Extracting Pediatric Information from Summaries of Product Characterics with a Large Language Model and No-Code

Area of Science:

• Pediatric pharmacology
• Health informatics
• Natural Language Processing in Medicine

Background:

• Accurate pediatric medication information is critical due to higher risks of dosing errors in children compared to adults.
• Existing methods for extracting pediatric drug data from regulatory documents can be time-consuming and require specialized expertise.

Purpose of the Study:

• To develop and evaluate an automated method for extracting pediatric information, specifically indications, from Summaries of Product Characteristics (SPCs).
• To assess the feasibility of using a no-code platform with advanced language models for this task.

Main Methods:

• Utilized AirOps, a no-code visual editor, to build a processing chain incorporating the GPT-4o mini large language model.
• Processed 50 Summaries of Product Characteristics (SPCs) to extract sentences related to pediatric indications.
• Evaluated the performance using recall and precision metrics.

Main Results:

• The automated method achieved a 95% recall rate and a 78% precision rate in extracting relevant pediatric indication sentences.
• The large language model demonstrated reliability in classifying drugs based on their pediatric uses.
• A no-code approach enabled implementation by healthcare professionals without IT backgrounds.

Conclusions:

• The proposed automated, no-code method effectively extracts pediatric drug information from SPCs with high accuracy.
• This approach enhances the accessibility and efficiency of identifying crucial pediatric medication data for clinical practice and research.
• Leveraging large language models in no-code platforms democratizes advanced data extraction for medical professionals.