An umbrella review of meta-analyses was carried out according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) reporting guidelines, as described in Checklists 1 and 2. The research questions used to guide this umbrella review were as follows:

The eligibility criteria were as follows: studies that used a meta-analytic method, meta-analyses about VR intervention in clinical nursing practice, and meta-analyses published in English. The exclusion criteria were as follows: meta-analyses that collected ≤2 intervention studies and meta-analyses that had no 95% CI and heterogeneity data.

We conducted the umbrella review by searching Web of Science, Embase (Ovid), Cochrane Library, PubMed, and relevant reference lists. Eligible studies were published between December 1, 2012, and September 1, 2023. The searches were rerun on September 20, 2023, to identify any recent publications. We reran the searches before the submission of this paper, and no extra literature was found. We searched for publications that included the following terms (including variations of these terms) in the title, abstract, and keywords list: virtual reality, VR, virtual environment, immersive, nursing, care, meta-analysis, and review. We also searched the reference lists of the most recent systematic reviews and meta-analyses. The literature retrieval strategy is shown in Multimedia Appendix 1.

One researcher conducted the electronic database searches, eliminated duplicates and titles that were clearly outside the scope of the umbrella review, and then uploaded the remaining citations to NoteExpress version 3.7.09258 (Aegean Software Corp). Two reviewers independently examined the remaining full-text articles to identify those that met the inclusion and exclusion criteria. If there were multiple meta-analyses with the same research objectives and outcome indicators, the one with the highest quality score was selected. Any disagreements were resolved through discussion with a third reviewer.

Data were extracted and managed independently by 2 reviewers using a predefined extraction form. Any concerns were discussed with a third reviewer.

The following data were extracted: (1) characteristic indicators of meta-analyses (first author, year of publication, study design, study period, and number of component primary studies); (2) characteristics of primary studies (trial design, number of participants, and sectionalization); (3) population information (diagnosis and sample size); (4) VR intervention information (VR platform, population, and intervention course); and (5) statistical summaries (outcomes and effect measures with 95% CIs and heterogeneity).

We did not reanalyze the other data or primary studies included in the meta-analyses because of the clinical and statistical heterogeneity between the study objectives and outcome indicators of the meta-analyses, and many of the articles did not provide the original data of the original clinical studies. As a result, a descriptive analysis was conducted to encapsulate the impact of VR on clinical nursing practice over the past decade. The process of executing this descriptive analysis involved presenting research results, such as participant details, outcomes, sample sizes, and study designs. These data were meticulously extracted into a predefined Excel (Microsoft Corp) form by YH and XY. Following this, the two authors conducted a thorough review and verification of the collected data to ensure their accuracy and reliability. In instances of disagreement, a consensus was achieved through discussion. In particular, the theme—discerning the relationships between VR interventions and patient outcomes—was summarized.

AMSTAR (A Measurement Tool to Assess Systematic Reviews) 2 [22] and the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) checklist [23] were used to independently evaluate the methodological quality of the selected meta-analyses. Two researchers separately evaluated the evidence strength of meta-analyses. If there was disagreement, another researcher was asked to make a judgment.

We reviewed the full texts and supplementary materials of included meta-analyses. Two researchers extracted the estimated pooled effect and heterogeneity of each outcome reported in the meta-analyses. The estimated pooled effect, along with its 95% CI, for each included meta-analysis was extracted. We used the I² metric to assess heterogeneity (<25%: might show no heterogeneity; 25%-50%: might show moderate heterogeneity; 50%-75%: might show substantial heterogeneity; 75%-100%: considerable heterogeneity), and heterogeneity and P values (significant at P<.05) were used to assess publication bias.

The protocol for this umbrella review was originally registered with PROSPERO in December 2022 (registration number: CRD42022381382). No ethical approval was needed, as we used data from published studies.

The electronic literature search identified 768 records from 4 databases and the reference lists of included reviews. We screened the titles and abstracts of 634 records after removing 134 duplicate records. A total of 260 reviews remained after the titles and abstracts were screened for inclusion against predefined criteria. After reading the full texts, 74 articles were finally selected. Further details can be found in Figure 1.

The included meta-analyses were meta-analyses of randomized controlled trials (RCTs), cross-over studies, pre-post studies, interrupted time series studies, quasi-controlled trials, case reports, controlled trials, or controlled clinical trials, and details of all included meta-analyses are listed in Tables 1 and 2.

