Innovations in dementia screening: a systematic review and meta-analysis of virtual reality assessments.

Background: The rising prevalence of dementia, driven by aging populations, presents a global public health challenge. Pharmacological treatments offer limited benefits unless initiated during pre-symptomatic stages, emphasizing the need for early, accurate, and cost-effective screening methods. This review investigates the diagnostic accuracy and feasibility of virtual reality-based assessments for identifying mild cognitive impairment, a prodromal stage of dementia. Methods: A systematic literature search of PubMed, PsycINFO, and IEEE Xplore was conducted to identify studies evaluating the diagnostic performance of virtual reality tools for mild cognitive impairment detection. Pooled sensitivity and specificity were calculated through meta-analysis, and methodological quality was assessed using the QUADAS-2 tool. The review adhered to PRISMA guidelines. Results: Twenty-nine studies met the inclusion criteria. Virtual reality-based assessments demonstrated pooled sensitivity and specificity of 0.883 and 0.887, respectively. Among 13 studies using machine learning, pooled sensitivity was 0.888 and specificity 0.885. Further sub-group analysis was done according to immersion degree, MCI-subtype and reference standard. Tools integrating machine learning with EEG or movement data showed particular promise. Conclusion: The findings of this meta-analysis indicate that virtual reality (VR) assessments have a promising level of accuracy for the detection of mild cognitive impairment (MCI). Nevertheless, the results are preliminary, and their interpretation warrants caution due to the substantial methodological heterogeneity observed among the included studies. Despite the potential of VR as a cost-effective solution for dementia screening, its implementation faces notable barriers, including the requirement for specialized personnel and the absence of clear data regarding software and support costs.