Review

What social media told us in the time of COVID-19: a scoping

review
Shu-Feng Tsao, Helen Chen, Therese Tisseverasinghe, Yang Yang, Lianghua Li, Zahid A Butt

With the onset of the COVID-19 pandemic, social media has rapidly become a crucial communication tool for Lancet Digit Health 2021;
information generation, dissemination, and consumption. In this scoping review, we selected and examined peer- 3: e175–94

reviewed empirical studies relating to COVID-19 and social media during the first outbreak from November, 2019, to Published Online
January 28, 2021
November, 2020. From an analysis of 81 studies, we identified five overarching public health themes concerning the
https://doi.org/10.1016/
role of online social media platforms and COVID-19. These themes focused on: surveying public attitudes, identifying S2589-7500(20)30315-0
infodemics, assessing mental health, detecting or predicting COVID-19 cases, analysing government responses to the School of Public Health and
pandemic, and evaluating quality of health information in prevention education videos. Furthermore, our Review Health Systems (S-F Tsao MSc,
emphasises the paucity of studies on the application of machine learning on data from COVID-19-related social Prof H Chen PhD, Y Yang MSc,
Z A Butt PhD) and Faculty of
media and a scarcity of studies documenting real-time surveillance that was developed with data from social media on
Science (L Li BSc), University of
COVID-19. For COVID-19, social media can have a crucial role in disseminating health information and tackling Waterloo, Waterloo, ON,
infodemics and misinformation. Canada; Seneca Libraries,
Seneca College, King City, ON,
Introduction perception of disease exposure, resultant decision making, Canada (T Tisseverasinghe MLIS)
Correspondence to:
Severe acute respiratory syndrome coronavirus 2 and risk behaviours.9,10 As information on social media is
Dr Zahid A Butt, School of Public
(SARS-CoV-2), and the resulting COVID-19, is a substan­ generated by users, such information can be subjective or Health and Health Systems,
tial international public health issue. As of Jan 18, 2021, inaccurate, and frequently includes misinformation and University of Waterloo, Waterloo,
an estimated 95 million people worldwide had been conspiracy theories.11 Hence, it is imperative that accurate ON N2L 3G1, Canada
[email protected]
infected with the virus, with about 2 million deaths.1 As a and timely information is disseminated to the general
consequence of the pandemic, social media is becoming public about emerging threats, such as SARS-CoV-2. A
the platform of choice for public opinions, perceptions, systematic review explored the major approaches that were
and atti­tudes towards various events or public health used in published research on social media and emerging
policies regarding COVID-19.2 Social media has become infectious diseases.12 The review identified three major
a pivotal communication tool for governments, organ­ approaches: assessment of the public’s interest in, and
isations, and universities to dissem­ inate crucial infor­ responses to, emerging infectious diseases; examination
mation to the public. Numerous studies have already of organisations’ use of social media in communicating
used social media data to help to identify and detect emerging infectious diseases; and evaluation of the
outbreaks of infectious diseases and to interpret public accuracy of medical information that is related to emerging
attitudes, behaviours, and perceptions.3–6 Social media, infectious diseases on social media. However, this review
particularly Twitter, can be used to explore multiple facets did not focus on studies that used social media data to
of public health research. A systematic review identified track and predict outbreaks of emerging infectious
six categories of Twitter use for health research, namely diseases.
content analysis, surveillance, engage­ment, recruitment, Analysing and disseminating information from peer-
as part of an intervention, and network analysis of Twitter reviewed, published research can guide policy makers
users.5 However, this review included only broader and public health agencies to design interventions for
research terms, such as health, medicine, or disease, by accurate and timely knowledge translation to the general
use of Twitter data and did not focus on specific disease public. Therefore, keeping in view the limitations of
topics, such as COVID-19. Another article analysed tweets existing research that we have previously mentioned, we
on COVID-19 and identified 12 topics that were did a scoping review with the aim of understanding the
categorised into four main themes: the origin, source, roles that social media has had since the beginning of
effects on individuals and countries, and methods of the COVID-19 crisis. We investigated public attitudes
decreasing the spread of SARS-CoV-2.7 In this study, data and perceptions towards COVID-19 on social media,
were not available for tweets that were related to information about COVID-19 on social media, use of
COVID-19 before February, 2020, thereby missing the social media for prediction and detection of COVID-19,
initial part of the epidemic, and the data for tweets were the effects of COVID-19 on mental health, and
limited to between Feb 2 and March 15, 2020. government responses to COVID-19 on social media.
Social media can also be effectively used to communicate Our objective was to identify and analyse studies on
health information to the general public during a pan­ social media that were related to COVID-19 and focused
demic. Emerging infectious diseases, such as COVID-19, on five themes: infodemics, public attitudes, mental
almost always result in increased usage and consumption health, detection or prediction of COVID-19 cases,
of media of all forms by the general public for information.8 government responses to the pandemic, and quality of
Therefore, social media has a crucial role in people’s health information in videos.

www.thelancet.com/digital-health Vol 3 March 2021 e175

 Review

Panel: Variations of keywords and indexed terms that were used in the literature search 2405 articles identified from database search
1084 from PubMed
Keywords 1021 from Scopus
“COVID-19” 300 from PsycINFO

• “Betacoronavirus”, “severe acute respiratory syndrome”, “covid 2019”, “COVID-19”,

