Using regular expressions to extract information on pacemaker implantation procedures from clinical reports

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Data Management

Background:

• Surgical reports contain valuable clinical data for cardiac pacing and defibrillation devices.
• Manual data extraction for registries is time-consuming and prone to errors.
• Automated methods are needed to efficiently populate clinical registries.

Purpose of the Study:

• To evaluate natural language processing (NLP) methods for extracting clinical data from free-text surgical reports.
• To assess the feasibility of using NLP to populate a registry for cardiac pacing and defibrillation devices.
• To compare the performance of an NLP system against manual data abstraction.

Main Methods:

• Developed a name entity recognition (NER) system using GATE and regular expressions.
• Analyzed 232 surgical reports, performing manual abstraction and system-based information extraction.
• Evaluated the system using sensitivity, positive predictive value, and accuracy metrics.

Main Results:

• The NLP system achieved high sensitivity (>90%) and very high positive predictive value (>95%) in extracting diverse data types.
• Extracted information included numeric, text, and combined data formats.
• The system's precision exceeded that of the original manually populated registry database.

Conclusions:

• The developed NLP tool, based on GATE open-source components, offers a robust method for information extraction in specialized domains like pacemaker reports.
• This approach enhances the efficiency and accuracy of populating clinical registries.
• Further research is recommended to extend the application of this tool to broader clinical domains.