks using a system is effortless (Davis, 1989). Over time, the TAM has gone through major revisions (Lee et al., 2003). Even the original authors had extended it to TAM2, adding more constructs to evaluate user acceptance (Venkatesh and Davis, 2000). Researchers have also combined other frameworks and added their questions and constructs considering the existing TAM as a base, to understand the users’ view better (Lee et al., 2019; Fussell and Truong, 2021).

Ethical aspects play an important role for the acceptance of software. Numerous articles have already investigated ethical aspects for VR use in general (e.g., Kuntze et al., 2002), in health-impaired individuals in general (e.g., Kellmeyer et al., 2019), or in specific health domains (e.g., Lewis and Griffin, 1997). The focus of the majority of papers has been on ethical aspects of virtual reality in psychology/psychotherapy (e.g., Yellowlees et al., 2012; Kellmeyer, 2018; Marloth et al., 2020) and in rehabilitation (Lewis and Griffin, 1997; Kellmeyer, 2018). The identified aspects of ethical relevance are manifold. In addition to aspects that concern the specifics of VR, differential ethical challenges arise depending on the health care setting (e.g., Lewis and Griffin, 1997) and patient groups (e.g., Kellmeyer et al., 2019). These papers are mainly theoretical. In contrast, empirical research on health professionals’ ethical evaluation of VR is scarce. A general assessment is desirable regarding ethically relevant principles, such as respect for autonomy or justice (e.g., Beauchamp and Childress, 2019), and how they are positively or negatively affected by VR from the perspectives of employees in health care areas. Positive and negative expectations need to be investigated for a wide range of health care areas, such as psychotherapy and palliative care. This will help to identify critical areas of VR usage, to get an impression of the awareness for ethical issues in the professional group and to assess the acceptance of the platform based on ethical criteria. This is particularly important since VR usage in healthcare is on the rise.

In the introduction, we described the target of understanding the prerequisites for healthcare professionals to autonomously operate and create VR applications. Literature on this question is missing to date. While there are regular acceptance studies for various soft- and hardware products, until now they have never focused on healthcare professionals’ views on VR. Therefore, we address an important gap. In order to proceed as systematically as possible, our study is based on two pillars: the Technology Acceptance Model (TAM) and the envisioned software platform. Thus, we have concrete categories available as well as novel ideas to be presented to potential users. In this way, people from the healthcare sector are encouraged to share their opinions and suggestions regarding our proposals.

First, we had to solve the problem of measuring the acceptance toward a software platform that does not exist yet. Moreover, we had to assume that for some of the potential respondents, the topic of VR is little or not known at all. Therefore, the subject of our evaluation was a mockup trailer, which is supposed to bring our ideas closer to the potential users from the healthcare sector. It should give them an idea of what using and creating VR applications might look like in their everyday work. The use of a trailer and surveying at an early stage also have the advantage that the results are less dependent on details. Instead, at a more general level, the survey asks what the professional group’s needs are when using VR and how they would envision using it professionally.

First, the trailer shows examples of VR applications in the healthcare sector, i.e., a virtual knee rehabilitation exercise and an example for exposure therapy targeting pyrophobic patients. Moreover, the trailer depicts potential features of the envisioned software platform. The first one would be a supervisor monitor that allows the supervisor of a healthcare application to keep track of the current state of the patient and adapt the virtual environment accordingly. Next, the trailer shows how users of the software platform can create their own VR applications by choosing from standard templates of virtual environments. Those templates are then adapted by creating virtual objects and by altering the sequence of events and interactions. Prototypical user interfaces and example interactions are shown for these operations. Another idea that is depicted in the trailer is the possibility to scan people via the smartphone and thus create custom avatars of a patient as described in Wenninger et al. (2020). So with the fire-exposure therapy and the knee rehabilitation exercise we chose two fundamentally different areas of application, which can nevertheless be presented in a relatively short duration and in an easily understandable way. However, both concepts are transferable to other areas and are thus intended to stimulate further ideas. By showing possible features like creating an avatar, we wanted to show viewers possibilities of VR that they might not have thought of on their own. Combined with expertise in their respective fields, we hope that participants will be stimulated to think about how such opportunities might be applied in their field. Hence, the mockup trailer introduces different facets of VR healthcare applications and gives the viewer a quick and concise impression of it.

