Health care providers, particularly during the COVID-19 crisis, have reported acting in ways that are contrary to their moral values, integrity, and professional commitments because of constraints in their work environments [1,2]. Moral suffering can ensue when a health care provider’s moral foundation is threatened or violated by witnessing or participating in decisions or actions that degrade their integrity [3]. Moral distress might arise when health care providers are unable to translate their moral choices into action because of internal or external constraints [4].

Moral injury, a more extreme form of moral distress, arises in high-stakes situations wherein integrity and conscience have been violated or one’s moral core is eroded [5]. As these events and other workplace stressors accumulate, health care providers’ capacities to provide the level of care they desire can be diminished. There has been a growing attempt to conceptualize moral distress and injury in different health care professions, but the numerous definitions can confuse the issue [6,7]. The concept of moral injury has existed for much longer than its first conceptualization by Shay [7] concerning the military, where it was defined as a betrayal of moral character, usually as a result of the actions of a person in a position of authority. There is a large body of literature within the military context, and moral injury has conceptually evolved to include betrayal of one’s moral values, and furthermore, as a spiritual, psychological, and somatic response to a series of morally injurious events [5]. While some may contend that actions by military members during a war require forethought based on training, others suggest that actions are made on rapid and subconscious decision-making and that moral injury occurs only when individuals question their actions after the fact [8].

However, moral injury is not well defined in health care professions. Cartolovini et al [9] argue that moral injury occurs in health care providers when they encounter traumatic events that involve high-stakes situations beyond their control in which they perpetuate, fail to prevent, or bear witness to actions that transgress deeply held moral beliefs [10]. A related concept, moral distress, was first described within the nursing context by Jameton [11] as “… when one knows the right thing to do, but institutional constraints make it nearly impossible to pursue the right course of action.” Additional studies [12] have provided more context to this definition and have described causes of moral distress as beyond the organization to specific clinical situations and internal caregiver dynamics. Moral distress and injury likely occur along a continuum, with moral injury at the extreme end [13].

The COVID-19 pandemic has necessitated fundamental changes to health care provision. As we continue to learn about the virus and differing courses of disease, health care providers have often willingly accepted inherent risks during the pandemic to care for their patients. However, the need for rapid and changing contingency planning, shortages in personal protective equipment, resource shortages, and the inability to provide the ethical caring called for by professional codes of ethics have placed health care providers in what has been described as a “war zone [14].” The sudden and unprecedented similarities between civilian health care and military scenarios have potentially resulted in moral distress and possibly moral injury. It is nearly impossible to anticipate the course of the virus and make predictions about its physiological outcomes; it is also challenging to predict the course of mental health needs of health care providers, but it is known that there will be needs.

Research has concentrated on treatment for moral injury, either in the military realm or the civilian realm. Gilligan [15] discussed moral injury as “a shattering of trust that compromised our ability to love” and would seem to suggest the health care providers who suffer moral injury during extreme circumstances such as a pandemic may lose the ability for empathy that is required by professional values. Moral injury has been found to not generally respond to typical treatment of posttraumatic stress disorder, even though most treatment focuses on the posttraumatic stress disorder symptomatology [5]. Being able to preserve or restore integrity requires that health care providers have the opportunity for moral repair by strengthening their moral compass and restoring their integrity. Having a moral compass has been identified as a modifiable factor effective in healing from trauma [16]. When one’s moral foundation is clear and accessible, health care providers have a potential opportunity to heal the wounds caused by the COVID-19 crisis and an additional resource that they can leverage when ethical challenges arise.

In this paper, we propose using digital platforms, specifically virtual reality and mobile apps, to develop an understanding of the continuum of moral distress that leads to moral injury in health care workers. Virtual reality technology will be used as a digital platform to simulate an intensive care unit during the pandemic. Our aim is to determine the feasibility of creating a digital environment and implementing digital interventions to better understand the phenomenon of moral distress by developing platforms to examine moral distress and monitor participants over time in order to develop a better understanding of the continuum of moral distress to moral injury.

As of March 20, 2021, the research ethics board at St. Michael’s Hospital, Unity Health Toronto has approved the study, and the reference number is UHTDTS25377. The clinical trial is registered with ClinicalTrial.gov with the Registration/identifier number NCT05001542.

We will recruit health care workers to participate in our feasibility trial that will be composed of pretesting, 2D educational video, posttesting, and longitudinal monitoring (Figure 1). During pre- and posttest components, we will pilot unique digital technology platforms (virtual reality environments) to identify circumstances that have the potential to cause moral distress; examine interactions at the individual, team, and organizational levels; perform detailed concept mapping to understand features of moral distress and injury and to aid the design of potential interventions. The educational video will be in 2D; however, the rest of the study components will be 3D virtual reality. The video will be presented on the virtual reality headset after the pretest and before the posttest, and is a brief educational video about moral injury. After the posttest, for 2 months, we will longitudinally monitor participants through the collection of active and passive data through our novel mobile platform.