“corona-19”, “n-cov”, “novel coronavirus”, “sars-cov”, or “wuhan 2019” 670 duplicates removed
“Social media”
• “Twitter”, “tweet”, “retweet”, “facebook”, “weibo”, “sina”, “youtube”, “webcast”, “user 1735 articles eligible for title and abstract
comment”, “online post”, “online discussion”, “social network”, “social media”, “online
community”, or “mobile app”
1434 non-empirical articles excluded
Indexed terms
“COVID-19”
• “Betacoronavirus”, “coronavirus infections”, “severe acute respiratory syndrome”, 301 articles qualified for full-text review

“coronavirus disease 2019”, “COVID-19”, or “covid-19”

“Social media” 220 articles excluded
117 excluded for not directly using social
• “Social networking”, “social media”, “mobile applications”, “blogging”, “social media data
networking (online)”, “online social network”, “webcast”, “mobile applications”, 103 excluded for reporting only
“mobile computing”, or “social network” development of advanced
mathematical models by use of fuzzy
logic, simulations, etc

Methods
Overview 81 articles selected
Studies exploring the use of social media relating to
COVID-19 were reviewed by use of the scoping review Figure 1: Preferred Reporting Items for Systematic Reviews and Meta-
methods of Arksey and O’Malley13 and Levac and Analyses flow chart of article extraction from the literature search
colleagues.14 We followed the five-step scoping review
protocol and the Preferred Reporting Items for Systematic the majority of paragraph codes. Next, quotes were
Reviews and Meta-Analyses extension for scoping reviews. sorted under each code, applying Ose’s method.15 Braun
and Clark’s thematic analysis method was used and
Data Sources involved searching for the text that matched the
Exploratory searches were done on COVID-19 Open identified predictors (ie, codes) from the quantitative
Research Dataset Challenge and Google Scholar in analysis and discovering emergent codes that were
April, 2020. These searches helped to define the Review relevant to either the study objective or identified in the
scope, develop the research questions, and determine relevant literature review.16 Finally, we categorised the
eligibility criteria. After such activity, MEDLINE and codes into main themes. These codes and themes were
PubMed, Scopus, and PsycINFO were selected for this compared and clarified by S-FT, ZAB, and YY to draw
Review because they include peer-reviewed literature in conclusions around the main themes. S-FT is fluent in
the fields of medicine, behavioural sciences, psychology, English and Mandarin. The secondary reviewer (ZAB)
health-care systems, and clinical sciences. Variations of is fluent in English, and the tertiary reviewer and
the key search terms can be found in the panel. Since the domain expert (YY and HC) are both fluent in English
start of the current pandemic, COVID-19 articles were and Mandarin. Any discrepancies among reviewers
reviewed and published at an unprecedently rapid rate, were discussed with the research team to reach
with numerous publications that were available ahead consensus.
of print referred to as preprints or articles in press. In
this Review, we consider peer-reviewed preprints to be Results
equivalent to published peer-reviewed articles, and With the application of appropriate search filters, a
relevant articles were screened accordingly. total of 2405 articles were retrieved from the identified
databases: PubMed (1084 articles), Scopus (1021 articles),
Screening procedure and PsycINFO (300 articles). Among these, 670 duplicates
Mainly, the primary reviewer (S-FT) screened title and were excluded. Of the remaining 1735 articles, 1434 were
abstract for each article to decide whether an article met deemed to be non-empirical, such as comments, editorial
the inclusion criteria. If the criteria were confirmed, essays, letters, opinions, and reviews. These exclusions
then the article was included; otherwise, it was left 301 articles for a full-text review on the basis of the
excluded. Paragraphs in articles were assigned a code screening results of titles and abstracts. After the full-text
representing one of the five themes (eg, I for infodemic), review, 81 articles were included in this scoping review
then a code was assigned to the article on the basis of (figure 1).

