Design considerations, architecture, and use of the Mini-Sentinel distributed data system

Area of Science:

• Health Informatics
• Public Health Surveillance
• Database Systems

Background:

• The US Food and Drug Administration (FDA) sought to develop an active surveillance system for medical product safety.
• Existing data systems required adaptation for a large-scale, distributed approach.

Purpose of the Study:

• To describe the design, implementation, and use of a large, multiorganizational distributed database.
• To support the Mini-Sentinel Pilot Program for active surveillance of medical product safety.
• To inform the development of a national active surveillance system.

Main Methods:

• A common data model was designed and applied by participating organizations to their source data.
• Data extraction, transformation, and loading (ETL) processes were standardized.
• Centralized program development and local execution facilitated data characterization and evaluation.
• A secure portal and distributed querying software were implemented for data access and analysis.

Main Results:

• The Mini-Sentinel Distributed Database included 99,260,976 health plan members as of July 2011.
• The database contained 316,009,067 person-years of observation time.
• All data partners successfully executed distributed code and returned findings.

Conclusions:

• Building a large, multiorganizational distributed data system is feasible.
• Organizations can retain data possession while contributing to an active surveillance system.
• This approach facilitates robust monitoring of medical product safety.