Characterizing Cancer Symptom Experiences and Related Psychosocial Concerns Through Chinese Social Media Narratives: Computational Text Analysis.
- 1Department of Epidemiology and Biostatistics, School of Public Health, Anhui Medical University, Hefei, Anhui, China.
- 2School of Nursing, Anhui Medical University, No. 26 Xinglin Road, Feixi County, Hefei, Anhui, 231200, China, 86 055165161009.
- 3Nursing International Collaboration Research Center of Anhui Province, Hefei, Anhui, China.
Abstract
Background
Patients with cancer frequently experience complex and multidimensional symptom burdens. Standardized patient-reported outcome measures provide structured approaches to symptom assessment but may not capture the full range of symptom-related and psychosocial concerns expressed in patients' narratives. Patient-generated social media data offer a complementary source for examining how these concerns are expressed online.Objective
This study aimed to characterize cancer symptom-related narratives in Chinese social media posts using a conventional symptom assessment framework and to identify additional experiential themes through neural topic modeling.Methods
A computational text analysis integrating large language model (LLM) annotation and neural topic modeling was conducted. A corpus of 26,086 first-person patient narratives was collected from Xiaohongshu and Zhihu between November 2013 and February 2026 using cancer-related search terms. First-person narratives and 6 SPPADE (sleep disturbance, pain, physical function impairment, anxiety, depression, and low energy or fatigue) symptom domains were identified using GPT-4o-mini (OpenAI) and validated against independent human coding. Symptom occurrence, co-occurrence, multiplicity, and coping strategies were characterized, with exploratory stratified analyses by cancer type. BERTopic modeling was applied to identify latent themes, platform sensitivity analyses, and temporal distribution were examined.Results
For first-person narrative identification, human raters showed 89.2% agreement (Cohen κ=0.784); the LLM achieved an F₁-score of 81.2% against the adjudicated reference standard. For SPPADE extraction, mean Cohen κ values for LLM-human agreement were 0.714 (SD 0.114) and 0.754 (SD 0.095) across the 6 symptom domains. Overall, 13,896 of 26,086 (53.3%) posts contained at least 1 identifiable symptom. Pain was the most frequently identified symptom (5783/26,086, 22.2%), followed by physical function impairment (4153/26,086, 15.9%) and depression (4050/26,086, 15.5%). Pain and physical function impairment formed the most frequent co-occurring symptom pair (n=1407). Coping strategies were described in 7104/26,086 (27.2%) posts; medication-based approaches were most prominent, particularly for pain and sleep disturbance, while social and psychological support were more common for anxiety and depression. BERTopic modeling identified 13 latent themes, including cancer treatment trajectories, health care system interaction, existential distress, family involvement, and financial burden, highlighting experiential concerns that may be underrepresented in conventional symptom assessment. The proportion of posts containing at least 1 SPPADE symptom was similar between Xiaohongshu and Zhihu (53.2% vs 54.1%; P=.35), although individual symptom distributions differed across platforms.Conclusions
Chinese social media posts contained multidimensional symptom-related narratives alongside psychosocial and experiential concerns that may be less visible in structured symptom assessment. Patient-generated social media data may therefore provide a complementary source for understanding cancer symptom experiences and supportive care concerns. However, differences in platform composition, uneven temporal coverage, and the nonrepresentativeness of social media users limit generalizability, and the clinical utility of integrating such data into routine symptom assessment requires further evaluation.
Related
- Characterizing cognitive concerns in subjective cognitive decline using large language models and electronic health record notes
- Lived experiences of psychosocial and quality of life outcomes for adults newly diagnosed with breast cancer in interior British Columbia: qualitative research
- Negotiating Mortality: A Qualitative Systematic Review and Thematic Synthesis of the Psychosocial Experiences of Pancreatic Cancer Across the Illness Trajectory.
- Patient-Reported Experiences of Long-Term Recovery and Persistent Psychosocial Burden in NMDAR Encephalitis.
