An adaptable architecture for patient cohort identification from diverse data sources

Area of Science:

• Biomedical Informatics
• Clinical Trial Management
• Data Science

Background:

• Identifying suitable patient cohorts for clinical trials is challenging due to heterogeneous data sources.
• Existing methods often struggle with temporal reasoning and semantic/structural data differences.

Purpose of the Study:

• To define and validate an architecture for patient cohort identification from multiple, heterogeneous data sources.
• To develop a query model supporting temporal reasoning and independent expression of eligibility criteria.

Main Methods:

• An architecture with a query model that pre- and post-processes queries to handle data heterogeneity.
• Separation of clinical fact extraction from reasoning processes.
• Implementation of a specific query model instance.

Main Results:

• Demonstrated wide applicability of the specific query model instance.
• Successfully accessed three diverse data warehouses to determine patient counts.
• Showcased feasibility in handling temporal reasoning and data heterogeneity.

Conclusions:

• The proposed architecture justifies implementation effort by supporting temporal reasoning and heterogeneous data sources.
• The query model requires one-time implementation, with lightweight adaptors for additional data sources.
• The system successfully implemented a query model for complex eligibility criteria across diverse data warehouses.