The study was carried out as part of the “Brain Death/Death by Neurologic Criteria” (BD/DNC) examination course at the University of Münster during the winter term 2025-2026. The course is scheduled for the seventh semester of the medical curriculum. A VR-based digital simulation depicting a clinical scenario for BD/DNC examination served as the research environment [13-15]. The VR environment replicates a realistic intensive care unit in which participants can interact with a virtual patient and perform clinical examinations. The course is part of a block week focusing on organ donation and has been delivered in a VR-based format since 2021. Further information about the course structure and procedures can be found in Junga et al [13,16]. Every student received a technical and medical preparation video podcast to overcome knowledge and computer literacy inequalities.

Although participation in the course was mandatory, participation in the study was voluntary. Students who declined to participate did not experience any disadvantages. Students enrolled in the course were recruited and assigned to 1 of 2 conditions through quasi-random allocation by room. Rooms were preassigned to either a control condition (“NoSmell”), in which participants performed the standard brain death examination, or an intervention condition (“Smell”), in which the same simulation was augmented with an olfactory stimulus. Participants selected their room without knowledge of the assigned condition or about the assignment at all. Inclusion criteria included enrollment in medical school, course attendance, and willingness to participate in the study.

The study was conducted in accordance with the Declaration of Helsinki and was approved by the responsible ethics committee (Ethik-Kommission Westfalen-Lippe; application 2022‐736-b-S; approved on November 9, 2022). Participation in the study was voluntary and independent of course participation, completion, and assessment, and declining to participate entailed no disadvantage. Written informed consent was obtained from all participants prior to allocation through a digital survey form. All study data were collected and analyzed in pseudonymized form using 6-character study codes, and no directly identifying information was retained at the analysis stage. No financial or nonfinancial compensation was provided for study participation.

The required sample size was determined a priori using G*Power (version 3.1; Heinrich Heine University Düsseldorf) [17]. For the primary dichotomous outcomes (hand disinfection and glove use, each coded as true or false), a medium effect size (w=0.30) [18] was assumed, and a χ²₁ test for a 2×2 contingency table was specified. With an α level of .05 and a statistical power of 80%, the required total sample size was 88. To account for potential dropouts, a minimum of 90 participants was targeted for recruitment.

In accordance with the recommendations of the Commission for Hospital Hygiene and Infection Prevention (KRINKO) at the Robert Koch Institute, hand disinfection is mandatory before patient contact or contact with the patient’s environment (eg, bed, side table, and catheter bag), before aseptic procedures, and after contact with potentially contaminated material or the patient and their surroundings. The use of medical gloves is recommended for “protection of the wearer against contamination with blood, secretions, and excretions, including pathogens, and indirectly to interrupt chains of infection” [19].

For the purposes of this study, the functionalities for using gloves and a hand disinfectant dispenser were specifically implemented and enabled within the VR environment (Figure 1A, B). To use the dispenser, at least one hand had to be placed underneath it, and a hissing sound was generated to confirm the interaction. To use the gloves, the participant’s hands had to touch the glove box, after which the gloves were attached automatically.

The software was further modified so that, in addition to interactions associated with a potential risk of exposure to bodily fluids (eg, gag reflex, tracheal reflex, and caloric-induced nystagmus), the examination procedure deliberately involved contact with a simulated extensive skin abrasion in the facial area (Figure 1C, D). This design rendered glove use mandatory for hygiene, in addition to hand disinfection, which is strictly required in intensive care settings.

As demonstrated in Junga et al [8], the representation of the hands significantly affects glove usage. Hence, the application was modified to provide the avatar with a bare hand model, as opposed to the previous one, which was equipped with gloves by default.

The objectives stated in Table 1 were identified in accordance with the recommendations for hand hygiene issued by the Robert Koch Institute [19]. Primary and explorative outcomes were analyzed separately. All actions were automatically tracked and documented by the software. Before and after the course, participants completed a survey on demographics and prior VR experience. No item assessed scent perception or other manipulation-check questions.

All end points were operationalized as binary variables. Hand disinfection was considered to have been performed if the participant activated the virtual disinfectant dispenser with at least one hand. Glove use was recorded if the participant interacted with the glove box using at least one hand, thereby triggering the glove application within the virtual environment.

