Data Sources

Nathan Hall

2024-04-10

Introduction

Real-world data (RWD) encompasses a vast array of information collected from a variety of sources outside traditional clinical research settings. This data, when analyzed, generates real-world evidence (RWE) that can inform healthcare decisions, policy-making, and patient care.

The Observational Medical Outcomes Partnership (OMOP) was a collaborative initiative chaired by the U.S. Food and Drug Administration (FDA) aimed at improving the understanding of how observational studies can contribute to evaluating the safety and effectiveness of medications. It focused on developing methods, tools, and standards to enhance the quality and reliability of real-world evidence generated from healthcare data. The OMOP’s legacy continues to influence current research methodologies, particularly through the adoption and evolution of its Common Data Model (CDM) in projects like those undertaken by the Observational Health Data Sciences and Informatics (OHDSI) community.

The OMOP CDM standardizes observational healthcare data, enabling the harmonization of disparate data sources into a consistent format. This standardization encompasses structured data schemas, controlled vocabularies for medical terminologies, and unique identifiers for clinical concepts, ensuring that data from varied healthcare systems and studies can be compared and analyzed collectively. By transforming diverse datasets into a unified model, the OMOP CDM facilitates robust, scalable, and reproducible research across the global healthcare research community, enhancing the reliability and validity of findings derived from real-world data.

The Extract-Transform-Load (ETL) process is a critical data management procedure that prepares raw data for analysis by extracting it from various sources, transforming it into a consistent format that aligns with analysis requirements, and loading it into a destination system for use. In the context of generating real-world evidence (RWE) and ensuring adherence to a Common Data Model (CDM) like OMOP, the ETL process is indispensable. It enables the standardization of observational healthcare data—often fragmented across different systems and formats—into a unified structure. This standardization is essential not only for the interoperability of data across different healthcare databases but also for ensuring that analyses conducted on these data are reliable, scalable, and reproducible. For more information on the ETL process and how OHDSI tools can be utilizied to perform it for a particular data source, please visit this page.

Key Data Sources for RWE Analyses

Real-world data sources that adhere to the OMOP CDM are varied and rich, including, but not limited to:

  1. Electronic Health Records (EHRs): These are one of the primary sources of RWD, containing detailed patient health information from clinical encounters.
  2. Claims and Billing Data: These data sources provide information on healthcare utilization, including diagnoses, procedures, and medication prescriptions, primarily for billing purposes.
  3. Patient Registries: Disease or condition-specific registries offer valuable insights into patient outcomes, treatment patterns, and disease epidemiology.
  4. Wearable Devices and Mobile Health Applications: An emerging source of RWD, these technologies capture real-time data on patient health and activity levels.

Example OMOP-mapped Datasets

The sensitivity of observational health data underscores the need for robust privacy protection measures due to the potential identification of individuals and the disclosure of sensitive health information. This sensitivity arises from the richness of health data, including diagnoses, treatments, and demographic details, which can reveal personal health conditions and behaviors. Ensuring privacy in observational health data is paramount to maintain trust and compliance with regulations such as the Health Insurance Portability and Accountability Act (HIPAA). Despite these challenges, several publicly available data sources mapped to the OMOP CDM offer valuable insights for research while safeguarding patient privacy. Some examples include:

  1. HowOften
  2. Public Datasets from OHDSI
  3. SynPUFs