Abstract
Background
Delirium is a prevalent and serious neuropsychiatric syndrome associated with adverse clinical outcomes. This systematic review and meta-analysis aimed to evaluate visual impairment as a prognostic factor for delirium across various clinical settings.Methods
MEDLINE, Embase, and CENTRAL were systematically searched from inception to September 2025. Title/Abstract/full-text screening, and data extraction were performed by two independent reviewers. Eligible studies included those reporting both visual impairment and delirium. Random-effects meta-analyses were performed to pool odds ratios (ORs). Risk of bias was assessed using the Quality in Prognosis Studies (QUIPS) tool, and the certainty of the evidence was evaluated using the GRADE approach. This review was prospectively registered (PROSPERO; CRD420251269018).Results
Forty-four studies involving 26,230 participants were included. In the primary analysis of 16 studies reporting multivariable-adjusted estimates, visual impairment was associated with increased odds of delirium (OR 2.02, 95% CI 1.29-3.16; I2 = 82.8%). In the secondary analysis including all available unadjusted estimates, the association was larger (OR 2.40, 95% CI 1.81-3.18; I2 = 87.7%). Although pooled estimates varied numerically across clinical settings, meta-regression did not show statistically significant differences by setting. Certainty of evidence was rated low, with uncertainty primarily related to substantial between-study heterogeneity.Discussion
Visual impairment is associated with increased odds of delirium across diverse adult populations and care settings; however, the observational evidence does not establish causality. Key limitations include heterogeneous definitions of visual impairment and delirium, variable adjustment for confounders, and potential residual confounding. These findings identify visual impairment as a prognostic factor rather than an established modifiable cause of delirium; whether treating visual impairment reduces delirium incidence was not evaluated and requires assessment within a causal framework.