Characterizing Tweet Volume and Content About Common Health Conditions Across Pennsylvania: Retrospective Analysis

Biases in using social media data for public health surveillance: A scoping review

OBJECTIVES

A landscape scan of the methods that are used to either assess or mitigate biases when using social media data for public health surveillance, through a scoping review.

MATERIALS AND METHODS

Following best practices, we searched two literature databases (i.e., PubMed and Web of Science) and covered literature published up to July 2021. Through two rounds of screening (i.e., title/abstract screening, and then full-text screening), we extracted study objectives, analysis methods, and the methods used to assess or address the different biases from the eligible articles.

RESULTS

We identified a total of 2,856 articles from the two databases. After the screening processes, we extracted and synthesized 20 studies that either assessed or mitigated biases when leveraging social media data for public health surveillance. Researchers have tried to assess or address several different types of biases such as demographic bias, keyword bias, and platform bias. In particular, we found 11 studies that tried to measure the reliability of the research findings from social media data by comparing them with other data sources.

DISCUSSION AND CONCLUSION

We synthesized the types of biases and the methods used to assess or address the biases in studies that use social media data for public health surveillance. We found very few studies, despite the large number of publications using social media data, considered the various bias issues that are present from data collection to analysis methods. Overlooking bias can distort the study results and lead to unintended consequences, especially in the field of public health surveillance. These research gaps warrant further investigations more systematically. Strategies from other fields for addressing biases can be introduced for future public health surveillance systems that use social media data.

Unraveling racial disparities in asthma emergency department visits using electronic healthcare records and machine learning

OBJECTIVE

Hospital emergency department (ED) visits by asthmatics differ based on race and season. The objectives of this study were to investigate season- and race-specific disparities for asthma risk, and to identify environmental exposure variables associated with ED visits among more than 42,000 individuals of African American (AA) and European American (EA) descent identified through electronic health records (EHRs).

METHODS

We examined data from 42,375 individuals (AAs = 14,491, EAs = 27,884) identified in EHRs. We considered associated demographic (race, age, gender, insurance), clinical (smoking status, ED visits, FEV1%), and environmental exposures data (mold, pollen, and pollutants). Machine learning techniques, including random forest (RF), extreme gradient boosting (XGB), and decision tree (DT) were used to build and identify race- and -season-specific predictive models for asthma ED visits.

RESULTS

Significant differences in ED visits and FEV1% among AAs and EAs were identified. ED visits by AAs was 32.0% higher than EAs and AAs had 6.4% lower FEV1% value than EAs. XGB model was used to accurately classify asthma patients visiting ED into AAs and EAs. Pollen factor and pollution (PM2.5, PM10) were the key variables for asthma in AAs and EAs, respectively. Age and cigarette smoking increase asthma risk independent of seasons.

CONCLUSIONS

In this study, we observed racial and season-specific disparities between AAs and EAs asthmatics for ED visit and FEV1% severity, suggesting the need to address asthma disparities through key predictors including socio-economic status, particulate matter, and mold.

Using Patient-Generated Health Data From Twitter to Identify, Engage, and Recruit Cancer Survivors in Clinical Trials in Los Angeles County: Evaluation of a Feasibility Study

BACKGROUND

Failure to find and attract clinical trial participants remains a persistent barrier to clinical research. Researchers increasingly complement recruitment methods with social media-based methods. We hypothesized that user-generated data from cancer survivors and their family members and friends on the social network Twitter could be used to identify, engage, and recruit cancer survivors for cancer trials.

OBJECTIVE

This pilot study aims to examine the feasibility of using user-reported health data from cancer survivors and family members and friends on Twitter in Los Angeles (LA) County to enhance clinical trial recruitment. We focus on 6 cancer conditions (breast cancer, colon cancer, kidney cancer, lymphoma, lung cancer, and prostate cancer).

METHODS

The social media intervention involved monitoring cancer-specific posts about the 6 cancer conditions by Twitter users in LA County to identify cancer survivors and their family members and friends and contacting eligible Twitter users with information about open cancer trials at the University of Southern California (USC) Norris Comprehensive Cancer Center. We reviewed both retrospective and prospective data published by Twitter users in LA County between July 28, 2017, and November 29, 2018. The study enrolled 124 open clinical trials at USC Norris. We used descriptive statistics to report the proportion of Twitter users who were identified, engaged, and enrolled.

