This nonrandomized controlled simulation study with a pretest-posttest design evaluated a neonatal resuscitation gamification program using immersive VR [18]. Prelicensure nursing students were divided into intervention and control groups. The study assessed outcomes such as neonatal resuscitation nursing knowledge, problem-solving skills, clinical reasoning ability, self-confidence in practical performance, anxiety levels, and learning motivation. The simulation group presented lower anxiety levels, compared to the VR and control groups. Anxiety was measured with the STAI (State-Trait Anxiety Inventory), a commonly used questionnaire to assess an individual’s tendency to suffer anxiety. Limitations include the inability to assess long-term effects through follow-up surveys and the lack of measurement for actual intervention competency reinforcement, relying solely on self-reported questionnaires.

This prospective pre-post study compared health care providers’ tracheostomy care training using immersive VR with head-mounted displays and web-based modules versus traditional text materials [19]. According to a personalized Likert-scale questionnaire, most providers in the VR group found that interactive visual demonstrations improved comprehension and reduced anxiety. Limitations included a small sample size, a short follow-up period, and reliance on self-reported feedback rather than quantitative measures of skill acquisition and patient outcomes. Larger, long-term studies with objective assessments are needed to evaluate the efficacy of VR training in improving tracheostomy care competency.

An unblinded, randomized, crossover simulation-based study evaluated whether augmented reality cardiopulmonary resuscitation (AR-CPR) improves chest compression performance in nonpediatric-specialized community emergency departments [20]. Participants performed chest compression with and without AR-CPR guidance in random order. Qualitative interviews suggested AR-CPR could be usable without device orientation, effective at cognitive offloading, and capable of reducing anxiety while boosting confidence. However, limitations included not excluding individuals with corrective eye lenses, which might affect the AR experience and the lack of feedback during non-AR-CPR cycles, unlike real-world scenarios where feedback devices are common.

The prospective cohort pre and post study evaluated a new VR game, designed to teach safe and unsafe behaviors regarding universal precautions on needlestick and sharp injury prevention among incoming medical and nursing interns in Taiwan [21]. The game focused on making correct safety choices. Many participants reported reduced anxiety about preventing these injuries in the personalized Likert-scale questionnaires. However, the study’s reliance on self-reported questionnaires could introduce reporting bias, and trainees might report behaviors aligning with their training, potentially skewing results.

A 2-stage sequential mixed-methods feasibility study assessed the impact of an immersive VR sepsis game on preregistration nurses [22]. The study examined its effect on self-efficacy and perceptions of its acceptability and applicability in nursing simulation education. In the first stage, pre and postintervention self-efficacy scores were collected from 19 preregistration nurses using the Nursing Anxiety and Self-Confidence with Clinical Decision-Making Scale (NASC-CDM). The second stage used a descriptive qualitative approach to explore perceptions of the game. Results showed a significant 23.4% decrease in anxiety. Limitations included the small sample size, the novelty of the educational approach, and the measurement of self-efficacy at a single time point.

This randomized pretest-posttest study compared the effectiveness of 2 maternal newborn clinical simulation scenarios: VR clinical simulation and face-to-face high-fidelity manikin simulation [23]. Although no statistically significant differences were found in student knowledge and self-confidence between the 2 modalities, anxiety scores, measured by the NASC-CDM, were higher for students in the VR simulation. Limitations included a small sample size, potential intervening variables, such as student motivation, interest, and technological competence, and a lack of orientation to the VR platform, which may have influenced their responses.

This randomized control trial assessed the educational effectiveness of practical exercises using intravenous simulators with VR/haptics technologies [24]. First-year nursing students were assigned randomly to three groups: Group A (conventional intravenous arm), Group B (VR/haptics intravenous simulator), and Group C (both intravenous arm and simulator). Group C scored highest in venipuncture procedures, while Group B excelled in injections. Group C completed venipuncture faster than Group B and slightly quicker than Group A. State-trait anxiety was measured using a Visual Analogue Scale (VAS) before and after the intervention. All groups showed reduced anxiety postvenipuncture, with no significant differences. Limitations included insufficient practice time due to curriculum constraints and the study’s single-school, single-country setting, which limits generalizability and cultural diversity.

After analyzing the risk of bias of the 7 articles included in the systematic review, summarized in Table 4, 1 [18] had a high risk in the randomization process because the participants were assigned systematically to the intervention and control groups. In total, 2 studies [21,22] lacked a control group and conducted pre and postintervention evaluations. The remaining 4 studies [19,20,23,24] demonstrated a low risk of bias.

The risk of deviations from the intended intervention raised some concerns in 3 studies due to the motivation of the intervention group related to their interest in the technology [19,23,24]. No risk was detected due to missing outcome data. However, some concerns arose in studies [18,20,21], regarding outcome measurement, as only the intervention group was evaluated in certain aspects. Finally, no risk was identified in any of the studies concerning the selection of reported results. In conclusion, all studies presented some concerns, but none was deemed to have a high risk of bias overall.

All of the papers had a common general learning objective: simulating and performing procedures. However, the clinical foci covered in the selected papers differed.

