From Classic to Cutting Edge: Addressing Data Quality Challenges in Electronic Health Record Data

Background: Electronic medical record (EHR) data has the potential to provide access to many of the rich clinical details that administrative claims data lack. Despite the great potential, data quality challenges have made it difficult to realize this vision. This symposium will focus on addressing those challenges.

Objectives: The main objectives of the symposium are to highlight a range of approaches to addressing data quality concerns that arise in the use of EHR data.

Description: We will present four case-studies from real-world data followed by 30 minutes for panel discussion and audience Q&A. Target audience: researchers who are involved in design and/or conduct of studies using EHR data.

(1) Use of multiple imputation (MI) to address incomplete data on body mass index in the CPRD (15m); Michele Jonsson Funk. Discuss assumptions, diagnostics, necessary inputs and selection of predictors for MI using findings from work conducted using the CPRD. Approaches for more complex settings when multiple covariates have missing data will be discussed.

(2) Selection of data partners based on completeness of in-hospital clinical data in the National Covid Cohort Collaborative (N3C) (15m); Hythem Sidky. Present findings from N3C regarding assessment of variability in data quality and completeness across 65 data partners using 4 different common data models. Discuss key factors in the selection of a subset of those sites for a study of real-world effectiveness of COVID treatment. Considerations related to study-specific critical data elements, time-varying missing data, sample size impacts, continuity of care, and internal validity will be highlighted.

(3) Improving data quality from routine oncology care for clinical trial readiness (15m); Anne-Marie Meyer. Present framework for integrating different technology approaches across longitudinal data pipeline. Highlight both Syntactic and Semantic Tools available for data quality checks along the data journey and at key normalization and transformation points (e.g., clinical phenotyping).

(4) Improving data quality by increasing their utility to healthcare providers in the Observational Pharmaco-Epidemiology Research and Analysis (OPERA) database (15m); Jennifer Fusco. Discuss data quality improvement through aligning the needs of researchers and practitioners. Share techniques for achieving high quality data through payer reporting, smartphone tools, and quality metrics to support reporting requirements, quality improvement initiatives, and patient care needs.