This paper is in the following e-collection/theme issue:

Games for Medical Education and Training (95) Systemic Racism and Racial Bias in Health Care and Society (85) Systemic Racism and Racial Bias in Health Care and Society (85) Clinical Integration of XR and Spatial Computing (32) Usability of Apps and User Perceptions of mHealth (974) Extended Reality, Virtual Reality and Virtual Worlds (560) Games and Gamification for Health (650) e-Learning and Medical Education (691) Clinical Integration of XR and Spatial Computing (32) Usability of Apps and User Perceptions of mHealth (974) Extended Reality, Virtual Reality and Virtual Worlds (560) Games and Gamification for Health (650) e-Learning and Medical Education (691) Validity, Reliability, and Bias (93)

Published on in Vol 12 (2024)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/51310, first published .
Smartphone-Based Virtual and Augmented Reality Implicit Association Training (VARIAT) for Reducing Implicit Biases Toward Patients Among Health Care Providers: App Development and Pilot Testing

Smartphone-Based Virtual and Augmented Reality Implicit Association Training (VARIAT) for Reducing Implicit Biases Toward Patients Among Health Care Providers: App Development and Pilot Testing

Smartphone-Based Virtual and Augmented Reality Implicit Association Training (VARIAT) for Reducing Implicit Biases Toward Patients Among Health Care Providers: App Development and Pilot Testing

Authors of this article:

Jiabin Shen1 Author Orcid Image ;   Alex J Clinton1 Author Orcid Image ;   Jeffrey Penka2 Author Orcid Image ;   Megan E Gregory3 Author Orcid Image ;   Lindsey Sova4 Author Orcid Image ;   Sheryl Pfeil5 Author Orcid Image ;   Jeremy Patterson6 Author Orcid Image ;   Tensing Maa7 Author Orcid Image
Article Authors Cited by (1) Tweetations (4) Metrics

    1Department of Psychology, University of Massachusetts Lowell, , Lowell, MA, , United States

    2LittleSeed, Inc, , Columbus, OH, , United States

    3Department of Health Outcomes & Biomedical Informatics, College of Medicine, University of Florida, , Gainesville, FL, , United States

    4Center for Advancement of Team Science, Analytics, and Systems Thinking in Health Services and Implementation Science Research, College of Medicine, Ohio State University, , Columbus, OH, , United States

    5College of Medicine, Ohio State University, , Columbus, OH, , United States

    6Advanced Computing Center for Arts and Design, Ohio State University, , Columbus, OH, , United States

    7Center for Clinical Excellence, Nationwide Children’s Hospital, , Columbus, OH, , United States

    Corresponding Author:

    Jiabin Shen, PhD


    Background: Implicit bias is as prevalent among health care professionals as among the wider population and is significantly associated with lower health care quality.

    Objective: The study goal was to develop and evaluate the preliminary efficacy of an innovative mobile app, VARIAT (Virtual and Augmented Reality Implicit Association Training), to reduce implicit biases among Medicaid providers.

    Methods: An interdisciplinary team developed 2 interactive case-based training modules for Medicaid providers focused on implicit bias related to race and socioeconomic status (SES) and sexual orientation and gender identity (SOGI), respectively. The simulations combine experiential learning, facilitated debriefing, and game-based educational strategies. Medicaid providers (n=18) participated in this pilot study. Outcomes were measured on 3 domains: training reactions, affective knowledge, and skill-based knowledge related to implicit biases in race/SES or SOGI.

    Results: Participants reported high relevance of training to their job for both the race/SES module (mean score 4.75, SD 0.45) and SOGI module (mean score 4.67, SD 0.50). Significant improvement in skill-based knowledge for minimizing health disparities for lesbian, gay, bisexual, transgender, and queer patients was found after training (Cohen d=0.72; 95% CI −1.38 to −0.04).

    Conclusions: This study developed an innovative smartphone-based implicit bias training program for Medicaid providers and conducted a pilot evaluation on the user experience and preliminary efficacy. Preliminary evidence showed positive satisfaction and preliminary efficacy of the intervention.

    JMIR Serious Games 2024;12:e51310

    doi:10.2196/51310

    Keywords

    implicit bias; health care; Medicaid; virtual reality; augmented reality; smartphone; mHealth; mobile app; innovative; implicit bias training program; sexual orientation; sexual orientations; gender identity; gender identities; gender preferences; gender preference; efficacy; health care providers; health care provider; socioeconomic; mobile application; training; XR; extended reality 


    Implicit Bias in Health Care Settings

    Defined as unconscious associations or negative evaluations of a person or group of people on the basis of nonrelevant characteristics [1], implicit biases have been found to be prevalent among the general population against “marginalized’ groups such as those from minority racial, ethnic, or socioeconomic backgrounds [2]. Implicit biases, which present in health care settings as irrational and unconscious perceptions, stereotypes, or prejudices among health care providers when interacting with patients, are especially concerning [3]. Past research has found that implicit bias in health care settings is associated with a decrease in overall quality of care, with impacts including increased risk of misdiagnosis [4-7], inaccurate patient pain perception [8,9], differential treatment recommendations for patients who belong to sexual orientation or gender identity minority groups [10-12], and negative perceptions of patients from racial minority backgrounds [13-17]. Implicit biases may also exist during interactions between health care professionals, such as selection bias when choosing candidates for future health care practitioner residency [18], which may have wider implications for the quality and safety of patient care. Furthermore, such implicit biases have been found within adult and pediatric health care settings [19] across medical conditions including ADHD, asthma, cardiology, and child abuse, which could affect quality of care for these vulnerable populations [20-22].