Such a mockup trailer has previously been used to give study participants an impression of a new technology (e.g., Haugstvedt and Krogstie, 2012). The biggest advantage of this method is that it provides a good overview that conveniently can be distributed to a large group of people, accordingly supporting gathering larger samples than individual presentations could. Figure 1 shows the screenshots of above mentioned features from the mockup trailer. The trailer is available in an English and a German version. The English version of this trailer can be viewed at https://www.youtube.com/watch?v=HWXp6ke93_w.

For the actual study, we adopted the TAM that we already introduced in the related work section (Davis, 1989; Lee et al., 2003). From this model, we adapted categories and corresponding questions as the basis for our study. These three categories are the perceived usefulness of the platform, the perceived ease of use, and the user’s attitude toward the platform. We adopted Likert-type questions from the TAM model that capture these categories. Furthermore, we added additional open-ended questions to the categories in order to gather valuable feedback on the prerequisites for the acceptance or rejection of the software platform and VR in general. Moreover, we wanted to get specific feedback on the idea of a supervisor monitor, which allows to surveil and control a VR session. In an additional question, we wanted to find out what kind of support potential users are hoping for when creating their own VR application. Then, questions were asked about ethical aspects of VR use. One question was on overall positive expectations regarding the effect of VR on people and society. Ultimately, expectations and concerns regarding the use of VR in different health areas were inquired.

That means in summary, six questions were asked about the perceived usefulness, four questions were asked about the perceived ease of use, four questions were asked about the attitude toward using the envisioned software platform, two questions were asked about a supervisor monitor, one question was asked about assistance, and three questions were asked on ethical topics. The final survey also included a demographics section. An overview of the categories and the asked questions can be found in Table 1.

The survey was conducted using LimeSurvey. It is an online tool that allows users to create their own surveys from a wide variety of question types and edit them at will. Before the actual survey started, participants confirmed their voluntary participation and the use of their anonymized data for research purposes. With the help of LimeSurvey, we were also able to integrate the trailer directly into the survey. That means participants first watched the trailer and then answered the questions right after. The tool was hosted on a university server to protect the corresponding data. Closed-ended questions were answered on a 7-point Likert scale. Survey participants could skip questions or stop answering at any time if they wanted. Before the actual survey started, we asked the participants for permission to store their data anonymously. The language of the survey was German. The link to the survey was distributed to potential users via email, either directly or via mail distribution lists. Each invitation email was also accompanied by a request to forward the survey to colleagues. The link was also shared in Facebook groups to which only persons of the respective professional groups had access. In addition, a professional association’s website invited participation in the survey. In this way, we ensured that we surveyed a broad user group. At the same time, we were able to comply with pandemic containment regulations, which is of course a priority when working together with people from the healthcare sector.

The analysis of the quantitative data was done using SPSS. For the qualitative analysis, we transferred all the answers to MAXQDA ⁸ , which is a software that helps with transcribing, ordering, and analyzing qualitative data. In this software, an entry for each subject was created which included all their respective statements. Subsequently, we developed a code system according to a methodology following Mayring (2015). In an iterative process, we analyzed the statements and grouped them by content. This created categories, which we also grouped again until a hierarchical system of categories and statements emerged. The coding was checked a second time after some weeks to uncover inconclusive categorizations. With this procedure, qualitative data could be converted into quantitative data. The frequency with which certain topics were mentioned can serve as one indicator of their importance.

A total of 102 participants took part in the survey. In order to ensure the voluntary nature of the survey, it was possible not to answer each question. This resulted in different numbers of cases for the various questions. The average age of the participants was 42.4 years and varied between 24 and 73. Of the participants, 58 were female, 42 were male, and 2 did not provide this information. Just under half of the participants in the overall sample were working in psychotherapy, 24 were working in physical therapy, and 11 were working in training and teaching. The rest of the participants were distributed among research, development, consulting, prevention, and other discrete medical fields. Another important piece of information to us was the extent to which participants were using technology to support their daily activities. Forty participants had prior experience with VR. Of these forty participants, 16 had experience with VR applications that were related to medical applications such as physical therapy, exposure therapy, various simulators, or anatomy studies. The rest of the participants that had prior VR experience had seen demos during exhibitions, worked on research projects, or just experienced common entertainment applications. Merely eight of the participants use VR frequently and only three of them for their work. In addition, 58 participants regularly use computers to support their work. Due to the voluntary nature of the study and also because the ethical aspects were addressed at the end of the survey, up to 54 valid answers were provided for the ethical questions.