In 3 of the 6 rooms, an olfactory stimulus was applied by placing 5 drops (approximately 0.25 mL) of an alcohol-based hand disinfectant (desmanol care, Schülke & Mayr GmbH) onto the facial interface of the VR headset, specifically at the upper edge of the nasal cutout (Figure 2). This application was repeated immediately before each new rotation (<3 min) to avoid differences in smell intensity.

Due to the use of additional hygienic face masks, there was no direct skin contact with the students’ faces, but the fluid was placed less than 5 mm from the bridge of the nose without any visible manipulation of the HMD. Participants were blinded to both the presence and the purpose of the olfactory intervention. The use of hand disinfectants and gloves in the virtual environment was automatically tracked by the system, and the data were exported after course completion.

The VR scenario was delivered using the BD/DNC Examination software (version December 2025; University of Münster) [13-15] on a Valve Index HMD with Knuckles controllers (Valve Corporation). Data were collected using LimeSurvey and analyzed with R (version 4.5.2; R Foundation for Statistical Computing) using the tidyverse [20] and broom [21] packages.

In the winter term 2025-2026, 132 individuals were assessed for eligibility, of which 130 responded to the consent request. Of these, 35 were excluded: 29 refused consent, and the consent status of 6 remained unclear. The remaining 95 participants were allocated to the control group (n=43) or the intervention group (n=52). All 95 allocated participants completed the VR simulation and were analyzed according to their assigned condition (intention-to-treat analysis). Six participants in the intervention group did not provide demographic data and were therefore excluded from the complete-case sensitivity analyses only (complete-case: n=89, comprising 43 control and 46 intervention participants). The minor imbalance in group sizes reflects the quasi-random allocation by room, as final numbers were determined only after all sessions had been completed. The flow of participants through the study is shown in Figure 3.

As shown in Table 2, participants had a mean age of 23.6 (SD 2.6; range 21‐35) years. Age did not differ significantly between conditions (24.0 vs 23.2 y; t_67.8=1.51; P=.14). Female students (59/89, 66.3%) were overrepresented in the sample, consistent with the sex composition of medical students in Germany [22]. However, the sex distribution did not differ significantly between groups (32/43, 74.4% vs 27/46, 58.7% female participants; χ²₁=2.46; P=.12).

In the subsequent survey about prior HMD experience, only 3 participants reported owning a personal VR headset, all of whom were male. However, the majority of students (52/89, 58.4%; regularly+once or rarely) had previously used a VR headset at least once (for further information, see Table 3).

The primary hypothesis, that olfactory cues would increase hand hygiene behavior, was partially supported. Hand disinfection increased significantly under olfactory exposure, whereas glove use showed no significant effect. Furthermore, exposure to the scent of disinfectants increased postcontact hand disinfection, whereas no effects were observed for initial hand disinfection or initial glove use. Because these behavioral changes were observed within a simulated environment, they correspond to Kirkpatrick level 3 (behavior) within simulation and serve as a proxy for, rather than direct evidence of, transfer to real-world clinical practice.

To contextualize these findings, it is important to note that not all hand hygiene behaviors are triggered by the same cues. While initial hand disinfection, which directly targets patient safety, represents a highly trained and routinized behavior, which is done straight as the first action while entering the room, the decision to use gloves is typically reassessed in each situation. This assessment depends on perceived exposure risks, such as contact with bodily fluids—either overt (eg, wound exudate) or subtle (eg, sweat)—as well as the need to protect one’s own skin during wet procedures. These evaluations are guided by sensory impressions.

Our findings suggest that positively valenced cues, such as the clean scent of disinfectants, may reinforce reactive behaviors (eg, final hand disinfection) rather than initiating preventive routines, whereas other behaviors, such as glove use or initial hand hygiene, may require more salient or negatively valenced cues to be affected. Beyond this descriptive pattern, the dissociation between hand disinfection and glove use is itself theoretically informative. The scent of disinfectant carries a strong, repeated real-world association with the act of hand disinfection, whereas no comparable associative link exists for glove use. This selective effect suggests that the olfactory cue operates through specific stimulus-response priming rather than a generalized activation of patient safety behavior, implying that effective sensory augmentation may need to match cues to target behaviors based on learned associations. As glove removal was not possible within the simulation, no conclusions regarding postglove hand hygiene can be drawn from the present data. In line with the existing literature, our findings confirm that positively valenced olfactory cues can also exert a beneficial effect on hand hygiene behavior in VR-based simulations, similar to the effects reported by Birnbach et al and King et al [10,11]. Notably, in contrast to these prior studies, which used a citrus scent, the present study used the smell of a disinfectant as the olfactory stimulus.

