OM1 - Noise vs. Value: Novel Approaches to Addressing Data Gaps in Real-World Data Sources

Background: Gaps in data are a common problem in real-world data (RWD) sources. Key variables for RWD studies may be found only in unstructured clinical notes or missing entirely from sources such as EMRs. Multiple methods, including novel approaches using artificial intelligence (AI) can be implemented according to specific research needs, trading off between the volume of datapoints generated, and individual datapoint quality.


Objectives: Review novel approaches, including the use of AI-based methods, to address data gaps in RWD sources

Description: This presentation will review the strengths and limitations of clinician abstraction, drawing on case examples of RWD studies that have used these approaches. Dr. Joseph Zabinski will present on the use of AI-based prediction models in RWD studies. He will discuss how to determine which approach is fit for purpose for a particular study, including balancing statistical noise with data availability, and describe lessons learned from implementation of these techniques across a range of clinical areas. Audience members will be invited to raise questions throughout the presentations and share examples of missing data challenges and novel approaches from their own work. The audience will also be challenged to identify strengths and limitations of various approaches for different research purposes. Researchers who design RWD studies and stakeholders who evaluate the findings of RWD studies would benefit from attending.