    Existing Efforts to Reduce Implicit Bias in Health Care Settings

    In response, increasing efforts have been devoted to addressing the significant threat posed by implicit bias toward health care services and patient outcomes. The first type of interventional efforts focus on “environmental engineering,” with the goal to minimize mechanisms in health care settings that may give rise to biased interactions between health care professionals and patients. One example of this type of intervention is the development and implementation of automatic patient care prompts through electronic portals, where computer algorithms are standardized for all patients regardless of sociodemographic backgrounds, attempting to reduce opportunities for human interference (beyond the algorithm development phase) [23]. A second type of intervention uses cognitive rehearsal to walk practitioners through potentially harmful scenarios to practice their ideal response; this has shown promise at changing health care practitioner behavior to reduce bullying and workplace turnover [24,25]. While not widely used in combatting health care bias explicitly, the methodology shows a clear avenue for its application to bias training.

    A third type of intervention, which will also be the focus of this study, attempts to develop educational programs with the goal of improving knowledge and awareness of implicit bias among medical students or health care professionals, which can range from traditional educational seminars to experience-oriented storytelling, to highlight the importance of patient perspectives in daily practice [26-28]. Such efforts have so far yielded positive results where health care professionals were found to become more aware of their own biases and have resulted in improved communication between health care professionals and marginalized patient groups [29,30].

    Application of Augmented Reality–Based Medical Training

    Despite the promising results from educational programs in the existing literature, one limitation in existing approaches for implicit bias training is the lack of immersive learning experiences that may provide optimal learning outcomes and behavior changes. As a cutting-edge technology that prioritizes experiential learning, virtual reality (VR) and augmented reality (AR) could provide an ideal solution with immersive learning experiences for implicit bias training. For example, one recent study examined biases during interactions between virtual health care providers and virtual patients for medical triage training. Regardless of the skin tone of the avatar (ie, the health care provider), it took participants more time to initiate assistance and they were more likely to make errors when triaging dark-skinned virtual patients compared to light-skinned virtual patients [31].

    AR, as a more recent member of the x-reality technologies, is posed to offer an even better learning experience that combines the immersion provided by VR and tailored customization that adapts to users’ dynamic environments. Adoption of AR in medical education has been found in a wide range of medical branches from surgery (eg, laparoscopic procedure training) to anatomy [32]. Furthermore, because AR-based training is readily available on consumer-grade mobile devices such as smartphones and tablets, its mobility provides medical professionals with remote accessibility to training content regardless of their physical location (this advantage has been further acknowledged during the COVID-19 pandemic) [33]. However, despite the increasing adoption of AR in medical training, a recent systematic review has found little evidence on the availability of AR-based implicit bias training among health care professionals in the literature [33].

    This Study

    To address this important gap, this study aimed to develop a mobile training program, VARIAT (Virtual and Augmented Reality Implicit Association Training), specifically for improving the awareness of implicit biases among health care providers when interacting with patients in daily practice. The design considerations for developing this novel AR-based implicit bias training program are described, followed by a preliminary examination of initial user feasibility and learning outcomes, including user reactions; relevance to practice; and changes in knowledge, attitudes, and behavioral skills related to implicit bias before and after receiving the training program.


    Designing the VARIAT Program

    Overview of Technical Design Considerations

    The VARIAT program focused on delivering an immersive, interactive learning experience to the broadest possible audience in self-contained segments, allowing users to complete the training over time and a variety of sessions while retaining their progress across sessions. When building the 3D worlds for delivery on the broadest number and sizes of mobile devices, simplified, realistic, and familiar spaces were built, including offices, lobbies, and examination rooms where the learner could experience the simulations. Characters in the world were designed with exaggerated cartoon features to provide visual distinction with skin tone, hair, size, outfits, and accessories, while minimizing unnecessary details and maximizing ease of recognition for interaction on mobile-sized screens. The approach to world design addressed design and performance considerations, allowing production of additional characters and scenarios without significant technical overhead in either the creation process or the learner’s experience on their device. The dialogue and training content was presented via text.

    Hardware Requirements and Considerations

    One key goal of the VARIAT program was the need to maximize audience reach and minimize specialty equipment for learners to access the content. At the time of its development, most iPhones and Android devices had the cameras, accelerometers, and gyroscopes needed to provide users the ability to see into and navigate virtual worlds by simply holding up and moving their devices. When the VARIAT program was introduced, learners needed at least an iPhone 8 with iOS 13 or an Android device running Android 9 or higher.

    Software Requirements and Considerations

    The maturity of the mobile app environment offers many development tools and approaches for developing mobile apps. For the VARIAT program, the developers used the Unity game engine (Unity Technologies) for game content with ARKit (Apple Inc) and ARCore (Google LLC) for the augmented reality component and deployed both iOS and Android apps that were readily available in their respective app stores. Blender (Blender Foundation) was used for 3D modeling and animation, and Photoshop (Adobe) was used to create 2D assets.

    The learner downloads the app from the Apple App Store or Google Play on their device, and their progress is maintained on the device with evaluation, progress, and study data synchronized as the learner completes various modules. When synchronized, the data are stored and managed using Google Big Query Workspace, which produces data feeds for training evaluators and researchers.

    Overview of Content Design Considerations

    Training Framework of the VARIAT App

    The VARIAT game was designed based on the integration of evidence-based cognitive psychology with the latest simulation technologies, including VR (eg, a simulated experience of interacting with a virtual clinician-patient scenario that is vulnerable to implicit biases in a virtual environment using 6-degrees-of-freedom motion- and gesture-based interactions) and AR (eg, converting a user’s physical environment into a clinic’s waiting room for interactive experiences). The goal of the app is to improve awareness of implicit biases among Medicaid providers, to educate them on how these biases can lead to inequitable care, and to offer strategies and resources that may minimize health disparities. This mobile app can be installed on any Apple iOS and Android device and is designed to be completed in one sitting or in short segments.

    The game consists of 2 distinct but interconnected modules, targeting implicit biases within medical settings toward patients from minority racial backgrounds, with low socioeconomic status (SES), or from a minority sexual orientation or gender identity (SOGI) group. Learners enter an AR-based interactive role-playing game, in which they encounter a series of 6 scenarios. Each scenario takes approximately 5 minutes to complete and is related to the specific implicit bias being addressed in that particular module (Race/SES or SOGI). Each scenario within the module is designed to address specific issues related to disparities in medical settings that relate to the overall theme of the module. The primary outcome measure of this training program is to evaluate providers’ attitudes and beliefs on key concepts related to implicit biases and health disparities in a medical setting before and immediately following the training.

