Mining Temporal Patterns from Sequential Healthcare Data

Electronic health records (EHRs) can be represented as sequences of time-stamped events. Therefore, temporal patterns, such as transitions between clinical events over time, can be extracted using temporal mining techniques. This has the benefit of transforming large temporal data records into a clear and easily understandable knowledge base that could support clinical practice. However, EHR data poses many challenges due to the dynamic, heterogeneous and complex nature of health information. This project focuses on developing a methodological approach to extract patterns of transitions between adverse events after Left Ventricular Assist Device (LVAD) implant in patients with advanced heart failure.