Among all included meta-analyses, the most common participant groups investigated were neuronursing (25/74, 34%), pediatric nursing (13/74, 18%), surgical and wound care (11/74, 15%), oncological nursing (11/74, 15%), and older adult nursing (10/74, 14%) populations. The number of participants in each meta-analysis ranged from 103 to 3540, and 214 outcome indicators were found. The main outcome indicators in all included meta-analyses were pain (37/214, 17.3%), anxiety (36/214, 16.8%), cognitive function (17/214, 7.9%), balance (16/214, 7.5%), depression (16/214, 7.5%), motor function (12/214, 5.6%), and participation in life (12/214, 5.6%), among others. The distribution of populations and outcome indicators of VR application in nursing is shown in Multimedia Appendix 2.

Only a few papers reported adverse effects caused by using VR. The prevalence of adverse effects in these studies ranged from 4.76% to 50%, and these adverse effects included nausea, vomiting, sickness, dizziness, fatigue, pain, and the risk of losing balance.

There were various platforms used for VR. We summarize the typical VR technology platforms used for different populations and outcome indicators in Table 3. The most commonly used VR platforms were Pico VR glasses, head-mounted displays, the Nintendo Wii, and the Xbox Kinect. The duration of VR intervention ranged from 2 weeks to 12 months (the most commonly reported durations were ≥4 weeks), and the length and frequency of VR sessions ranged from 5 to 100 minutes and from 1 to 10 times per week, respectively.

Clinical applications of VR in nursing practice predominately concentrate on neurology, pediatrics, oncology, surgical and wound care, and gerontology. Extant evidence indicates that VR interventions confer benefits for alleviating patient anxiety, pain, and depression, especially in pediatrics, among patients with breast cancer, and in wound care processes. VR is also a promising intervention for enhancing cognitive function in neurology patients. However, research on the utility of VR for improving motor function, balance, memory, and attention remains equivocal.

Many meta-analyses suggested that VR in nursing is useful for relieving anxiety and pain in patients, especially in pediatrics, among patients with breast cancer, and in wound care processes. Chen et al [62] found that a VR operation before surgery helped people familiarize themselves with the environment and understand the preoperative preparation procedures, thereby effectively reducing anxiety and improving compliance. Additionally, VR is a promising intervention for procedural pain [18,30], as reported in a study by Addab et al [92]. As a nondrug distraction intervention in nursing, VR can reduce the side effects of pain, depression, and anxiety drugs. However, heterogeneity was substantial in oncology studies and surgical and wound care studies. More high-quality studies are needed to help define the effectiveness of VR in these fields [93]. Further, VR can potentially be used to interfere with cognitive function, as it has been shown to affect cognitive plasticity and neuroplasticity. One study demonstrated plasticity in patients diagnosed with mild cognitive impairment and confirmed that repeated VR situational interactive training improved the excitability of the remaining nerve cells, promoted the functional reorganization of the damaged brain area, and resulted in the formation of new neural circuits, thereby improving the patients’ cognition [94]. In a systematic review and meta-analysis that included 21 original studies (1149 participants), Chen et al [61] suggested that compared with a controlled group, VR training increased cognition because VR training increased the 1-naphthylacetic acid to creatine ratio in the hippocampus, which indicated that VR could improve the levels of hippocampal metabolites in patients with poststroke cognitive impairment, thereby promoting neuronal repair and improving cognitive function. However, several studies indicated that the effects of VR in nursing on cognition need to be clarified in more well-designed RCTs with large sample sizes because of small effect sizes, insufficient outcome indicators, and the low quality of current evidence [34,50,56].

