All enrolled patients will undergo 72 hours of monitoring of cerebral oxygenation with near-infrared spectroscopy.

Multivariate linear regression will be used to characterize the association between poor cerebral perfusion (as measured using duration of time (minutes) outside of MAPOPT, mean rSO2, and duration of disturbed cerebral autoregulation) and delirium severity to determine if poor cerebral perfusion is an independent predictor of delirium severity. We will estimate the unadjusted effect of each individual predictor on delirium severity (i.e., cumulative CAM-7 scores per patient). The simultaneous multivariate regression model will adjust for the following covariates (sex, a history of hypertension, a history of alcohol abuse, total sedative dose (in midazolam equivalents), total narcotic dose (fentanyl equivalents), severity of illness (APACHE II scores), pre-existing cognitive impairment (CDR score), length of ICU stay, and blood urea nitrogen. The multivariable model will provide the adjusted regression coefficients after controlling for all predictors included in the model.

To assess the physiological determinants of rSO2, multiple linear regression will be performed using the patient average of each variable over the 72 hour data collection period. The following predictors will be included in the regression model: heart rate (HR), arterial oxygen saturation (SpO2), MAP, arterial blood gas data (i.e., pH, partial pressure of oxygen, and partial pressure of carbon dioxide), central venous oxygen saturation, and hemoglobin concentration (Hb). In addition, the multivariate model will control for the following covariates associated with cerebral perfusion: sex, age, as well as total sedative, narcotic, and vasopressor dosing. Simultaneous multiple linear regression with adjustment for all aforementioned covariates will be implemented. A within patient repeated measures analysis will also be performed by using the linear mixed effects model with 6 observations per subject reflecting each 12 hour period during the 72 hour data collection period.

We define the optimal MAP as the MAP at which there is no statistically significant correlation between MAP and rSO2. The time outside the optimal MAP will be calculated, and will be assessed as an independent predictor of delirium, and 3- and 12-month cognitive outcomes.