    Race/SES Module

    The first module within the VARIAT mobile game is the race/SES module, which consists of 3 scenarios dealing with issues of racial bias, transportation and food instability, and implicit bias. The first scenario in this module addresses issues of racial bias within a health care setting and prompts the user to consider how issues of racial and ethnic identity could impact treatment recommendations and the resulting care for patients of minority groups. The next scenario is designed to promote self-reflection on how socioeconomic factors like unreliable transportation or housing could affect a patient’s ability to show up for health care visits or comply with care recommendations by medical providers. The last scenario is designed to help users understand how implicit bias from medical providers could impact patient perception and negatively impact patient care.

    Each scenario contains prompts where the user is asked to make a choice about the “case” presented within the VARIAT AR game. The user is then given information about the scenario and resources for how to better understand the specific issues for each scenario with the goal to educate them on how to improve practitioner behavior as it relates to the theme of the module. A summary of the scenarios and objectives for this module can be found in Multimedia Appendix 1.

    SOGI Module

    The second module in the VARIAT mobile game is the SOGI module, which consists of 3 scenarios dealing with issues of SOGI implicit bias, inclusivity for patient care settings, and lesbian, gay, bisexual, transgender, queer (LGBTQ+) patient considerations. The first scenario helps portray the way that microaggressions and implicit biases in patient-provider communication can promote negative disparities in treatment for SOGI minority patients. The second scenario asks users to design their own patient waiting room and helps educate and guide users on what considerations should be made to ensure a medical setting is a safe and welcoming environment for SOGI minority patients. The last scenario in the module helps users recognize the harmful effects of biased behavior toward LGBTQ+ patients and offers space for self-reflection on how to reduce enacted bias for this patient group.

    The scenarios in this module also contain prompts for users to answer to better assess their understanding of key concepts or takeaways from each scenario. The SOGI module places an increased emphasis on self-reflection as the scenarios are designed to help users draw parallels to their own experiences in medical practice through reflective exercises following the conclusion of each scenario within the module. A summary of the scenarios and objectives for this module can be seen in Multimedia Appendix 2.

    User Workflow

    On start-up, users are given some brief instructions on how to prepare themselves for immersion in the VARIAT AR game. Users are then instructed to select any available module to start engaging with the content within. Once a module is selected, participants are placed in a virtual hospital setting and can check on the various patients within. When selected in the AR game, these patients display information on their illnesses and present the user with additional narratives about the patients from the “staff” in the AR game. Users are then given different decision options on what to do for each patient’s individual case. After helping these patients, the user is provided with information and resources that relate to the content of the module. Unbiased choices “score” higher than choices that are considered to have been influenced by implicit biases toward marginalized patients. After completing the tasks in their module, the users are given a summary of their scores for that module with feedback on how to improve, and additional information to support that improvement relative to the context of their scenario. After completing a module, users are sent back to the home screen, where they can replay the same module or select a new module to explore. A depiction of the app layout, user experience, and scenario prompts is presented in Multimedia Appendix 3.

    User Experience and Preliminary Efficacy of the VARIAT Program

    Participants and Procedure

    Eighteen clinicians (n=12 female) who were predominantly White (non-Hispanic) participated in the VARIAT training. Physicians comprised 8 of the 18 (44%) participants, and 12 of the 18 participants had more than 5 years’ experience in health care. The most common workplace setting was hospitals, with private practices, health care system–affiliated clinics, and other workplace settings reported as well. Most participants estimated that Medicaid patients comprised more than 30% of their total caseload, with reported ages of patients seen varying between children, adults, and older adults. Demographic information is reported in Table 1 for the total number of participants (N=18), participants who participated only in the race/SES module (n=7), participants who participated in only the SOGI module (n=5), and participants who completed both the race/SES and SOGI modules (n=6). Participants were recruited through professional networks and were eligible for the study if they were Medicaid providers.

    Table 1. Demographic information for the participants.
    VariablesOverall (n=18), n (%)Race/SESa only (n=7), n (%)SOGIb only (n=5), n (%)Race/SES and SOGI (n=6), n (%)
    Race/ethnicity
    White (non-Hispanic)16 (88)6 (85)5 (100)5 (83)
    Non–Whitec2 (11)1 (14)0 (0)1 (16)
    Gender
    Male6 (33)2 (28)2 (40)2 (33)
    Female12 (66)5 (71)3 (60)4 (66)
    Prefer not to say0 (0)0 (0)0 (0)0 (0)
    Profession
    Medical resident3 (16)2 (28)0 (0)1 (16)
    Nurse0 (0)0 (0)0 (0)0 (0)
    Fully credentialed physician8 (44)3 (42)2 (40)3 (50)
    Social worker5 (27)1 (14)2 (40)2 (33)
    Other2 (11)1 (14)1 (20)0 (0)
    Work setting
    Health care system–affiliated clinic2 (11)1 (14)1 (20)0 (0)
    Hospital9 (50)3 (42)3 (60)3 (50)
    Private practice2 (11)1 (14)0 (0)1 (16)
    Other4 (22)2 (28)1 (20)1 (16)
    Missing1 (5)0 (0)0 (0)1 (16)
    Experience in work setting
    Less than 1 year0 (0)0 (0)0 (0)0 (0)
    1-5 years6 (33)3 (42)1 (20)2 (33)
    6-10 years3 (16)2 (28)1 (20)0 (0)
    11-15 years2 (11)1 (14)0 (0)1 (16)
    16-20 years0 (0)0 (0)0 (0)0 (0)
    21-25 years2 (11)1 (14)1 (20)0 (0)
    26-30 years2 (11)0 (0)1 (20)1 (16)
    31 years or more2 (11)0 (0)1 (20)1 (16)
    Missing1 (5)0 (0)0 (0)1 (16)
    Percentage of Medicaid patients seen
    Less than or equal to 30%4 (22)2 (28)1 (20)1 (16)
    Greater than 30%12 (66)5 (71)4 (80)3 (50)
    I do not see Medicaid patients1 (5)0 (0)0 (0)1 (16)
    Missing1 (5)0 (0)0 (0)1 (16)
    Age of patientsd
    Children10 (55)5 (71)3 (60)2 (33)
    Adults11 (61)0 (0)2 (40)2 (33)
    Older adults8 (44)5 (71)2 (40)1 (16)
    I do not see Medicaid patients1 (5)0 (0)0 (0)1 (16)

    aSES: socioeconomic status.

    bSOGI: sexual orientation and gender identity.

    cCombined category.

    dMultiple answers selected.