In contrast, Schmidtke et al [12] were unable to demonstrate a consistent effect of olfactory priming on hand hygiene behavior in a real-world hospital entrance setting, suggesting that the effectiveness of olfactory cues may depend on contextual specificity and clearly defined patient contact situations. To our knowledge, no prior studies have investigated the influence of olfactory cues on medical glove use, indicating a relevant gap in the current literature.

Beyond the behavior-specific effects, these results raise broader questions regarding the role of sensory modalities in VR-based patient safety training. Visual and auditory cues are inherently embedded in contemporary VR systems and can be rendered digitally with high fidelity. In contrast, olfactory stimuli cannot be generated virtually and must originate from the physical world. Thus, smell represents the only deliberately introduced real sensory cue in this study, effectively crossing the boundary between virtual and physical experience—a boundary we refer to as the virtual threshold.

This concept relates to several established constructs in the simulation and cognitive psychology literature. Presence and embodiment, defined above, both depend on sensory congruence between the simulated scene and learners’ real-world expectations. Ecological validity, in turn, refers to the correspondence between the simulation and the clinical situation it is intended to prepare for. From a memory perspective, the principle of context-dependent retrieval predicts that behaviors learned and habitually performed under specific sensory conditions are more readily activated when those conditions are reinstated. Within this framework, the disinfectant scent can be understood as a sensory retrieval cue that reinstantiates clinically learned associations within an otherwise digitally generated environment, thereby narrowing the gap between virtual and clinical contexts.

Despite its central role in clinical practice, olfaction has largely been neglected in VR-based medical education. As demonstrated in this study, olfactory cues can influence safety-relevant behavior even when visual and auditory cues alone are insufficient. Evidence on olfactory augmentation in immersive medical simulation remains sparse. The most recent systematic review was published in 2016 by Kent et al [23] and primarily investigated olfactory cues in non-VR simulation contexts, leaving the role of real olfactory input in immersive VR-based medical education largely unexplored.

Insights from psychology and entertainment-oriented VR research suggest that olfactory cues can enhance emotional salience, contextual memory, and behavioral realism [24-26]. However, these findings have rarely been translated into medical education, where multisensory perception is fundamental to safe clinical action.

This study should be interpreted in light of both theoretical mechanisms and methodological constraints relevant to immersive medical education. Immersive VR aims not only to convey procedural knowledge but also to elicit self-relevant clinical behavior. In this context, it is important to distinguish between immersion (objective system properties), presence (the subjective feeling of “being there”), and embodiment, which refers to the experience of the virtual body as one’s own. Embodiment is commonly conceptualized along the dimensions of body ownership, agency, and self-location and has been identified as a key determinant of behavior in virtual environments, particularly when actions involve self-protection or bodily risk [27-29].

Although olfactory cues do not directly modify visual body representation, they may indirectly reinforce embodiment by strengthening the coupling between bodily actions and real-world physiological expectations. In clinical settings, smells associated with contamination, cleanliness, or disinfection are closely linked to bodily vulnerability and habitual protective routines. Introducing such cues from the physical world into an otherwise fully virtual environment may therefore increase the salience of the virtual body as a potential site of risk. This mechanism could help shift learner behavior from abstract rule compliance toward self-protective, behaviorally realistic responses, such as postcontact hand disinfection. This proposed mechanism was not directly tested in the present study and remains a hypothesis for future investigation.

The relatively low baseline rate of hand disinfection in the control group warrants explicit consideration. Several factors plausibly contribute. First, the novelty of the VR environment may divert attention from routine behaviors. Second, the environmental cues that normally prompt hand hygiene in real settings, such as ambient scent, tactile feedback from the dispenser, and the physical proximity of cleansed surfaces, are absent or only weakly present in a purely audiovisual VR scene. Third, learners may operate with a reduced perceived consequence in an obviously simulated context. Under this reading, the olfactory intervention does not enhance hygiene behavior beyond real-world baselines but rather restores cue-triggered behavior under impoverished sensory conditions. This interpretation does not diminish the practical value of the finding because, if VR environments inherently suppress hygiene-relevant behavior, sensory augmentation is precisely what is required to elicit ecologically valid training responses.

