Project: miniTUBA Demo for Clinical Dataset
Investigators: miniTUBA team at University of Michigan. Created at: 2007-02-07 15:28. Last updated at: 2007-02-14 15:54.
Description:
The following represents a potential clinical scenario and dataset which miniTUBA could be used to evaluate. An infinite number of variables can be included in the dataset, but for the purposes of this demonstration only twenty variables will be analyzed. This data is synthetic, and has been generated for the purposes of demonstrating the features available in the miniTUBA system.
Demonstration Scenario:
A group of investigators are interested in determining how a cadre of clinical variables and inflammatory mediators can be used to predict which patients in the surgical intensive care unit (SICU) are likely to develop sepsis. Severe sepsis continues to carry a mortality rate approaching 50%, and many sepsis therapeutic trials have failed due to the complex and dynamic nature of this syndrome. Thus, dynamic Bayesian analysis provides a unique means of analyzing multiple influences which may change over time, and how they impact an individual SICU patient’s likelihood of developing sepsis on any given day.
A cohort of 500 patients were followed daily in the SICU from admission to discharge. Of these patients, 122 met criteria for sepsis. A number of clinical variables and inflammatory mediators were measured as shown in the data entry demonstration below. For the purposes of this demonstration, a two day interval (Markov lag) was chosen for analysis; meaning, the relationships shown would be expected to occur with a two day lag (i.e. if a patient received a transfusion, their WBC would be impacted two days later). This lag was chosen to allow for a period which seemed clinically relevant to provide information which would allow for a possible intervention before the deleterious outcome developed, however, any time interval (Markov lag) can be chosen for analysis.
With an interval of two days selected for analysis, miniTUBA identified the following most likely causal relationships between variables studied and the development of sepsis as shown in the networks below. Probability tables, graphic representation of the relationships in 2-D and 3-D scatter plots, and predictions are also provided. Please explore the many features of miniTUBA using this synthetic dataset.
Status: Approved
Operations:
Analyses: Refresh
|