    Questionnaires were administered to participants remotely through the VARIAT app, and data collection took place from March to June 2020.

    Measures

    User Experience Measures

    Users’ reactions to both the race/SES and SOGI modules of the VARIAT program were assessed by asking participants about their perception of the modules after the test. After experiencing the AR simulation, users were asked questions designed to test their engagement with the AR experience, such as if they felt a sense of “being there” in the AR experience or how real they found the AR experience to be. These answers were scored on a scale of 1 to 5, with higher scores indicating stronger agreement. Participants were also asked questions about how they might apply the AR experience to their job with questions such as “How do you think this training will help you on the job (Mark all that apply)?” with different response items to assess perceived benefits from the training. These items were scored using dichotomous coding for each option (0 for not applicable and 1 for applicable).

    Preliminary Efficacy Measures

    Training outcomes were reported through changes in affective knowledge and changes in skill-based knowledge measured by comparing pre-post test responses. Affective knowledge (items assessing how participants expect their perceptions to impact their patients) was measured by agreement with items that were adapted from the California Brief Multicultural Competence Scale [31]. Example items include the following: “I am aware of how my own values might affect my patients” or “I am aware of institutional barriers that affect patients.” Skill-based knowledge was assessed differently for the race/SES and SOGI modules, with questions referring to each respective population focused on in the module. Race/SES skill-based knowledge was measured by rating participant agreement with the following internally developed statements: “I am confident that I can recognize the role that implicit bias plays in leading to inequitable care for patients of low socioeconomic status,” and “I am confident that I can apply strategies and use resources to minimize health care disparities for patients with low socioeconomic status.” SOGI skill-based knowledge was measured similarly, with “race/SES population” being replaced with “LGBTQ+ population” in the skills-based questions. Training outcomes were reported for each module separately, with the race/SES module (n=13) and SOGI module (n=11) consisting of all participants that completed each module. All measures were scored on a scale of 1 to 5, with higher scores representing stronger agreement.

    Data Analysis Plan

    All analyses were conducted using SPSS Statistics (version 27.0; IBM Corp). Demographic characteristics were described using frequencies and percentages for the categorical variables. Demographic characteristics were reported across 4 participant groupings: participants who took only the Race/SES module, participants who took only the SOGI module, participants who took both the Race/SES and SOGI modules, and an overall group of all unique participants.

    After testing for normality using the Shapiro-Wilk test, the training reactions and pre-post skills and attitude outcome data were found to not be normally distributed (P<.001). As a result, we used nonparametric tests for analyzing these 2 outcome domains. For the usability data, we used the Wilcoxon signed-rank test to measure the continuous reaction items and report the mean and SD for participants who used both the race/SES module and the SOGI module. The categorical training reaction items were reported using frequencies and percentages. For analyzing the skills and attitudes outcome data, the Wilcoxon matched-pairs signed-rank test was used to report the mean, SD, effect size (Cohen d), and 95% CI for pre-post changes in scores. The scores for each module were analyzed separately for all participants who took each respective module. Given that some participants completed both modules (n=6), there is a small amount of overlap in participant representation across all reported outcome data. All data and study materials will be made available on request.

    Ethical Considerations

    The Ohio State University (OSU) Institutional Review Board has determined this study was exempt from review according to the Policy on Human Subjects Research of the OSU Human Research Protection Program.


    User Experience (Training Reactions)

    For perception of the AR experience, participants who received training in the race/SES and SOGI modules reported similar ratings for the overall AR experience. Participants reported positive feelings of “being there” (race/SES module: mean score 4.62, SD 1.56; SOGI module: mean score 3.91, SD 1.97) and high relevance of the AR training to their respective jobs (race/SES module: mean score 4.75, SD 0.45; SOGI module: mean score 4.67, SD 0.50) across both modules. Participants across both modules perceived the AR experience as being “a little” realistic, with the SOGI participants reporting less realism on average (mean score 2.91, SD 1.64) compared to the race/SES participants (mean score 3.77, SD 1.83). For the reported intention to apply the AR experience to their jobs, only the participants who received training in the race/SES module responded to this item. On average, these participants reported that they were less likely to apply the AR experience to their jobs (mean score 2.31, SD 1.11).

    Assessing the perceived benefits of the AR experience to the participants’ jobs revealed that the race/SES and SOGI modules had some key differences in support across beliefs. Participants from both the race/SES and SOGI modules reported varying levels of positive agreement that the experience could improve their relationships with their patients (8/11, 73% SOGI participants; 8/13, 62% race/SES participants) and avoid undesirable events in patient care (8/11, 73% SOGI participants; 8/13, 62% SES participants). Conversely, 9 of 11 SOGI participants (82%) showed adequate agreement with the belief that the training would help improve tailored care and 7 of 11 participants (64%) believed that the training would improve patient satisfaction. For the race/SES participants, 7 of 13 (54%) showed moderate agreement with beliefs about improving tailored care, while 6 of 13 (46%) agreed that the training could improve patient satisfaction. The race/SES participants showed higher agreement with the belief that the module would improve their community resources (9/13, 69%) compared to the SOGI module participants (5/11, 45%). A detailed summary of user experience findings is reported in Table 2.

