Health Misinformation Detection in the Social Web: An Overview and a Data Science Approach

Racial and Demographic Disparities in Susceptibility to Health Misinformation on Social Media: National Survey-Based Analysis

BACKGROUND

Social media platforms have transformed the dissemination of health information, allowing for rapid and widespread sharing of content. However, alongside valuable medical knowledge, these platforms have also become channels for the spread of health misinformation, including false claims and misleading advice, which can lead to significant public health risks. Susceptibility to health misinformation varies and is influenced by individuals' cultural, social, and personal backgrounds, further complicating efforts to combat its spread.

OBJECTIVE

This study aimed to examine the extent to which individuals report encountering health-related misinformation on social media and to assess how racial, ethnic, and sociodemographic factors influence susceptibility to such misinformation.

METHODS

Data from the Health Information National Trends Survey (HINTS; Cycle 6), conducted by the National Cancer Institute with 5041 US adults between March and November 2022, was used to explore associations between racial and sociodemographic factors (age, gender, race/ethnicity, annual household income, marital status, and location) and susceptibility variables, including encounters with misleading health information on social media, difficulty in assessing information truthfulness, discussions with health providers, and making health decisions based on such information.

RESULTS

Over 35.61% (1740/4959) of respondents reported encountering "a lot" of misleading health information on social media, with an additional 45% (2256/4959) reporting seeing "some" amount of health misinformation. Racial disparities were evident in comparison with Whites, with non-Hispanic Black (odds ratio [OR] 0.45, 95% CI 0.33-0.6, P<.01) and Hispanic (OR 0.54, 95% CI 0.41-0.71, P<.01) individuals reporting lower odds of finding deceptive information, while Hispanic (OR 1.68, 95% CI 1.48-1.98, P<.05) and non-Hispanic Asian (OR 1.96, 95% CI 1.21-3.18, P<.01) individuals exhibited higher odds in having difficulties in assessing the veracity of health information found on social media. Hispanic and Asian individuals were more likely to discuss with providers and make health decisions based on social media information. Older adults aged ≥75 years exhibited challenges in assessing health information on social media (OR 0.63, 95% CI 0.43-0.93, P<.01), while younger adults (18-34) showed increased vulnerability to health misinformation. In addition, income levels were linked to higher exposure to health misinformation on social media: individuals with annual household incomes between US $50,000 and US $75,000 (OR 1.74, 95% CI 1.14-2.68, P<.01), and greater than US $75,000 (OR 1.78, 95% CI 1.20-2.66, P<.01) exhibited greater odds, revealing complexities in decision-making and information access.

CONCLUSIONS

This study highlights the pervasive presence of health misinformation on social media, revealing vulnerabilities across racial, age, and income groups, underscoring the need for tailored interventions.

Are there 34,000 human emotions? Deconstructing patterns of scientific misinformation

BACKGROUND

Scientific misinformation is a much-discussed topic, and the COVID-19 crisis has highlighted the importance of reliability in science and research. However, limiting misinformation is complicated because of the growing number of communication channels, in which scientific and nonscientific content are often mixed.

METHODS

This case study combines the examination of references, online observation, and a content and frequency analysis to investigate the dissemination of scientific misinformation in the interplay of different genres and media.

RESULTS

Using the example of the claimed existence of 34,000 human emotions, this study demonstrates how questionable statements are spread in science, popular science, and pseudoscience, making it particularly challenging to track and correct them.

CONCLUSIONS

The findings highlight epistemic authority, trust, and injustice within and between scientific and nonscientific communities. The author argues that, in the digital age, researchers should defend and monitor scientific principles beyond academia.

Online Health Search Via Multidimensional Information Quality Assessment Based on Deep Language Models: Algorithm Development and Validation

BACKGROUND

Widespread misinformation in web resources can lead to serious implications for individuals seeking health advice. Despite that, information retrieval models are often focused only on the query-document relevance dimension to rank results.

OBJECTIVE

We investigate a multidimensional information quality retrieval model based on deep learning to enhance the effectiveness of online health care information search results.

