🤖 AI Summary
This study addresses the challenges of fragmented medical data, storage-centric architectures, and redundant reconstruction of clinical states for downstream applications by proposing PRomop—a longitudinal patient health record representation grounded in the OMOP Common Data Model (CDM) version 5.4, extended with oncology-specific elements. PRomop compresses each patient’s complete clinical history into a single-row, 286-column decision-ready format through flat projection and precomputes key clinical states to enable efficient reuse while preserving full compatibility with OMOP standards. Deployed across two oncology institutions, the approach supports eligibility screening for 19,500 clinical trials; empirical evaluation on synthetic data demonstrates a 23.9-fold acceleration in trial matching and a 30- to 200-fold reduction in query complexity.
📝 Abstract
Objective: Health systems and biopharma face a persistent gap between storing patient data and acting on it. Records are fragmented across providers, structured for storage rather than decision-making, and each downstream application independently reconstructs patient clinical state. We present PRomop (PatientRecord on OMOP), an open-source longitudinal patient record designed to close that gap.
Materials and Methods: PRomop extends the OMOP Common Data Model (CDM 5.4) with oncology extensions and introduces PatientRecord, a flattened projection that collapses each patient's longitudinal history into a single decision-ready row of 286 columns. Clinical state--including lines of therapy, disease status, and normalized biomarkers--is derived once during projection and materialized for reuse by analytics, clinical trial matching, and standard-of-care evaluation. We report production deployments and a controlled benchmark on synthetic data.
Results: PRomop is deployed by two independently governed oncology organizations--the HealthTree Foundation (14,000 blood-cancer patients) and CancerBot (3,500)--supporting trial matching across 19,500 actively recruiting trials in five cancer types. A representative 20-criterion eligibility query requiring 27-39 joins on raw OMOP requires none against PatientRecord (analytical estimate: 30x-200x reduction). In a benchmark using 100 Synthea-generated breast-cancer patients, eligibility screening averaged 0.92 ms versus 20.7 ms from raw OMOP, a 23.9x speedup.
Discussion and Conclusion: PatientRecord provides a shared, decision-ready foundation for downstream applications, eliminating repeated clinical-state derivation while preserving OMOP conformance. PRomop demonstrates that a flattened projection over a standards-based longitudinal record is a practical, deployed architecture for analytics, AI/ML, clinical trial matching, and decision support.