    Table 2. User experience (training reaction) outcomes.
    VariablesRace/SESa (n=13)SOGIb (n=11)Effect size (Cohen d)
    Augmented reality experience scores, mean (SD)
    Feeling of ”being there”c4.6 (1.6)3.9 (2.0)0.40
    Realism of augmented realityc3.7 (1.8)2.9 (1.6)0.49
    Relevance to jobd4.8 (0.5)4.7 (0.5)0.17
    Intention to apply augmented reality experienced2.3 (1.1)N/AeN/A
    Participants reporting applicability to job, n (%)f
    Improve relationship with patients8 (62)8 (73)N/A
    Improve patient satisfaction6 (46)7 (64)N/A
    Improve tailored care7 (54)9 (82)N/A
    Avoid undesirable events8 (62)8 (73)N/A
    Improve community resources9 (69)5 (45)N/A
    Other benefit0 (0)1 (9)N/A

    aSES: socioeconomic status.

    bSOGI: sexual orientation and gender identity.

    cMeasured on a scale from 1=not at all to 7=very much.

    dMeasured on a scale from 1=strongly disagree to 5=strongly agree.

    eN/A: not applicable.

    fMultiple answers selected.

    Preliminary Efficacy

    For the skills questions, there was no significant difference in pre-post scores assessing the changes in awareness of implicit bias for patients of varying race/SES groups (pre: mean 4.31, SD 0.48; post: mean 4.46, SD 0.52; d=0.22; 95% CI −0.77 to 0.33) or the ability to manage health disparities caused by race/SES group (pre: mean 3.85, SD 0.56; post: mean 4.31, SD 0.48; d=0.52; 95% CI −1.10 to 0.07) . This pattern was true for measuring awareness of implicit bias for LGBTQ+ patients (pre: mean 4.36, SD 0.51; post: mean 4.73, SD 0.47; d=0.54; 95% CI −1.16 to 0.11). For minimizing health disparities realted to LGBTQ+ status, there was a significant difference between pre- and posttest scores (pre: mean 3.91, SD 0.94; post: mean 4.64, SD 0.51; d=0.72; 95% CI −1.38 to −0.04) with participants scoring closer to “strongly agree” after experiencing the AR experience.

    For the attitudinal questions, there were nonsignificant improvements in the race/SES module in assessing how personal values affected patients (pre: mean 3.92, SD 0.95; post: mean 4.31, SD 0.84; d=0.44; 95% CI −1.01 to 0.14), how institutional barriers affect patients (pre: mean 4.23, SD 0.60; post: mean 4.31, SD 0.48; d=0.12; 95% CI −0.66 to 0.43), and participants’ ability to identify reactions based on stereotypes (pre: mean 4.15, SD 0.56; post: mean 4.38, SD 0.51; d=0.53; 95% CI −1.10 to 0.07). For the SOGI module, changes from pre to posttraining were also nonsignificant for all attitudinal items (pre: mean 4.18, SD 0.87; post: mean 4.45, SD 0.52; d=0.30; 95% CI −0.90 to 0.31), institutional barrier items (pre: mean 4.36, SD 0.67; post: mean 4.36, SD 0.67; d=0.00; 95% CI −0.59 to 0.59), and items related to identifying stereotypical reactions (pre: mean 4.36, SD 0.51; post: mean 4.36, SD 0.92; d=0.00; 95% CI −0.59 to 0.59). A summary of preliminary efficacy findings for each module can be found in Table 3.

    Table 3. Race/socioeconomic status (SES) pre- and posttest skills and attitude outcomes (n=13).
    VariablesPretest score, mean (SD)Posttest score, mean (SD)Cohen d (95% CI)
    Skills questions
    Implicit bias (race/SES)4.3 (0.5)4.5 (0.5)0.22 (–0.77 to 0.33)
    Minimize health disparities (race/SES)3.9 (0.6)4.3 (0.5)0.52 (–1.10 to 0.07)
    Attitudinal questions
    How my values affect patients3.9 (1.0)4.3 (0.5)0.44 (–1.01 to 0.14)
    How institutional barriers affect patients4.2 (0.6)4.3 (0.5)0.12 (–0.66 to 0.43)
    Identify reactions based on stereotypes4.2 (0.6)4.4 (0.5)0.53 (–1.10 to 0.07)

    This study developed a VR and AR implicit association training program for Medicaid providers based on cognitive psychology and the latest mobile simulation technologies. Designed to improve awareness of implicit biases related to patients’ SES and sexual orientation/gender identity, learners are able to complete six 5-minute interactive role-playing scenarios on their smartphones. Results of pilot user experience research among 18 participants found adequate acceptability and preliminary efficacy (ie, a nonsignificant increase in most outcomes) of the VARIAT program. These findings are consistent with recent literature in cognitive psychology about the possible benefits of AR interventions for health care providers [34-37].

    While researchers have spent the last 20 years attempting to reduce implicit bias [38-41], few attempts have been made to integrate the latest immersive technologies, such as AR and VR, with provider-level implicit bias training. For example, a recent meta-analysis of 492 interventions on implicit biases found only a handful of studies attempting to change implicit bias used any kind of VR or AR [42]. Narrowing down to implicit bias training in the health care setting, another recent literature review found few studies that focused on addressing bias at the provider level [43-46]. Therefore, while implicit bias in health care more broadly has been long recognized as a prominent issue [3], there is an important gap in research that develops technology-assisted training programs so that such programs can be more readily available for health care providers and so that implicit bias training can be received at a time and location that works best for them rather than having to attend in-person training sessions. The VARIAT program reported in this study addresses this critical literature gap by offering a convenient and publicly available program that can be integrated into medical training for health care professionals interacting with Medicaid patients, whose training may have important beneficial impacts on patients from disadvantaged backgrounds and those who experience reduced access to high quality of care due to multiple individual and societal barriers [47]. For example, the VARIAT program is brief and can be completed on a mobile device during “fragmented” time windows that fit within the often-chaotic work schedule of medical professionals. Therefore, medical institutions may consider integrating the VARIAT training as a regular refresh of lengthier and more comprehensive in-person or on-site bias-reduction training for their health care professional teams.

