The Taub Faculty of Computer Science Events and Talks

Machine learning made many recent advances in science and technology, specifically in healthcare information technology. Electronic Health Records (EHR) data store the healthcare information. EHR data consists of many features. It is highly complicated, noisy and includes many outliers and missing values. It also contains time-dependent information such as vital sign measurements, diagnosis, treatment, etc. Therefore, basic machine learning models have poor performance on this data, as they do not utilize the time-series aspect of the data. We hypothesize that by using the Transformer, one of the most recent advancements in machine learning, we can better model EHR data, as its primary goal is to process time-dependent data. Our model uses data in the OMOP Common data model form. Specifically, we model data related to ICU stays for predicting Blood Stream Infection (BSI), a critical condition with a mortality rate above 30%. Early prediction and antibiotic treatment of BSI are essential, as it reduces mortality and morbidity significantly. We developed a Transformer-based architecture. This model was inspired by the work of Kodialam et al, a Transformer model which uses OMOP common data model to model time-dependent healthcare information data. We aim to utilize a proprietary dataset of about 200,000 ICU stays to identify hidden structures and perform risk prediction for the BSI condition in advance.