The envisioned software platform is intended to serve as a tool that people can use to create VR environments without major hurdles. From the demographic data we collected, we found that only three of the participants use VR for their work. For most of the other participants, VR is new territory. Especially here it is important to see if the potential users even consider the concept of the envisioned software platform as useful for their own work. For this exact purpose, the TAM offers questions about the perceived usefulness of a product. Thus, participants were asked whether the envisioned software platform could improve their work effectiveness, enhance their work performance, increase their work productivity, and whether it could prove useful for their work.

An exact overview of the distribution of responses in this category can be found in Table 2. On a Likert scale of 1−7, the mean score in this category for 64 participants was 4.4, indicating that they regard the envisioned software platform as a potentially useful tool for their work (t (63) = 2.00, p = 0.050).

Next, we wanted to find out if potential users can imagine that the envisioned software platform could be useable for them. Therefore, we asked two questions from the TAM that focus on the perceived ease of use of a product. Here we asked how easy the potential users thought it was to use the envisioned software platform and how easy they imagined the learning process of the platform.

An overview of the responses to this category can be found in Table 3. On a Likert scale of 1-7, the mean of responses for all questions about the perceived ease of use was 5.30 (t (68) = 9.86, p < 0.001). When asked if potential users would find it easy to learn how to use the envisioned software platform, 54 of 69 people responded with a 5 (somewhat agree) or higher. This indicates that users tend to be confident that they can master the use of a software platform that allows them to create their own healthcare applications.

Next we wanted to know whether the respondents would actually use the platform to support their work - in essence, we wanted to find out about the attitudes toward using the platform. To understand this, two questions were asked. With one, we wanted to know if the participants thought that introducing the envisioned software platform into their daily work routine was a good idea, and with the other, if they were positive about using the software platform. These questions were also adopted from the TAM.

On a Likert scale of 1-7, the mean of all 65 responses was 5.66 (t (64) = 11.45, p < 0.001)–indicating that participants agreed that the envisioned software platform can actually help them in their job. The results are depicted in Table 4.

The planned implementation of a supervisor monitor was also part of our acceptance study. For this purpose, the participants once again saw a short clip of the trailer showing a prototype of the monitor.

If people from the healthcare sector should really be able to integrate VR into their everyday work, then they need to be supported as well as possible. To better understand how assistance should look like we asked potential users what kind of support they would want when creating a virtual environment.

Users want continuous information about the next steps to take (n = 8 (36,4%)). There should therefore be as linear a process as possible when creating VR applications, with the next step always being clearly indicated. This support was requested across the whole process, e.g., support in designing the virtual environment, or help in finding suitable applications and templates.

Again, some of the participants took the opportunity to express that they would like templates provided to them (n = 7 (31.9%)). Frequently used standard applications should be identified and made available as a basis for further adjustments. Other features that were requested are the explanation of the different functions, the highlighting of often-used applications, or the linking to further help.

While few actionable responses were provided here, the answers to previous questions also offer many insights into what type of support is necessary when people want to create their own VR applications. More on this is outlined in the discussion.

The questions on ethical issues covered two topics: an evaluation of how several ethical principles are affected as well as an assessment of benefits and concerns expected for different health areas. Due to the novelty of the research, a principle-based approach provides an excellent starting point for an ethical assessment. This approach relates to a range of ethical principles that were previously defined as being important with regard to innovative health technologies (Walker and Morrissey, 2014). Principles considered relevant were quality of life as the key outcome of all health care activities (Musschenga, 1997), respect for autonomy and justice drawn from the four principles of biomedical ethics (Beauchamp and Childress, 2019), safety/security, privacy, participation, and a positive self-conception delineated in a model for the ethical evaluation of age-appropriate assistive systems, ambient-assisted living (AAL), and other health-related technologies (Manzeschke et al., 2015; Nelles et al., 2016), as well as transparency (e.g., Turilli and Floridi, 2009). The term support substituted the principle of beneficence (Beauchamp and Childress, 2019) because it was considered more concrete and tangible. Due to the wide range of areas in which VR is used, it was considered important to compare benefits and concerns for VR use in different health care areas. Examining professionals’ views of benefits and concerns reveals areas where VR use requires particular caution, for example for vulnerable persons with psychiatric symptoms like psychosis (Rizzo et al., 2003) or adolescents and children (Kaimara et al., 2021). For all questions, 7-point scales were utilized. The endpoints for the question on the agreement to how VR would support the selected ethical principles were 1-totally disagree, 7-totally agree. For the questions on expectations, the endpoints were 1-very low expectations/morally very low concerns; 7-very high expectations/morally very high concerns. T- and p-values are reported to indicate whether the single means deviate significantly from the mean of the scale to appraise whether the evaluation is above or below the average.

