Prognostic indication can be associated directly with statistical classification and prediction, and as more data becomes available, it is tempting to believe that we can only get better results. However, this really depends on whether data can be equated with information. The EPIC system, within Addenbrookes, is a unique system within the UK, where all hospital records have been digitised and converted into machine readable formats, including all related image data. The wealth of data is almost unimaginable, but converting this into useful information is a much more difficult task. In addition the CENTER-TBI co-ordinates data collection on head injury patients internationally, and the development of prognostic indicators for outcome is a key aspect of this programme. While this project will focus on the generics of integrating such big data into statistical prognosis, a particular focus of this project will be to integrate clinical, demographic and imaging data into a prediction algorithm for heart attack and stroke. We will take a mainly statistical approach to this, but will also integrate relevant machine learning techniques.
Current algorithms do not take account of imaging data (such as calcium scores derived from heart CT scans). Our approach would be a potential improvement of the QRISK scoring system that is the basis for the widely accepted JBS3 risk calculator. We will test the addition of calcium scoring (vs. standard practice) in a prospective study by using data freely available on the EPIC platform in Cambridge (see letter from Addenbrooke’s hospital CIO). We will determine the most effective means of combining calcium scoring with clinical and demographic data.
This project could appeal to a candidate with a background in mathematical statistics or machine learning.