|Friday, February 21|
|CS10 Not Your Usual Assumptions||
Fri, Feb 21, 1:30 PM - 3:00 PM
Statistical approach for prediction, validation and creation of a simple score: Application to a neurocritical care study. (302760)*Jay N. Mandrekar, Mayo Clinic
Keywords: Prediction, Validation, ROC, Withdrawal of Life Sustaining Measures
Patients admitted to neurocritical care units often have devastating neurologic conditions and are likely candidates for organ donation after cardiac death. Given the uncertainty in the time to death, improving the prediction of time to death after withdrawal of life sustaining measures (WLSM) based on pre-WLSM clinical factors is crucial to have a positive impact on the rates of organ donation. Two goals of this research were: 1) to identify variables that predict time to death after WLSM specifically for patients with severe brain injury allowing for organ donation, and develop a predictive model, and 2) to validate the predictive model and develop a score incorporating these variables to reliably identify potential donor patients. The first part of the presentation will focus on how we arrived at a pool of factors associated with earlier time to death using a retrospective database. Next, we will discuss the validation of these identified factors using data from a multicenter prospective study. Logistic regression and ROC curves were used in this analysis. Future research possibilities for improved clinical utility of such score will also be discussed.