Discovering and Analyzing Trend-Event Patterns on Clinical Data

Abstract

Pattern mining, even with suitable temporal extensions, has been proposed as a methodological tool for mining clinical data. It allows healthcare stakeholders to derive new knowledge from overwhelming amount of healthcare and clinical data. However, many methods proposed so far focus on mining of temporal rules which describe relationships between data sequences or instantaneous events, without considering the presence of more complex temporal patterns into the dataset. These patterns, such as trends of a certain vital sign, are often very interesting for clinicians to better understand their data. Moreover, it is really interesting to discover if some sort of event, such as a drug administration, is capable of changing these trends and how.In this paper, we propose a new kind of temporal patterns, called trend-event patterns (TE-Ps), that focus on events and their influence on trends that can be retrieved from some measures, such as vital signs. With TE-Ps we can express concepts such as “The administration of paracetamol on a patient with an INCREASING temperature leads to a DECREASING trend in temperature after such administration occurs”.We also provide an original algorithm to mine such kind of temporal patterns, and we developed a tool for retrieving TE-Ps. Moreover, we proposed the multidimensional modeling for TE-Ps through the creation of a data-cube.Finally, we discuss some first results we obtained by preprocessing and mining ICU data from MIMIC III database. These results are also analyzed in some aggregate forms through OLAP analysis.