COVID-19 Health Beliefs Regarding Mask Wearing and Vaccinations on Twitter: Deep Learning Approach.

Background: Amid the global COVID-19 pandemic, a worldwide infodemic also emerged with large amounts of COVID-19-related information and misinformation spreading through social media channels. Various organizations, including the World Health Organization (WHO) and the Centers for Disease Control and Prevention (CDC), and other prominent individuals issued high-profile advice on preventing the further spread of COVID-19. Objective: The purpose of this study is to leverage machine learning and Twitter data from the pandemic period to explore health beliefs regarding mask wearing and vaccines and the influence of high-profile cues to action. Methods: A total of 646,885,238 COVID-19-related English tweets were filtered, creating a mask-wearing data set and a vaccine data set. Researchers manually categorized a training sample of 3500 tweets for each data set according to their relevance to Health Belief Model (HBM) constructs and used coded tweets to train machine learning models for classifying each tweet in the data sets. Results: In total, 5 models were trained for both the mask-related and vaccine-related data sets using the XLNet transformer model, with each model achieving at least 81% classification accuracy. Health beliefs regarding perceived benefits and barriers were most pronounced for both mask wearing and immunization; however, the strength of those beliefs appeared to vary in response to high-profile cues to action. Conclusions: During both the COVID-19 pandemic and the infodemic, health beliefs related to perceived benefits and barriers observed through Twitter using a big data machine learning approach varied over time and in response to high-profile cues to action from prominent organizations and individuals.

A step towards intelligent EBSD microscopy: machine-learning prediction of twin activity in MgAZ31.

Although microscopy is often treated as a quasi-static exercise for obtaining a snapshot of events and structure, it is clear that a more dynamic approach, involving real-time decision making for guiding the investigation process, may provide deeper insights, more efficiently. On the other hand, many applications of machine learning involve the interpretation of local circumstances from experience gained over many observations; that is, machine learning potentially provides an ideal solution for more efficient microscopy. This paper explores the potential for informing the microscope's observation strategy while characterising critical events. In particular, the identification of regions likely to experience twin activity (twin interaction with grain boundary) in AZ31 magnesium is attempted, from only local information. EBSD-based observations in the neighbourhoods of twin activity are fed into a machine-learning environment to inform the future search for such events, and the accuracy of the resultant decisions is quantified relative to the number of prior observations. The potential for utilising different types of local information, and their resultant value in the prediction process, is also assessed. After applying an attribute selection filter, and various other machine-learning tools, a decision-tree model is able to classify likely neighbourhoods of twin activity with 85% accuracy. The resultant framework provides the first step towards an intelligent microscopy for efficient observation of stochastic events during in situ microscopy campaigns. LAY DESCRIPTION: One role of artificial intelligence is to predict future events after learning from many previous observations. In materials science, various phenomena (such as crack nucleation) are difficult to predict because they have been insufficiently observed. Furthermore, observation is difficult, precisely because their location cannot be predicted, leading to a chicken and egg conundrum. This paper applies machine learning to the search for twin nucleation sites in a magnesium alloy, in an attempt to guide the observation of twin nucleation events in a microscope based on previous observations. As more data is obtained, the accuracy of the location prediction will increase. In the current case, the machine-learning tool achieved 85% accuracy for predicting the location of twin interactions with grain boundaries after several thousand observations. The resultant framework provides the first step towards an intelligent microscopy for efficient observation of stochastic events during in situ microscopy campaigns.

Epidemiology from Tweets: Estimating Misuse of Prescription Opioids in the USA from Social Media.

