Heterogeneity introduced by EHR system implementation in a de-identified data resource from 100 non-affiliated...

Area of Science:

• Biomedical Informatics
• Health Services Research
• Digital Health

Background:

• Aggregate electronic health record (EHR) data from multiple sources is valuable for research, including digital phenotyping.
• Analyzing aggregate EHR data poses unique challenges compared to single-institution data.

Purpose of the Study:

• To investigate the impact of EHR implementation factors on aggregate data quality and variability.
• To identify specific factors contributing to data inconsistencies in a large, de-identified EHR dataset.

Main Methods:

• Utilized the Cerner Health Facts dataset, comprising de-identified EHR data from 100 independent health systems.
• Examined the influence of ancillary module usage, data continuity, ICD version adoption, and clinical documentation prompts.

Main Results:

• Significant variation exists in the utilization of EHR ancillary modules and data contribution across facilities.
• Inconsistent adoption of ICD-10 and variable use of fields like "discharge disposition" were observed.
• Documentation of clinical events, such as smoking history, showed fluctuations related to implementation factors like Meaningful Use.

Conclusions:

• Variability in aggregate EHR data is often attributable to diverse implementation practices, not simply missing information.
• Researchers must account for EHR system configurations and adoption timelines when analyzing aggregate EHR data.
• Understanding these implementation factors is crucial for accurate interpretation of digital phenotyping and other research using aggregate EHR data.