The expected outcomes of our protocol include (1) using virtual reality technology to design a virtual reality environment to simulate intensive care unit settings during the pandemic and determine the feasibility of understanding moral distress and injury for potential future implementations of digital interventions and (2) using acute physiological monitoring within a virtual reality for simulation of a hospital environment where health care workers have to make challenging moral decisions in complex moral simulated situations, and using a novel mobile platform for longitudinal monitoring by collecting passive and active data to understand the contribution of moral distress to the experience in real-time. The information from these digital platforms will influence each other and aid in the development of models to predict and respond to risks associated with moral distress and injury.

The compound intervention is composed of the virtual reality simulation-based educational intervention and data collection of mental health and moral injury surveys. The virtual reality simulation-based educational intervention contains a moral injury educational video based on the Moral Injury Guide developed by the Ottawa PTSD Center for Excellence [17].

The latter component of the compound intervention evaluates the effectiveness of the virtual reality simulation-based educational intervention through mental health surveys. These surveys include the Perceived Stress Scale (PSS), Moral Injury Outcome Scale (MIOS), Igroup presence questionnaire (IPQ), and Adapted Moral Injury Symptom Scale: Health care Professionals Version (MISS-HP). PSS is a self-report instrument that evaluates level of stress. MIOS and MISS-HP are tools that evaluate the effect of moral injury, and the IPQ is a questionnaire that evaluates the experience of presence during virtual reality simulations. We will use this information to improve our compounded intervention for use on digital platforms iteratively.

This is a multilayer feasibility trial (Figure 2). The in-person session (Figure 3), begins with participant arrival at St. Michael’s Hospital Simulation Centre. This is followed by experimental set-up and prebrief. The prebrief is held to inform the participant of the purpose of the session, to collect presurvey data, and obtain written consent. The experimental set-up involves the attachment of the virtual reality equipment and physiological sensors. The debrief is hosted after the experiment and is composed of a semistructured interview, open-ended questions, and an exit survey. The purpose of the debrief is to obtain qualitative feedback on the session to better understand the participant experience.

The virtual reality session will be used for short-term monitoring of participants for acute distress. After the intervention, participants will use a mobile app (developed by the Digital intervention/intelligence Group) for long-term distress monitoring.

During the session hosted at St. Michael’s Hospital Simulation Centre, data (CSV files, text files, audio, and video files from the session) will be securely uploaded to a password-protected drive. Only specific members of the project have access to the data. Responses on paper questionnaires will be input to a spreadsheet and stored on the encrypted drive. During the longitudinal monitoring part of the study, all active and passive data will be stored on the participant’s phone (if storage space is available) until those data are synced back to the back-end server. Data are transferred over a TLS layer; hence, it is encrypted by RSA Asymmetric Encryption Algorithm.

Study data will be stored for 10 years. No personal information will be collected through the mobile component of the study. Demographic data (ie, age, sex, type of job at Unity Health Toronto, diagnosis, and COVID-19 status) will be collected separately from data collected by the app. A study participant ID will link these data for final analysis.

The server and the dashboard will be hosted on the same virtual machine server, and the database will be hosted on a different database server (Multimedia Appendix 1).

The first stage of the development involved defining the context of the scenario, with the support of a subject-matter expert, and determining how it connects to the domains betrayal of the team, transgression of moral values, emotional response, and support and reframing in relation to guilt and loss of trust in colleagues and shame.

An overview of the scenario was then outlined, describing the location, main characters, key points (eg, the removal of the ventilator), and the actions and reactions of each character. Based on this document, 2 activities were performed in parallel: the definition of the scenario script with the lines and actions of all characters and the development of the 3D assets (characters, objects, and environment) of the virtual reality scenario. Both activities were developed iteratively, with discussions and demonstrations during weekly meetings. The final version of the script includes 7 nonplayable characters (Table 1) and 4 conversations (Table 2), not including the initial tutorial.

For the virtual environment (Figure 4 and Figure 5), royalty-free models and scenarios were purchased and used as a starting point. They were combined and modified to fit the project's needs. Subject-matter experts provided feedback on key elements to make the environment appear closer to the reality of a Canadian hospital environment. The goal was to represent a standard intensive care unit with individual rooms for patients and a nursing station.

To represent the nonplayable characters and the participant, royalty-free models were purchased as starting points. Based on these models, a customizable character system was implemented that allowed us to quickly modify external elements (such as clothing or masks) or the character’s characteristics, such as height and weight (Figure 6).

For the user interface, 2 semitranslucent panels are displayed as spatial elements in the virtual reality environment (Figure 7). The dialogue panel shows current nonplayable character’s photograph, name, and the text version of the dialogue being spoken. The interaction panel displays available choices (for example, the option Continue to advance to the following dialogue or a list of choices).

To interact with the panel, participants need to move one of the controllers, point to the desired choice, and click any buttons under the index finger or thumb. The participant can see a blue ray being casted from their index finger tip within the virtual reality environment, which turns yellow when hovering over the interactive panel (Figure 7). As additional visual feedback, dialogue choices include a highlight effect when the ray hovers over each.