BACKGROUND: The misuse of prescription opioids (MUPO) is a leading public health concern. Social media are playing an expanded role in public health research, but there are few methods for estimating established epidemiological metrics from social media. The purpose of this study was to demonstrate that the geographic variation of social media posts mentioning prescription opioid misuse strongly correlates with government estimates of MUPO in the last month. METHODS: We wrote software to acquire publicly available tweets from Twitter from 2012 to 2014 that contained at least one keyword related to prescription opioid use (n = 3,611,528). A medical toxicologist and emergency physician curated the list of keywords. We used the semantic distance (SemD) to automatically quantify the similarity of meaning between tweets and identify tweets that mentioned MUPO. We defined the SemD between two words as the shortest distance between the two corresponding word-centroids. Each word-centroid represented all recognized meanings of a word. We validated this automatic identification with manual curation. We used Twitter metadata to estimate the location of each tweet. We compared our estimated geographic distribution with the 2013-2015 National Surveys on Drug Usage and Health (NSDUH). RESULTS: Tweets that mentioned MUPO formed a distinct cluster far away from semantically unrelated tweets. The state-by-state correlation between Twitter and NSDUH was highly significant across all NSDUH survey years. The correlation was strongest between Twitter and NSDUH data from those aged 18-25 (r = 0.94, p < 0.01 for 2012; r = 0.94, p < 0.01 for 2013; r = 0.71, p = 0.02 for 2014). The correlation was driven by discussions of opioid use, even after controlling for geographic variation in Twitter usage. CONCLUSIONS: Mentions of MUPO on Twitter correlate strongly with state-by-state NSDUH estimates of MUPO. We have also demonstrated that a natural language processing can be used to analyze social media to provide insights for syndromic toxicosurveillance.

Validating Machine Learning Algorithms for Twitter Data Against Established Measures of Suicidality.

BACKGROUND: One of the leading causes of death in the United States (US) is suicide and new methods of assessment are needed to track its risk in real time. OBJECTIVE: Our objective is to validate the use of machine learning algorithms for Twitter data against empirically validated measures of suicidality in the US population. METHODS: Using a machine learning algorithm, the Twitter feeds of 135 Mechanical Turk (MTurk) participants were compared with validated, self-report measures of suicide risk. RESULTS: Our findings show that people who are at high suicidal risk can be easily differentiated from those who are not by machine learning algorithms, which accurately identify the clinically significant suicidal rate in 92% of cases (sensitivity: 53%, specificity: 97%, positive predictive value: 75%, negative predictive value: 93%). CONCLUSIONS: Machine learning algorithms are efficient in differentiating people who are at a suicidal risk from those who are not. Evidence for suicidality can be measured in nonclinical populations using social media data.

Tracking suicide risk factors through Twitter in the US.

BACKGROUND: Suicide is a leading cause of death in the United States. Social media such as Twitter is an emerging surveillance tool that may assist researchers in tracking suicide risk factors in real time. AIMS: To identify suicide-related risk factors through Twitter conversations by matching on geographic suicide rates from vital statistics data. METHOD: At-risk tweets were filtered from the Twitter stream using keywords and phrases created from suicide risk factors. Tweets were grouped by state and departures from expectation were calculated. The values for suicide tweeters were compared against national data of actual suicide rates from the Centers for Disease Control and Prevention. RESULTS: A total of 1,659,274 tweets were analyzed over a 3-month period with 37,717 identified as at-risk for suicide. Midwestern and western states had a higher proportion of suicide-related tweeters than expected, while the reverse was true for southern and eastern states. A strong correlation was observed between state Twitter-derived data and actual state age-adjusted suicide data. CONCLUSION: Twitter may be a viable tool for real-time monitoring of suicide risk factors on a large scale. This study demonstrates that individuals who are at risk for suicide may be detected through social media.

Using Twitter for breast cancer prevention: an analysis of breast cancer awareness month.

BACKGROUND: One in eight women will develop breast cancer in her lifetime. The best-known awareness event is breast cancer awareness month (BCAM). BCAM month outreach efforts have been associated with increased media coverage, screening mammography and online information searching. Traditional mass media coverage has been enhanced by social media. However, there is a dearth of literature about how social media is used during awareness-related events. The purpose of this research was to understand how Twitter is being used during BCAM. METHODS: This was a cross-sectional, descriptive study. We collected breast cancer- related tweets from 26 September - 12 November 2012, using Twitter's application programming interface. We classified Twitter users into organizations, individuals, and celebrities; each tweet was classified as an original or a retweet, and inclusion of a mention, meaning a reference to another Twitter user with @username. Statistical methods included ANOVA and chi square. For content analysis, we used computational linguistics techniques, specifically the MALLET implementation of the unsupervised topic modeling algorithm Latent Dirichlet Allocation. RESULTS: There were 1,351,823 tweets by 797,827 unique users. Tweets spiked dramatically the first few days then tapered off. There was an average of 1.69 tweets per user. The majority of users were individuals. Nearly all of the tweets were original. Organizations and celebrities posted more often than individuals. On average celebrities made far more impressions; they were also retweeted more often and their tweets were more likely to include mentions. Individuals were more likely to direct a tweet to a specific person. Organizations and celebrities emphasized fundraisers, early detection, and diagnoses while individuals tweeted about wearing pink. CONCLUSIONS: Tweeting about breast cancer was a singular event. The majority of tweets did not promote any specific preventive behavior. Twitter is being used mostly as a one-way communication tool. To expand the reach of the message and maximize the potential for word-of-mouth marketing using Twitter, organizations need a strategic communications plan to ensure on-going social media conversations. Organizations may consider collaborating with individuals and celebrities in these conversations. Social media communication strategies that emphasize fundraising for breast cancer research seem particularly appropriate.