RESULTS

We analyzed 107,424 Twitter posts in English by 25,032 unique Twitter users in LA County for the 6 cancer conditions. We identified and contacted 1.73% (434/25,032) of eligible Twitter users (127/434, 29.3% cancer survivors; 305/434, 70.3% family members and friends; and 2/434, 0.5% Twitter users were excluded). Of them, 51.4% (223/434) were female and approximately one-third were male. About one-fifth were people of color, whereas most of them were White. Approximately one-fifth (85/434, 19.6%) engaged with the outreach messages (cancer survivors: 33/85, 38% and family members and friends: 52/85, 61%). Of those who engaged with the messages, one-fourth were male, the majority were female, and approximately one-fifth were people of color, whereas the majority were White. Approximately 12% (10/85) of the contacted users requested more information and 40% (4/10) set up a prescreening. Two eligible candidates were transferred to USC Norris for further screening, but neither was enrolled.

CONCLUSIONS

Our findings demonstrate the potential of identifying and engaging cancer survivors and their family members and friends on Twitter. Optimization of downstream recruitment efforts such as screening for digital populations on social media may be required. Future research could test the feasibility of the approach for other diseases, locations, languages, social media platforms, and types of research involvement (eg, survey research). Computer science methods could help to scale up the analysis of larger data sets to support more rigorous testing of the intervention.

TRIAL REGISTRATION

ClinicalTrials.gov NCT03408561; https://clinicaltrials.gov/ct2/show/NCT03408561.

Methadone and suboxone® mentions on twitter: thematic and sentiment analysis

BACKGROUND

According to the latest medical evidence, Methadone and buprenorphine-naloxone (Suboxone®) are effective treatments for opioid use disorder (OUD). While the evidence basis for the use of these medications is favorable, less is known about the perceptions of the general public about them.

OBJECTIVE

This study aimed to use Twitter to assess the public perceptions about methadone and buprenorphine-naloxone, and to compare their discussion contents based on themes/topics, subthemes, and sentiment.

METHODS

We conducted a descriptive analysis of a small and automatic analysis of a large volume of microposts ("tweets") that mentioned "methadone" or "suboxone". In the manual analysis, we categorized the tweets into themes and subthemes, as well as by sentiment and personal experience, and compared the information posted about these two medications. We performed automatic topic modeling and sentiment analysis over large volumes of posts and compared the outputs to those from the manual analyses.

RESULTS

We manually analyzed 900 tweets, most of which related to access (15.3% for methadone; 14.3% for buprenorphine-naloxone), stigma (17.0%; 15.5%), and OUD treatment (12.8%; 15.6%). Only a small proportion of tweets (16.4% for Suboxone® and 9.3% for methadone) expressed positive sentiments about the medications, with few tweets describing personal experiences. Tweets mentioning both medications primarily discussed MOUD broadly, rather than comparing the two medications directly. Automatic topic modeling revealed topics from the larger dataset that corresponded closely to the manually identified themes, but sentiment analysis did not reveal any notable differences in chatter regarding the two medications.

CONCLUSIONS

Twitter content about methadone and Suboxone® is similar, with the same major themes and similar sub-themes. Despite the proven effectiveness of these medications, there was little dialogue related to their benefits or efficacy in the treatment of OUD. Perceptions of these medications may contribute to their underutilization in combatting OUDs.

Expressed Symptoms and Attitudes Toward Using Twitter for Health Care Engagement Among Patients With Lupus on Social Media: Protocol for a Mixed Methods Study

BACKGROUND

Lupus is a complex autoimmune disease that is difficult to diagnose and treat. It is estimated that at least 5 million Americans have lupus, with more than 16,000 new cases of lupus being reported annually in the United States. Social media provides a platform for patients to find rheumatologists and peers and build awareness of the condition. Researchers have suggested that the social network Twitter may serve as a rich avenue for exploring how patients communicate about their health issues. However, there is a lack of research about the characteristics of lupus patients on Twitter and their attitudes toward using Twitter for engaging them with their health care.

