Speaker
Description
This work refers to the construction of models using machine learning algorithms for early prediction (during the first 24 hours of admission) of hospital mortality in patients with coronary artery disease through the use of clinical notes and structured clinical data (electronic health record -EHR). The aim is to effectively identify suitable models to predict early mortality and recognize risk factors. For unstructured EHR, n-gram models to extract feature in the clinical notes are explored. For structured data, different machine learning algorithms are evaluated and combined with different kind of information to identify risk factors. We also validate the model performance and compare its performance using reference scores. It is important to mention, that this study is focused on the analysis of the first 24 hours of admission, because during this time it is possible to identify invasive procedures or not, avoiding irreversible damage or sudden death.