When we asked the participants about the potential use of VR in their field, the answers were fairly broad (see Section 4.2.1). This is very important for the basic motivation behind the envisioned software platform. If experts from different fields can imagine a concrete use of VR for their work, then this speaks for the potential of a broad-based software platform that can support many of them. It should also be considered that many of the study participants had no previous contact with VR. The fact that we received a total of 88 mentions about possible use cases for VR shows that the mockup trailer was successful in getting our ideas across.

In terms of the perceived ease of use, the figures also show that potential users tend to feel able to operate the envisioned software platform. This also shows a positive basic attitude of the participants toward the platform. However, questions and statements relating to the user-friendliness and operation of the system should always be treated with caution in this case. After all, the basis of this evaluation was a video. The actual usability of a user interface can only be determined on the basis of actual prototypes and in controlled test environments. Only then can we really make statements about what intuitive use of the platform could look like.

It is particularly interesting that there is a certain trade-off in desired functions. The people demand a certain amount of customization, as the multitude of different clients put different requirements to the VR applications. At the same time, they want the envisioned software platform to be very simple and straightforward without too much effort. In general, healthcare applications in VR have to manage this balancing act. Therefore, when implementing the envisioned software platform, we should ensure that standard applications and templates are available to users. We should thus prevent users from being forced to build their own applications from scratch and thus being overwhelmed with a multitude of options. At the same time, these standard applications must be modifiable in the right places so that they cover as many use cases as possible. So when implementing VR healthcare applications, one should always have an overview of what the most commonly used default scenarios and settings are. These must then be adaptable in the simplest possible way.

For exposure therapy, this could mean that common fear stimuli are available (animals, height, confinement, presentation in front of an audience), but that their peculiarities can be adapted. For example, with a virtual audience for virtual speech training, one should be able to modify the number of virtual agents, their degree of realism, and their behavior. At the same time, there could also be standard environments in which such an exposition can take place, e.g., in a large hall (for more stress) or on a beach (for more relaxation).

For physical rehabilitation, this could mean that one can choose between standard muscle groups that should be trained, e.g., the arm/shoulder area or the leg/knee area. Standard exercises could comprise the continuous lifting, holding up, or rotation of those extremities, while the number of repetitions or the time of the exercise could be adjustable and visible for the immersed person. Adaptability can also come into its own when it comes to embodiment. As mentioned before, it is possible to create custom avatars of people quite comfortably with just a smartphone Wenninger et al. (2020). However, one could also provide standard avatars for those who find it too time-consuming to carry out a scan process. This could also be useful in some applications, as already the sense of embodiment toward a regular hand can be sufficient to alter emotional responses to virtual stimuli (Gall et al., 2021). It would also be conceivable here to exploit the Proteus Effect, for example by providing athletic avatars when physical exercise is required.

The most important thing when it comes to the learning process of autonomously creating and operating VR applications is simplicity. For the envisioned software platform this means that we should refrain from overwhelming users with too many options. A large proportion of the participants in this survey had little to no prior experience with VR. This also explains the wish for close support. The potential users, therefore, attach less importance to having a large number of functions and options available. Rather, they want a relatively straightforward and simple path to their own virtual application. Again, this is generally true for healthcare applications in VR. One should not try to cover every extraordinary use case, but rather make standard use cases available quickly and easily. If it takes too long to learn how to use the VR application and to prepare it, it makes it more or less useless for practical use in the healthcare sector. Their already stressful working day does not allow for too much additional effort. This does not only apply to the software itself, but also to the equipment or hardware. Again, the fact that many of the participants had less prior technical experience comes into play here. On the one hand, this means that the requirements for operating VR applications should not be too high. On the other hand, one should make suitable hardware recommendations or minimum requirements for the applications. In this way, one can save potential users from having to search for their own hardware solutions. This can apply to both computers and head-mounted displays.

The desire for appropriate language is also relevant in this context. The focus should be on the use of the native language. Technical terms could additionally be provided with their own explanations via mouseover effect or in separate documentation. To be absolutely sure, one should test a VR application beforehand with potential users in the healthcare sector to find out whether the descriptions are comprehensible and appropriate. If one wants to ensure an intuitive and smooth user interaction, this step is inevitable anyway.

