AI-Driven Prediction of Cancer Pain Episodes: A Hybrid Decision Support Approach

2025-12-18

AI Summary

Breakthrough pain (BTP) occurs frequently in lung cancer patients (up to 91%), necessitating proactive analgesic intervention through early prediction. To address this, we propose the first hybrid predictive framework synergizing machine learning (ML) and large language models (LLMs): XGBoost and LSTM models process structured, longitudinal medication data, while a fine-tuned medical LLM (a Med-PaLM variant) interprets unstructured clinical orders and free-text notes. Multimodal feature fusion integrates outputs from both modalities, and SHAP-based interpretability analysis ensures clinical transparency. Evaluated on 266 inpatients, the framework achieves predictive accuracy of 0.874 at 48 hours and 0.917 at 72 hours. LLM integration improves sensitivity by 8.6% (48-h) and 10.4% (72-h), significantly enhancing both clinical interpretability and timeliness of intervention.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Lung cancer patients frequently experience breakthrough pain episodes, with up to 91% requiring timely intervention. To enable proactive pain management, we propose a hybrid machine learning and large language model pipeline that predicts pain episodes within 48 and 72 hours of hospitalization using both structured and unstructured electronic health record data. A retrospective cohort of 266 inpatients was analyzed, with features including demographics, tumor stage, vital signs, and WHO-tiered analgesic use. The machine learning module captured temporal medication trends, while the large language model interpreted ambiguous dosing records and free-text clinical notes. Integrating these modalities improved sensitivity and interpretability. Our framework achieved an accuracy of 0.874 (48h) and 0.917 (72h), with an improvement in sensitivity of 8.6% and 10.4% due to the augmentation of large language model. This hybrid approach offers a clinically interpretable and scalable tool for early pain episode forecasting, with potential to enhance treatment precision and optimize resource allocation in oncology care.

Problem

• Predicts cancer pain episodes using AI and clinical data.
• Enhances sensitivity by integrating machine learning and language models.
• Provides interpretable forecasting to improve oncology pain management.

Innovation

• Hybrid machine learning and large language model pipeline
• Uses structured and unstructured electronic health record data
• Improves sensitivity and interpretability for pain prediction