Using K-Means Clustering to Identify Physician Clusters by Electronic Health Record Burden and Efficiency

Area of Science:

• Health Informatics
• Medical Documentation
• Physician Workflow Analysis

Background:

• Electronic health records (EHRs) are integral to modern medicine but contribute significantly to physician documentation burden.
• Assessing and mitigating this burden is crucial for physician well-being and healthcare system efficiency.

Purpose of the Study:

• To analyze physician metadata from the Epic EHR's Signal tool within the Montefiore Health System.
• To utilize cluster analysis to quantify EHR burden and efficiency among physicians.

Main Methods:

• Data from July 2020 were analyzed, reflecting a post-telemedicine operational period.
• Six Signal metrics were used: time on unscheduled days, pajama time, time outside of scheduled hours, turnaround time, proficiency score, and same-day visit closure.
• k-Means clustering was applied to group physicians based on these metrics.

Main Results:

• Physicians were clustered into distinct groups: one with higher efficiency and lower time outside scheduled hours (TOSH), and another with lower efficiency and higher TOSH.
• These clusters persisted even when stratified by physician sex and specialty.
• Intra-cluster analysis revealed homogeneity in physician metrics, with some variations noted in the higher-efficiency cluster when stratified by sex.

Conclusions:

• The identified clusters effectively represent the EHR efficiency-burden landscape at the Montefiore Health System.
• k-Means clustering of EHR data offers a scalable and adaptable method for assessing physician EHR burden.
• This approach enables health systems to identify documentation trends and implement targeted wellness interventions for physicians.