e176 www.thelancet.com/digital-health Vol 3 March 2021

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
Detection or prediction of COVID-19 cases
Li et al17 March China Google Trends, Keywords of coronavirus and Lag correlation Lag correlations showed a maximum correlation between
Baidu Search pneumonia were searched, and trend trend data and the number of diagnoses at 8–12 days
Index, and Sina data was collected from before for laboratory-confirmed cases and 6–8 days
Weibo Index Google Trends, Baidu Search Index, before for suspected cases
and Sina Weibo Index from
Jan 2 to Feb 20, 2020
Liu et al18 August China Sina Weibo Sina Weibo messages between Jan 20 Gathered data via Sina Older age (ie, >69 years), diffuse pneumonia, and
and Feb 15, 2020; 599 participants Weibo, then followed up hypoxaemia are factors that can help clinicians to identify
with telephone call; patients with COVID-19 who have poor prognosis;
statistical analysis taken with aggregated data from social media can also be
Fisher exact test; rates of comprehensive, immediate, and informative in disease
death calculated with prognosis
Kaplan-Meier method;
multivariate Cox regression
used to establish risk factors
for mortality
O’Leary and September USA Google Trends, Google Trends searches for coronavirus Regression analysis To model the number of cases, the current Wikipedia page
Storey19 Wikipedia, and and COVID-19 between Jan 21 and views, tweets from 1 week before, and Google Trends
Twitter April 5, 2020; Wikipedia page views for searches from 2 weeks before were used; to model of the
coronavirus and COVID-19 between number of deaths, each variable was taken from 1 week
Jan 12 and April 5, 2020; number of earlier than for cases
Twitter original tweets between
Jan 27 and April 5, 2020; numbers of
COVID-19 cases and deaths in the USA20
Peng et al21 June China Sina Weibo 1200 records Spatiotemporal distribution Older people (ie, >60 years) are at high risk of severe
of COVID-19 cases in the symptoms and have high prevalence in the COVID-19
main urban area of Wuhan, outbreak, and they account for >50% of the total number
China; kernel density of Sina Weibo help seekers; early transmission of
analysis; ordinary least COVID-19 in Wuhan, China, could be divided into three
square regression phrases: scattered infection, community spread, and full-
scale outbreak
Qin et al22 March China Baidu Search Social media search index for dry Subset selection; forward Case numbers of new suspected COVID-19 correlated
Index cough, fever, chest distress, selection; lasso regression; significantly with the lagged series of social media search
coronavirus, and pneumonia from ridge regression; elastic net index; social media search index could detect new
Dec 31, 2019, to Feb 9, 2020; data for suspected COVID-19 cases 6–9 days earlier than could
new suspected cases of COVID-19 laboratories
from Jan 20 to Feb 9, 2020
Zhu et al23 April China Sina Weibo 1101 Sina Weibo posts related to Descriptive statistics: Attention to COVID-19 was low until China openly
COVID-19 from Dec 31, 2019, to numbers and percentage; admitted human-to-human transmission on
Feb 12, 2020 time series analysis Jan 20, 2020; attention quickly increased and remained
high over time
Government responses
Basch et al24 April USA YouTube 100 most widely viewed videos Descriptive analysis: Percentage of each of the seven key prevention
uploaded in January, 2020 frequency, percentage, behaviours that are listed on the US Centers for Disease
mean, and standard Control and Prevention website that were covered in the
deviation 100 videos varied from 0% (eg, use a face mask for
protection if you are caring for the ill) to 31% (avoid close
contact with people who are sick); overall, videos that
covered at least one prevention behaviour accounted for
less than one-third of the 100 videos
Basch et al25 April USA YouTube 100 most widely viewed YouTube Descriptive analysis: <50% of videos in either sample covered any of the
videos as of Jan 31, 2020, and frequency, percentage, prevention behaviours that are recommended by the
March 20, 2020, with keyword of mean, and standard US Centers for Disease Control and Prevention
coronavirus in English, with English deviation
subtitles, or in Spanish
Khatri et al26 March Singapore Youtube 150 videos collected on Descriptive analysis: Mean DISCERN score for reliability was 3·12 of 5·00 for
Feb 1–2, 2020, with keywords of percentage and mean; English and 3·25 of 5·00 for Mandarin videos; mean
2019 novel coronavirus (50 videos), DISCERN score; Medical cumulative Medical Information and Content Index score
and Wuhan virus in English Information and Content of useful videos was 6·71 of 25·00 for English and 6·28 of
(50 videos) and Mandarin (50 videos) Index score 25·00 for Mandarin
(Table continues on next page)

www.thelancet.com/digital-health Vol 3 March 2021 e177

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Li et al27 March China Sina Weibo 36 746 Sina Weibo data from Linear regression; support Classified the information related to COVID-19 into seven
Dec 30, 2019, to Feb 1, 2020; a vector machine; Naive Bayes; types of situational information and their predictors
random sample of 3000 Sina Weibo natural language processing
posts as training dataset
Merkley et al28 April Canada Twitter and 33 142 tweets from 292 social media Linear regression No members of parliament from any party downplaying
Google Trends accounts of federal members of the pandemic; no association between Conservative Party
parliament from Jan 1 to vote share and Google search interest in the coronavirus
March 28, 2020; 87 Google search
trends for the search term coronavirus
in the first half (ie, days 1–14) and
second half (ie, days 15–31) of
March, 2020; a survey of
2499 Canadian citizens ≥18 years
from April 2 to April 6, 2020
Rufai and April USA Twitter 203 viral tweets from G7 world Qualitative design; content 166 of 203 of tweets were informative; 9·4% (19) were
Bunce29 leaders from Nov 17, 2019, to analysis morale-boosting; 6·9% (14) were political
March 17, 2020 with keywords
COVID-19 or coronavirus and a
minimum of 500 likes
Sutton et al30 September USA Twitter 690 accounts representing public χ² analyses; negative Systematic changes were made in message strategies over
health, emergency management, and binomial regression time and identified key features that affect message
elected officials and 149 335 tweets modelling passing, both positively and negatively; results have the
potential to aid in message design strategies as the
pandemic continues, or in similar future events
Wang et al31 September USA Twitter 13 598 tweets related to COVID-19 Temporal analysis and 16 categories of message types were manually annotated;
from Jan 1 to April 27, 2020 networking analysis inconsistencies and incongruencies were identified in four
critical topics (ie, wearing masks, assessment of risks, stay
at home order, and disinfectant and sanitiser); network
analysis showed increased communication coordination
over time
Infodemics
Ahmed et al32 October UK Twitter 22 785 tweets and 11 333 Twitter Social network analysis; The most important drivers of the #FilmYourHospital
users with #FilmYourHospital from user analysis conspiracy theory are ordinary citizens; YouTube was the
April 13 to April 20, 2020 information source most linked to by users; the most
retweeted post belonged to a verified Twitter user
Ahmed et al33 May UK Twitter A subsample of 233 tweets from Descriptive statistics: 34·8% (81 of 233) of tweets linked 5G and COVID-19;
10 140 tweets collected from 19:44 h numbers, percentage; social 32·2% (75) of tweets denounced the conspiracy theory
UTC on Friday, March 27, 2020, to network analysis; content
10:38 h UTC on Saturday, analysis
April 4, 2020, were used for content
analysis
Brennen et al34 October UK Digital visual 96 samples of visuals from January to Qualitative coding Organised all findings into six trends: authoritative
media March, 2020 agency, virulence, medical efficacy, intolerance, prophecy,
satire; a small number of manipulated visuals, all were
produced by use of simple tools; no examples of so-called
deepfakes (ie, techniques that are used to make synthetic
videos that closely resemble real videos) or other
techniques that were based on artificial intelligence
Bruns et al35 August Australia Facebook 89 664 distinct Facebook posts from Time series; network analysis Substantially increased number of posts about 5G
Jan 1 to April 12, 2020 rumours on Facebook after March 19, 2020; network
analysis showed that coalitions of various groups were
brought together by conspiracy theories about COVID-19
and 5G technology
Galhardi et al36 October Brazil WhatsApp, Fake news collected from March 17 to Quantitative content WhatsApp is the main channel for sharing fake news,
Instagram, April 10, 2020, on the basis of data analysis followed by Instagram and Facebook
and Facebook from the Eu Fiscalizo app
(version 5.0.5)
Gallotti et al37 October Italy Twitter >100 million Tweets Developed an Infodemic Risk Before the rise of COVID-19 cases, entire countries had
Index measurable waves of potentially unreliable information,
posing a serious threat to public health
(Table continues on next page)

