Linking clinically rich data from a large electronic health record (EHR) database, the RPDR, to multiple claims data and other long-term outcome data, we have established a valuable dataset that can enable and enrich the analyses in the following use case scenarios:
- Pragmatic randomized trials – The linked data provide rich data to determine clinical phenotypes and enable efficient recruitment of eligible subjects. The database also provides comprehensive data for long-term outcome ascertainment even if they occur outside the Partners EHR system.
- Comparative effectiveness research – To compare the effectiveness and safety of new diabetes medications, claims data and EHR data can complement each other to provide a wealth of information. Claims data have sufficient size and full capture of both medication exposure and health outcomes, which may happen outside of the Partners EHR system. EHR data provide information on BMI, Hemoglobin A1C results, duration of diabetes, and renal function estimated by laboratory test results. Together, large-scale populations can be studied with this detailed clinical information.
Validation of claims-based algorithm – Much of drug safety and effectiveness research is conducted using claims data. Many claims-based algorithms for outcome ascertainment or patient phenotyping need to be validated against the gold-standard, established via EHR through chart review and/or confirmation by testing/laboratory results. For example, we have used the linked EHR-claims data to develop and validate an algorithm to identify patients with reduced vs. preserved ejection fractions (EF) in claims data. The gold-standard was established based on echocardiogram or other cardiac imaging available in Partners EHR data.
- Personalized medicine (Highly stratified treatment effect evaluation) – To precisely characterize individual patients, we need the EHR system, which contains information on relevant genetic testing and biomarker profiles. To study how these factors influence effectiveness of medical interventions, we need claims data to assess longitudinal medication use patterns and clinical outcomes that frequently happen outside the Partners EHR system.
- Adherence patterns – EHR systems record the provider’s prescribing information. However, it is known that 20-60% of prescriptions are never filled and for many medications, adherence is often as low as 50% within 6 months of initiation. Claims data, which provides information on actual filling of prescriptions, provide a more complete picture of utilization patterns, switching, and discontinuation. This can be critical when disentangling the effect of the drug from optimal use and evaluating need and effectiveness of adherence improvement strategies.
- Natural history of disease – EHR data enable better characterization of natural history of diseases defined by biomarkers or genetic polymorphism, while claims data provide long-term follow-up to characterize how the disease state of these patients changes and how care patterns adjust.
- Burden of disease – Since claims data record all professional health services that resulted in insurance payments, they suit themselves well to describe longitudinal utilization patterns as well as associated costs. This allows us to assess the resource use and economic burden of specific diseases.