BackgroundIt remains unknown whether perivascular spaces (PVS) are associated with delirium in older hospitalized patients. We aimed to determine the association between magnetic resonance imaging (MRI)-visible PVS and the risk of delirium in a cohort of older patients.MethodsWe consecutively recruited older patients (≥70 years) admitted to the Geriatric Department of West China Hospital between March 2016 and July 2017, and their imaging data within one year before admission were reviewed retrospectively. PVS was rated on axial T2-weighted images in the basal ganglia (BG) and centrum semiovale (CS) using the validated semiquantitative 4-point ordinal scale. Delirium was screened within 24 h of admission and three times daily thereafter, using the confusion assessment method. Binary logistic regression analyses were performed to investigate the associations between PVS and delirium.ResultsAmong 114 included patients (mean age 84.3 years, 72.8% male), delirium occurred in 20 (17.5%). In patients with MRI examined within 6 months before admission, CS-PVS was found to be associated with delirium (odds ratio [OR] 3.88, 95% confidence interval [CI] 1.07-14.06, unadjusted; and OR 4.24, 95% CI 1.11-16.28, adjusted for age). The associations were enhanced and remained significant even after full adjustment of covariates (OR 7.16, 95% CI 1.16-44.32, adjusted for age, cognitive impairment, smoking, and Charlson Comorbidity Index). Similarly, the relationships between high CS-PVS and delirium were also strengthened after sequentially adjusting all variables of interest, with OR 4.17 (95% CI 1.04-16.73) in unadjusted model and OR 7.95 (95% CI 1.14-55.28) in fully-adjusted model. Adding CS-PVS to the established risk factors improved the risk reclassification for delirium (continuous net reclassification index 62.1%, P = 0.04; and integrated discrimination improvement 12.5%, P = 0.01).ConclusionsCS-PVS on MRI acquired 6 months earlier predicts subsequent delirium in older patients and may have clinical utility in delirium risk stratification to enable proactive interventions.