METHODS

In this study, we simulated online health information search scenarios with a topic set of 32 different health-related inquiries and a corpus containing 1 billion web documents from the April 2019 snapshot of Common Crawl. Using state-of-the-art pretrained language models, we assessed the quality of the retrieved documents according to their usefulness, supportiveness, and credibility dimensions for a given search query on 6030 human-annotated, query-document pairs. We evaluated this approach using transfer learning and more specific domain adaptation techniques.

RESULTS

In the transfer learning setting, the usefulness model provided the largest distinction between help- and harm-compatible documents, with a difference of +5.6%, leading to a majority of helpful documents in the top 10 retrieved. The supportiveness model achieved the best harm compatibility (+2.4%), while the combination of usefulness, supportiveness, and credibility models achieved the largest distinction between help- and harm-compatibility on helpful topics (+16.9%). In the domain adaptation setting, the linear combination of different models showed robust performance, with help-harm compatibility above +4.4% for all dimensions and going as high as +6.8%.

CONCLUSIONS

These results suggest that integrating automatic ranking models created for specific information quality dimensions can increase the effectiveness of health-related information retrieval. Thus, our approach could be used to enhance searches made by individuals seeking online health information.

Text-Mining and Video Analytics of COVID-19 Narratives Shared by Patients on YouTube

This study explores how individuals who have experienced COVID-19 share their stories on YouTube, focusing on the nature of information disclosure, public engagement, and emotional impact pertaining to consumer health. Using a dataset of 186 YouTube videos, we used text mining and video analytics techniques to analyze textual transcripts and visual frames to identify themes, emotions, and their relationship with viewer engagement metrics. Findings reveal eight key themes: infection origins, symptoms, treatment, mental well-being, isolation, prevention, government directives, and vaccination. While viewers engaged most with videos about infection origins, treatment, and vaccination, fear and sadness in the text consistently drove views, likes, and comments. Visuals primarily conveyed happiness and sadness, but their influence on engagement varied. This research highlights the crucial role YouTube plays in disseminating COVID-19 patient narratives and suggests its potential for improving health communication strategies. By understanding how emotions and content influence viewer engagement, healthcare professionals and public health officials can tailor their messaging to better connect with the public and address pandemic-related anxieties.

Educational Video Intervention to Improve Health Misinformation Identification on WhatsApp Among Saudi Arabian Population: Pre-Post Intervention Study

BACKGROUND

Health misinformation can adversely affect individuals' quality of life and increase the risk of mortality. People often fail to assess the content of messages before sharing them on the internet, increasing the spread of misinformation. The problem is exacerbated by the growing variety of digital information environments, especially social media, which presents as an effective platform for spreading misinformation due to its rapid information-sharing capabilities. Educational interventions have been developed to help consumers verify the validity of digital health information. However, tools designed to detect health misinformation on social media content have not been validated. Given the increased use of social media platforms, particularly WhatsApp, it is crucial to develop tools to help consumers assess the credibility of messages and detect misinformation.

OBJECTIVE

The main objective of this study is to develop and assess an educational tool aimed at educating consumers about detecting health misinformation on WhatsApp. The secondary objective is to assess the association between demographic factors and knowledge levels.

METHODS

The study used a single-arm, pre-post intervention design to evaluate the effectiveness of an educational video in improving participants' ability to detect health-related misinformation in WhatsApp messages. In the first phase, an educational video intervention was developed and validated. In the second phase, participants were invited to complete a web-based survey that consisted of pre-evaluation questions, followed by the educational video intervention. Subsequently, they were asked to answer the same questions as the postevaluation questions.

RESULTS

The web-based survey received 485 responses. The completion rate was 99.6% (n=483). Statistically significant associations existed between knowledge level and age, gender, employment, and region of residence (P<.05). The video intervention did elicit a statistically significant change in the participants' abilities to identify misinformation in WhatsApp messages (z=-6.887; P<.001). Viewing the video was associated with increased knowledge about the following concepts: checking the "forwarded" label (P<.001), looking for spelling and grammatical errors (P<.001), analyzing the facts (P=.03), checking links (P=.002, P=.001), and assessing the photos and videos (P<.001). There was a statistically significant difference in knowledge level before and after the intervention (P<.001).

