Generalizability of an EHR-Network Dataset to the United States for Cardiovascular Disease Conditions: Comparison of Cerner Real World Data with the National Inpatient Sample

Observational studies tracking disease patterns, clinical characteristics of patients, treatment patterns, and outcomes of patients provide critical insights into disease epidemiology, gaps in guideline-based care, and care disparities. Traditionally, data for these types of studies wasonly available via costly prospective registries1,2, or via insurance claims data, which are reflective of insured patients only and had limited detailed patient information 3,4. More recently, the widespread availability of electronic health record (EHR) data has begun to transform observational epidemiology. Such EHR data, when aggregated across many health systems, offers significant potential for observational research. The EHR often contains much more granular patient-level information than is available in claims datasets, and with improvements in natural language processing, new data elements are increasingly available. In the past several years, there have been many efforts to aggregate EHR data from large health systems (e.g. Optum), private entities (e.g. TriNetX), government-funded networks (PCORNet), EPIC5, and Cerner6. One example of an EHR-company-based data resource is Cerner RealWorldData (CRWD). While Cerner has a very large footprint in the US7, to date, the completeness, accuracy and representativeness of these data have not been evaluated.

This study sought to compare characteristics of patients hospitalized with three different cardiovascular conditions: myocardial infarction (MI), congestive heart failure (CHF), and stroke in hospitals participating in CWRD with those in the National Inpatient Sample (NIS), a nationally representative claims database of hospitalizations in the US.

留言 (0)

沒有登入
gif