JMIR Publications

ABSTRACT

Background:

Patients with cancer frequently experience complex and multidimensional symptom burdens. Standardised patient-reported outcome measures are constrained by predefined item sets and may not fully capture the broader range of symptom-related and psychosocial experiences expressed by patients. Patient-generated social media data offer a complementary viewpoint to examine how cancer symptoms are experienced and communicated in naturalistic settings.

Objective:

To characterise symptom narratives expressed by Chinese cancer patients on social media using a conventional assessment framework, and to identify additional experiential themes through neural topic modelling.

Methods:

A computational text analysis integrating large language model annotation and neural topic modelling. A corpus of 26,086 first-person patient narratives was collected from two major Chinese social media platforms, Xiaohongshu and Zhihu, between January 2013 and January 2026, using cancer-related search terms. Cancer symptoms were identified using GPT-4o-mini and validated against independent human coding. Symptom prevalence, co-occurrence patterns, and multiplicity were analysed across cancer types and disease stages. BERTopic modelling was applied to identify latent themes within patient narratives.

Results:

Overall, 53.3% of posts contained at least one identifiable symptom. Pain was the most prevalent symptom (22.2%), followed by physical function impairment (15.9%) and depression (15.5%). Pain and physical function impairment formed the most frequent co-occurring symptom pair (n = 1,407). Advanced-stage narratives showed greater symptom multiplicity than early-stage narratives, with 11.8% versus 7.2% containing three or more co-occurring symptoms. Medication-based coping predominated across narratives, whereas exercise-related coping appeared in only 0.3% of posts. Topic modelling identified 13 latent themes, including treatment navigation, healthcare system interaction, existential distress, family caregiver burden, and financial toxicity, concerns that are not routinely represented within conventional symptom assessment frameworks.

Conclusions:

Chinese patient-generated social media narratives revealed multidimensional cancer symptom experiences together with psychosocial concerns that may be underrepresented in routine symptom assessment. Integrating patient-generated data with structured symptom assessment may support more comprehensive and patient-centred oncology nursing care, particularly in identifying supportive care concerns that remain insufficiently addressed in clinical practice. Clinical Trial: Not registered.