Abstract
<title>Abstract</title> <p>Background Immersive virtual reality (VR) reliably produces large simulated-performance effects in health professions education, yet whether such effects reflect transferable competence or alignment between rehearsed scenarios and assessment criteria remains empirically untested. Meta-analyses disagree about whether VR benefits skills or knowledge more—an inconsistency a constructive-alignment account predicts if studies align scenario and assessment to differing degrees. We tested this within one adequately powered, blinded trial by locating the VR advantage at the assessment-component level. Methods In a cluster randomized controlled study, 32 intact teaching subgroups (442 second-year nursing students) were allocated 1:1 to immersive VR (Meta Quest Pro; three procedure-choreographed scenarios) or conventional low-fidelity simulation for urinary, nasogastric, and rectal catheterization; both arms then received an identical seminar. Outcomes were a blinded, double-rated three-station OSCE one week after training and a 30-item knowledge test administered immediately and at four weeks. Exploratory analyses decomposed the advantage by checklist item and compared performance and knowledge effect sizes; inference used cluster-robust wild bootstrap. Results The VR arm outperformed control on the composite OSCE (87.33 ± 5.05 versus 70.70 ± 8.51; difference 16.63, 95% CI 15.32–17.94; Hedges’ g = 2.38) and on knowledge (immediate difference 4.80 of 30, g = 1.26; four weeks 4.03, g = 1.04; ANCOVA-adjusted 3.23). The performance effect significantly exceeded the knowledge effect (Δg = 1.34; p < 0.001). The item-level advantage was selective (range − 0.11 to + 0.39): largest for documentation, positioning, verification, and asepsis (steps the scenarios rehearsed), whereas control outperformed VR on material preparation—the one step VR cannot rehearse. Knowledge effects tracked scenario dose (urinary g = 1.53; nasogastric 0.25; rectal 0.33). Conclusions Immersive VR produced a large but selective advantage that followed what its scenarios rehearsed, alongside a smaller, retained knowledge benefit. To our knowledge, this is the first adequately powered, blinded, within-trial demonstration that immersive-VR effects localize to rehearsed, assessed components—reversing on the one step VR cannot rehearse—helping reconcile why syntheses disagree about VR’s benefit to skills versus knowledge. Large VR effects should be read as partly alignment-dependent; VR is best positioned as a preparatory complement to physical practice, with assessment sampling beyond the rehearsed choreography.</p>