Clinical encounter heterogeneity and methods for resolving in networked EHR data: a study from N3C and RECOVER...

Area of Science:

• Health Informatics
• Data Science
• Clinical Data Management

Background:

• Electronic health record (EHR) data exhibit significant heterogeneity across institutions, hindering analysis and interpretability.
• Multisite EHR data integration amplifies challenges in data variance, impacting the usability of clinical encounter information.
• Existing methods inadequately address the inherent complexities and inconsistencies within raw EHR encounter data.

Purpose of the Study:

• To present a novel, generalizable method for resolving encounter heterogeneity in clinical data.
• To develop a technique for combining atomic encounters into composite "macrovisits" for improved analysis.
• To enhance the interpretability and usability of multisite EHR data.

Main Methods:

• Utilized harmonized data from 75 partner sites within the National Covid Cohort Collaborative (N3C) initiative.
• Developed two algorithms to refine atomic encounters into analyzable longitudinal clinical visits, termed macrovisits.
• Computed summary statistics to assess and modify data issues at both overall and site levels.

Main Results:

• Atomic inpatient encounters demonstrated significant disparities in length-of-stay (LOS) and measurement counts across sites.
• Aggregation of encounters into macrovisits led to a notable decrease in LOS and measurement variance.
• A subsequent algorithm specifically for hospitalized macrovisits further reduced overall data variability.

Conclusions:

• Clinical encounters represent a complex and heterogeneous EHR data component, with native issues often unaddressed by current methodologies.
• Foundational research and development are crucial for overcoming challenges in deriving value from complex EHR data.
• The presented method developments offer a generalizable foundation for manipulating and resolving EHR encounter data issues for future research.