Social media—a bridge connecting cancer survivors' problem identification and evidence application

Authors: Prof Yanni Wu¹, Chuhan Zhong¹

1. 1. JBI Nanfang Nursing Centre for Evidence-Based Practice, Nanfang Hospital, Southern Medical University

Introduction

Cancer survivors contend with numerous unexpected challenges on the road to recovery. After active treatment, they often continue to face a range of unmet needs—including persistent physical symptoms, psychological adaptations and a lack of accessible, easy-to-understand information.^[1-3] While evidence-based guidelines inform clinical practice, many day-to-day issues survivors encounter remain difficult to detect early until they escalate.^[4] How can we identify these real needs sooner and provide timely, evidence-based support? The answer may lie in a space where cancer survivors increasingly voice their experiences: social media. By using artificial intelligence (AI)-assisted systems wisely, we can transform these public digital platforms into a vital bridge linking real-time problem identification with evidence application—connecting statements like ‘I’m struggling right now’ with actionable insights on ‘what works’.^[5]

Background

The global population of cancer survivors continues to increase. According to the World Health Organization (WHO),  the annual number of new cases of cancer will reach 35 million by 2050.^[1] Many face long-term physical and psychological health issues such as fatigue, anxiety and sleep disorders after treatment ends.^[6-8]  Traditional monitoring methods—such as infrequent follow-up appointments and low response rate questionnaires—often fail to provide timely support.^[9] Subtle or early-stage symptoms (e.g. ‘occasionally feeling low energy’) are easily overlooked, only gaining attention after developing into more serious problems. This gap highlights the inadequacy of current systems in capturing survivors’ real-life experiences between clinical visits.^[10] There is a clear need for low-cost, continuous and patient-centred channels to capture lived experience. Social media, used by over 4 billion people globally, has shown great potential in addressing this gap. Active cancer communities on platforms like Weibo or Facebook have become modern forums where patients freely express joys, fears and frustrations.^[11] These spontaneous, unedited expressions represent a valuable resource for early problem detection. Furthermore, advances in Natural Language Processing (NLP) and Large Language Models (LLMs) now enable us to analyse this data meaningfully and ethically.^[12-14]

Core problems: Two major barriers for cancer survivors

Inefficient early symptom recognition
Many survivors experience subtle emotional shifts or intermittent fatigue during recovery,^[15,16] often dismissed as ‘normal’ post-treatment reactions. This perception delays intervention, creating a gap between the health issues survivors experience and the support they receive. The optimal window for early action is frequently missed.

Difficulty accessing tailored, evidence-based information
Due to geographical limits, uneven resource distribution and a lack of personalised communication channels, many survivors turn to social media for support.^[17] However, they often encounter an overwhelming mix of misinformation and evidence. Without professional training, distinguishing reliable information becomes highly challenging. This not only highlights unmet needs but also presents a responsibility: how can we ethically leverage social media to transform patient voices into precise, evidence-based support?

Solution: building an evidence-based ‘Identify-Match-Deliver’ loop

To address these barriers, our team has designed an innovative model that combines social media and AI.

Building an Emotional Lexicon (Identification)
We collected and analysed text from cancer-specific social media communities to create a lexicon capturing cancer survivors' authentic descriptions of physical and emotional states. In short, our lexicon contains words and phrases taken directly from patient posts to vividly capture their real-world feelings and experiences. Examples include: ‘helplessness’ and ‘bewilderment’ (conveying fear and uncertainty); ‘bone pain’ and ‘numbness’ (reflecting physical symptoms); and medical terms like ‘chemotherapy’ and ‘targeted therapy,’ all of which can carry profound emotional weight within the context of patients' personal narratives. This lexicon has been developed and validated in our recent study.^[18] It addresses the first barrier by enabling the early, passive detection of unmet needs through the conversion of subtle personal expressions into structured data.

Co-creating patient-friendly guidelines (Evidence-based)
We translated complex evidence into accessible, patient-tested guides. This step tackles the second barrier by curating and simplifying evidence, ensuring information is both accurate and practical.

AI-driven matching and delivery (Match & Push)
Using our proprietary AI match and push system, we automatically capture patient social media posts. Upon identifying potential needs, we match them to problem tags in the Emotional Lexicon and deliver corresponding evidence-based strategies from the guide, which enables timely, precise and personalised support.

Preliminary results: Making evidence-based care faster and more accessible

Although still in development, preliminary data indicate that our model shows high potential. Early user engagement has been positive. As one breast cancer survivor in our pilot test shared, ‘I posted late at night about feeling alone and scared of the pain coming back. To my surprise, the system picked up on my fear and sent me a small guide on managing fear of recurrence. It wasn't just a link—it felt like someone was actually listening.’ Feedback like this suggests that AI-driven, social media-based support is both feasible and acceptable. This approach facilitates a shift from passive, hospital-centric care to proactive, continuous and patient-centered support.

Challenges

Implementing such an innovative system is certainly no easy feat. We encountered challenges, including data privacy, algorithmic bias and ensuring cross-cultural applicability. Engaging patients and clinicians in the design process is crucial for building trust and relevance. We also recognise that improving the accuracy of early detection is another critical issue.

Next steps

Future work will focus on refining the Emotional Lexicon, integrating multimodal data (e.g. video, audio) and developing adaptive feedback systems to improve recommendation accuracy. We plan to scale the model and rigorously evaluate its long-term impact on patient outcomes and healthcare sustainability.

References

Institute of Medicine and National Research Council of the National Academies. From cancer patient to cancer survivor: Lost in transition. Washington, DC: National Academies Press; 2006: 534. https://www.nap.edu/catalog/11468/from-cancer-patient-to-cancer-survivor-lost-in-transition. Accessed Sep 24, 2025.

