Results and Methodological Implications of the Digital Epidemiology of Prescription Drug References Among Twitter Users: Latent Dirichlet Allocation (LDA) Analyses

Journal of Medical Internet Research, Год журнала: 2023, Номер 25, С. e48405 - e48405

Опубликована: Июль 28, 2023

Social media is an important information source for a growing subset of the population and can likely be leveraged to provide insight into evolving drug overdose epidemic. Twitter valuable trends, colloquial available potential users, how networks interactivity might influence what people are exposed they engage in communication around use.This exploratory study was designed investigate ways which unsupervised machine learning analyses using natural language processing could identify coherent themes tweets containing substance names.This involved harnessing data from Twitter, including large-scale collection brand name (N=262,607) street (N=204,068) prescription drug-related use (ie, processing) collected with visualization pertinent tweet themes. Latent Dirichlet allocation (LDA) coherence score calculations performed compare (eg, OxyContin) oxys) tweets.We found discussed differently depending on whether or used. Brand categories often contained political talking points border, crime, handling ongoing mitigation strategies). In contrast, names occasionally referenced misuse, though multiple social uses term Sonata) muddled topic clarity.Content corpus reflected discussion about itself less personal use. However, content notably more diverse resisted simple LDA categorization. We speculate this may reflect effective slang terminology clandestinely discuss activity. If so, straightforward digital difficult than previously assumed. This work has used surveillance detection harmful information. It also appropriate education dissemination persons engaged Twitter.

Язык: Английский

---

Analyzing Themes, Sentiments, and Coping Strategies Regarding Online News Coverage of Depression in Hong Kong: Mixed Methods Study

Journal of Medical Internet Research, Год журнала: 2025, Номер 27, С. e66696 - e66696

Опубликована: Фев. 13, 2025

Depression, a highly prevalent global mental disorder, has prompted significant research concerning its association with social media use and impact during Hong Kong's unrest COVID-19 pandemic. However, other mainstream media, specifically online news, been largely overlooked. Despite extensive conducted in countries, such as the United States, Australia, Canada, to investigate latent subthemes, sentiments, coping strategies portrayed depression-related landscape Kong remains unexplored. This study aims uncover subthemes presented news coverage of depression Kong, examine sentiment conveyed assess whether have provided for individuals experiencing depression. used natural language processing (NLP) techniques, namely Dirichlet allocation topic modeling Valence Aware Dictionary Sentiment Reasoner (VADER) analysis, fulfill first second objectives. Coping were rigorously assessed manually labeled designated categories by content analysis. The was collected from February 2019 May 2024 websites latest portrayal depression, particularly after In total, 2435 articles retained data analysis screening process. A total 7 identified based on results. Societal system, law enforcement, recession, lifestyle, leisure, health issues, US politics subthemes. Moreover, overall exhibited slightly positive sentiment. correlations between scores indicated that societal revealed negative tendencies, while remainder leaned toward substantially lacking; however, emphasizing information skills resources individual adjustment cope emerged priority focus. pioneering mixed methods approach where NLP underlying news. Content also performed available strategies. findings this enhance our understanding how is through preferable being mitigate potential readers discussed. Future encouraged address mentioned implications limitations, recommendations apply advanced techniques new issue case or language.

Язык: Английский

---

Geospatial vaccine misinformation risk on social media: Online insights from an English/Spanish natural language processing (NLP) analysis of vaccine-related tweets

Social Science & Medicine, Год журнала: 2023, Номер 339, С. 116365 - 116365

Опубликована: Ноя. 10, 2023

Misinformation is known to affect norms, attitudes, and intentions engage with healthy behaviors. Evidence strongly supports that Spanish speakers may be particularly affected by misinformation its outcomes, yet current insights into the scope scale of primarily ethnocentric, greater emphasis on English-language design.

