Discovering the Temporal Interactions between Clinical Observations

The explosion of clinical information provided by the advent of electronic health records (EHRs) offers an exciting opportunity to substantially improve the quality of health care. Updated throughout each patient´s health care, EHRs document a wide variety of clinical observations, such as the patient´s diagnoses, risk factors, medications, and test results at various points in the patient´s history. This allows for the secondary use of EHRs to provide substantial information about the way that patients´ clinical pictures and therapies have evolved over time. In this abstract, we outline our ongoing work towards harnessing the chronological, natural language text in EHRs in order to model how individual patients´ clinical pictures and therapies have evolved. Moreover, we show our experiments on improving predictive accuracy by discovering latent groups of "similar" patients.