OBJECTIVE

This study has two objectives: (1) to conduct a content analysis of Twitter data published by users (in English) in the United States between September 1, 2017 and October 31, 2018 to identify patients who publicly discuss their lupus condition and to assess their expressed health themes and (2) to conduct a cross-sectional survey among these lupus patients on Twitter to study their attitudes toward using Twitter for engaging them with their health care.

METHODS

This is a mixed methods study that analyzes retrospective Twitter data and conducts a cross-sectional survey among lupus patients on Twitter. We used Symplur Signals, a health care social media analytics platform, to access the Twitter data and analyze user-generated posts that include keywords related to lupus. We will use descriptive statistics to analyze the data and identify the most prevalent topics in the Twitter content among lupus patients. We will further conduct self-report surveys via Twitter by inviting all identified lupus patients who discuss their lupus condition on Twitter. The goal of the survey is to collect data about the characteristics of lupus patients (eg, gender, race/ethnicity, educational level) and their attitudes toward using Twitter for engaging them with their health care.

RESULTS

This study has been funded by the National Center for Advancing Translational Science through a Clinical and Translational Science Award. The institutional review board at the University of Southern California (HS-19-00048) approved the study. Data extraction and cleaning are complete. We obtained 47,715 Twitter posts containing terms related to "lupus" from users in the United States published in English between September 1, 2017 and October 31, 2018. We included 40,885 posts in the analysis. Data analysis was completed in Fall 2020.

CONCLUSIONS

The data obtained in this pilot study will shed light on whether Twitter provides a promising data source for garnering health-related attitudes among lupus patients. The data will also help to determine whether Twitter might serve as a potential outreach platform for raising awareness of lupus among patients and implementing related health education interventions.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/15716.

COVID-19 pandemic impact on people with rheumatic and musculoskeletal diseases: Insights from patient-generated health data on social media

Objectives

During the COVID-19 pandemic, much communication occurred online, through social media. This study aimed to provide patient perspective data on how the COVID-19 pandemic impacted people with rheumatic and musculoskeletal diseases (RMDs), using Twitter-based patient-generated health data (PGHD).

Methods

A convenience sample of Twitter messages in English posted by people with RMDs was extracted between March 1, and July 12, 2020 and examined using thematic analysis. Included were Twitter messages that mentioned keywords and hashtags related to both COVID-19 (or SARS-CoV-2) and select RMDs. The RMDs monitored included inflammatory-driven (joint) conditions (Ankylosing Spondylitis, Rheumatoid Arthritis, Psoriatic Arthritis, Lupus/Systemic Lupus Erythematosus, and Gout).

Results

The analysis included 569 tweets by 375 Twitter users with RMDs across several countries. Eight themes emerged regarding the impact of the COVID-19 pandemic on people with RMDs: (1) lack of understanding of SARS-CoV-2/COVID-19; (2) critical changes in health behaviour; (3) challenges in healthcare practice and communication with healthcare professionals; (4) difficulties with access to medical care; (5) negative impact on physical and mental health, coping strategies; (6) issues around work participation, (7) negative effects of the media; (8) awareness-raising.

Conclusion

The findings show that Twitter serves as a real-time data source to understand the impact of the COVID-19 pandemic on people with RMDs. The platform provided “early signals” of potentially critical health behaviour changes. Future epidemics might benefit from the real-time use of Twitter-based PGHD to identify emerging health needs, facilitate communication, and inform clinical practice decisions.

Perspectives Toward Seeking Treatment Among Patients With Psoriasis: Protocol for a Twitter Content Analysis

BACKGROUND

Psoriasis is an autoimmune disease estimated to affect more than 6 million adults in the United States. It poses a significant public health problem and contributes to rising health care costs, affecting people's quality of life and ability to work. Previous research showed that nontreatment and undertreatment of patients with psoriasis remain a significant problem. Perspectives of patients toward seeking psoriasis treatment are understudied. Social media offers a new data source of user-generated content. Researchers suggested that the social network Twitter may serve as a rich avenue for exploring how patients communicate about their health issues.

OBJECTIVE

The objective of this study is to conduct a content analysis of Twitter posts (in English) published by users in the United States between February 1, 2016, and October 31, 2018, to examine perspectives that potentially influence the treatment decision among patients with psoriasis.

