Data-driven modeling of clinical pathways using electronic health records

Abstract

This paper presents results of an ongoing research on dynamic modeling of clinical pathways with data assimilation. The research is aimed at data-driven prediction model of clinical pathways while a patient stays at a hospital. The current model implementation includes patients' clustering according to their movements (clinical paths) in a hospital. Also, we considered diversity of patients' clinical pathways and explored their parameters to understand causes of these differences using a classifier. An example of pathway classification, a decision tree with an accuracy of 81.2%, is described. In the future, we plan to improve the model by incorporating assimilation of patient's information, such as medical history and laboratory screening and update the patient pathway if necessary.

Ähnliche Arbeiten