Perception of Online Communities towards the Use of Cancer Immunotherapy: A Data Mining Study of 3.6 Million Web-based Posts from Social Media Platforms
Refereed conference paper presented and published in conference proceedings
全文
沒有全文檔案提供 |
其它資訊
摘要Purpose of the study: Immunotherapy has become a game changer in cancer treatment, yet, patients may have concerns and deficit knowledge regarding its use. This study analyzed the perspectives of online communities towards the use of immunotherapy.
Methods: A total of 4.9 million posts were extracted from Facebook, Twitter, Reddit, and 16 health forums from November to December 2022. The posts were first pre-processed using Spark NLP, an Apache Spark Natural Language Processing library in Python. Subsequently, topic modeling using BERTopic was performed. Topic diversity and coherence metrics were computed to determine the optimal number of topics. After identifying the 30 most important keywords for each topic, labels were assigned to the topics based on domain knowledge and clinical consensus; related topics were further classified into themes.
Results and discussions: 3.6 million posts were available for model building and evaluation after data cleaning. BERTopic performed best with 14 topics (topic diversity: 87.76%; topic coherence: 80.21%). They were further classified into four positive themes (8 topics), one neutral theme (2 topics), and three negative themes (4 topics). The positive themes included feelings of hope and positivity towards immunotherapy, the perceived superior effectiveness of immunotherapy over conventional chemotherapy, and its synergistic effects with other treatments. The negative themes were related to the extortionate cost and unpredictable side effects of immunotherapy, and how it caused lifestyle disruptions. Another emerging theme is the use of complementary therapies (e.g., massage, meditation, and herbal medicine) and self-treatments to relieve symptoms and side effects of immunotherapy.
Conclusions: The social media is a valuable source of real-world data reflecting patients’ perceptions. We found that the online communities tended to have positive expectations towards immunotherapy, especially combination therapies. Our findings reinforce the need for quality supportive care for patients to alleviate their fear of side effects and lifestyle disruption.
Methods: A total of 4.9 million posts were extracted from Facebook, Twitter, Reddit, and 16 health forums from November to December 2022. The posts were first pre-processed using Spark NLP, an Apache Spark Natural Language Processing library in Python. Subsequently, topic modeling using BERTopic was performed. Topic diversity and coherence metrics were computed to determine the optimal number of topics. After identifying the 30 most important keywords for each topic, labels were assigned to the topics based on domain knowledge and clinical consensus; related topics were further classified into themes.
Results and discussions: 3.6 million posts were available for model building and evaluation after data cleaning. BERTopic performed best with 14 topics (topic diversity: 87.76%; topic coherence: 80.21%). They were further classified into four positive themes (8 topics), one neutral theme (2 topics), and three negative themes (4 topics). The positive themes included feelings of hope and positivity towards immunotherapy, the perceived superior effectiveness of immunotherapy over conventional chemotherapy, and its synergistic effects with other treatments. The negative themes were related to the extortionate cost and unpredictable side effects of immunotherapy, and how it caused lifestyle disruptions. Another emerging theme is the use of complementary therapies (e.g., massage, meditation, and herbal medicine) and self-treatments to relieve symptoms and side effects of immunotherapy.
Conclusions: The social media is a valuable source of real-world data reflecting patients’ perceptions. We found that the online communities tended to have positive expectations towards immunotherapy, especially combination therapies. Our findings reinforce the need for quality supportive care for patients to alleviate their fear of side effects and lifestyle disruption.
著者Xingyue WU, Chun Sing LAM, Herbert Ho-fung LOONG, Keary Rui ZHOU, Chun-Kit NGAN, Yin Ting CHEUNG
會議名稱9th Nursing Symposium on Cancer Care cum 1st Genomic and Genetic Nursing Forum
會議開始日22.05.2024
會議完結日24.05.2024
會議地點The Nethersole School of Nursing
會議國家/地區香港
出版年份2024
語言美式英語
關鍵詞social media, cancer, immunotherapy, perception, data mining