Distributed Drug Safety Analysis Using EHR Data from German University Hospitals: The POLAR_MI ETL Pipeline and FHIR Application
德国大学医院电子病历药物警戒数据的分布式分析方法：POLAR_MI ETL（Extract-Transform ...
Summary
The Medical Informatics Initiative (MII) conducted a large-scale, distributed analysis using data from various university hospitals to detect risks associated with polypharmacy. Researchers developed an ETL pipeline based on HL7 FHIR standards and employed decentralized statistical methods to integrate and analyze vast clinical datasets from multiple institutions.
Details
This study was part of the POLAR_MI project under the Medical Informatics Initiative (MII). The goal was to detect drug-related risks stemming from polypharmacy, utilizing Electronic Health Record (EHR) data sourced from various university hospitals. Key challenges included inconsistencies in FHIR resource mapping across different Data Integration Centers (DICs), issues with timestamp reliability, and overall data heterogeneity. To address these, the research team developed a distributed, privacy-preserving analysis pipeline combining local ETL processing (Extract, Transform, Load) and central random-effects meta-analysis, based on the HL7 FHIR Core Data Set (CDS) specification. This methodology successfully analyzed approximately 788,000 inpatient encounter records from ten participating hospitals. The findings demonstrated the feasibility of drug risk assessment using a decentralized analysis model leveraging DICs and FHIR standards. This outcome is expected to be applicable as a distributed analysis model for other countries and healthcare systems.
Original content copyright by respective publishers