Leveraging Data Pipeline and LLM to Advance Patient Safety Event Studies
View abstract on PubMed
Summary
This summary is machine-generated.This study introduces an ETL-LLM pipeline to standardize medical device report (MDR) data analysis, improving event categorization accuracy and efficiency for better patient safety research.
Area Of Science
- Medical Device Safety
- Health Informatics
- Natural Language Processing
Background
- Medical device report (MDR) data extraction from the MAUDE database often lacks clear methodologies, hindering reproducibility.
- Inconsistent data processing affects the reliability of patient safety research.
Purpose Of The Study
- To develop and demonstrate an Extract-Transform-Load (ETL) pipeline coupled with a Large Language Model (LLM) for analyzing MDR narratives.
- To enhance the accuracy and efficiency of categorizing adverse events reported in medical device data.
- To explore the application of this approach in patient safety research.
Main Methods
- Utilized the OpenFDA API and a custom MAUDE ETL pipeline to standardize MDR data extraction and transformation.
- Employed a Large Language Model (LLM) to analyze free-text narratives within the processed MDRs.
- Demonstrated the ETL-LLM approach using MDRs specifically for endoscopic mucosal resection devices.
Main Results
- The ETL-LLM pipeline standardizes MDR data, enabling more consistent analysis.
- LLM-driven analysis of free-text narratives improves the accuracy and efficiency of event categorization.
- The approach shows potential for broader application in analyzing diverse medical device data.
Conclusions
- The developed ETL-LLM pipeline offers a robust method for analyzing complex MDR data.
- Standardized data processing and LLM analysis are crucial for advancing data-driven patient safety research.
- Further expansion of MDR sample size and diversity is recommended for enhanced research capabilities.