This study investigates clinician workflows in electronic health records (EHR) using a novel combination of context-based embedding and graph-based dimensionality reduction techniques to EHR-based audit log sequences. We identified distinct clinical task groups, suggesting the potential for semi-automated, unsupervised methods for characterizing EHR-based workflow patterns.
Keywords: Electronic health records; action sequence; audit logs; clinical workflow; context embedding; task identification; unsupervised learning.