Effectiveness of virtual reality disaster education on the performance of nursing students: a systematic review and meta-analysis

Disaster education is essential in nursing students to prepare students for crises. Virtual reality (VR) technology offers a promising training approach. To estimate the difference in performance, time to complete performance, cognitive learning, and self-efficacy of nursing students in VR disaster education compared to traditional education only. PubMed, EMBASE, CENTRAL, CINAHL, and ProQuest Dissertations & Theses Global were searched from inception to 22 January 2025. We included randomized controlled trials (RCTs) and non-RCTs that investigated the effect of VR disaster education on performance, time to complete performance, cognitive learning, and self-efficacy. The Risk Of Bias In Non-randomized Studies of Interventions, Version 2 (ROBINS-I V2) was used to assess risk of bias. A random-effects model was used to pool the effect estimates across studies. I 2 was used to quantify statistical heterogeneity among included non-RCTs. The protocol of this study was registered in PROSPERO (CRD420250652860). Of the 307 articles identified, two quasi-experimental studies (non-RCTs) were included ( n = 220). The time to complete the performance was significantly lower in the VR disaster education compared to the control group (mean difference [MD]: -45.18 s; 95% confidence interval [CI]; -52.06 to -38.30), with no evidence of heterogeneity (I 2 = 0%). No significant difference was observed between VR and the control group for performance score (MD: -0.14; 95%CI; -1.98 to 1.70; I 2 = 96%) and cognitive learning (MD: 3.56; 95%CI; -3.12 to 10.25; I 2 = 99%), both showing considerable heterogeneity. No safety concern was reported in any of the included studies. VR-based disaster education has the potential to improve completion times in decontamination scenarios, which may contribute to enhanced nursing students’ efficiency in emergency response. However, its effects on performance scores and cognitive learning remain inconclusive, due to non-significant pooled estimates and considerable heterogeneity across studies. Further high-quality RCTs with longer intervention durations are needed to validate these findings.