CONCLUSIONS

This study developed and evaluated the effectiveness of an educational video intervention to improve health misinformation identification on WhatsApp among the Saudi Arabian population. The results indicate that educational videos can be valuable tools for improving participants' abilities to identify misinformation. The outcomes of this research can contribute to our understanding of what constitutes an effective tool for enhancing health misinformation awareness. Such interventions may be particularly useful in combating misinformation among Arabic-speaking populations on WhatsApp, which may ultimately improve eHealth literacy. Limiting the prevalence and impact of misinformation allows people to make better-informed health decisions.

Evaluating the perceptions of pesticide use, safety, and regulation and identifying common pesticide-related topics on Twitter

Synthetic pesticides are important agricultural tools that increase crop yield and help feed the world's growing population. These products are also highly regulated to balance benefits and potential environmental and human risks. Public perception of pesticide use, safety, and regulation is an important topic necessitating discussion across a variety of stakeholders from lay consumers to regulatory agencies since attitudes toward this subject could differ markedly. Individuals and organizations can perceive the same message(s) about pesticides differently due to prior differences in technical knowledge, perceptions, attitudes, and individual or group circumstances. Social media platforms, like Twitter, include both individuals and organizations and function as a townhall where each group promotes their topics of interest, shares their perspectives, and engages in both well-informed and misinformed discussions. We analyzed public Twitter posts about pesticides by user group, time, and location to understand their communication behaviors, including their sentiments and discussion topics, using machine learning-based text analysis methods. We extracted tweets related to pesticides between 2013 and 2021 based on relevant keywords developed through a "snowball" sampling process. Each tweet was grouped into individual versus organizational groups, then further categorized into media, government, industry, academia, and three types of nongovernmental organizations. We compared topic distributions within and between those groups using topic modeling and then applied sentiment analysis to understand the public's attitudes toward pesticide safety and regulation. Individual accounts expressed concerns about health and environmental risks, while industry and government accounts focused on agricultural usage and regulations. Public perceptions are heavily skewed toward negative sentiments, although this varies geographically. Our findings can help managers and decision-makers understand public sentiments, priorities, and perceptions and provide insights into public discourse on pesticides. Integr Environ Assess Manag 2023;19:1581-1599. © 2023 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Public Health Policy, Political Ideology, and Public Emotion Related to COVID-19 in the U.S

Social networks, particularly Twitter 9.0 (known as X as of 23 July 2023), have provided an avenue for prompt interactions and sharing public health-related concerns and emotions, especially during the COVID-19 pandemic when in-person communication became less feasible due to stay-at-home policies in the United States (U.S.). The study of public emotions extracted from social network data has garnered increasing attention among scholars due to its significant predictive value for public behaviors and opinions. However, few studies have explored the associations between public health policies, local political ideology, and the spatial-temporal trends of emotions extracted from social networks. This study aims to investigate (1) the spatial-temporal clustering trends (or spillover effects) of negative emotions related to COVID-19; and (2) the association relationships between public health policies such as stay-at-home policies, political ideology, and the negative emotions related to COVID-19. This study employs multiple statistical methods (zero-inflated Poisson (ZIP) regression, random-effects model, and spatial autoregression (SAR) model) to examine relationships at the county level by using the data merged from multiple sources, mainly including Twitter 9.0, Johns Hopkins, and the U.S. Census Bureau. We find that negative emotions related to COVID-19 extracted from Twitter 9.0 exhibit spillover effects, with counties implementing stay-at-home policies or leaning predominantly Democratic showing higher levels of observed negative emotions related to COVID-19. These findings highlight the impact of public health policies and political polarization on spatial-temporal public emotions exhibited in social media. Scholars and policymakers can benefit from understanding how public policies and political ideology impact public emotions to inform and enhance their communication strategies and intervention design during public health crises such as the COVID-19 pandemic.

Knowledge and attitude toward oral health behavior of overseas students during the COVID-19 pandemic

BACKGROUND

The COVID-19 pandemic has impacted overseas students, including their oral health. Due to movement restrictions, limited living allowances, dental treatment costs, and health insurance fees, overseas students might be more concerned about their oral health. The objective of the present study was to determine the association of knowledge and attitude toward oral healthcare behavior of overseas university students staying in Thailand between January 2020 to July 2022 and explore the experiences of their oral health problems.

