Heart failure (HF) is the #1 cause of hospitalization in the US. Yet HF can be tough to study, with many recent trials stopping early without results, or completed with suspect data, making conclusions impossible. Recent successful trials in HF required thousands of patients, and took more than 5 years to complete. While Real-World Evidence (RWE) research using claims data is possible for some diseases, HF is too heterogeneous, and requires a more complete picture from an electronic health record (EHR). With recent advances, aggregated EHR may be mature enough to be considered as alternative to prospective trials.
We describe our efforts to simulate a ‘virtual trial’ in a selected HF cohort with one such aggregated EHR data set. Our team used data from diagnoses, procedures, labs, medications, and other information extracted from clinician notes. We first applied medical guideline definitions, then worked through scenarios to adjust for missing data. I will describe our results with multiple methods to simulate a clinical trial, and characterize potential applications for other areas of study.