Enhanced Screening and Research Data Collection via Automated EHR Data Capture and Early Identification of Sepsis

Area of Science:

• Clinical Informatics
• Critical Care Medicine
• Health Data Science

Background:

• Clinical research, particularly in sepsis, necessitates extensive longitudinal data collection.
• Electronic health records (EHRs) offer potential for identifying sepsis patients and facilitating recruitment.
• Extracting reliable and usable data from EHRs presents significant challenges.

Purpose of the Study:

• To explore infrastructures for capturing EHR data for sepsis research.
• To apply criteria for identifying sepsis patients within EHR systems.
• To assess the feasibility of abstracting EHR data for future clinical studies.

Main Methods:

• A prospective feasibility study was conducted to locate and capture EHR data.
• Data was extracted from Health Level Seven (HL7) interfaces into a prototype database.
• Patient identification for sepsis was evaluated at two time points and over a 2-month period in an intensive care unit (ICU).

Main Results:

• Most selected EHR parameters were accessible through an iterative abstraction process.
• Sepsis patients were identified in the ICU using four data interfaces.
• Retrospective application of criteria showed lower sensitivity for direct ICU admissions with sepsis; classification accuracy (Kappa .39) was fair compared to manual review.

Conclusions:

• EHR data abstraction for sepsis research is feasible but requires further refinement for customizable reports.
• Barriers in data extraction are confirmed, highlighting the complexity of using EHR data for sepsis classification.
• Researchers should collaborate with IT departments to electronically apply research criteria for improved screening at ICU admission.