Overall, 2 of the studies focused on reducing anxiety around newborn care. The first was with an immersive VR neonatal resuscitation gamification program, enabling hands-on experience in a VR environment [18]. The second, a simulation-based pilot study used AR to teach pediatric chest compression [20].

In total, 2 studies used serious games to teach and reduce anxiety in students. Occupational needle stick or sharp injury prevention was sought through a game of right and wrong choices for safe or unsafe universal precaution behaviors [21]. The other used a 3D, computer-based simulation sepsis game [22].

In 3 studies, they opted for a procedure-specific training, Cobbett and Snelgrove-Clarke [23], evaluated VR clinical simulation in preeclampsia and Group B Streptococcus scenarios. Another study divided the first-year nursing students into 3 groups, learning how to practice injections. A VR intravenous training simulator using haptic skills was evaluated, along with its effect on reducing students’ anxiety toward this procedure [24]. Finally, in [19] used a combination of head-mounted display VR and smartphone application VR in an environment for tracheostomy-related materials. The VR teachings on how to dress the stoma or handle emergencies such as aspiration, aimed to reduce anxiety among health care providers.

This analysis reveals significant methodological differences among the articles, particularly in their approaches to study design, including quasi-experimental, randomized controlled trials (RCTs), mixed methods, pre-post, and crossover designs.

RCTs ([20,23,24]) provide stronger evidence due to their randomization and ability to control the relationship between the XR interventions and outcomes. Quasi-experimental ([18]) and pre-post ([20,22,23]) designs are valuable for exploratory or feasibility research, but are less robust due to biases and limitations in establishing causality. Mixed-methods designs ([20,22]) add depth to the analysis by incorporating qualitative perspectives but lack the quantitative generalizability of RCTs. Finally, crossover designs ([20]) offer a unique advantage by enabling within-subject comparisons, which may result in more reliable conclusions about XR effectiveness.

The studies demonstrate that the choice of study design is dependent on the research goals, feasibility, and the need to balance internal validity, external validity, and depth of analysis. RCTs are the most robust and reliable for establishing causality, making them the gold standard for intervention studies. However, quasi-experimental designs and cohort studies provide practical alternatives for real-world contexts where randomization is not feasible. Meanwhile, mixed-methods designs offer depth and insights into user experiences, making them ideal for exploratory and feasibility studies, though their findings may be less generalizable. Finally, crossover designs effectively reduce variability and improve validity but require careful management to avoid carryover effects.

All the selected articles included nurses, midwives, or nursing and midwifery students of any kind. In total, 428 individuals participated in the studies, and 332 were nursing students, mixed with other medical students. Only 2 studies included 96 professional health care providers [19,20].

Exclusion criteria varied across the studies. In one case, they excluded those with previous clinical or VR experience, recruiting 88 prelicensure nursing students (VR group=31, simulation group=28, control group=29) [18]. The groups were homogeneous in sex, satisfaction with the nursing major, clinical practice training, and demand for XR education, but differed in anxiety levels.

Notably, 2 studies included third-year nursing students, with 56 and 19 participants, respectively [22,23]. The age ranges and gender proportions were similar: 20‐44 years, mostly female (84%), with no previous degree (81%) in the former study, and 25‐45 years, with 74% female in the latter. No statistical differences were found between the types of simulation groups.

One invited both 50 medical and 59 nursing interns [21]. Nursing interns were aged 17‐22 years, and medical interns were 20‐29 years. Female representation was 85% in nursing and 52% in medicine. The previous deep occupational experience was 34% in nursing and 61% in medicine.

In one study, 36 professional nurses (18 per group, AR and no AR) evenly distributed by age, sex, clinical role, and experience participated, with participants excluded based on their medical specialty [20]. Also involving 60 professional health care providers, including physicians, nurses, and respiratory therapists, in the study they randomly divided participants into regular and intervention groups, excluding those with incomplete training or questionnaires [19]. A similar exclusion criteria was applied to the same number of nursing students, ensuring homogeneity in age, gender, anxiety, and intravenous knowledge [24].

In total, 6 articles used VR interventions, while one used AR [20]. Among included studies, 3 used a Head-Mounted device [18-20], 3 relied on computer-based simulations [22-24], and 1used a mobile device app.

The duration of the training sessions varied minimally across the studies, ranging from a 10-minute VR session [24], to a 50-minute gamification program [18]. Notably, only one conducted two sessions [23], while the others performed a single XR experiment.

Throughout the studies, common elements were analyzed. The 7 articles evaluated anxiety using various methods, such as personalized Likert scale questionnaires [19,21]. Among them, 2 of the articles [22,23] useda specific tool to measure anxiety pretest and posttest, the NASC-CDM, which aims to provide insight into the emotional aspects of clinical decision-making, which can impact nursing performance and patient care outcomes [25].

