Longitudinal Data Discontinuity in Electronic Health Records and Consequences for Medication Effectiveness Studies

Area of Science:

• Health Informatics
• Real-World Evidence Research
• Biostatistics

Background:

• Electronic health record (EHR) discontinuity introduces information bias in real-world evidence (RWE) studies.
• A validated algorithm exists to identify patients with high EHR continuity.

Purpose of the Study:

• To evaluate if an EHR continuity algorithm can mitigate bias in RWE studies.
• To assess the impact of EHR discontinuity on bias in drug effect estimation.

Main Methods:

• Utilized Medicare beneficiaries (>=65 years) from 2007-2014 across four RWE cohorts.
• Linked EHR data with claims data to quantify bias from external healthcare encounters.
• Calculated percentage bias in effect estimates comparing EHR-only versus linked EHR-claims data.

Main Results:

• Patients in the lowest quartile of EHR continuity captured significantly fewer outcome events (21-81%).
• Bias ranged from 6-99% for short-term and 62-112% for long-term outcomes.
• Bias was more pronounced with nonuser vs. active comparators; no significant heterogeneity by continuity found.

Conclusions:

• Excluding patients with low algorithm-predicted EHR continuity is advisable in EHR-based RWE studies.
• Low EHR continuity leads to unreliable treatment effect estimates due to incomplete outcome capture.
• Improving EHR continuity enhances the validity of real-world evidence research.