Язык: Английский

---

Public attention and psychological trends towards waste reduction: A large-scale data analysis based on social media

Journal of Cleaner Production, Год журнала: 2024, Номер 467, С. 142873 - 142873

Опубликована: Июнь 11, 2024

Язык: Английский

---

Understanding public perceptions and discussions on diseases involving chronic pain through social media: cross-sectional infodemiology study

BMC Musculoskeletal Disorders, Год журнала: 2024, Номер 25(1)

Опубликована: Июль 22, 2024

Abstract Background Chronic pain is a highly prevalent medical condition that negatively impacts quality of life and associated with considerable functional disability. Certain diseases, such as fibromyalgia, headache, paraplegia, neuropathy, multiple sclerosis, manifest chronic pain. Objective The aim this study to examine the number type tweets (original or retweet) related pain, well analyze emotions compare societal impact diseases under study. Methods We investigated posted between January 1, 2018, December 31, 2022, by Twitter users in English Spanish, generated retweets. Additionally, were extracted from these their diffusion was analyzed. Furthermore, topics most frequently discussed collected. Results A total 72,874 analyzed, including 44,467 28,407 Spanish. Paraplegia represented 23.3% 16,461 classified tweets, followed headache fibromyalgia 15,337 (21.7%) 15,179 (21.5%) respectively. Multiple sclerosis 14,781 (21%), fewest neuropathy 8,830 (12.5%). results showed primary "fear" "sadness." reach through retweets, those headaches showing highest interest interaction among users. Conclusion Our underscore potential leveraging social media for better understanding patients suffering its on society. Among encountered are treatment, symptoms, causes disease. Therefore, it relevant inform patient prevent misconceptions regarding illness.

Язык: Английский

---

Navigating the Digital Neurolandscape: Analyzing the Social Perception of and Sentiments Regarding Neurological Disorders through Topic Modeling and Unsupervised Research Using Twitter

Information, Год журнала: 2024, Номер 15(3), С. 152 - 152

Опубликована: Март 8, 2024

Neurological disorders represent the primary cause of disability and secondary mortality globally. The incidence prevalence most notable neurological are growing rapidly. Considering their social public perception by using different platforms like Twitter can have a huge impact on patients, relatives, caregivers professionals involved in multidisciplinary management disorders. In this study, we collected analyzed all tweets posted English or Spanish, between 2007 2023, referring to headache disorders, dementia, epilepsy, multiple sclerosis, spinal cord injury Parkinson’s disease search engine that has access 100% publicly available tweets. aim our work was deepen understanding addressing three major objectives: (1) analyzing number temporal evolution both Spanish discussing (dementias, disease, injury, epilepsy disorders); (2) determining main thematic content posts interest they generated temporally topic modeling; (3) sentiments associated with topics were previously collected. Our results show dementias were, far, common whose treatment discussed Twitter, included diseases patients claims increase awareness, support research, activities ameliorate development existent/potential treatments approaches significant showing negative emotions fear, anger sadness, some also demonstrating positive joy. Thus, study shows not only is an important active platform implicated dissemination normalization but these entities quite inequitable, greater intervention more accurate information figures media could help convey better current state, project future for general public.

Язык: Английский

---

Information Disorders in the Chilean and Spanish Press: A Comparison Using Thematic Modelling

Journalism and Media, Год журнала: 2024, Номер 5(1), С. 148 - 162

Опубликована: Янв. 27, 2024

This article focuses on the role of information disorders in media coverage cancer as a growing public health problem both sides Atlantic. Taking examples Chile and Spain, we analysed news (n = 5522) published by major digital newspaper outlets countries between 2020 2022 to explore elements contextual disorders, over- and/or under-representation mentions sources actors, latent topics journalistic systems. To achieve these objectives, employed topic modelling coherence techniques. The results revealed high number references institutional, administrative, political followed issuers strategic communication and, less frequently, patients’ associations. discourses differed their underlying topics, with risk factors psycho-social being most frequently addressed Spanish press geo-political institutional contexts mentioned Chilean press. advances research, however, was common closes identifying future challenges communication.

