Biases in Electronic Health Records Data for Generating Real-World Evidence: An Overview.

Area of Science:

• Health Informatics
• Clinical Research Methodology
• Real-World Evidence Generation

Background:

• Electronic Health Records (EHR) are increasingly recognized as a valuable data source for clinical research.
• EHRs provide large volumes of real-time data from real-world clinical settings.
• Secondary use of EHR data presents both opportunities and challenges for research.

Purpose of the Study:

• To review the secondary uses of EHR data in clinical research.
• To identify opportunities and data deficiencies limiting causal inference.
• To provide a comprehensive overview of biases in EHR data generation and research processes.

Main Methods:

• Review of literature on secondary uses of EHR data.
• Identification and categorization of biases at healthcare system and research levels.
• Analysis of data deficiencies and potential biases impacting causal relationships.

Main Results:

• EHR data offers significant potential but is subject to numerous deficiencies and biases.
• Biases arise at the healthcare system level (access to care, data documentation) and research level (data extraction, analysis, interpretation).
• Selection and information bias are prominent issues limiting the validity of causal inferences.

Conclusions:

• Secondary use of EHR data requires extreme caution due to inherent biases.
• Careful consideration of data deficiencies and bias sources is crucial for valid research.
• The potential for causal inference from EHR data is limited without rigorous bias mitigation strategies.