    Furthermore, among the studies that focused on mitigating health care provider biases, few documented detailed feasibility and efficacy data [48-51]. This study is among the first in the literature to measure both positive provider reactions and efficacy outcomes at multiple levels, including user experience with AR, perceived utility in users’ professional work, and perceived attitudes toward patients and skills in mitigating implicit biases at work. It was interesting to find that although the study participants perceived relatively high levels of immersion (“being there”), AR realism, and job relevance from the VARIAT training, they expressed low levels of intention to apply this experience to their daily work. One possible explanation for this discrepancy might be the challenges of translating learned knowledge to behavioral changes, as commonly seen in educational interventions, potentially due to the limited scenarios provided by the training compared to the broad variations in participants’ own daily work experiences. The collection of both pre- and postintervention efficacy outcomes further allowed us to measure the potential interventional effects of each of the VARIAT training modules (race/SES and SOGI). However, it should be noted that this paper focused primarily on sharing with the scientific community the development processes and design considerations of a novel implicit bias training program for Medicaid providers. Therefore, caution should be applied when interpreting the preliminary results of this pilot user experience study.

    Study Limitations

    There are several important limitations to this study. First, the current iteration of the VARIAT program is being delivered on mobile devices. This training program might elicit different user experiences and efficacy outcomes should it be delivered on other platforms such as through an immersive VR headset. Second, the study sample for this user experience testing study was small and potentially unbalanced. Larger sample sizes and a more rigorous study design (eg, a randomized controlled trial) should be used in future research to formally evaluate the efficacy of the VARIAT program with sufficient statistical power and without inflating the type II error rate [52,53]. Third, the present version of the VARIAT program only consisted of 2 modular domains for implicit bias training, race/SES and SOGI, with only 3 training scenarios for each module due to limitations on study resources and team expertise. Further, although these modules were developed by an interdisciplinary team of clinicians and researchers, patient communities were not involved in the design process. Future research will invite patient advisory groups into the development and refinement process of additional modules and scenarios for VARIAT to provide training in more comprehensive implicit bias domains during clinician-patient interactions. Fourth, this study used only self-reported measures developed by the study team to assess the efficacy outcomes, which may not be able to accurately measure biases that are inherently “implicit.” Future efficacy trials of the VARIAT program (and interventions alike) should incorporate validated implicit bias assessment tools such as the Implicit Association Test (IAT), which has been increasingly used by health care professionals in the existing literature [54]. Finally, several limitations of the study design should be noted. For example, this study did not restrict or record the number of times participants were allowed to undergo the training, which may have impacted usability and efficacy outcomes. Additionally, this study used an immediate pre-post training design. A more distant posttraining evaluation should be conducted to allow for examination of the impact of the modules on biases over time.

    Conclusions

    This study presents a novel intervention (VARIAT) that uses immersive mobile technology to improve awareness of implicit bias related to race/SES and SOGI among Medicaid providers. This publicly available training program has found a promising avenue for future research and practice in reducing implicit bias in health care workplaces. Future research should be conducted to formally evaluate the VARIAT program with large samples and implicit bias testing measures, as well as incorporate additional training domains to provide impactful benefits to both health care professionals and their patients.

    Acknowledgments

    This Medicaid Equity Simulation Project is funded by the Ohio Department of Medicaid and administered by the Ohio Colleges of Medicine Government Resource Center. The views expressed in this report are solely those of the authors and do not represent the views of the state of Ohio or federal Medicaid programs. Research efforts of the first (JS) and second (AJC) authors for this paper were partly supported by the Richard K. and Nancy L. Donahue Humanities Ethics Fellowship (University of Massachusetts Lowell).

    Conflicts of Interest

    J Penka is the founder and CEO and J Patterson is the cofounder of LittleSeed Inc. Both authors are members of the board of directors of LittleSeed Inc.

    Multimedia Appendix 1

    Race and socioeconomic status (SES) - module 1.

    DOCX File, 24 KB
    Multimedia Appendix 2

    Sexual orientation gender identity (SOGI) - module 2.

    DOCX File, 23 KB
    Multimedia Appendix 3

    User workflow.