Язык: Английский

---

Understanding Public Perceptions and Discussions on Fibromyalgia through Social Media: Cross-Sectional Infodemiology Study

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Март 7, 2024

Background: Fibromyalgia is a prevalent condition of unknown etiology, characterized by generalized chronic pain that leads to disability, and has significant direct indirect costs. The objective this study was examine the content key aspects tweets pertaining diseases associated with pain. Methods: We investigated published between January 1, 2018, December 31, 2022, English Spanish-speaking Twitter users, as well generated retweets. Additionally, emotions were extracted from these their dissemination analysed. Similarly, topics users most frequently address compiled. Results: In total, 72,874 analysed in both (44,467) Spanish (28,407). Paraplegia represented 23.3%, 16,461 classifiable tweets, followed headache fibromyalgia, 15,337 (21.7%) 15,179 (21.5%) respectively. Multiple sclerosis 14,781 (21%), while lowest number neuropathy, totaling 8,830 (12.5%). findings revealed primary "fear" "sadness". Furthermore, scope impact through retweets, those related headaches being highest interest having greatest interaction among users. Conclusions: Our contribute understanding role social media plays fostering public awareness improving patients' comprehension treatment. Moreover, results are likely be applicable other countries whose languages not covered our study.

Язык: Английский

---

Digital Epidemiology of Prescription Drug References on X (Formerly Twitter): Neural Network Topic Modeling and Sentiment Analysis (Preprint)

Опубликована: Фев. 28, 2024

Data from the social media platform X (formerly Twitter) can provide insights into types of language that are used when discussing drug use. In past research using latent Dirichlet allocation (LDA), we found tweets containing “street names” prescription drugs were difficult to classify due similarity other colloquialisms and lack clarity over how terms used. Conversely, “brand name” references more amenable machine-driven categorization. This study sought use next-generation techniques (beyond LDA) natural processing reprocess data automatically cluster groups topics differentiate between street- brand-name sets. We also aimed analyze differences in emotional valence 2 sets relationship engagement on sentiment. Twitter application programming interface collect contained street brand name a within tweet. Using BERTopic combination with Uniform Manifold Approximation Projection k-means, generated for street-name corpus (n=170,618) (n=245,145). Valence Aware Dictionary Sentiment Reasoner (VADER) scores whether had positive, negative, or neutral sentiments. Two different logistic regression classifiers predict sentiment label each corpus. The first model tweet’s metrics topic ID label, while second those features addition top 5000 largest term-frequency–inverse document frequency score. BERTopic, identified 40 set 5 set, which generalized 8 discussion, respectively. Four general themes discussion referenced use, From VADER scores, both corpora inclined toward positive Adding vectorized tweet text increased accuracy our models by around 40% compared did not incorporate corpora. was able well. As LDA, names similar than names. could only be logically applied because high prevalence non–drug-related data. Brand-name either discussed positively negatively, few posts having emotionality. machine learning models, alone enough label; added context needed understand emotionality

Язык: Английский

---

Digital Epidemiology of Prescription Drug References on X (Formerly Twitter): Neural Network Topic Modeling and Sentiment Analysis

Journal of Medical Internet Research, Год журнала: 2024, Номер 26, С. e57885 - e57885

Опубликована: Авг. 23, 2024

Background Data from the social media platform X (formerly Twitter) can provide insights into types of language that are used when discussing drug use. In past research using latent Dirichlet allocation (LDA), we found tweets containing “street names” prescription drugs were difficult to classify due similarity other colloquialisms and lack clarity over how terms used. Conversely, “brand name” references more amenable machine-driven categorization. Objective This study sought use next-generation techniques (beyond LDA) natural processing reprocess data automatically cluster groups topics differentiate between street- brand-name sets. We also aimed analyze differences in emotional valence 2 sets relationship engagement on sentiment. Methods Twitter application programming interface collect contained street brand name a within tweet. Using BERTopic combination with Uniform Manifold Approximation Projection k-means, generated for street-name corpus (n=170,618) (n=245,145). Valence Aware Dictionary Sentiment Reasoner (VADER) scores whether had positive, negative, or neutral sentiments. Two different logistic regression classifiers predict sentiment label each corpus. The first model tweet’s metrics topic ID label, while second those features addition top 5000 largest term-frequency–inverse document frequency score. Results BERTopic, identified 40 set 5 set, which generalized 8 discussion, respectively. Four general themes discussion referenced use, From VADER scores, both corpora inclined toward positive Adding vectorized tweet text increased accuracy our models by around 40% compared did not incorporate corpora. Conclusions was able well. As LDA, names similar than names. could only be logically applied because high prevalence non–drug-related data. Brand-name either discussed positively negatively, few posts having emotionality. machine learning models, alone enough label; added context needed understand emotionality

Язык: Английский

---