e178 www.thelancet.com/digital-health Vol 3 March 2021

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Islam et al38 October Bangladesh Fact-checking 2311 infodemic reports related to Descriptive analysis; spatial Misinformation that is fuelled by rumours, stigma, and
agency COVID-19 between Dec 31, 2019, and distribution analysis conspiracy theories can have potentially severe
websites, April 5, 2020 implications on public health if prioritised over scientific
Facebook, guidelines; governments and other agencies should
Twitter, and understand the patterns of rumours, stigma, and
websites for conspiracy theories that are related to COVID-19 and
television circulating globally so that they can develop appropriate
networks and messages for risk communication
newspapers
Kouzy et al39 March Lebanon Twitter 673 English tweets collected on Descriptive statistics; bar 153 (24·8%) of 617 tweets had misinformation;
Feb 27, 2020; 617 tweets after chart; χ² statistic to calculate 107 (17·3%) had unverifiable information;
exclusion of tweets that were p value (2-sided; p=0·05 misinformation rate higher in informal individual or
humorous or not serious significance threshold) for the group accounts than in formal individual or group
association between account accounts (33·8% [123 of 364] vs 15·0% [30 of 200],
or tweet characteristics and p<0·0010)
the presence of
misinformation or
unverifiable information
about COVID-19
Moscadelli August Italy Fake news and 2102 articles between Dec 31, 2019, Social media trend analysis Links containing fake news were shared 2 352 585 times,
et al40 corresponding and April 30, 2020 by use of BuzzSumo accounting for 23·1% (2 352 585 of 10 184 351) of total
verified news shares of all reviewed articles
that was
circulated in
Italy
Pulido et al41 April Spain Twitter 942 valid tweets between Feb 6 and Communicative content Misinformation was tweeted more but retweeted less
Feb 7, 2020 analysis than tweets based on scientific evidence; tweets based on
scientific evidence had more engagement than
misinformation
Rovetta and August Italy Google Trends 2 million Google Trends queries and Classification of infodemic Globally, growing interest exists in COVID-19, and
Bhagavathula42 and Instagram Instagram hashtags from Feb 20 to monikers (ie, a term, query, numerous infodemic monikers continue to circulate on
May 6, 2020 hashtag, or phrase that the internet
generates or feeds fake news,
misinterpretations, or
discrimination); computed
the mean peak volume with
a 95% CI
Uyheng and October USA and Twitter 12·0 million tweets from 1·6 million Hate speech score assigned Analysis showed idiosyncratic relationships between bots
Carley43 Philippines users from the USA and 15·0 million to each tweet by use of and hate speech across datasets, emphasising different
tweets from 1·0 million users from machine learning algorithm; network dynamics of racially charged toxicity in the USA
the Philippines between March 5 and bot scores were assigned to and political conflicts in the Philippines; bot activity is
March 19, 2020 each user via BotHunter linked to hate in both countries, especially in
algorithm; social media communities that are dense and isolated from others
analysis via ORA software;
network analysis via
centrality analysis; cluster
analysis via Leiden algorithm
Mental health
Gao et al44 April China Sina Weibo Online survey on Wenjuanxing Multivariable logistic Social media exposure was frequently positively
platform from Jan 31 to Feb 2, 2020; regression associated with high odds of anxiety (odds ratio 1·72,
with 4872 Chinese citizens aged 95% CI 1·31–2·26) and combination of depression and
≥18 years from 31 provinces and anxiety (odds ratio 1·91, 95% CI 1·52–2·41)
autonomous regions in China
Li et al45 March China Sina Weibo Sina Weibo posts from 17 865 active Sentiment analysis; paired Negative emotions and sensitivity to social risks
Sina Weibo users between Jan 13 and sample t-test increased; scores of positive emotions and life satisfaction
Jan 26, 2020 decreased after outbreak declaration
(Table continues on next page)