An exploration of social circles and prescription drug abuse through Twitter.

BACKGROUND: Prescription drug abuse has become a major public health problem. Relationships and social context are important contributing factors. Social media provides online channels for people to build relationships that may influence attitudes and behaviors. OBJECTIVE: To determine whether people who show signs of prescription drug abuse connect online with others who reinforce this behavior, and to observe the conversation and engagement of these networks with regard to prescription drug abuse. METHODS: Twitter statuses mentioning prescription drugs were collected from November 2011 to November 2012. From this set, 25 Twitter users were selected who discussed topics indicative of prescription drug abuse. Social circles of 100 people were discovered around each of these Twitter users; the tweets of the Twitter users in these networks were collected and analyzed according to prescription drug abuse discussion and interaction with other users about the topic. RESULTS: From November 2011 to November 2012, 3,389,771 mentions of prescription drug terms were observed. For the 25 social circles (n=100 for each circle), on average 53.96% (SD 24.3) of the Twitter users used prescription drug terms at least once in their posts, and 37.76% (SD 20.8) mentioned another Twitter user by name in a post with a prescription drug term. Strong correlation was found between the kinds of drugs mentioned by the index user and his or her network (mean r=0.73), and between the amount of interaction about prescription drugs and a level of abusiveness shown by the network (r=0.85, P<.001). CONCLUSIONS: Twitter users who discuss prescription drug abuse online are surrounded by others who also discuss it-potentially reinforcing a negative behavior and social norm.

Results on mining NHANES data: a case study in evidence-based medicine.

The National Health and Nutrition Examination Survey (NHANES), administered annually by the National Center for Health Statistics, is designed to assess the general health and nutritional status of adults and children in the United States. Given to several thousands of individuals, the extent of this survey is very broad, covering demographic, laboratory and examination information, as well as responses to a fairly comprehensive health questionnaire. In this paper, we adapt and extend association rule mining and clustering algorithms to extract useful knowledge regarding diabetes and high blood pressure from the 1999-2008 survey results, thus demonstrating how data mining techniques may be used to support evidence-based medicine.

Tweaking and tweeting: exploring Twitter for nonmedical use of a psychostimulant drug (Adderall) among college students.

BACKGROUND: Adderall is the most commonly abused prescription stimulant among college students. Social media provides a real-time avenue for monitoring public health, specifically for this population. OBJECTIVE: This study explores discussion of Adderall on Twitter to identify variations in volume around college exam periods, differences across sets of colleges and universities, and commonly mentioned side effects and co-ingested substances. METHODS: Public-facing Twitter status messages containing the term "Adderall" were monitored from November 2011 to May 2012. Tweets were examined for mention of side effects and other commonly abused substances. Tweets from likely students containing GPS data were identified with clusters of nearby colleges and universities for regional comparison. RESULTS: 213,633 tweets from 132,099 unique user accounts mentioned "Adderall." The number of Adderall tweets peaked during traditional college and university final exam periods. Rates of Adderall tweeters were highest among college and university clusters in the northeast and south regions of the United States. 27,473 (12.9%) mentioned an alternative motive (eg, study aid) in the same tweet. The most common substances mentioned with Adderall were alcohol (4.8%) and stimulants (4.7%), and the most common side effects were sleep deprivation (5.0%) and loss of appetite (2.6%). CONCLUSIONS: Twitter posts confirm the use of Adderall as a study aid among college students. Adderall discussions through social media such as Twitter may contribute to normative behavior regarding its abuse.