The virtual reality scenarios were created using a cross-platform game engine (Unity 3D, Unity Technologies) that has been adopted, not only by the game industry, but also by film, architecture, and education industries [18]. The virtual reality scenario is played using a consumer-grade head-mounted display (Oculus Quest 2, MetaQuest) (Figure 8) that can track the user’s head movements with 6 degrees of freedom and track the user’s movements either in stationary mode (sitting in a chair) or in room mode (recommended room size is up to 2 m × 2 m), allowing the user to physically walk around within the defined boundaries. In addition, the system includes 2 handheld motion-tracked controllers that allow the user to interact with the environment and with virtual reality interface components. Our virtual reality scenarios will adopt stationary-mode tracking and use the handheld motion-tracked controllers for more accurate tracking.

Currently, Oculus Quest and Oculus Quest 2 are not Health Insurance Portability and Accountability Act–compliant. However, no data will be stored in the virtual reality device, and the Oculus framework is not used to develop our virtual reality app. The Oculus Quest 2 is connected to a virtual reality-capable PC laptop using a USB-C cable.

We plan to collect subjective data such as qualitative data from surveys and open-ended feedback. As mentioned in the previous sections, the MISS-HP, IPQ, PSS, and MIOS will be answered throughout the experiment. At the end of the virtual reality session, an exit survey composed of open-ended and 3 multiple choice questions for general feedback.

Overall, the MIOS and PSS are evaluated at the beginning and end of the study for a global assessment of distress (Multimedia Appendix 2), whereas the MISS-HP is collected for a local assessment of distress (Multimedia Appendix 3).

During the virtual reality session, electrocardiogram, electrodermal activity, photoplethysmography, and respiration impedance sensors will be attached to the participant (Figure 9), and these data will be collected with an acquisition and analysis system (MP160, Biopac).

We will create a personalized digital phenotype profile, an objective parameter which will help track individual trends over time which will assist in identifying the efficacy of digital interventions and exposure to immersive virtual reality environments, for each participant using primarily heart rate, respiratory rate, oxygen saturation, and skin conductance (BioNomadix, Biopac) (Table 3). These signal features will be paired with reported psychological measures (scales) for a holistic understanding. Both cardiovascular and cerebrovascular dynamics can be captured with these data by deriving features such as heart rate, respiratory rate, and heart rate variability (which is related to the autonomic nervous system and affect states [19]). The autonomic nervous system is the control system that acts largely unconsciously and regulates bodily functions, including heart rate and respiratory rate. This system is the primary mechanism in control of the fight-or-flight response [19]. We expect that participants in stressful situations will have specific personalized digital phenotype profiles.

Since participants are required to come to the St. Michael’s Hospital Simulation Centre to complete the virtual reality and wearable components of the study, the study will not be anonymous. Participants will only start using the mobile app after they participate in the virtual reality simulation component of the study. The study coordinator will provide an ID and password by email after participants provide written consent, logistical information (preferred time slot), and demographic information to the study coordinator by email.

Smartphones are ubiquitous in our daily lives and incorporate a suite of sensors, such as accelerometer and GPS. In addition, digital platforms can be used to collect active data. This results in a multimodal system on a smartphone device. To objectively analyze the data, preprocessing will include Kalman filtering, GPS data cleaning, and Wavelet transforms, and feature extraction techniques, such as multiview biclustering [27] and statistical and structural features from time-series signals [28], will be used for machine learning (decision tree, logistic regression, and support vector machine) to find associations between passive and active data. These models are simple, consume low power, and have the capability of explainable artificial intelligence to understand our machine learning models better [29]. We plan to extend the machine learning models to long short-term memory and convolutional neural networks due to the popularity and effectiveness of the models in mobile data analysis [28].

Signals (Figure 11; Multimedia Appendix 4) will be collected with a sampling frequency of 2000 Hz. The software used to collect the data from Biopac performs all internal calculations to the accuracy defined by the IEEE format for double-precision floating-point numbers and stores the results of those calculations in double-precision floating-point format; this format assigns 8 bytes to all numbers in calculations or resulting from calculations [30]. Biopac equipment has an interface to interact with the virtual reality scenario; thus, we have markers for the sessions, such as the start and end of the session and when the user answers a question.

Numeric programming computing software will be used to conduct the analysis. The signals will be collected in a raw format. Preprocessing, feature extraction, and machine learning will be applied to the data. Preprocessing involves techniques such as low-pass filters, high-pass filters, and notch filters to remove noise. The filtered signal will then be passed along to extract features. Feature extraction will involve extracting attributes of the signals to be better represented. Techniques for feature extraction include handcrafted features, such as statistical values. Lastly, machine learning techniques such as decision tree, random forest, and logistic regression will be used for classification. These techniques are chosen due to their ability to manage small sample sizes. Once familiar with the newly acquired data set, we will reflect on a new approach for analysis.