As far as the method of learning is concerned, there should definitely be documentation on the envisioned software platform. However, to further ease the burden on potential users, the suggestion of interactive tutorials is quite interesting. If the envisioned software platform provides the standard applications already mentioned, then these could also be used relatively easily as the basis for a tutorial. For each demo application provided, there could be a tutorial that shows where one has to change which parameters and what you achieve with them. This idea is also transferable to all VR applications that are to be used autonomously by people from the healthcare sector.

If healthcare professionals are to be truly capable of creating and operating VR applications on their own, they need support in different dimensions. Based on the answers to the survey, there are a few points that can be addressed. Based on our ideas for a software platform, an assistance system seems to be a really interesting solution. This system could offer personalized assistance, based on the experience of the user, the environment that needs to be created, or the type of treatment that is aimed for. The exchange of ideas with other users and communication with experts can also be supported by the assistance system. Ideas and approaches of other users on a certain topic could be shared and then, if relevant, suggested by the assistance system. The templates created by other users can also be shared if they have been published for reuse. The templates could be categorized via a tagging system and thus be found and tested more quickly by others. These tags could refer to diseases, forms of treatment, or medical specialties. This would also greatly facilitate an active search for relevant templates by potential users and optimize possible recommendations by the system. Thus, we could also fulfill the collaborative character that some of the participants have been asking for.

The desire of the potential users that some of the applications should also function without direct supervision, i.e. be carried out by the clients themselves, is quite interesting. In some areas, this could be an interesting approach to further relieve staff in the healthcare sector. People who want to make their VR applications more attractive might keep this option in mind. Suggestions in this direction came mainly from participants in the field of physiotherapy. This is probably less relevant for the treatment of mental disorders, as more intensive care is sometimes necessary here. For our envisioned software platform this means that it should allow users to decide which form of application they want to create. Either a form of application that is controlled by a supervisor or the form in which the clients themselves are responsible for the progress of the application. Automated logs could then still inform caregivers and clients of the achieved progress. In physiotherapy, this could result in some kind of scoreboard. Repetitions of lifting exercises or endurance exercises, e.g., of a treadmill or ergometer training, could serve as evidence of continuous training. Thus, the supervisor would have a convenient overview and clients could draw additional motivation from the introduction of gamification elements.

To make a VR application as attractive as possible for usage in healthcare, it should support the minimization of potential negative consequences and thus also take weight off the shoulders of a supervisor. This includes the definition of an exact play area to prevent falls. To prevent the symptoms of cybersickness, applications should give warnings when the frame rate is too low or the latency is too high. The dynamic restriction of the field of view (Groth et al., 2021; Teixeira and Palmisano, 2021) or the adaption of navigation velocity and acceleration (Plouzeau et al., 2018; Chardonnet et al., 2021) are possible automatic methods that can reduce the hazards of cybersickness. In this way, the supervisor could be further relieved.

The specification or recommendation of certain hardware can also help to ensure minimum quality. Another starting point is the collection of physiological data. It can help to detect cybersickness and general discomfort (Cebeci et al., 2019; Islam et al., 2021). Here, too, the automatic recording, analysis, and visualization of the user’s condition can help to relieve the supervisor. More on this when we talk about the supervisor monitor in the next subsection.

The idea of implementing a supervisor monitor also seems to be a suitable measure to give people from the healthcare sector the necessary support and security to integrate VR into their work. Frequently requested was information about the current view of the VR users and information about their physiological state. Wearable sensors are particularly interesting here, as they have a very short setup time. Parameters like heart rate and skin conductivity can be conveniently captured with wristbands such as the Empatica E4 ⁹ , that has been used before in VR studies (Šalkevicius et al., 2019). Gaze and pupil behavior information could also be captured relatively conveniently using head-mounted displays. Eye trackers from tobii ¹⁰ or Pupil Labs ¹¹ are commonly used to augment VR headsets. Different characteristics of experience can thus be classified automatically, e.g. stress (Ham et al., 2017; Robitaille and McGuffin, 2019), anxiety (Šalkevicius et al., 2019; Bălan et al., 2020), or cognitive workload (Currie et al., 2019). Corresponding classification results can then either be used to directly control the virtual environment, can be visualized for the supervisor, or be used to determine the progress over multiple sessions. The various options that exist here are discussed in Halbig and Latoschik (2021). In any case, the capture, analysis, and visualization of physiological data can help to give the supervisor more confidence in dealing with VR. In practice, of course, one has to weigh how much and to what extent this information is displayed at once. Dangers and unusual physiological signals must be presented in such a way that they can be perceived immediately, e.g. via acoustic signals and a salient layout. So, if possible, not all details should be shown to the supervisor, such as raw physiological or movement data. Rather, the supervisor should be shown the interpretation of the data (e.g., high or low stress) by default. Raw data could be displayed on request. The data and its interpretation could also be part of automated logs of the specific sessions.

