Real-Time Collection and Analysis of Big EHR Data to Enable Meaningful, Prospective Comparison of Medical Interventions

• Activity Number: 68 - Government Health Statistics
• Type: Contributed
• Date/Time: Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
• Sponsor: Government Statistics Section
• Author(s): Cynthia Hau* and Sarah Leatherman and Nilla Majahalme and Amanda Guski and Jade Riotto and Ryan Ferguson
• Companies: MAVERIC, Boston VA Healthcare System and MAVERIC, Boston VA Healthcare System and MAVERIC, Boston VA Healthcare System and MAVERIC, Boston VA Healthcare System and MAVERIC, Boston VA Healthcare System and MAVERIC, Boston VA Healthcare System
• Keywords: Electronic health records ; Big data ; Comparative effectiveness research ; Data quality

Abstract

Capturing data from electronic health records (EHRs) is widely promoted for comparative effectiveness research (CER). While CER has a critical role in advancing evidence-based practice, the use of EHRs can be difficult due to questionable data quality. To support a robust EHR-based clinical trial comparing thiazide diuretics in 13,500 US veterans from 50 VA medical centers, an integrative analytic approach was designed to retrieve EHRs from the VA Corporate Data Warehouse, external patient encounters from the Centers for Medicare and Medicaid Services database, and vital records from National Center of Health Statistics. We will share our successful method and highlight the need for statistical collaboration to ensure meaningful use of EHRs with applied statistics in clinical informatics. We will also reveal our analytics workflow to overcome potential data quality issues when applying diagnostic and predictive models to multiple large repositories including: multipath extraction of structured and textual data from distinct EHR platforms, data-driven extraction controls to monitor changes in the EHR over time, and synthetic analysis for unique outcome and safety events detection.