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

IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS 1 AI-Driven Prediction of Cancer Pain Episodes: A Hybrid Decision Support Approach Yipeng Zhuang, Member, IEEE, Yifeng Guo, Member, IEEE, Yuewen Li, Yuheng Wu, Philip Leung-Ho Yu, Tingting Song, Zhiyong Wang, Kunzhong Zhou, Weifang Wang, and Li Zhuang 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 unstruc- tured 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 tem- poral 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 clini- cally interpretable and scalable tool for early pain episode forecasting, with potential to enhance treatment precision and optimize resource allocation in oncology care. Index Terms— Lung cancer pain, machine learning, large language models, electronic health records, pain predic- tion, analgesic ladder, clinical decision support I. INTRODUCTION Lung cancer remains the leading cause of cancer-related mortality globally, with pain affecting 42.2% of patients and 91% reporting moderate-severe intensity [1]. The World Health Organization (WHO) recommends a stepwise analgesic This work was supported by the National Natural Science Foundation of China (8226110412 and 82460540), the Key Project of the Basic Joint Special Program of Yunnan Provincial Science and Technology Department - Kunming Medical University (202401AY070001-014), the Construction of First-Class Discipline Team of Kunming Medical Uni- versity (2024XKTDPY08), and the Medical and Health Talent Special Project of the ”Support Program for Developing Talents in Yunnan” (Grant No. CZ0096-901895). (Y. Zhuang, Y. Guo, and Y. Li contributed equally to this work as co-first authors.) (Corresponding author: L. Zhuang.) Y. Zhuang is with the Department of Orthopaedics and Traumatology, The University of Hong Kong, Hong Kong (email: yipengzh@hku.hk). Y. Guo, and P.L.H. Yu are with the Department of Statistics and Actuarial Science, Faculty of Science, The University of Hong Kong, Hong Kong (email: gyf9712@connect.hku.hk; plhyu@eduhk.hk). Y. Li, T. Song, Z. Wang, K. Zhou, W. Wang, and L. Zhuang are with the Rehabilitation and Palliative Medicine Department, Peking University Cancer Hospital Yunnan Hospital, The Third Affiliated Hospital of Kunming Medical University, Kunming, China (email: 20211312@kmmu.edu.cn; medical- song@163.com; 20201184@kmmu.edu.cn; 20231409@kmmu.edu.cn; 20231408@kmmu.edu.cn; zhuangli@kmmu.edu.cn). Y. Wu is with the Department of Biomedical Engineering, City Univer- sity of Hong Kong, Hong Kong (email: yuhengwu7-c@my.cityu.edu.hk). ladder for cancer pain management, progressing from non- opioid analgesics (e.g., NSAIDs) to weak (e.g., codeine) and strong opioids (e.g., morphine) [2]. However, despite these guidelines, effective pain control remains challenging due to limitations in provider training and restricted access to opioids in many settings [3]. Predicting pain episodes in hospitalized patients holds sig- nificant clinical value by enabling timely analgesic adjust- ments, preventing breakthrough pain, and optimizing resource utilization. This is especially critical in inpatient oncology, where lung cancer patients experience among the highest rates of moderate to severe pain. Delays in pain control can lead to escalated opioid use, reduced functional status, and diminished quality of life. In clinical practice, a 72-hour observation window is often used to evaluate the effectiveness of pain management strategies, consistent with WHO and institutional guidelines [4]–[6]. Machine learning (ML) methods have been used to forecast cancer pain and other symptoms using structured EHR data. For example, in musculoskeletal pain, baseline disability and early symptom changes can predict healthcare utilization [7]; in migraine, ML has been used to classify pain trajecto- ries based on self-reported features [8]; and in end-of-life care, structured electronic health record (EHR) models have forecasted pain at hospitalization and shift levels [9]. ML models, including logistic regression, random forests, and neural networks, have also been applied to predict cancer- related symptoms such as pain, fatigue, and depression [10], [11]. However, most existing approaches treat symptoms and medicatio

This content is AI-processed based on open access ArXiv data.