Another interesting approach is the ability to intervene in what is happening in the virtual world via the supervisor monitor. Especially the attenuation or amplification of stimuli plays an important role in applications that are supposed to trigger fear or stress. Feedback from potential users also speaks in favor of a kind of emergency off switch on the display that allows the situation to be completely resolved. Here, more control and security can be brought to the supervisor.

The possibility to take notes and save them for the respective client and session could provide basic support for interviews and observations. Such a digital notepad could then also be helpful when documenting progress across multiple sessions. In addition, one has all the information about the stimuli that were set in the virtual environment. Automated logs can then be used to store information about the stimuli and the client’s behavior. In this way, it should be possible for therapists and trainers to analyze the progress of the treatments.

Overall, the results demonstrate that professionals expect positive outcomes for their patients with regard to ethical principles and for most health areas. Nevertheless, negative outcomes are mostly perceived for transparency and justice. Doubts with regard to transparency refer to the possibility that users might unconsciously be manipulated by VR (Kool, 2016; Spiegel, 2018) or that wrong information about VR effectiveness is given (Madary and Metzinger, 2016). With regard to justice, only privileged groups could benefit from VR (Madary and Metzinger, 2016). These issues were previously mentioned in the literature. Future research needs to address the reasons for these concerns and derive conclusions to avoid these negative effects. Besides, psychiatry/psychotherapy and palliative care are the areas with the highest ethical concerns. These disciplines have previously been identified as ethically critical (e.g., Rizzo et al., 2003; Weijers and DiSilvestro, 2017), and the concerns should be addressed adequately. VR has already been proven beneficial in these areas, at least in certain contexts (e.g., Dellazizzo et al., 2020; Johnson et al., 2020). Accordingly, it is important to reflect on these concerns and to carefully consider potential applications, but also to resolve concerns by informing about proven positive effects and positive perceptions by users.

Although this study provides many useful information, it also has limitations. Probably the most severe is that the findings of this study are based on a video and not on actual interaction with a prototype. It is therefore very difficult to make statements about what a functioning interaction with our envisioned software platform would look like. Nevertheless, it is important to collect the opinion of potential users at the very beginning. In addition, the video was also a suitable means to convey ideas of VR and its connection with healthcare to a wide range of people.

Another point to mention here is the fact that this is an online study. Interviews would have made it possible to get more in-depth information. In addition, ambiguities could have been explained. However, the decision to conduct the study online also had its advantages. First, it allowed us to reach more different people as we were able to send invitations around easily. These could then spread further and reach many experts to whom we would not have had access at all. Second, this also enabled us to comply with the pandemic restrictions and get in contact with healthcare personnel at the same time.

In the course of the discussion, we elucidated how the knowledge gained in the present study can be used to develop VR applications that respect more of the requirements of healthcare practice. However, there is still a lot of unfinished work here, especially considering the concrete implementation of the discussed ideas. While it became clear that some kind of supervisor monitor should be helpful to spur the integration of VR into healthcare practice, it remains unclear how exactly this should look. It needs to be clarified how this monitor represents the current state of the immersed person, may it be the location in 3D space or the physiological status. This is accompanied by the question of how a supervisor would alter the sequence of events and how possible dangers are communicated perceptibly and understandably via the monitor.

Further research potential lies in the question of how exactly the support looks like that is received by a person that creates a VR environment. An AI could be designed that understandably supports laymen to create their own virtual applications. The exact solution of the trade-off between adaptability of such applications and low effort in their creation process is another unknown. While we have already shown possible approaches, the details of which features should be customizable and which not remain unclear. Just as with all the other open points, progress will only be made here with concrete user studies. Only a series of iterative tests with concrete interfaces and actual users can provide accurate information on what the implementation of the discussed features would look like. User studies with healthcare professionals and clients using demonstrator applications are planned as a next step.