A total of 2 studies assessed state-trait anxiety. One study distinguished between temporary anxiety influenced by environmental factors and the more stable, underlying trait anxiety [24]. Transient anxiety was measured using a Visual Analogue Scale, which consists of a 10-cm horizontal line with marked points corresponding to different levels of worry, such as “Not worried at all,” “Worried a little bit” and “Very worried,”. The chosen point on the VAS was converted to a numerical score for pre and posttest comparison. This analysis was performed and compared pre and post test. The other study [18] used the STAI, a 20-item questionnaire assessing state anxiety. Each item is rated on a 4-point Likert scale (1=not at all; 4=very much so), with a total score ranging from 20 to 80 (not 15 to 75 as stated in the original). Scores of 30 or lower indicate low or no anxiety, while scores of 31 or higher indicate high anxiety. The STAI’s Cronbach α was 0.93 at its development and 0.90 in this study.

Finally, in [20] they assessed anxiety by performing qualitative individual interviews while evaluating the rate and depth of chest compressions to provide AR feedback and therefore measure performance simultaneously.

Other aspects of the studies were evaluated, including knowledge pre and post test related to preeclampsia and group B strep, a Simulation Completion Questionnaire [23], knowledge related to neonatal resuscitation [18], or an multiple choice questionnaire on tracheostomy care and skills [19]. Another qualitative approach explores student nurses’ perceptions of the game [22], while others evaluated only the performance of the subjects, regardless of the knowledge acquisition [20,21].

After assessing and analyzing the results, most of the articles reported positive outcomes concerning anxiety reduction, as seen in the summary of anxiety outcomes in Table 5.

In total, 4 studies reported decreased anxiety levels after the XR intervention. In [20], participant feedback supported that AR-CPR could be effective for cognitive offloading, reduction in performer anxiety, and increase in performer confidence in the care delivered. In the case of [21], 68% of nursing and 58% of medical interns reported that the extended reality intervention significantly decreased their anxiety about occupational needle injury prevention, and also in [22] the participants reported a 23.4% decrease in anxiety. In addition, in [19] at baseline, there were no significant differences between the intervention and regular groups. However, at follow-up, the intervention group showed significantly higher agreement with statements about increased familiarity (83% vs 76%, P=.04), enhanced confidence (92% vs 74%, P=.001), and reduced anxiety (93% vs 75%, P=.002). The VR intervention effectively improved familiarity, boosted confidence, and reduced anxiety in tracheostomy-related skills compared to the regular training.

One study [24] reported no statistical differences among the 3 groups. State anxiety and VAS for anxiety decreased in all groups after venipuncture.

A total of 2 studies reported higher anxiety in the intervention group. In [18] the anxiety score of the 3 groups decreased from pre-intervention to post-intervention (VR group: 59.14 (SD 9.62) to 56.72 (7.50); simulation group: 66.46 (8.60) to 57.65 (SD 10.35); and control group: 59.50 (8.35) to 57.65 (6.86). However, the VR group did not show significant improvement compared to the simulation and control groups, with the highest difference observed in the simulation one.

Also, anxiety scores were higher for students in the VR clinical simulation than for those in the face-to-face simulation [23]. However, the new technology rather than the VR simulation itself might have caused the increased anxiety. In addition, students were more comfortable with the mannequin, as they had previous exposure to this type of exercise.

The authors consider that the studies provided improvement in knowledge [18], demonstrated feasibility [19], improved educational quality [20], presented a promising pedagogical approach [22], and were educationally and cost-effective, as they allowed for repeated simulations [23]. This characteristic among others, contributes to the reduction of anxiety [24].

Nevertheless, even if the results were positive, all the studies included in this review performed a small sample, single-school or medical center, single-country sessions, which is a significant limitation as it is not very representative. The authors suggest using larger sample sizes with random selection to increase generalizability and conducting multicenter studies.

Furthermore, in 3 studies, the evaluation method was qualitative and self-reported, which might also lead to report bias [19-21]. In addition, the long-term effect was not evaluated in any of the articles and often the evaluation was only made at one-time point [22].

Several limitations were related to the use of new technologies, in the study by [23], students preferred face-to-face simulations, citing the similarities to practicing in a “real” situation and the immediate debriefing. Students who did not like the VR clinical simulation often cited technological issues, such as “online program was slow”, or “didn’t know where to find things. Similarly, in [24] they noted that the results might have been affected by the students’ IT expertise, while [18] pointed out that the unfamiliar VR environment could be a factor. In [22], they suggested that students more comfortable with technology might have been more willing to participate in the experiments.

An orientation activity built into the study design would be useful so that when the XR scenario is presented, students will not be focused on learning the software. Furthermore, the studies were often performed in a single session; multiple sessions would increase students’ familiarity with the tools. On the other hand [18], suggests that it would be better to establish an MR environment, using hand tracking and physically practicing the skills, rather than using the HMD controllers. This approach would also reduce the learner’s burden owing to unfamiliar environments, enabling experiences similar to reality.

Finally, none of the studies they engaged with the focus group (nurses and midwives) to identify and prioritize their real needs. A better approach would have been to perform a systematic conception method to translate their needs into tangible project objectives and features aligned with user expectations, rather than imposing the technological approach without working and thinking about the solution with them first.