Research is currently being undertaken to expand the window of efficiency for medical treatment through pre-symptomatic diagnosis. This is achieved through an observational clinical study. Blood is taken from consenting elective surgery patients from pre-surgery to treatment end. Some of these patients go on to develop sepsis (3.8%) and the majority recover without developing sepsis. Blood is taken daily. The diagnosis of sepsis has a level of variation between clinicians and hospitals and consensus is reached via a clinical advisory panel where the level of disagreement is analysed. The bloods are stored and then shipped to a laboratory where the RNA or transcriptomic signature is measured by microarray and quantitative methods. The data is retrieved, pre-processed, normalised and undergoes statistical modelling. This then predicts when a patient is likely to go on to develop sepsis or not. At every point of this process from patient to statistical result there is an associated error or accuracy. There are different data types present and not all of the error points can be considered independent. In order to give the clinician confidence in using this process to assist at point of care, we need to be able to propagate the errors through the complex process to provide an overall uncertainty measurement.
Cumming, J., Riseth, A., & Williams, J. (2016). Understanding the accuracy of pre-symptomatic diagnosis of sepsis. [No known commissioning body]