METHODS

User-generated Twitter posts that include keywords related to psoriasis will be analyzed using text classifiers to identify themes related to the research questions. We will use Symplur Signals, a health care social media analytics platform, to access the Twitter data. We will use descriptive statistics to analyze the data and identify the most prevalent topics in the Twitter content among people with psoriasis.

RESULTS

This study is supported by the National Center for Advancing Translational Science through a Clinical and Translational Science Award award. Study approval was obtained from the institutional review board at the University of Southern California. Data extraction and cleaning are complete. For the time period from February 1, 2016, to October 31, 2018, we obtained 95,040 Twitter posts containing terms related to "psoriasis" from users in the United States published in English. After removing duplicates, retweets, and non-English tweets, we found that 75.51% (52,301/69,264) of the psoriasis-related posts were sent by commercial or bot-like accounts, while 16,963 posts were noncommercial and will be included in the analysis to assess the patient perspective. Analysis was completed in Summer 2020.

CONCLUSIONS

This protocol paper provides a detailed description of a social media research project including the process of data extraction, cleaning, and analysis. It is our goal to contribute to the development of more transparent social media research efforts. Our findings will shed light on whether Twitter provides a promising data source for garnering patient perspective data about psoriasis treatment decisions. The data will also help to determine whether Twitter might serve as a potential outreach platform for raising awareness of psoriasis and treatment options among patients and implementing related health interventions.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/13731.

Assessment of Antipsychotic Medications on Social Media: Machine Learning Study

Background: Antipsychotic medications are the first-line treatment for schizophrenia. However, non-adherence is frequent despite its negative impact on the course of the illness. In response, we aimed to investigate social media posts about antipsychotics to better understand the online environment in this regard. Methods: We collected tweets containing mentions of antipsychotic medications posted between January 1st 2019 and October 31st 2020. The content of each tweet and the characteristics of the users were analyzed as well as the number of retweets and likes generated. Results: Twitter users, especially those identified as patients, showed an interest in antipsychotic medications, mainly focusing on the topics of sexual dysfunction and sedation. Interestingly, paliperidone, despite being among one of the newest antipsychotics, accounted for a low number of tweets and did not generate much interest. Conversely, retweet and like ratios were higher in those tweets asking for or offering help, in those posted by institutions and in those mentioning cognitive complaints. Moreover, health professionals did not have a strong presence in tweet postings, nor did medical institutions. Finally, trivialization was frequently observed. Conclusion: This analysis of tweets about antipsychotic medications provides insights into experiences and opinions related to this treatment. Twitter user perspectives therefore constitute a valuable input that may help to improve clinicians' knowledge of antipsychotic medications and their communication with patients regarding this treatment.

Insights From Twitter Conversations on Lupus and Reproductive Health: Protocol for a Content Analysis

Background

Systemic lupus erythematosus (SLE) is the most common form of lupus. It is a chronic autoimmune disease that predominantly affects women of reproductive age, impacting contraception, fertility, and pregnancy. Although clinic-based studies have contributed to an increased understanding of reproductive health care needs of patients with SLE, misinformation abounds and perspectives on reproductive health issues among patients with lupus remain poorly understood. Social networks such as Twitter may serve as a data source for exploring how lupus patients communicate about their health issues, thus adding a dimension to enrich our understanding of communication regarding reproductive health in this unique patient population.

Objective

The objective of this study is to conduct a content analysis of Twitter data published by users in English in the United States from September 1, 2017, to October 31, 2018, in order to examine people’s perspectives on reproductive health among patients with lupus.

Methods

This study will analyze user-generated posts that include keywords related to lupus and reproductive health from Twitter. To access public Twitter user data, we will use Symplur Signals, a health care social media analytics platform. Text classifiers will be used to identify topics in posts. Posts will be classified manually into the a priori and emergent categories. Based on the information available in a user’s Twitter profile (ie, username, description, and profile image), we will further attempt to characterize the user who generated the post. We will use descriptive statistics to analyze the data and identify the most prevalent topics in the Twitter content among patients with lupus.

Results

This study has been funded by the National Center for Advancing Translational Science (NCATS) through their Clinical and Translational Science Awards program. The Institutional Review Board at the University of Southern California approved the study (HS-18-00912). Data extraction and cleaning are complete. We obtained 47,715 Twitter posts containing terms related to “lupus” from users in the United States, published in English between September 1, 2017, and October 31, 2018. We will include 40,885 posts in the analysis, which will be completed in fall 2020. This study was supported by funds from the has been funded by the National Center for Advancing Translational Science (NCATS) through their Clinical and Translational Science Awards program.