The table summarises the 81 articles that were selected novel framework called Social Media and Public Health
on COVID-19 and social media. All articles were written Epidemic and Response (SPHERE) and developed a
in English. Data from Twitter (45 articles) and Sina Weibo modified version of SPHERE framework to organise the
(16 articles) were undoubtedly the most frequently themes for our scoping review (figure 2).98 Themes were
studied. To categorise these chosen articles, we adopted a identified through reviewers’ consensus based on our

www.thelancet.com/digital-health Vol 3 March 2021 e179

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Prevention education in videos
Hakimi and September USA YouTube 49 of the first 100 videos on YouTube Codified video content; Most videos did not describe labelling storage containers,
Armstrong46 with the most views that were assessed by use of Cohen’s κ; 69% (34 of 49) of videos encouraged the use of oils or
identified by the search term DIY descriptive statistics perfumes to enhance hand sanitiser scent, and 2% (1) of
hand sanitiser; 51 videos were calculated; assessed by χ² test videos promoted the use of colouring agents to be more
excluded because they were not in with 2-sided p value <0·05 as attractive for use among children specifically; significantly
English or not related to the search the threshold for significance increased mean number of daily calls to poison control
term centres regarding unsafe paediatric exposure to hand
sanitiser since the first confirmed patient with COVID-19 in
the USA (p<0·0010); significantly increased mean number
of daily calls in March, 2020, compared with the previous
2 years (p<0·0010)
Hernández- June Spain YouTube 129 videos in Spanish with the terms Univariate analysis; multiple Information from YouTube in Spanish on basic measures
García and prevencion coronavirus and logistic regression model to prevent COVID-19 is usually not complete and differs
Giménez- prevencion COVID19 according to the type of authorship (ie, mass media,
Júlvez47 health professionals, individual users, or others)
Moon and August South YouTube 105 most viewed YouTube videos Modified DISCERN index; 37·14% (39 of 105) of videos contained misleading
Lee48 Korea from Jan 1 to April 30, 2020 Journal of the American Medical information; independent user-generated videos showed
Association Score benchmark the highest proportion of misleading information
criteria; Global Quality Score; at 68·09% (32 of 47); misleading videos had more likes,
Title–Content Consistency fewer comments, and longer running times than did
Index; Medical Information useful videos; transmission and precautionary measures
and Content Index were the most frequently covered content
Ozdede and July–August Turkey YouTube The top 116 English language videos Precision indices and total High number of views on dentistry YouTube videos
Peker49 with at least 300 views video information and related to COVID-19; quality and usefulness of these
quality index scores were videos are moderate
calculated
Yüce et al50 July Turkey YouTube 55 English videos about COVID-19 Modified DISCERN Only two (3·6%) of 55 videos were good quality, whereas
control procedures for dental instrument; descriptive 24 (43·6%) videos were poor quality
practices collected on March 31, 2020, statistics
between 9:00 h and 18:00 h
Public attitudes
Abd-Alrazaq April Qatar Twitter 2·8 million English tweets Word frequencies of single Identified 12 topics and grouped into four themes;
et al7 (167 073 unique tweets from (ie, unigrams) and double average sentiment positive for ten topics and negative for
160 829 unique users) from Feb 2 to words (ie, bigrams); two topics
March 15, 2020 sentiment analysis; mean
number of retweets, likes,
and followers for each topic;
interaction rate per topic;
LDA for topic modelling
Al-Rawi et al51 November Canada Twitter Over 50 million tweets referencing Mixed method: analysed Identified five major themes in the analysis: morbidity fears,
#Covid-19 and #Covid19 for more emoji use by each gender health concerns, employment and financial issues, praise
than 2 months in early 2020 category; the top 600 emojis for front-line workers, and unique gendered emoji use;
were manually classified on most emojis are extremely positive across genders, but
the basis of their sentiment discussions by women and gender minorities are more
negative than by men; when discussing particular topics
(eg, financial and employment matters, gratitude, and
health care), there are many differences; use of several
unique gender emojis to express specific issues (eg, coffin,
skull, and siren emojis were used more often by men than
by other genders when discussing fears and morbidity,
whereas the use of the folded hands emoji as a thankful
gesture for front-line workers was found more often in
discussions by women than by other genders and the bank
emoji was noted only in women’s discussions)
(Table continues on next page)

modified SPHERE framework. We identified six themes: Social media as contagion and vector
infodemics, public attitudes, mental health, detecting or According to WHO, the term infodemic, a combination
predicting COVID-19 cases, government responses, and of information and epidemic, refers to a fast and
quality of health information in prevention education widespread dissemination of both accurate and
videos. inaccurate information about an epidemic, such as

