Evaluation of Documentation Patterns of Trainees and Supervising Physicians Using Data Mining.

Area of Science:

• Medical Informatics
• Graduate Medical Education
• Health Services Research

Background:

• Electronic Health Records (EHR) offer rich data for analyzing healthcare quality and resident education.
• Natural Language Processing (NLP) and data mining can extract valuable insights from EHR data.
• Understanding resident and faculty documentation practices is crucial for optimizing clinical workflows and education.

Purpose of the Study:

• To analyze inpatient documentation practices of residents and attending physicians using EHR data.
• To objectively measure the workflow of resident physicians and their supervising faculty.
• To demonstrate the utility of data mining as an assessment tool in graduate medical education.

Main Methods:

• Retrospective observational study of deidentified patient notes from a multispecialty university physician group.
• Utilized a novel automated data mining technology (Healthcare Smartgrid) to extract patient note variables.
• Analyzed 26,802 consecutive patient notes over a 7-month period.

Main Results:

• Residents predominantly entered notes between 8 am and 4 pm.
• Attending physicians provided teaching attestations within 24 hours for only 73% of records.
• Surgical residents were more likely to document before noon, while nonsurgical faculty were more timely with attestations.

Conclusions:

• EHR data successfully quantified resident physician workflow and faculty oversight.
• Findings have implications for improving physician oversight in resident clinical work.
• Data mining models are effective assessment tools in graduate medical education.