Multi-Task Learning for Extracting Menstrual Characteristics from Clinical Notes

TL;DR

The paper tackles the challenge of extracting detailed menstrual characteristics from unstructured clinical notes by introducing a retrieval-augmented, multi-task prompt-based learning framework. Using GatorTron, it compares supervised fine-tuning, in-context learning, and prompt-based approaches, and demonstrates that a hybrid retrieval step consistently improves performance. The authors propose Multi-Task Prompt-Based Learning (MTPBL), which jointly trains across five attributes with shared representations, achieving strong generalization and an average macro-F1 around 0.90 on the test set. This approach advances automated menstrual health data extraction from clinical notes, offering a practical pathway to enhance women’s health research and clinical decision support. Future work includes reducing manual prompt engineering, improving note segmentation, and validating on larger, multi-institutional datasets to ensure robustness and scalability.

Abstract

Menstrual health is a critical yet often overlooked aspect of women's healthcare. Despite its clinical relevance, detailed data on menstrual characteristics is rarely available in structured medical records. To address this gap, we propose a novel Natural Language Processing pipeline to extract key menstrual cycle attributes -- dysmenorrhea, regularity, flow volume, and intermenstrual bleeding. Our approach utilizes the GatorTron model with Multi-Task Prompt-based Learning, enhanced by a hybrid retrieval preprocessing step to identify relevant text segments. It out- performs baseline methods, achieving an average F1-score of 90% across all menstrual characteristics, despite being trained on fewer than 100 annotated clinical notes. The retrieval step consistently improves performance across all approaches, allowing the model to focus on the most relevant segments of lengthy clinical notes. These results show that combining multi-task learning with retrieval improves generalization and performance across menstrual charac- teristics, advancing automated extraction from clinical notes and supporting women's health research.

Figures (3)

• Figure 1: Number of tokens, before and after retrieval of relevant information in both, train and test dataset.
• Figure 2: NLP Pipeline for extracting and categorizing menstrual characteristics from unstructured clinical notes used in the MTPBL + retrieval approach.
• Figure 3: Average F1-scores for validation and testing using the different classification approaches presented in this work.