Abstract
BACKGROUND: To detect the differences in physical symptoms between depressed and undepressed patients with breast cancer (BC), including common symptoms, co-occurring symptoms, and symptom clusters based on texts derived from social media and expressive writing. METHODS: A total of 1830 texts from social media and expressive writing were collected. The Chi-square test was used to compare the frequency of physical symptoms between depressed and undepressed patients with BC. Symptom lexicon of BC and K-medoids Clustering were used for mining physical symptoms and cluster analysis. RESULTS: The common physical symptoms reported by texts included general pains (59.38%), fatigue (26.60%), vomiting (24.82%), swelling of limbs (21.69%), difficulty sleeping (21.56%), nausea (16.78%), alopecia (15.14%), loss of appetite (13.78%), dizziness (11.60%), and concentration problems (11.19%). The frequency of difficulty sleeping (depressed 28.40%; undepressed 18.16%; P = 0.002) in depressed patients was higher than undepressed patients with BC. High co-occurrence was observed in both commonly mentioned symptoms and those less commonly mentioned but frequently co-occurring with them. There were 5 symptom clusters identified in depressed patients and 6 symptom clusters in undepressed patients. Pain-related symptom cluster and gastrointestinal symptom cluster were both identified in the depressed and undepressed patients. The novel immune system impairment symptom cluster consisting of bleeding and fever was found in the undepressed patients. CONCLUSIONS: This study found that difficulty sleeping was reported more frequently, and identified difficulty sleeping-pain symptom cluster in depressed patients. The novel immune system impairment symptom cluster in undepressed patients was detected. Healthcare providers can provide targeted care to depressed and undepressed patients based on these differences. These findings demonstrate that social media can provide new perspectives on symptom experiences. The combination of digital tools and traditional clinical tools for symptom management in follow-up has great potential in the future. CLINICAL TRIAL NUMBER: Not applicable.