METHODS

A cross-sectional study was conducted using an online survey in English operated through the Google platform by convenience sampling among overseas Chulalongkorn University students. A newly developed self-administered questionnaire on knowledge and attitude toward oral health-related behavior and experiences in oral health problems was completed voluntarily. Descriptive statistics, Chi-square test, t-test, ANOVA, and Pearson correlations were employed using IBM SPSS version 29.

RESULTS

Of 311 overseas students, 55.6% were male. The average age of students was 27.5 ± 4.5 years. 68.81% of students were from ASEAN countries, and 73.31% studied in non-health science programs. The study fields, health and non-health sciences, were associated with knowledge score (p < 0.001) and attitude score (p = 0.004), whereas the type of health insurance had an association with behavior score (p = 0.014) and the student's perspective about dental visits (p = 0.014). Three hundred fifty-nine cases of oral health problems were experienced by 47.3% of overseas students. These problems consisted primarily of tooth hypersensitivity (21.2%), gingivitis (15.3%), caries (14%), cracked or broken tooth (10%), severe toothache (9%), fallen out filling (8%), and wisdom tooth pain (7.8%). There was an association between oral healthcare behavior and oral health problems (p < 0.001), and a negative correlation was found between behavior score and the number of oral health problems (p < 0.001, r=-0.204).

CONCLUSION

The oral healthcare habits of overseas university students correlated positively with knowledge and attitude. A negative correlation was observed between behavior and the number of oral health problems. Furthermore, studying in health science programs impacted students' knowledge and attitude toward oral health, while dental treatment coverage insurance affected decisions for dental visits.

Explainable online health information truthfulness in Consumer Health Search

Introduction

People are today increasingly relying on health information they find online to make decisions that may impact both their physical and mental wellbeing. Therefore, there is a growing need for systems that can assess the truthfulness of such health information. Most of the current literature solutions use machine learning or knowledge-based approaches treating the problem as a binary classification task, discriminating between correct information and misinformation. Such solutions present several problems with regard to user decision making, among which: (i) the binary classification task provides users with just two predetermined possibilities with respect to the truthfulness of the information, which users should take for granted; indeed, (ii) the processes by which the results were obtained are often opaque and the results themselves have little or no interpretation.

Methods

To address these issues, we approach the problem as an ad hoc retrieval task rather than a classification task, with reference, in particular, to the Consumer Health Search task. To do this, a previously proposed Information Retrieval model, which considers information truthfulness as a dimension of relevance, is used to obtain a ranked list of both topically-relevant and truthful documents. The novelty of this work concerns the extension of such a model with a solution for the explainability of the results obtained, by relying on a knowledge base consisting of scientific evidence in the form of medical journal articles.

Results and discussion

We evaluate the proposed solution both quantitatively, as a standard classification task, and qualitatively, through a user study to examine the "explained" ranked list of documents. The results obtained illustrate the solution's effectiveness and usefulness in making the retrieved results more interpretable by Consumer Health Searchers, both with respect to topical relevance and truthfulness.

Automatic detection of health misinformation: a systematic review

The spread of health misinformation has the potential to cause serious harm to public health, from leading to vaccine hesitancy to adoption of unproven disease treatments. In addition, it could have other effects on society such as an increase in hate speech towards ethnic groups or medical experts. To counteract the sheer amount of misinformation, there is a need to use automatic detection methods. In this paper we conduct a systematic review of the computer science literature exploring text mining techniques and machine learning methods to detect health misinformation. To organize the reviewed papers, we propose a taxonomy, examine publicly available datasets, and conduct a content-based analysis to investigate analogies and differences among Covid-19 datasets and datasets related to other health domains. Finally, we describe open challenges and conclude with future directions.