    DOCX File, 959 KB
    1. Blair IV, Steiner JF, Havranek EP. Unconscious (implicit) bias and health disparities: where do we go from here? Perm J. 2011;15(2):71-78. [CrossRef] [Medline]
    2. Xu K, Nosek B, Greenwald A. Psychology data from the Race Implicit Association Test on the Project Implicit Demo website. J Open Psychol Data. Mar 18, 2014;2(1):e3. [CrossRef]
    3. FitzGerald C, Hurst S. Implicit bias in healthcare professionals: a systematic review. BMC Med Ethics. Mar 1, 2017;18(1):19. [CrossRef] [Medline]
    4. Bhat A, Mahajan V, Wolfe N. Implicit bias in stroke care: a recurring old problem in the rising incidence of young stroke. J Clin Neurosci. Mar 2021;85:27-35. [CrossRef] [Medline]
    5. Chapman EN, Kaatz A, Carnes M. Physicians and implicit bias: how doctors may unwittingly perpetuate health care disparities. J Gen Intern Med. Nov 2013;28(11):1504-1510. [CrossRef] [Medline]
    6. Dovidio JF, Eggly S, Albrecht TL, Hagiwara N, Penner L. Racial biases in medicine and healthcare disparities. TPM Test Psychom Methodol Appl Psychol. 2016;23(4):489-510. [CrossRef]
    7. Hall WJ, Chapman MV, Lee KM, et al. Implicit racial/ethnic bias among health care professionals and its influence on health care outcomes: a systematic review. Am J Public Health. Dec 2015;105(12):e60-e76. [CrossRef] [Medline]
    8. Saluja B, Bryant Z. How implicit bias contributes to racial disparities in maternal morbidity and mortality in the United States. J Womens Health (Larchmt). Feb 2021;30(2):270-273. [CrossRef] [Medline]
    9. Shen J, Carmichael A, Clinton AJ. A scoping review of research on potential impact of implicit bias in healthcare settings for children with acquired brain injuries. Rehabil Psychol. Feb 2024;69(1):36-44. [CrossRef] [Medline]
    10. Lelutiu-Weinberger C, Clark KA, Pachankis JE. Mental health provider training to improve LGBTQ competence and reduce implicit and explicit bias: a randomized controlled trial of online and in-person delivery. Psychol Sex Orientat Gend Divers. Dec 2023;10(4):589-599. [CrossRef] [Medline]
    11. McDowell MJ, Goldhammer H, Potter JE, Keuroghlian AS. Strategies to mitigate clinician implicit bias against sexual and gender minority patients. Psychosomatics. 2020;61(6):655-661. [CrossRef] [Medline]
    12. Sabin JA, Riskind RG, Nosek BA. Health care providers’ implicit and explicit attitudes toward lesbian women and gay men. Am J Public Health. Sep 2015;105(9):1831-1841. [CrossRef] [Medline]
    13. Wolsiefer KJ, Mehl M, Moskowitz GB, et al. Investigating the relationship between resident physician implicit bias and language use during a clinical encounter with Hispanic patients. Health Commun. Jan 2023;38(1):124-132. [CrossRef] [Medline]
    14. Ivers NN, Johnson DA, Rogers JL. The association between implicit racial bias and mindfulness in mental health practitioners. J Couns Dev. Jan 2021;99(1):11-23. [CrossRef]
    15. Hagiwara N, Elston Lafata J, Mezuk B, Vrana SR, Fetters MD. Detecting implicit racial bias in provider communication behaviors to reduce disparities in healthcare: challenges, solutions, and future directions for provider communication training. Patient Educ Couns. Sep 2019;102(9):1738-1743. [CrossRef] [Medline]
    16. Maina IW, Belton TD, Ginzberg S, Singh A, Johnson TJ. A decade of studying implicit racial/ethnic bias in healthcare providers using the implicit association test. Soc Sci Med. Feb 2018;199:219-229. [CrossRef] [Medline]
    17. Sabin JA, Rivara FP, Greenwald AG. Physician implicit attitudes and stereotypes about race and quality of medical care. Med Care. Jul 2008;46(7):678-685. [CrossRef] [Medline]
    18. Maxfield CM, Thorpe MP, Desser TS, et al. Awareness of implicit bias mitigates discrimination in radiology resident selection. Med Educ. Jul 2020;54(7):637-642. [CrossRef] [Medline]
    19. Johnson TJ, Winger DG, Hickey RW, et al. Comparison of physician implicit racial bias toward adults versus children. Acad Pediatr. Mar 2017;17(2):120-126. [CrossRef] [Medline]
    20. Sabin JA, Greenwald AG. The influence of implicit bias on treatment recommendations for 4 common pediatric conditions: pain, urinary tract infection, attention deficit hyperactivity disorder, and asthma. Am J Public Health. May 2012;102(5):988-995. [CrossRef] [Medline]
    21. Rojas M, Walker-Descartes I, Laraque-Arena D. An experimental study of implicit racial bias in recognition of child abuse. Am J Health Behav. May 1, 2017;41(3):358-367. [CrossRef] [Medline]
    22. Amdani S, Bhimani SA, Boyle G, et al. Racial and ethnic disparities persist in the current era of pediatric heart transplantation. J Card Fail. Sep 2021;27(9):957-964. [CrossRef] [Medline]
    23. Simon SL, Clay D, Chandrasekhar J, Duncan CL. Gender bias in pediatric psychology. Clin Pract Pediatr Psychol. 2021;9(1):82-95. [CrossRef]
    24. Seibel LM, Fehr FC, Sarwal S, Panchuk HJ. Impact of cognitive rehearsal training over time for new registered nurses. J Contin Educ Nurs. May 2022;53(5):203-211. [CrossRef] [Medline]
    25. Kang J, Jeong YJ. Effects of a smartphone application for cognitive rehearsal intervention on workplace bullying and turnover intention among nurses. Int J Nurs Pract. Dec 2019;25(6):e12786. [CrossRef] [Medline]
    26. Teal CR, Shada RE, Gill AC, et al. When best intentions aren't enough: helping medical students develop strategies for managing bias about patients. J Gen Intern Med. May 2010;25 Suppl 2(Suppl 2):S115-S118. [CrossRef] [Medline]
    27. Gonzalez CM, Walker SA, Rodriguez N, Noah YS, Marantz PR. Implicit bias recognition and management in interpersonal encounters and the learning environment: a skills-based curriculum for medical students. MedEdPORTAL. 2021;17:11168. [CrossRef] [Medline]
    28. Sukhera J, Watling CJ, Gonzalez CM. Implicit bias in health professions: from recognition to transformation. Acad Med. May 2020;95(5):717-723. [CrossRef] [Medline]
    29. McMichael B, Nickel A, Duffy EA, et al. The impact of health equity coaching on patient’s perceptions of cultural competency and communication in a pediatric emergency department: an intervention design. J Patient Exp. Dec 2019;6(4):257-264. [CrossRef] [Medline]
