Text mining of patient-reported concerns during palliative chemotherapy: evolving patterns and implications for personalised supportive care
Nobumichi Takeuchi, Saiko Kurosawa, Sonomi Yoshida, Misato ItoObjectives
To analyse patient-reported concerns regarding palliative chemotherapy using text mining and to identify how these concerns evolve according to treatment phase and patient characteristics.
Methods
We conducted a retrospective cross-sectional text-mining study of free-text responses from 479 patients with advanced solid tumours treated at a single institution between 2015 and 2025. A total of 584 responses were analysed. Text data were processed using KH Coder (version 3), employing morphological analysis and co-occurrence network analysis based on the Jaccard coefficient. Words appearing two or more times were included and a threshold of Jaccard coefficient≥0.1 was applied. Frequently occurring words were categorised into four domains according to the nature of the concerns. Analyses were stratified by treatment phase, age and family structure. Sensitivity analyses were performed to assess robustness.
Results
‘Adverse effects’ were the most frequently reported concern across all groups. Symptom-related concerns decreased after treatment initiation, whereas concerns regarding treatment value and continuation increased. Practical concerns declined over time, while uncertainty-related concerns remained relatively stable. Co-occurrence network analysis showed a shift from broad and multifaceted concerns before treatment to more structured and treatment-focused concerns during treatment. Findings were consistent across sensitivity analyses.
Conclusions
Patient concerns during palliative chemotherapy are dynamic and structured, evolving according to treatment phase and social context. Free-text analysis provides clinically meaningful insights beyond structured questionnaires and supports the need for tailored, ongoing communication addressing both medical and psychosocial concerns.