To systematically characterize how cancer patients experience and articulate psychosocial issues using advanced natural language processing techniques.
Approach:
Study Design: The study proposes an integrated, domain-adapted pipeline combining TopicGPT with network analysis to examine psychosocial issues in cancer patients.
Methodology: Utilizes TopicGPT to generate coherent themes from patient narratives and maps them at the sentence level to identify psychosocial challenges.
Network Analysis: Constructs a network where topics are nodes and their co-occurrences form edges, modeling relationships between different psychosocial issues.
Key Findings:
Cancer burden estimated at 20 million new cases in 2022, with 1 in 5 individuals developing cancer.
69.9% of cancer patients survive at least 5 years post-diagnosis.
Traditional research methods are limited in scope and scalability, particularly for sensitive topics.
TopicGPT shows superior performance in extracting nuanced themes compared to traditional topic models.
Interpretation:
Limitations:
Applications of TopicGPT in psycho-oncology remain largely unexplored.
The qualitative corpora from online health communities may vary in quality and structure.
