This document discusses analyzing patient data to predict mortality risk and describes the steps taken:
1) A training and validation set were created from over 17 million patient cases to calculate likelihood ratios and predict 6-month mortality.
2) The most and least deadly diseases were identified based on their likelihood ratios, with things like intracranial hemorrhage being very deadly and disorders like uterine myomas being less so.
3) Sensitivity and specificity of the predictions were calculated to determine the usefulness of the model.