Conclusions

The findings from this study will provide pilot data on the use of Twitter among patients with lupus. Our findings will shed light on whether Twitter is a promising data source for learning about reproductive health issues expressed among patients with lupus. The data will also help to determine whether Twitter can serve as a potential outreach platform for raising awareness of lupus and reproductive health and for implementing relevant health interventions.

International Registered Report Identifier (IRRID)

DERR1-10.2196/15623

Eating Disorder Awareness Campaigns: Thematic and Quantitative Analysis Using Twitter

Background

Health awareness initiatives are frequent but their efficacy is a matter of controversy. We have investigated the effect of the Eating Disorder Awareness Week and Wake Up Weight Watchers campaigns on Twitter.

Objective

We aimed to examine whether the Eating Disorder Awareness Week and Wake Up Weight Watchers initiatives increased the volume and dissemination of Twitter conversations related to eating disorders and investigate what content generates the most interest on Twitter.

Methods

Over a period of 12 consecutive days in 2018, we collected tweets containing the hashtag #wakeupweightwatchers and hashtags related to Eating Disorder Awareness Week (#eatingdisorderawarenessweek, #eatingdisorderawareness, or #EDAW), with the hashtag #eatingdisorder as a control. The content of each tweet was rated as medical, testimony, help offer, awareness, pro-ana, or anti-ana. We analyzed the number of retweets and favorites generated, as well as the potential reach and impact of the hashtags and the characteristics of contributors.

Results

The number of #wakeupweightwatchers tweets was higher than that of Eating Disorder Awareness Week and #eatingdisorder tweets (3900, 2056, and 1057, respectively). The content of tweets was significantly different between the hashtags analyzed (P<.001). Medical content was lower in the awareness campaigns. Awareness and help offer content were lower in #wakeupweightwatchers tweets. Retweet and favorite ratios were highest in #wakeupweightwatchers tweets. Eating Disorder Awareness Week achieved the highest impact, and very influential contributors participated.

Conclusions

Both awareness campaigns effectively promoted tweeting about eating disorders. The majority of tweets did not promote any specific preventive or help-seeking behaviors.

Harnessing the power of social media: how can it help in axial spondyloarthritis research?

PURPOSE OF REVIEW

Axial spondyloarthritis (axSpA) is a chronic inflammatory rheumatic disease that is relatively unknown among the general public. Most patients with axSpA are young or middle-aged adults and more likely to use some social media. This review highlights trends in the application of social media and different ways in which these tools do already or may benefit clinical research, delivery of care, and education in rheumatology, particularly in the field of axSpA.

RECENT FINDINGS

This article discusses four areas in the biomedical field that social media has infused with novel ideas: (i) the use of patient-generated health data from social media to learn about their disease experience, (ii) delivering health education and interventions, (iii) recruiting study participants, and (iv) reform, transfer, and disseminate medical education. We conclude with promising studies in rheumatology that have incorporated social media and suggestions for future directions.

SUMMARY

Rheumatologists now have the opportunity to use social media and innovate on many aspects of their practice. We propose further exploration of multiple ways in which social media might help with the identification, diagnosis, education, and research study enrollment of axSpA patients. However, standardization in study design, reporting, and managing ethical and regulatory aspects will be required to take full advantage of this opportunity.

Infodemiology and Infoveillance: Scoping Review

Background

Web-based sources are increasingly employed in the analysis, detection, and forecasting of diseases and epidemics, and in predicting human behavior toward several health topics. This use of the internet has come to be known as infodemiology, a concept introduced by Gunther Eysenbach. Infodemiology and infoveillance studies use web-based data and have become an integral part of health informatics research over the past decade.

Objective

The aim of this paper is to provide a scoping review of the state-of-the-art in infodemiology along with the background and history of the concept, to identify sources and health categories and topics, to elaborate on the validity of the employed methods, and to discuss the gaps identified in current research.

Methods

The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed to extract the publications that fall under the umbrella of infodemiology and infoveillance from the JMIR, PubMed, and Scopus databases. A total of 338 documents were extracted for assessment.