However, several limitations constrain the interpretation and generalizability of these findings. First, the olfactory intervention was limited to a single, positively connoted scent associated with cleanliness. While appropriate for feasibility and safety, this does not reflect the full range of clinically relevant olfactory cues. Future studies should include aversive or negatively valenced odors, which may exert stronger or qualitatively different effects on risk perception and behavior.

Second, participant allocation was constrained by faculty-assigned groups, resulting in limited group sizes and minor imbalances. Although the sample size was sufficient to detect the main effects, these constraints may have reduced statistical precision and may have limited subgroup analyses. Six participants were excluded from the analyses due to missing data, all of whom were in the intervention group, reducing the sample from 95 to 89. Sex distribution showed a nonsignificant imbalance (P=.12). Because allocation was quasi-random by room rather than by individual-level randomization, the primary intention-to-treat analysis was complemented by covariate-adjusted sensitivity analyses, and the intervention effect on hand disinfection remained robust across these models. Because participants were allocated within 6 rooms, clustering at the room level cannot be fully excluded, and with only 6 clusters, formal multilevel modeling was not feasible. Third, the behavioral effects observed were behavior-specific: olfactory augmentation influenced overall and postcontact hand disinfection but not initial hand hygiene or glove use. This suggests that olfactory cues may preferentially reinforce reactive, situationally triggered behaviors rather than preventive or purely rule-based routines.

Fourth, the study was conducted within a single VR clinical scenario, which limits extrapolation to other procedures, specialties, or safety behaviors. Broader evaluation across multiple scenarios and end points is required to determine the generalizability of the effects of olfactory augmentation. Finally, technical constraints of the VR tracking system limited the granularity of behavioral measurement. Certain interaction sequences, such as returning to patient contact after hand disinfection outside the tracked examinations, could not be reliably detected, potentially leading to an overestimation of compliance. Future implementations should incorporate more fine-grained temporal and interaction tracking to better capture hygiene-critical behavior.

Fifth, no formal manipulation check was conducted to verify whether participants consciously perceived the olfactory stimulus. We therefore cannot determine whether the effect operates through conscious recognition, such as identifying the disinfectant scent and responding accordingly, or through subconscious priming. Future studies should include a brief postsimulation item assessing scent perception, which would enable stratified analyses of participants who did and did not consciously notice the scent and help adjudicate between conscious and subconscious mechanisms.

A further question concerns the transferability of these findings to experienced clinicians. Our sample comprised medical students with limited clinical experience, whose hygiene behaviors are not yet fully automated. Experienced practitioners might rely on highly routinized motor patterns that are less susceptible to environmental cues, a ceiling interpretation, or they might instead have more strongly encoded cue-behavior associations and therefore respond more readily to congruent sensory primes. Both predictions are plausible and yield different design implications for VR-based continuing education, and which one holds is an empirical question for future studies.

Taken together, these findings and limitations highlight both the promise and the current constraints of olfactory augmentation in VR-based medical education. Future research should systematically vary olfactory valence, improve behavioral tracking fidelity, optimize randomization procedures, and expand scenario diversity. From an educational perspective, the results suggest that olfactory cues may serve as a targeted design element to promote behaviorally realistic, safety-relevant actions in immersive simulations, particularly when the educational goal is not knowledge acquisition alone but the elicitation of authentic clinical behavior.

This study provides empirical evidence that olfactory cues—an integral yet previously neglected component of clinical environments—can meaningfully enhance behavioral outcomes in immersive VR-based medical education. The introduction of a positively connoted disinfectant scent led to a 19.9 percentage point increase in hand hygiene behavior during simulation training, suggesting that real-world sensory augmentation can strengthen safety-relevant actions beyond purely visual and auditory immersion.

In contrast, glove use was not significantly influenced by the olfactory intervention, indicating that different hygiene behaviors may rely on distinct cognitive mechanisms. While hand disinfection appears susceptible to context-driven sensory priming, glove use may depend more strongly on rule-based or procedural decision processes.

Taken together, these findings highlight the potential of integrating physical sensory cues into digital simulation environments to enhance behavioral realism and patient safety training. Future research should investigate the effects of diverse olfactory stimuli across scenarios and learner populations to further elucidate the mechanisms and generalizability of sensory augmentation in VR-based medical education.