Abstract:
|
A trajectory in the medical world, is the sequence of events that occur during the life of a
patient. In the recent years, these trajectories have been stored in the Electronical Health
Records and many of the health organizations have databases with the clinical history of all their
patients. The trajectories can be summarized in a trajectory graph which shows the different
paths the trajectory of a patient may take. The graph contains events on its nodes and the edges
contain the temporal relations. Previous works focused in the exploration of trajectory graphs
only allow one event at each node, thus losing information and potentially mixing different
groups of patients. In this work, we have developed a new procedure to extract the trajectory
graphs that allows having several events in a single node of the graph. This procedure has been
tested in two real world datasets: one related to diagnostics at hospital admissions, and the other
on prescriptions in intensive care units. |