Results

Of the 338 studies, the vast majority (n=282, 83.4%) were published with JMIR Publications. The Journal of Medical Internet Research features almost half of the publications (n=168, 49.7%), and JMIR Public Health and Surveillance has more than one-fifth of the examined studies (n=74, 21.9%). The interest in the subject has been increasing every year, with 2018 featuring more than one-fourth of the total publications (n=89, 26.3%), and the publications in 2017 and 2018 combined accounted for more than half (n=171, 50.6%) of the total number of publications in the last decade. The most popular source was Twitter with 45.0% (n=152), followed by Google with 24.6% (n=83), websites and platforms with 13.9% (n=47), blogs and forums with 10.1% (n=34), Facebook with 8.9% (n=30), and other search engines with 5.6% (n=19). As for the subjects examined, conditions and diseases with 17.2% (n=58) and epidemics and outbreaks with 15.7% (n=53) were the most popular categories identified in this review, followed by health care (n=39, 11.5%), drugs (n=40, 10.4%), and smoking and alcohol (n=29, 8.6%).

Conclusions

The field of infodemiology is becoming increasingly popular, employing innovative methods and approaches for health assessment. The use of web-based sources, which provide us with information that would not be accessible otherwise and tackles the issues arising from the time-consuming traditional methods, shows that infodemiology plays an important role in health informatics research.

Machine Learning and Natural Language Processing for Geolocation-Centric Monitoring and Characterization of Opioid-Related Social Media Chatter

IMPORTANCE

Automatic curation of consumer-generated, opioid-related social media big data may enable real-time monitoring of the opioid epidemic in the United States.

OBJECTIVE

To develop and validate an automatic text-processing pipeline for geospatial and temporal analysis of opioid-mentioning social media chatter.

DESIGN, SETTING, AND PARTICIPANTS

This cross-sectional, population-based study was conducted from December 1, 2017, to August 31, 2019, and used more than 3 years of publicly available social media posts on Twitter, dated from January 1, 2012, to October 31, 2015, that were geolocated in Pennsylvania. Opioid-mentioning tweets were extracted using prescription and illicit opioid names, including street names and misspellings. Social media posts (tweets) (n = 9006) were manually categorized into 4 classes, and training and evaluation of several machine learning algorithms were performed. Temporal and geospatial patterns were analyzed with the best-performing classifier on unlabeled data.

MAIN OUTCOMES AND MEASURES

Pearson and Spearman correlations of county- and substate-level abuse-indicating tweet rates with opioid overdose death rates from the Centers for Disease Control and Prevention WONDER database and with 4 metrics from the National Survey on Drug Use and Health for 3 years were calculated. Classifier performances were measured through microaveraged F1 scores (harmonic mean of precision and recall) or accuracies and 95% CIs.

RESULTS

A total of 9006 social media posts were annotated, of which 1748 (19.4%) were related to abuse, 2001 (22.2%) were related to information, 4830 (53.6%) were unrelated, and 427 (4.7%) were not in the English language. Yearly rates of abuse-indicating social media post showed statistically significant correlation with county-level opioid-related overdose death rates (n = 75) for 3 years (Pearson r = 0.451, P < .001; Spearman r = 0.331, P = .004). Abuse-indicating tweet rates showed consistent correlations with 4 NSDUH metrics (n = 13) associated with nonmedical prescription opioid use (Pearson r = 0.683, P = .01; Spearman r = 0.346, P = .25), illicit drug use (Pearson r = 0.850, P < .001; Spearman r = 0.341, P = .25), illicit drug dependence (Pearson r = 0.937, P < .001; Spearman r = 0.495, P = .09), and illicit drug dependence or abuse (Pearson r = 0.935, P < .001; Spearman r = 0.401, P = .17) over the same 3-year period, although the tests lacked power to demonstrate statistical significance. A classification approach involving an ensemble of classifiers produced the best performance in accuracy or microaveraged F1 score (0.726; 95% CI, 0.708-0.743).

CONCLUSIONS AND RELEVANCE

The correlations obtained in this study suggest that a social media-based approach reliant on supervised machine learning may be suitable for geolocation-centric monitoring of the US opioid epidemic in near real time.