The meta-analysis results showed that VR in nursing probably increased the exercise capacity of individuals undergoing neurologic rehabilitation when compared with usual nursing [60]. In contrast, many meta-analyses showed that there was little confidence that VR in nursing improves physical outcomes in older adults when compared with usual nursing [29,57]. The lack of allocation concealment and the absence of assessor blinding were the main causes of bias. As such, the final grade of the evidence was low, and the conclusions were hard to accept [37]. Some researchers thought that the results of VR therapy showed a trend of gait and upper limb function improvement in individuals [95,96], but Laver et al [31] believed that VR video gaming was not more beneficial than conventional nursing approaches in improving upper limb function, as did Voinescu et al [5]. Further, heterogeneity in protocols, VR task performance, and the different characteristics of the participants could have influenced the results [39]. Zhang et al [54] and Lei et al [13] pointed out that VR training helped to improve balance when compared with usual care. Similarly, Gates et al [97] and Saragih et al [98] concluded that VR training helped to improve balance when compared with usual nursing. All of these included studies did not analyze which subtypes of patients can benefit more from VR training in terms of improved balance, did not assess the long-term effects of VR on balance and mobility, and often included a relatively small number of patients and inadequate control groups [99]. The results need to be interpreted with caution due to the high heterogeneity, small sample sizes, and different outcome measures. In terms of improving memory, even after excluding studies with high heterogeneity, several meta-analyses still concluded that the effects of VR nursing measures on patients’ instantaneous memory, long-term memory, and attention were not statistically significant [31,54,56]. The possible reasons for this are that VR training duration was too short to result in significant improvements [56] and only 6 studies (within the meta-analyses included in this umbrella review) with small samples were used to evaluate the effects of VR on attention.

It is also important to classify and summarize the VR platforms and frequencies of VR use that are commonly used in different nursing scenarios because of their influence on nursing outcomes. Meta-regression analyses showed that the number of VR sessions and the frequency of VR training had statistically significant impacts on balance scores [51]. There are few studies that explain why VR devices and scenes are chosen in a particular context, and there is no comparative analysis of the effects of using different VR devices and scenes in the same session. With regard to these biases, the included studies claimed that heterogeneity was due to the use of VR hardware with multiple types of VR software and variation in the duration of VR therapy sessions [93], which resulted in patients experiencing different levels of immersion, presence, and interactivity. Future research on VR in nursing should focus on providing a detailed report on VR device and platform instructions, the length and frequency of VR sessions, and the duration of VR intervention.

Previous reviews on VR in nursing mostly discussed the impact of VR on nursing education and training at different levels [100,101]. However, the role of VR in clinical nursing practice is also very important. Therefore, we conducted this umbrella review, which aimed to effectively synthesize the combined evidence from meta-analyses that assessed the effects of nurses using VR technology on patients’ health outcomes.

This review followed the PRISMA reporting guidelines; thus, it can be viewed as a transparent and reproducible review. The findings in our review are more likely to be comprehensive and provide evidence for further VR application in clinical nursing practice. This study has a low risk of publication bias because we searched all relevant databases by using a reliable search strategy to identify as many eligible studies as possible. Moreover, we controlled the risk of selection bias by having 2 independent authors perform the research process, including the study selection, data extraction, and quality assessment processes.

First, the quality of many included meta-analyses was not high. Therefore, the conclusions should be considered with care. Second, our study only reports on meta-analyses published in English, which may have resulted in language bias. Third, a quantitative data synthesis was not conducted in this review, given the significant clinical heterogeneity of the included studies in terms of study designs, intervention and population characteristics, and outcome measurements. This was one of the main reasons for the inconclusive research evidence found in this review. Follow-up studies can search for more literature, without strict language restrictions, and try to conduct high-quality quantitative syntheses to establish more reliable conclusions.

VR games have distinct clinical advantages, as they offer challenging and interesting environments [54]. Authoritative conclusions on the applications of VR in nursing might promote the application of VR games. It is suggested that nursing researchers should improve the quality of research; carry out large-scale, multicenter randomized controlled studies; and make authoritative conclusions on VR in nursing. In addition, the research outcomes of VR application in nursing are mostly subjective indicators, such as anxiety, depression, and satisfaction, and more objective outcome indicators, such as serological indicators, should be mentioned in the future. In the existing research, VR platforms, the duration of VR intervention, and the length and frequency of VR sessions are different. We hope that further research considers establishing standard operation protocols for VR intervention in specific populations.

Our study comprehensively summarized and indicated the potentially beneficial role of VR intervention in enhancing the management of pain, depression, anxiety, and cognition in neurology, pediatrics, oncology, surgical and wound care, and gerontology; the effects of VR intervention on improving motor function, balance, memory, and attention remain equivocal. However, these findings should be interpreted with caution due to the unsatisfactory quality of the included studies. It is recommended that more research of rigorous methodological quality is necessary to further determine the role of VR in promoting health care outcomes.