[2] Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, Jemal A. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024 May-Jun;74(3):229-263. doi: 10.3322/caac.21834. Epub 2024 Apr 4. PMID: 38572751.

[3] Spies, E., Flynn, J. A., Oliveira, N. G., Karmalkar, P., & Gurulingappa, H. (2024). Artificial intelligence-enabled social media listening to inform early patient-focused drug development: perspectives on approaches and strategies. Frontiers in digital health, 6, 1459201. https://doi.org/10.3389/fdgth.2024.1459201

[4] Marwah, R., Mishra, S., Gross, B., Couturiaux, S., Calara, R., Sabate Estrella, E. J., & Hogea, C. (2025). Social Media Insights Into Disease Burden in Patients and Caregivers of Myelodysplastic Syndrome: Subcohort Analysis of High-Risk Patients. Journal of medical Internet research, 27, e65460. https://doi.org/10.2196/65460

[5] Samira Daswani et al. Clinical utility and patient engagement with a novel digital platform for cancer care navigation.. J Clin Oncol 43, e13713-e13713(2025).DOI:10.1200/JCO.2025.43.16_suppl.e13713

[6] Schmidt, M. E., Goldschmidt, S., Hermann, S., & Steindorf, K. (2022). Late effects, long-term problems and unmet needs of cancer survivors. International journal of cancer, 151(8), 1280–1290. https://doi.org/10.1002/ijc.34152

[7] Wu, Z., Luo, F., Wang, S., Hu, X., & Chen, M. (2025). Digital Interventions and Mental Health Outcomes in Patients With Cancer: Systematic Review and Meta-Analysis. JMIR cancer, 11, e64754. https://doi.org/10.2196/64754

[8] Kline, R. M., Arora, N. K., Bradley, C. J., Brauer, E. R., Graves, D. L., Lunsford, N. B., McCabe, M. S., Nasso, S. F., Nekhlyudov, L., Rowland, J. H., Schear, R. M., & Ganz, P. A. (2018). Long-Term Survivorship Care After Cancer Treatment - Summary of a 2017 National Cancer Policy Forum Workshop. Journal of the National Cancer Institute, 110(12), 1300–1310. https://doi.org/10.1093/jnci/djy176

[9] Anhang Price, R., Quigley, D. D., Hargraves, J. L., Sorra, J., Becerra-Ornelas, A. U., Hays, R. D., Cleary, P. D., Brown, J., & Elliott, M. N. (2022). A Systematic Review of Strategies to Enhance Response Rates and Representativeness of Patient Experience Surveys. Medical care, 60(12), 910–918. https://doi.org/10.1097/MLR.0000000000001784

[10] Schiavi, M., Costi, S., Barbieri, I., Ghirotto, L., Fugazzaro, S., Bressi, B., Paltrinieri, S., Luminari, S., & Contri, A. (2024). Identifying unmet needs in cancer survivorship by linking patient-reported outcome measures to the International Classification of Functioning, Disability and Health. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer, 32(12), 835. https://doi.org/10.1007/s00520-024-09019-8

[11] Watson J. Social Media Use in Cancer Care. Semin Oncol Nurs. 2018 May;34(2):126-131. doi: 10.1016/j.soncn.2018.03.003. Epub 2018 Apr 2. PMID: 29622519.

[12] Chen D, Parsa R, Swanson K, Nunez J, Critch A, Bitterman DS, et al. Large language models in oncology: a review. BMJ Oncology. 2025;4:e000759. https://doi.org/10.1136/bmjonc-2025-000759

[13] Caglayan, A., Slusarczyk, W., Rabbani, R. D., Ghose, A., Papadopoulos, V., & Boussios, S. (2024). Large Language Models in Oncology: Revolution or Cause for Concern?. Current oncology (Toronto, Ont.), 31(4), 1817–1830. https://doi.org/10.3390/curroncol31040137

[14] Chen, D., Alnassar, S. A., Avison, K. E., Huang, R. S., & Raman, S. (2025). Large Language Model Applications for Health Information Extraction in Oncology: Scoping Review. JMIR cancer, 11, e65984. https://doi.org/10.2196/65984

[15] Bower, J. Cancer-related fatigue—mechanisms, risk factors and treatments. Nat Rev Clin Oncol, 11, 597–609 (2014). https://doi.org/10.1038/nrclinonc.2014.127

[16] Krebber, A. M., Buffart, L. M., Kleijn, G., Riepma, I. C., de Bree, R., Leemans, C. R., Becker, A., Brug, J., van Straten, A., Cuijpers, P., & Verdonck-de Leeuw, I. M. (2014). Prevalence of depression in cancer patients: a meta-analysis of diagnostic interviews and self-report instruments. Psycho-oncology, 23(2), 121–130. https://doi.org/10.1002/pon.3409

[17] Dee EC, Muralidhar V, Butler SS, Yu Z, Sha ST, Mahal BA, Nguyen PL, Sanford NN. General and Health-Related Internet Use Among Cancer Survivors in the United States: A 2013-2018 Cross-Sectional Analysis. J Natl Compr Canc Netw. 2020 Nov 2;18(11):1468-1475. doi: 10.6004/jnccn.2020.7591. PMID: 33152707.

[18]Li C, Fu J, Lai J, Sun L, Zhou C, Li W, Jian B, Deng S, Zhang Y, Guo Z, Liu Y, Zhou Y, Xie S, Hou M, Wang R, Chen Q, Wu Y. Construction of an Emotional Lexicon of Patients With Breast Cancer: Development and Sentiment Analysis. J Med Internet Res. 2023 Sep 12;25:e44897. doi: 10.2196/44897. PMID: 37698914; PMCID: PMC10523220.

To link to this article - DOI: https://doi.org/10.70253/KQHH9307

Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.