e180 www.thelancet.com/digital-health Vol 3 March 2021

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Arpaci et al52 July Turkey Twitter 43 million tweets between March 22 Evolutionary clustering Unigram terms appear more frequently than bigram and
and March 30, 2020 analysis trigram (ie, triple words) terms; during the epidemic,
many tweets about COVID-19 were distributed and
attracted widespread public attention; high-frequency
words (eg, death, test, spread, and lockdown) indicated
that people were afraid of being infected and people who
were infected were afraid of death; people agreed to stay
at home due to fear of spread and called for physical
distancing since they became aware of COVID-19
Barrett et al53 August USA Twitter 188 tweets about Thematic analysis 90% (169 of 188) of tweets opposed calculated ageism,
Governor Dan Patrick’s statement on whereas only 5% (9) supported it and 5% (10) conveyed no
March 23, 2020, about generational position; opposition centred on moral critiques, political–
self-sacrifice. economic critiques, assertions of the worth of older adults
(eg, >60 years), and public health arguments; support
centred on individual responsibility and patriotism
Boon-Itt and November Thailand Twitter 107 990 English tweets related to Sentiment analysis; topic Sentiment analysis showed a predominantly negative
Skunkan54 COVID-19 between Dec 13, 2019, modelling by use of LDA feeling towards the COVID-19 pandemic; topic modelling
and March 9, 2020 revealed three themes relating to COVID-19 and the
outbreak: the COVID-19 pandemic emergency, how to
control COVID-19, and reports on COVID-19
Budhwani and May USA Twitter 16 535 tweets about Chinese virus or Descriptive analysis; spatial Nearly 10 times increase at the national level; all
Sun55 China virus between March 9 and analysis 50 states had an increase in the number of tweets
March 15, 2020, 177 327 tweets exclusively mentioning Chinese virus or China virus
between March 19 and instead of coronavirus disease, COVID-19, or
March 25, 2020 coronavirus; mean 0·38 tweets referencing Chinese virus
or China virus were posted per 10 000 people at the
state level in the preperiod (ie, March 9–15, 2020), and
4·08 of these stigmatising tweets were posted in the
postperiod (ie, March 19–25, 2020), also indicating a
10 times increase
Chang et al56 November Taiwan 10 news 1·07 million Chinese texts from Deductive analysis Online news promoted negativity and drove emotional
websites, Dec 30, 2019, to March 31, 2020 social posts; stigmatising language that was linked to the
11 discussion COVID-19 pandemic showed an absence of civic
forums, responsibility that encouraged bias, hostility, and
1 social discrimination
network,
2 principal
media sharing
networks
Chehal et al57 July India Twitter 29 554 tweets during the second Sentiment analysis by use of A positive approach in the second lockdown but a
lockdown (ie, April 15–May 3, 2020); the National Research negative approach in the third lockdown
47 672 tweets during the third Council of Canada Emotion
lockdown (May 4–17, 2020) Lexicon
Chen et al58 September China Sina Weibo 1411 posts pertinent to COVID-19 Descriptive analysis; Media richness (ie, potential information load, where low
taken from Healthy China, an official hypothesis testing richness is only text and high richness is not only text)
Sina Weibo account of the National negatively predicted citizen participation via government
Health Commission of China, from social media, but dialogic loop (ie, stimulation of public
Jan 14 to March 5, 2020 dialogue, provision of the dialogue channel, and response
to public feedback in a timely manner) facilitated
engagement
Damiano and August USA Twitter 600 English tweets from the USA Frequencies; χ² statistics Neutral sentiment; tweets about COVID-19 risks and
Allen were selected: 300 from emotional outrage accounted for <50% (135 of 600);
Catellier59 February, 2020, and 300 from few tweets were related to blame
March, 2020
Darling- September USA Twitter 339 063 tweets from non-Asian Local polynomial regression; Implicit Americanness Bias steadily decreased from
Hammond respondents of the Project Implicit interrupted time-series 2007 to 2020; when media entities began using
et al60 Asian Implicit Association Test from analyses stigmatising terms, such as Chinese virus, starting from
2007–20 and were broken into March 8, 2020, Implicit Americanness Bias began to
two datasets: the first dataset was increase; such bias was more pronounced among
from Jan 1, 2007, to Feb 10, 2020; conservative individuals than among non-conservative
the second data set was from individuals
Feb 11 to March 31, 2020
(Table continues on next page)