    30. Long A, Jennings J, Bademosi K, et al. Storytelling to improve healthcare worker understanding, beliefs, and practices related to LGBTQ + patients: a program evaluation. Eval Program Plann. Feb 2022;90:101979. [CrossRef] [Medline]
    31. Zipp SA, Craig SD. The impact of a user’s biases on interactions with virtual humans and learning during virtual emergency management training. Educ Technol Res Dev. Dec 2019;67(6):1385-1404. [CrossRef]
    32. Tang KS, Cheng DL, Mi E, Greenberg PB. Augmented reality in medical education: a systematic review. Can Med Educ J. Mar 2020;11(1):e81-e96. [CrossRef] [Medline]
    33. Dhar P, Rocks T, Samarasinghe RM, Stephenson G, Smith C. Augmented reality in medical education: students’ experiences and learning outcomes. Med Educ Online. Dec 2021;26(1):1953953. [CrossRef] [Medline]
    34. Chiang DH, Huang CC, Cheng SC, et al. Immersive virtual reality (VR) training increases the self-efficacy of in-hospital healthcare providers and patient families regarding tracheostomy-related knowledge and care skills: a prospective pre-post study. Medicine (Baltimore). Jan 14, 2022;101(2):e28570. [CrossRef] [Medline]
    35. Croghan IT, Hurt RT, Aakre CA, et al. Virtual reality for health care professionals during a pandemic: a pilot program. J Prim Care Community Health. 2022;13:21501319221086716. [CrossRef] [Medline]
    36. Hayakawa J, Barrows J, See S, Schomberg J. Effects of classical music virtual reality on pediatric healthcare worker compassion fatigue. J Nurs Adm. May 1, 2022;52(5):280-285. [CrossRef] [Medline]
    37. Beverly E, Hommema L, Coates K, et al. A tranquil virtual reality experience to reduce subjective stress among COVID-19 frontline healthcare workers. PLoS One. 2022;17(2):e0262703. [CrossRef] [Medline]
    38. Barnes-Holmes D, Barnes-Holmes Y, Power P, Hayden E, Milne R, Stewart I. Do you really know what you believe? Developing the implicit relational assessment procedure (IRAP) as a direct measure of implicit beliefs. Ir Psychol. 2006;32(7):169-177. URL: https:/​/www.​researchgate.net/​publication/​241832168_Do_You_Really_Know_What_You_Believe_Developing_the_Implicit_Relational_Assessment_Procedure_IRAP_as_a_Direct_Measure_of_Implicit_Beliefs
    39. Greenwald AG, Poehlman TA, Uhlmann EL, Banaji MR. Understanding and using the implicit association test: III. Meta-analysis of predictive validity. J Pers Soc Psychol. Jul 2009;97(1):17-41. [CrossRef] [Medline]
    40. Lai CK, Marini M, Lehr SA, et al. Reducing implicit racial preferences: I. A comparative investigation of 17 interventions. J Exp Psychol Gen. Aug 2014;143(4):1765-1785. [CrossRef] [Medline]
    41. Whitford DK, Emerson AM. Empathy intervention to reduce implicit bias in pre-service teachers. Psychol Rep. Apr 2019;122(2):670-688. [CrossRef] [Medline]
    42. Forscher PS, Lai CK, Axt JR, et al. A meta-analysis of procedures to change implicit measures. J Pers Soc Psychol. Sep 2019;117(3):522-559. [CrossRef] [Medline]
    43. Ricks TN, Abbyad C, Polinard E. Undoing racism and mitigating bias among healthcare professionals: lessons learned during a systematic review. J Racial Ethn Health Disparities. Oct 2022;9(5):1990-2000. [CrossRef] [Medline]
    44. Gran-Ruaz S, Feliciano J, Bartlett A, Williams MT. Implicit racial bias across ethnoracial groups in Canada and the United States and Black mental health. Can Psychol. 2022;63(4):608-622. [CrossRef]
    45. Vo JB, Gillman A, Mitchell K, Nolan TS. Health disparities: impact of health disparities and treatment decision-making biases on cancer adverse effects among Black cancer survivors. Clin J Oncol Nurs. Oct 1, 2021;25(5):17-24. [CrossRef] [Medline]
    46. Newcomb AB, Rothberg S, Zewdie M, et al. Managing patient bias: teaching residents to navigate racism and bias in the workplace. J Surg Educ. 2021;78(6):1791-1795. [CrossRef] [Medline]
    47. Hsiang WR, Lukasiewicz A, Gentry M, et al. Medicaid patients have greater difficulty scheduling health care appointments compared with private insurance patients: a meta-analysis. Inquiry. 2019;56:46958019838118. [CrossRef] [Medline]
    48. Alexis C, White AM, Thomas C, Poweski J. Unpacking reactions to white privilege among employees of an academic medical center. J Cult Divers. 2019;26(1). URL: https://www.proquest.com/docview/2201638179 [Accessed 2024-03-04]
    49. Crawford C. From the American Academy of Family Physicians: the Everyone Project unveils Implicit Bias Training Guide. Ann Fam Med. 2020;18(2):182. [CrossRef]
    50. Nelson SC, Prasad S, Hackman HW. Training providers on issues of race and racism improve health care equity. Pediatr Blood Cancer. May 2015;62(5):915-917. [CrossRef] [Medline]
    51. Sherman MD, Ricco J, Nelson SC, Nezhad SJ, Prasad S. Implicit bias training in a residency program: aiming for enduring effects. Fam Med. Sep 2019;51(8):677-681. [CrossRef] [Medline]
    52. Guo Y, Logan HL, Glueck DH, Muller KE. Selecting a sample size for studies with repeated measures. BMC Med Res Methodol. Jul 31, 2013;13:100. [CrossRef] [Medline]
    53. Overall JE, Starbuck RR. Sample size estimation for randomized pre-post designs. J Psychiatr Res. 1979;15(1):51-55. [CrossRef] [Medline]
    54. Sukhera J, Wodzinski M, Rehman M, Gonzalez CM. The Implicit Association Test in health professions education: a meta-narrative review. Perspect Med Educ. Oct 2019;8(5):267-275. [CrossRef] [Medline]

    ‎
    AR: augmented reality
    LGBTQ+: lesbian, gay, bisexual, transgender, queer
    SES: socioeconomic status
    SOGI: sexual orientation and gender identity
    VARIAT: Virtual and Augmented Reality-based Implicit Association Training
    VR: virtual reality

    Edited by Gunther Eysenbach, Tiffany Leung; submitted 29.07.23; peer-reviewed by Elizabeth Chuang, Olivia Joseph, Jong Long Guo; final revised version received 24.10.23; accepted 12.12.23; published 07.03.24.

    Copyright

    © Jiabin Shen, Alex J Clinton, Jeffrey Penka, Megan E Gregory, Lindsey Sova, Sheryl Pfeil, Jeremy Patterson, Tensing Maa. Originally published in JMIR Serious Games (https://games.jmir.org), 7.3.2024.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Serious Games, is properly cited. The complete bibliographic information, a link to the original publication on https://games.jmir.org, as well as this copyright and license information must be included.