- Effectiveness of psychosocial and therapeutic interventions for children and adolescents with adverse childhood experiences: A systematic review and meta‐analysis
BibTeX
@article{yu2026characterizing,
title = {{Characterizing Cancer Symptom Experiences and Related Psychosocial Concerns Through Chinese Social Media Narratives: Computational Text Analysis.}},
author = {Huan Yu and Yin Wang and Jiaqi Ding and Ying Wang and Yi Zhang and Yaxi Wang and Jingfang Hong},
journal = {Journal of medical Internet research},
year = {2026},
month = oct,
doi = {10.2196/102862},
url = {https://europepmc.org/article/MED/42902828},
note = {CC BY},
}
RIS
TY - JOUR
TI - Characterizing Cancer Symptom Experiences and Related Psychosocial Concerns Through Chinese Social Media Narratives: Computational Text Analysis.
AU - Huan Yu
AU - Yin Wang
AU - Jiaqi Ding
AU - Ying Wang
AU - Yi Zhang
AU - Yaxi Wang
AU - Jingfang Hong
PY - 2026
DA - 2026/10/20
JO - Journal of medical Internet research
DO - 10.2196/102862
UR - https://europepmc.org/article/MED/42902828
AB - Background Patients with cancer frequently experience complex and multidimensional symptom burdens. Standardized patient-reported outcome measures provide structured approaches to symptom assessment but may not capture the full range of symptom-related and psychosocial concerns expressed in patients' narratives. Patient-generated social media data offer a complementary source for examining how these concerns are expressed online.Objective This study aimed to characterize cancer symptom-related narratives in Chinese social media posts using a conventional symptom assessment framework and to identify additional experiential themes through neural topic modeling.Methods A computational text analysis integrating large language model (LLM) annotation and neural topic modeling was conducted. A corpus of 26,086 first-person patient narratives was collected from Xiaohongshu and Zhihu between November 2013 and February 2026 using cancer-related search terms. First-person narratives and 6 SPPADE (sleep disturbance, pain, physical function impairment, anxiety, depression, and low energy or fatigue) symptom domains were identified using GPT-4o-mini (OpenAI) and validated against independent human coding. Symptom occurrence, co-occurrence, multiplicity, and coping strategies were characterized, with exploratory stratified analyses by cancer type. BERTopic modeling was applied to identify latent themes, platform sensitivity analyses, and temporal distribution were examined.Results For first-person narrative identification, human raters showed 89.2% agreement (Cohen κ=0.784); the LLM achieved an F₁-score of 81.2% against the adjudicated reference standard. For SPPADE extraction, mean Cohen κ values for LLM-human agreement were 0.714 (SD 0.114) and 0.754 (SD 0.095) across the 6 symptom domains. Overall, 13,896 of 26,086 (53.3%) posts contained at least 1 identifiable symptom. Pain was the most frequently identified symptom (5783/26,086, 22.2%), followed by physical function impairment (4153/26,086, 15.9%) and depression (4050/26,086, 15.5%). Pain and physical function impairment formed the most frequent co-occurring symptom pair (n=1407). Coping strategies were described in 7104/26,086 (27.2%) posts; medication-based approaches were most prominent, particularly for pain and sleep disturbance, while social and psychological support were more common for anxiety and depression. BERTopic modeling identified 13 latent themes, including cancer treatment trajectories, health care system interaction, existential distress, family involvement, and financial burden, highlighting experiential concerns that may be underrepresented in conventional symptom assessment. The proportion of posts containing at least 1 SPPADE symptom was similar between Xiaohongshu and Zhihu (53.2% vs 54.1%; P=.35), although individual symptom distributions differed across platforms.Conclusions Chinese social media posts contained multidimensional symptom-related narratives alongside psychosocial and experiential concerns that may be less visible in structured symptom assessment. Patient-generated social media data may therefore provide a complementary source for understanding cancer symptom experiences and supportive care concerns. However, differences in platform composition, uneven temporal coverage, and the nonrepresentativeness of social media users limit generalizability, and the clinical utility of integrating such data into routine symptom assessment requires further evaluation.
SN - 1439-4456
SN - 1438-8871
ER -