www.thelancet.com/digital-health Vol 3 March 2021 e181

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Das and July India Twitter 410 643 tweets with #IndiaLockdown National Research Council of For the broad corpus-level analysis, the context of
Dutta61 and #IndiafightsCorona from Canada lexicon for corpus- positiveness was substantially higher than were negative
March 22 to April 21, 2020 level emotion mining; sentiments; however, positive sentiment trends were
sentimentr from open source similar to negative sentiment trends in terms of topics
R software for sentiment covered when the analysis was done at individual tweet
analysis to create additional level; the results showed that the discussion of COVID-19
sentiment scores; LDA for in India on Twitter contains slightly more positive
topic models; Natural sentiments than negative sentiments
Language Toolkit to develop
sentiment-based topic models
De Santis July Italy Twitter 1 044 645 tweets A general purpose Energy evolution through time was monitored; daily hot
et al62 methodological framework, topics were identified (eg, COVID-19, Walter Ricciardi’s
grounded on a biological retweet of an anti-Trump tweet from Michael Moore,
metaphor and on a chain of Gabriele Gravina’s argument against suspension of Italian
NLP and graph analysis football, increased COVID-19 cases in Italy, high case
techniques numbers in Lombardy, Italy, and an interview of
Matteo Salvini about COVID-19 topics by Massimo Giletti)
Dheeraj63 May–June India Reddit 868 posts related to COVID-19 Fetching the articles: Python Of 868 posts on Reddit that were related to COVID-19
Reddit Application articles, 50% (434) were neutral, 22% (191) were positive,
Programming Interface and 28% (243) were negative
Wrapper; data preprocessing:
Reddit Application
Programming Interface and
Natural Language Toolkit
library
Essam and August Egypt Twitter 1 920 593 tweets with corona, Thematic analysis The dominant themes that were closely related to
Abdo64 coronavirus, or COVID-19 keywords coronavirus tweets included the outbreak of the pandemic,
from Feb 1 to April 30, 2020 metaphysics responses, signs and symptoms in confirmed
cases, and conspiracies; the psycholinguistic analysis
showed that tweeters maintained high amounts of
affective talk (ie, expression of feelings), which was loaded
with negative emotions and sadness; Linguistic Inquiry
and Word Count’s psychological categories of religion and
health dominated the Arabic tweets discussing the
pandemic situation
Yin FL et al65 March China Sina Weibo Sina Weibo posts from Dec 31, 2019, Multiple-information Model reproduction ratio declined from 1·78 to 0·97,
to Feb 7, 2020 susceptible-discussing- showing that the peak of posts had passed but the topic
immune model was still on social media afterwards with a decreased
number of posts
Gozzi et al66 October Italy, UK, News, 227 768 web-based news articles Linear regression; topic Collective attention was mainly driven by media coverage
USA, and YouTube, from Feb 7 to May 15, 2020; modelling by use of LDA rather than epidemic progression, rapidly became
Canada Reddit, and 13 448 YouTube videos from Feb 7 to saturated, and decreased despite media coverage and
Wikipedia May 15, 2020; 107 898 English user COVID-19 incidence remaining high; Reddit users were
posts and 3 829 309 comments on generally more interested in health, data regarding the
Reddit from Feb 15 to May 15, 2020; new disease, and interventions needed to halt the
278 456 892 views of Wikipedia spreading with respect to media exposure than were users
pages that were related to COVID-19 of other platforms
from Feb 7 to May 15, 2020
Green et al67 July USA Twitter 19 803 tweets from Democrats and Random forest Democrats discussed the crisis more frequently—
11 084 tweets from Republicans emphasising public health and direct aid to US workers—
between Jan 17 and March 31, 2020 whereas Republicans placed greater emphasis on national
unit, China, and businesses
Han et al68 April China Sina Weibo 1 413 297 Sina Weibo messages, Time series analysis; Public response was sensitive to the epidemic and notable
including 105 330 texts with kernel density estimation; social events, especially in urban agglomerations
geographical location information, Spearman correlation; LDA
from 00:00 h on Jan 9, 2020, to model; random forest
00:00 h on Feb 11, 2020 algorithm
Jelodar et al69 June China Reddit 563 079 English comments related to Topic modelling by use of The results showed a novel application for NLP based on a
COVID-19 from Reddit between LDA and probabilistic latent long short term memory model to detect meaningful
Jan 20 and March 19, 2020 semantic analysis; sentiment latent topics and sentiment–comment classification on
classification by use of issues related to COVID-19 on social media
recurrent neural network
(Table continues on next page)

e182 www.thelancet.com/digital-health Vol 3 March 2021

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Jimenez- April Mexico Twitter A random sample of Qualitative content The most common types of tweets were personal opinions
Sotomayor 351 of 18 128 tweets were analysed classification (31·9% [112 of 351]), followed by informative tweets
et al70 from March 12 to March 21, 2020 (29·6% [104]), jokes or ridicule (14·2% [50]), and personal
accounts (13·4% [47]); 72 of 351 tweets were most likely
intended to ridicule or offend someone and 21·1% (74) had
content implying that the life of older adults (ie, referred to
in tweets as “elderly”, “older”, and “boomer”) was less
valuable than that of younger people or downplayed the
relevance of COVID-19
Kim71 August South Twitter 27 849 individual tweets about Binary logistic regression; Social network size was a negative predictor of incivility
Korea COVID-19 between Feb 10 and semantic network analysis
Feb 14, 2020
Kurten and August Belgium Twitter 373 908 tweets and retweets from Time series; network Notable COVID-19 events immediately increased the
Beullens72 Feb 25 to March 30, 2020 bigrams; emotion lexicon; number of tweets; most topics focused on the need for
LDA EU collaboration to tackle the pandemic
Kwon et al73 October USA Twitter 259 529 unique tweets containing the Trending analysis; Early facets of physical distancing appeared in Los Angeles
word coronavirus between Jan 23 and spatiotemporal analysis (CA, USA), San Francisco (CA, USA), and Seattle (WA, USA);
March 24, 2020 social disruptiveness tweets were most retweeted, and
intervention implementation tweets were most favourited
Lai et al74 October USA Reddit 522 comments from an Ask Me Content analysis The highest number of posts were about symptoms
Anything session on COVID-19 on (27% [141 of 522]), followed by prevention (25% [131]);
March 11, 2020, from 14:00 h to symptoms was the most common intended topic for
16:00 h EST further discussions (28% [94 of 337])
Li et al75 April China Sina Weibo 115 299 Sina Weibo posts from Linear regression model; Positive correlation between the number of Sina Weibo
Dec 23, 2019, to Jan 30, 2020; qualitative content analysis posts and the number of reported cases, with
11 893 of them were collected from ten COVID-19 cases per 40 posts; posts grouped into
Dec 31, 2019, to Jan 20, 2020, for four themes
qualitative analysis; total daily cases
of COVID-19 in Wuhan, China, were
obtained from the Chinese National
Health Commission
Li et al76 September USA Twitter 155 353 unique English tweets related Content analysis Peril of COVID-19 was mentioned the most often,
to COVID-19 that were posted from followed by content about marks (ie, cues to identify
Dec 31, 2019, to March 13, 2020 members of a stigmatised group: flu-like symptoms,
personal protective equipment, Asian origin, and health-
care providers and essential workers), responsibility, and
group labelling; information on conspiracy theories was
more likely to be included in tweets about group labelling
and responsibility than in tweets about COVID-19 peril
Lwin et al77 May Singapore Twitter 20 325 929 tweets from Sentiment analysis Public emotions shifted strongly from fear to anger over
7 033 158 unique users from Jan 28 to the course of the pandemic, while sadness and joy also
April 9, 2020 surfaced; anger shifted from xenophobia at the beginning
of the pandemic to discourse around the stay-at-home
notices; sadness was emphasised by the topics of losing
friends and family members, whereas topics that were
related to joy included words of gratitude and good
health; emotion-driven collective issues around shared
public distress experiences of the COVID-19 pandemic are
developing and include large-scale social isolation and the
loss of human lives
Ma et al78 July China WeChat Top 200 accounts from Jan 21 to Simple linear regression; For non-medical institution accounts in the model, report
Jan 27, 2020 multiple linear regression; and story types of articles had positive effects on whether
content analysis users followed behaviours; for medical institution accounts,
report and science types of articles had a positive effect
Medford et al79 June USA Twitter 126 049 English tweets from Temporal analysis; sentiment The hourly number of tweets that were related to COVID-19
53 196 unique users with matching analysis; topic modelling by starkly increased from Jan 21, 2020, onwards; fear was the
hashtags that were related to use of LDA most common emotion and was expressed in 49·5%
COVID-19 from Jan 14 to Jan 28, 2020 (62 424 of 126 049) of all tweets; the most common
predominant topic was the economic and political effect
Mohamad80 June Brunei Twitter, 30 individual profiles from Instagram, Qualitative content analysis Five narratives of local responses to physical distancing
Instagram, and Twitter, and TikTok practices were apparent: fear, responsibility, annoyance,
TikTok fun, and resistance
(Table continues on next page)

