Natural language processing to identify documented pain preceding radiation therapy for bone metastases

Area of Science:

• Oncology
• Medical Informatics
• Health Services Research

Background:

• Radiotherapy (RT) is vital for managing cancer symptoms, particularly pain associated with bone metastasis (BM).
• Effective symptom documentation is crucial for timely and appropriate treatment initiation.
• Understanding symptom reporting patterns can reveal disparities in care.

Purpose of the Study:

• To characterize symptom documentation, focusing on pain, before BM diagnosis and RT initiation for BM.
• To identify demographic factors associated with pain documentation in patients receiving RT for BM.
• To leverage natural language processing (NLP) for detailed clinical data extraction.

Main Methods:

• A cohort of 1,061 patients receiving RT for BM (2013-2023) was analyzed.
• NLP techniques were employed to extract CTCAE-encoded symptoms from clinical notes within 30 days of BM diagnosis and RT initiation.
• Logistic regression analyzed associations between demographics and pain documentation.

Main Results:

• BM-related pain documentation increased significantly from before BM diagnosis (52.5%) to before RT (91.6%).
• Multiracial/other race patients showed decreased pain documentation pre-diagnosis compared to white patients (OR=0.61).
• Women demonstrated increased pain documentation pre-RT compared to men (OR=1.48).

Conclusions:

• A relative increase in pain documentation was observed for women and multiracial/other race patients between BM diagnosis and RT initiation.
• This suggests potential differences in RT decision-making based on symptom trajectory and demographics.
• Interventions are needed to ensure equitable access to RT for bone metastasis.