Patient Voices in Dialysis Care: Sentiment Analysis and Topic Modeling Study of Social Media Discourse
Uloženo v:
| Vydáno v: | Journal of Medical Internet Research vol. 27 (2025), p. e70128 |
|---|---|
| Hlavní autor: | |
| Další autoři: | , |
| Vydáno: |
Gunther Eysenbach MD MPH, Associate Professor
|
| Témata: | |
| On-line přístup: | Citation/Abstract Full Text + Graphics Full Text - PDF |
| Tagy: |
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
MARC
| LEADER | 00000nab a2200000uu 4500 | ||
|---|---|---|---|
| 001 | 3222369421 | ||
| 003 | UK-CbPIL | ||
| 022 | |a 1438-8871 | ||
| 024 | 7 | |a 10.2196/70128 |2 doi | |
| 035 | |a 3222369421 | ||
| 045 | 2 | |b d20250101 |b d20251231 | |
| 100 | 1 | |a Shankar, Ravi | |
| 245 | 1 | |a Patient Voices in Dialysis Care: Sentiment Analysis and Topic Modeling Study of Social Media Discourse | |
| 260 | |b Gunther Eysenbach MD MPH, Associate Professor |c 2025 | ||
| 513 | |a Journal Article | ||
| 520 | 3 | |a Background:Patients with end-stage kidney disease undergoing dialysis face significant physical, psychological, and social challenges that impact their quality of life. Social media platforms such as X (formerly known as Twitter) have become important outlets for these patients to share experiences and exchange information.Objective:This study aimed to uncover key themes, emotions, and challenges expressed by the dialysis community on X from April 2006 to August 2024 by leveraging natural language processing techniques, specifically sentiment analysis and topic modeling.Methods:We collected 12,976 publicly available X posts related to dialysis using the platform’s application programming interface version 2 and Python’s Tweepy library. After rigorous preprocessing, 58.13% (7543/12,976) of the posts were retained for analysis. Sentiment analysis using the Valence Aware Dictionary and Sentiment Reasoner (VADER) model, which is a rule-based sentiment analyzer specifically attuned to social media content, classified the emotional tone of posts. VADER uses a human-curated lexicon that maps lexical features to sentiment scores, considering punctuation, capitalization, and modifiers. For topic modeling, posts with <50 tokens were removed, leaving 53.81% (4059/7543) of the posts, which were analyzed using latent Dirichlet allocation with coherence score optimization to identify the optimal number of topics (k=8). The analysis pipeline was implemented using Python’s Natural Language Toolkit, Gensim, and scikit-learn libraries, with hyperparameter tuning to maximize model performance.Results:Sentiment analysis revealed 49.2% (3711/7543) positive, 26.2% (1976/7543) negative, and 24.7% (1863/7543) neutral sentiment posts. Latent Dirichlet allocation topic modeling identified 8 key thematic clusters: medical procedures and outcomes (722/4059, 17.8% prevalence), daily life impact (666/4059, 16.4%), risks and complications (621/4059, 15.3%), patient education and support (544/4059, 13.4%), health care access and costs (499/4059, 12.3%), symptoms and side effects (442/4059, 10.9%), patient experiences and socioeconomic challenges (406/4059, 10%), and diet and fluid management (162/4059, 4%). Cross-analysis of topics and sentiment revealed that negative sentiment was highest for daily life impact (580/666, 87.1%) and socioeconomic challenges (145/406, 35.8%), whereas the education and support topic exhibited more positive sentiment (250/544, 46%). Topic coherence scores ranged from 0.38 to 0.52, with the medical procedures topic showing the highest semantic coherence. Intertopic distance mapping via multidimensional scaling revealed conceptual relationships between identified themes, with lifestyle impact and socioeconomic challenges clustering closely. Our longitudinal analysis demonstrated evolving discourse patterns, with technology-related discussions increasing by 24% in recent years, whereas financial concerns remained consistently prominent.Conclusions:This study provides a comprehensive, data-driven understanding of the complex lived experiences of patients undergoing dialysis shared on social media. The findings underscore the need for more holistic, patient-centered care models and policies that address the multidimensional challenges illuminated by patients’ voices. | |
| 653 | |a Hemodialysis | ||
| 653 | |a Patient education | ||
| 653 | |a Discourse | ||
| 653 | |a Analysis | ||
| 653 | |a Patient-centered care | ||
| 653 | |a Datasets | ||
| 653 | |a Patients | ||
| 653 | |a Health care expenditures | ||
| 653 | |a Data mining | ||
| 653 | |a Kidney diseases | ||
| 653 | |a Medical personnel | ||
| 653 | |a Social networks | ||
| 653 | |a Chronic illnesses | ||
| 653 | |a User generated content | ||
| 653 | |a Health care access | ||
| 653 | |a Side effects | ||
| 653 | |a Emotions | ||
| 653 | |a Virtual communities | ||
| 653 | |a Mapping | ||
| 653 | |a Automation | ||
| 653 | |a Global health | ||
| 653 | |a Coherence | ||
| 653 | |a COVID-19 | ||
| 653 | |a Dialysis | ||
| 653 | |a Quality of life | ||
| 653 | |a Clinical outcomes | ||
| 653 | |a Challenges | ||
| 653 | |a Artificial intelligence | ||
| 653 | |a Sentiment analysis | ||
| 653 | |a Diet | ||
| 653 | |a Social media | ||
| 653 | |a Pandemics | ||
| 653 | |a Activities of daily living | ||
| 653 | |a Health education | ||
| 653 | |a Optimization | ||
| 653 | |a Natural language processing | ||
| 653 | |a Topics | ||
| 653 | |a Clustering | ||
| 653 | |a Caregivers | ||
| 653 | |a Libraries | ||
| 653 | |a Peritoneal dialysis | ||
| 653 | |a Semantics | ||
| 653 | |a Multidimensional scaling | ||
| 653 | |a Measures | ||
| 653 | |a Punctuation | ||
| 653 | |a Mass media effects | ||
| 653 | |a Everyday life | ||
| 653 | |a Models | ||
| 653 | |a Discourse analysis | ||
| 653 | |a Programming languages | ||
| 653 | |a Valence | ||
| 653 | |a Mass media images | ||
| 653 | |a Health services | ||
| 653 | |a Mental health services | ||
| 653 | |a Rules | ||
| 653 | |a Education | ||
| 653 | |a Allocation | ||
| 653 | |a Sociolinguistics | ||
| 653 | |a Drug effects | ||
| 700 | 1 | |a Xu, Qian | |
| 700 | 1 | |a Bundele, Anjali | |
| 773 | 0 | |t Journal of Medical Internet Research |g vol. 27 (2025), p. e70128 | |
| 786 | 0 | |d ProQuest |t Library Science Database | |
| 856 | 4 | 1 | |3 Citation/Abstract |u https://www.proquest.com/docview/3222369421/abstract/embedded/6A8EOT78XXH2IG52?source=fedsrch |
| 856 | 4 | 0 | |3 Full Text + Graphics |u https://www.proquest.com/docview/3222369421/fulltextwithgraphics/embedded/6A8EOT78XXH2IG52?source=fedsrch |
| 856 | 4 | 0 | |3 Full Text - PDF |u https://www.proquest.com/docview/3222369421/fulltextPDF/embedded/6A8EOT78XXH2IG52?source=fedsrch |