Effects of Visual Communication Design Accessibility (VCDA) Guidelines for Low Vision on Public and Open Government Health Data

Since 2019, the Korean government's investments in making data more accessible to the public have grown by 337%. However, open government data, which should be accessible to everyone, are not entirely accessible to people with low vision, who represent an information-vulnerable class. Emergencies, such as the COVID-19 pandemic, decrease face-to-face encounters and inevitably increase untact encounters. Thus, the information gap experienced by low-vision people, who are underprivileged in terms of information, will be further widened, and they may consequently face various disadvantages. This study proposed visual communication design accessibility (VCDA) guidelines for people with low vision. Introduced screens enhanced by accessibility guidelines were presented to 16 people with low vision and 16 people with normal vision and the speed of visual information recognition was analyzed. No statistically significant difference (p > 0.05) was found due to the small sample size; however, this study's results approached significance with improved visual recognition speed for people with low vision after adopting VCDA. As a result of the intervention, the visual information recognition speed of both normal and low-vision people improved. Thus, our results can help improve information recognition speed among people with normal and low vision.

Detecting and classifying online health misinformation with 'Content Similarity Measure (CSM)' algorithm: an automated fact-checking-based approach

Information dissemination occurs through the 'word of media' in the digital world. Fraudulent and deceitful content, such as misinformation, has detrimental effects on people. An implicit fact-based automated fact-checking technique comprising information retrieval, natural language processing, and machine learning techniques assist in assessing the credibility of content and detecting misinformation. Previous studies focused on linguistic and textual features and similarity measures-based approaches. However, these studies need to gain knowledge of facts, and similarity measures are less accurate when dealing with sparse or zero data. To fill these gaps, we propose a 'Content Similarity Measure (CSM)' algorithm that can perform automated fact-checking of URLs in the healthcare domain. Authors have introduced a novel set of content similarity, domain-specific, and sentiment polarity score features to achieve journalistic fact-checking. An extensive analysis of the proposed algorithm compared with standard similarity measures and machine learning classifiers showed that the 'content similarity score' feature outperformed other features with an accuracy of 88.26%. In the algorithmic approach, CSM showed improved accuracy of 91.06% compared to the Jaccard similarity measure with 74.26% accuracy. Another observation is that the algorithmic approach outperformed the feature-based method. To check the robustness of the algorithms, authors have tested the model on three state-of-the-art datasets, viz. CoAID, FakeHealth, and ReCOVery. With the algorithmic approach, CSM showed the highest accuracy of 87.30%, 89.30%, 85.26%, and 88.83% on CoAID, ReCOVery, FakeHealth (Story), and FakeHealth (Release) datasets, respectively. With a feature-based approach, the proposed CSM showed the highest accuracy of 85.93%, 87.97%, 83.92%, and 86.80%, respectively.

Evaluating online health information quality using machine learning and deep learning: A systematic literature review

Background

Due to the large volume of online health information, while quality remains dubious, understanding the usage of artificial intelligence to evaluate health information and surpass human-level performance is crucial. However, the existing studies still need a comprehensive review highlighting the vital machine, and Deep learning techniques for the automatic health information evaluation process.

Objective

Therefore, this study outlines the most recent developments and the current state of the art regarding evaluating the quality of online health information on web pages and specifies the direction of future research.

Methods

In this article, a systematic literature is conducted according to the PRISMA statement in eight online databases PubMed, Science Direct, Scopus, ACM, Springer Link, Wiley Online Library, Emerald Insight, and Web of Science to identify all empirical studies that use machine and deep learning models for evaluating the online health information quality. Furthermore, the selected techniques are compared based on their characteristics, such as health quality criteria, quality measurement tools, algorithm type, and achieved performance.

Results

The included papers evaluate health information on web pages using over 100 quality criteria. The results show no universal quality dimensions used by health professionals and machine or deep learning practitioners while evaluating health information quality. In addition, the metrics used to assess the model performance are not the same as those used to evaluate human performance.

Conclusions

This systemic review offers a novel perspective in approaching the health information quality in web pages that can be used by machine and deep learning practitioners to tackle the problem more effectively.

Breast cancer prevention and treatment misinformation on Twitter: An analysis of two languages

Objective

To determine the prevalence and types of misinformation on Twitter related to breast cancer prevention and treatment; and compare the differences between the misinformation in English and Malay tweets.

Methods

A total of 6221 tweets related to breast cancer posted between 2018 and 2022 were collected. An oncologist and two pharmacists coded the tweets to differentiate between true information and misinformation, and to analyse the misinformation content. Binary logistic regression was conducted to identify determinants of misinformation.

