Evaluating Unmeasured Confounding Factors in Claims Data Using Linked Electronic Health Records: A Proof-of-Principle...

Area of Science:

• Real-world evidence research
• Pharmacoepidemiology
• Health informatics

Background:

• Claims data analyses are susceptible to residual and unmeasured confounding.
• Factors poorly captured in claims, like laboratory results or physician notes, are often available in electronic health records (EHR).
• Robust linked EHR-claims data infrastructure is crucial for advanced drug safety surveillance.

Purpose of the Study:

• To demonstrate a process for evaluating confounding factors poorly captured in claims but measurable in EHR.
• To assess the potential of EHR data to mitigate confounding in drug safety surveillance.
• To evaluate the balance of risk factors for angioedema using linked claims-EHR data.

Main Methods:

• Utilized linked claims-EHR data from the Mass General Brigham site of the FDA Sentinel Initiative.
• Extracted a cohort previously used in a claims-based query comparing sacubitril-valsartan initiators to ACE inhibitors or ARBs.
• Characterized angioedema risk factors using EHR data and compared them to claims-based proxies.

Main Results:

• Claims-based proxies effectively balanced most risk factors measurable only in EHR data.
• The study served as a proof-of-principle for using EHR data to address confounding.
• Quantitative bias analysis was deemed unnecessary due to the achieved balance on key risk factors.

Conclusions:

• Linked EHR-claims data can enhance drug safety surveillance by addressing confounding.
• This approach can be used alongside other sensitivity analyses for residual confounding evaluation.
• A strong linked data infrastructure is vital for routine application in pharmacovigilance.