Understanding physicians' work via text analytics on EHR inbox messages

Area of Science:

• Health Informatics
• Medical Informatics
• Clinical Workflow Analysis

Background:

• Electronic Health Records (EHRs) generate substantial physician inbox messages.
• Understanding the drivers of this workload is crucial for optimizing primary care.
• Previous methods lacked the granularity to detail EHR inbox task composition.

Purpose of the Study:

• To develop and apply a text analytics methodology for analyzing primary care physicians' (PCPs') EHR inbox workload.
• To identify and quantify the specific work themes contributing to PCP inbox management.
• To assess the variability of workload drivers across physicians and practices.

Main Methods:

• Utilized 1 year (2018) of EHR inbox messages from the Epic system for 184 PCPs across 18 practices.
• Employed an advanced text analytics latent Dirichlet allocation model to categorize message content.
• Trained the model on physicians' inbox message texts to identify work themes and workload distribution.

Main Results:

• Identified 30 distinct work themes within EHR inbox messages, categorized as medical and administrative.
• Found that 50.8% of messages concerned medical issues and 34.1% administrative matters.
• Specific tasks included ambiguous diagnosis (13.6%), condition management (13.2%), paperwork (9.5%), and scheduling (17.6%), with significant inter-physician variability.

Conclusions:

• Advanced text analytics offers a reliable, data-driven method to analyze EHR inbox workload with high detail.
• This methodology can inform workflow redesign to reduce PCP burden and enhance care quality.
• Findings support improvements in cost, quality of care, and staff work satisfaction.