Results

There were 780 tweets related to breast cancer prevention and treatment, and 456 (58.5%) contain misinformation, with significantly more misinformation in Malay compared to English tweets (OR = 6.18, 95% CI: 3.45-11.07, p < 0.001). Other determinants of misinformation were tweets posted by product sellers and posted before the COVID-19 pandemic. Less misinformation was associated with tweets utilising official/peer-reviewed sources of information compared to tweets without external sources and those that utilised less reliable information sources. The top three most common content of misinformation were food and lifestyle, alternative medicine and supplements, comprising exaggerated claims of anti-cancer properties of traditional and natural-based products.

Conclusion

Misinformation on breast cancer prevention and treatment is prevalent on social media, with significantly more misinformation in Malay compared to English tweets. Our results highlighted that patients need to be educated on digital health literacy, with emphasis on utilising reliable sources of information and being cautious of any promotional materials that may contain misleading information. More studies need to be conducted in other languages to address the disparity in misinformation.

Associations between Public Fear of COVID-19 and Number of COVID-19 Vaccinations: A County-Level Longitudinal Analysis

Background and Purpose: A large number of COVID-19 infections and deaths and the ensuing socioeconomic problems created widespread public fear around COVID-19. Fear around COVID-19 greatly influences people’s attitudes towards receiving the COVID-19 vaccines. The purpose of this study is examining (a) the impact of the public fear of COVID-19 (PFC) on the number of COVID-19 vaccinations at the county level; (b) the interaction effect between the PFC and per capita income, unemployment rates, and COVID-19 vaccines incentive policies, on the number of COVID-19 vaccinations at the county level. Method: This is a longitudinal analysis across states in the U.S. by using county-level data of 2856 counties from 1 February to 1 July. Random-effects models were adopted to analyze the associations between the PFC and the number of COVID-19 vaccinations. Result: the PFC was positively associated with the number of COVID-19 vaccinations at county-level, as PFC increases from 0 to 300, the predicted vaccination number increases from 10,000 to 230,000. However, the associations were divergent when the PFC interacts with county-level per capita income, unemployment rates, and incentive policies. Conclusion: public fear is an important indicator for the county-level vaccination numbers of COVID-19. However, it is critical to consider public fear and socioeconomic factors when making policies that aim to increase COVID-19 vaccination rates.

A survey of uncover misleading and cyberbullying on social media for public health

Misleading health information is a critical phenomenon in our modern life due to advance in technology. In fact, social media facilitated the dissemination of information, and as a result, misinformation spread rapidly, cheaply, and successfully. Fake health information can have a significant effect on human behavior and attitudes. This survey presents the current works developed for misleading information detection (MLID) in health fields based on machine learning and deep learning techniques and introduces a detailed discussion of the main phases of the generic adopted approach for MLID. In addition, we highlight the benchmarking datasets and the most used metrics to evaluate the performance of MLID algorithms are discussed and finally, a deep investigation of the limitations and drawbacks of the current progressing technologies in various research directions is provided to help the researchers to use the most proper methods in this emerging task of MLID.

Algebraic Zero Error Training Method for Neural Networks Achieving Least Upper Bounds on Neurons and Layers

In the domain of artificial neural networks, it is important to know what their representation, classification and generalization capabilities are. There is also a need for time and resource-efficient training algorithms. Here, a new zero-error training method is derived for digital computers and single hidden layer networks. This method is the least upper bound on the number of hidden neurons as well. The bound states that if there are N input vectors expressed as rational numbers, a network having N − 1 neurons in the hidden layer and M neurons at the output represents a bounded function F: RD→RM for all input vectors. Such a network has massively shared weights calculated by 1 + M regular systems of linear equations. Compared to similar approaches, this new method achieves a theoretical least upper bound, is fast, robust, adapted to floating-point data, and uses few free parameters. This is documented by theoretical analyses and comparative tests. In theory, this method provides a new constructional proof of the least upper bound on the number of hidden neurons, extends the classes of supported activation functions, and relaxes conditions for mapping functions. Practically, it is a non-iterative zero-error training algorithm providing a minimum number of neurons and layers.