Evaluating social media's capacity to develop engaged audiences in health promotion settings: use of Twitter metrics as a case study.

Use of social media in health promotion and public health continues to grow in popularity, though most of what is reported in literature represents one-way messaging devoid of attributes associated with engagement, a core attribute, if not the central purpose, of social media. This article defines engagement, describes its value in maximizing the potential of social media in health promotion, proposes an evaluation hierarchy for social media engagement, and uses Twitter as a case study to illustrate how the hierarchy might function in practice. Partnership and participation are proposed as culminating outcomes for social media use in health promotion. As use of social media in health promotion moves toward this end, evaluation metrics that verify progress and inform subsequent strategies will become increasingly important.

"Right time, right place" health communication on Twitter: value and accuracy of location information.

BACKGROUND: Twitter provides various types of location data, including exact Global Positioning System (GPS) coordinates, which could be used for infoveillance and infodemiology (ie, the study and monitoring of online health information), health communication, and interventions. Despite its potential, Twitter location information is not well understood or well documented, limiting its public health utility. OBJECTIVE: The objective of this study was to document and describe the various types of location information available in Twitter. The different types of location data that can be ascertained from Twitter users are described. This information is key to informing future research on the availability, usability, and limitations of such location data. METHODS: Location data was gathered directly from Twitter using its application programming interface (API). The maximum tweets allowed by Twitter were gathered (1% of the total tweets) over 2 separate weeks in October and November 2011. The final dataset consisted of 23.8 million tweets from 9.5 million unique users. Frequencies for each of the location options were calculated to determine the prevalence of the various location data options by region of the world, time zone, and state within the United States. Data from the US Census Bureau were also compiled to determine population proportions in each state, and Pearson correlation coefficients were used to compare each state's population with the number of Twitter users who enable the GPS location option. RESULTS: The GPS location data could be ascertained for 2.02% of tweets and 2.70% of unique users. Using a simple text-matching approach, 17.13% of user profiles in the 4 continental US time zones were able to be used to determine the user's city and state. Agreement between GPS data and data from the text-matching approach was high (87.69%). Furthermore, there was a significant correlation between the number of Twitter users per state and the 2010 US Census state populations (r ≥ 0.97, P < .001). CONCLUSIONS: Health researchers exploring ways to use Twitter data for disease surveillance should be aware that the majority of tweets are not currently associated with an identifiable geographic location. Location can be identified for approximately 4 times the number of tweets using a straightforward text-matching process compared to using the GPS location information available in Twitter. Given the strong correlation between both data gathering methods, future research may consider using more qualitative approaches with higher yields, such as text mining, to acquire information about Twitter users' geographical location.

There's an app for that: content analysis of paid health and fitness apps.

BACKGROUND: The introduction of Apple's iPhone provided a platform for developers to design third-party apps, which greatly expanded the functionality and utility of mobile devices for public health. OBJECTIVE: This study provides an overview of the developers' written descriptions of health and fitness apps and appraises each app's potential for influencing behavior change. METHODS: Data for this study came from a content analysis of health and fitness app descriptions available on iTunes during February 2011. The Health Education Curriculum Analysis Tool (HECAT) and the Precede-Proceed Model (PPM) were used as frameworks to guide the coding of 3336 paid apps. RESULTS: Compared to apps with a cost less than US $0.99, apps exceeding US $0.99 were more likely to be scored as intending to promote health or prevent disease (92.55%, 1925/3336 vs 83.59%, 1411/3336; P<.001), to be credible or trustworthy (91.11%, 1895/3336 vs 86.14%, 1454/3349; P<.001), and more likely to be used personally or recommended to a health care client (72.93%, 1517/2644 vs 66.77%, 1127/2644; P<.001). Apps related to healthy eating, physical activity, and personal health and wellness were more common than apps for substance abuse, mental and emotional health, violence prevention and safety, and sexual and reproductive health. Reinforcing apps were less common than predisposing and enabling apps. Only 1.86% (62/3336) of apps included all 3 factors (ie, predisposing, enabling, and reinforcing). CONCLUSIONS: Development efforts could target public health behaviors for which few apps currently exist. Furthermore, practitioners should be cautious when promoting the use of apps as it appears most provide health-related information (predisposing) or make attempts at enabling behavior, with almost none including all theoretical factors recommended for behavior change.