Leveraging Centralized Health System Data Management and Large Language Model-Based Data Preprocessing to Identify...

Area of Science:

• Oncology
• Health Informatics
• Public Health

Background:

• Unplanned radiation therapy interruptions (RTI) compromise cancer treatment quality.
• Identifying risk factors for RTI is essential for improving patient outcomes.

Purpose of the Study:

• To evaluate the use of a centralized electronic health record warehouse and large language model (LLM) for data preprocessing.
• To facilitate the identification of risk factors associated with radiation therapy interruptions (RTI).

Main Methods:

• Analysis of demographic, behavioral, clinical, and neighborhood data for 2,130 radiotherapy patients.
• Measurement of treatment interruptions as missed days, adjusted for weekends/holidays.
• Multinomial logistic regression to identify factors associated with moderate (2-4 days) and severe (≥5 days) RTI.

Main Results:

• Moderate RTI (15.8%) linked to genitourinary/prostate cancer and Medicaid coverage.
• Severe RTI (7.7%) associated with marital status, head/neck/gynecologic cancers, Medicaid, radiation dose, and neighborhood social vulnerability.
• LLM-based preprocessing enabled efficient identification of these associations.

Conclusions:

• Automated data preprocessing effectively identified key factors associated with RTI.
• Marital status, disease site, Medicaid coverage, and social vulnerability are significant predictors of RTI.
• Data-driven risk assessment and intervention strategies are needed to maintain cancer treatment quality.