www.thelancet.com/digital-health Vol 3 March 2021 e183

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Nguyen et al81 September USA Twitter 3 377 295 US tweets that were related Support vector machine was Proportion of negative tweets referencing Asians
to race from November, 2019, to used for sentiment analysis increased by 68·4%; proportion of negative tweets
June, 2020 referencing other racial or ethnic minorities was stable;
common themes that emerged during the content
analysis of a random subsample of 3300 tweets included:
racism and blame, anti-racism, and effect on daily life
Odlum et al82 June USA Twitter 2 558 474 Tweets from Jan 21 to Clustering algorithm; NLP; 15 topics (in four themes) were identified; positive
May 3, 2020 network diagrams sentiments, cohesively encouraging online discussions,
and behaviours for COVID-19 prevention were uniquely
observed in African American Twitter communities
Park et al83 May South Twitter 43 832 unique users and Network analysis; content Spread of information was faster in the COVID-19
Korea 78 233 relationships on Feb 29, 2020 analysis network than in the other networks; tweets containing
medically framed news articles were more popular than
were tweets that included news articles adopting
non-medical frames
Pastor84 April Philippines Twitter Tweets were collected on NLP for sentiment analysis Negative sentiments increased over time in lockdown
three Tuesdays in March, 2020,
since lockdown in Philippines
Samuel et al85 June USA Twitter 900 000 tweets from February to Sentiment analysis packages; For short tweets, classification accuracy was 91% with
March, 2020 textual analytics; machine Naive Bayes whereas accuracy was 74% with logistic
learning classification regression; both methods showed weaker performance for
methods: Naive Bayes and longer tweets
logistic regression
Samuel et al86 August USA Twitter 293 597 tweets, 90 variables Textual analytics to analyse For the reopening of the US economy, there was more
public sentiment support; positive sentiment support than there was negative
sentiment analysis by use of support; developed a novel sentiment polarity based
R package Syuzhet public sentiment scenarios framework
(version 1.0.6)
Su et al87 June China and Sina Weibo and 850 Sina Weibo users with posts Wilcoxon tests Individuals focused more on home and expressed a high
Italy Twitter published from Jan 9 to Feb 5, 2020; level of cognitive process after a lockdown in both
14 269 tweets from 188 unique Wuhan, China, and Lombardy, Italy; level of stress
Twitter users from Feb 23 to decreased, and the attention to leisure increased in
March 21, 2020 Lombardy, Italy, after the lockdown; attention to group,
religion, and emotions became more prevalent in Wuhan,
China, after the lockdown
Thelwall and May UK Twitter 3 038 026 English tweets from Word frequency comparison; Women were more likely to tweet about the virus in the
Thelwall88 March 10 to March 23, 2020 χ² analysis context of family, physical distancing, and health care,
whereas men were more likely to tweet about sports
cancellations, the global spread of the virus, and political
reactions
Wang et al89 July China Sina Weibo 999 978 randomly selected Sina Unsupervised Bidirectional People were concerned about four aspects regarding
Weibo posts that were related to Encoder Representations from COVID-19: the virus origin, symptoms, production
COVID-19 from Jan 1 to Feb 18, 2020 Transformers model: classify activity, and public health control
sentiment categories; Term
Frequency-Inverse Document
Frequency model: summarise
the topics of posts; trend
analysis; thematic analysis
Wicke and September Ireland Twitter 203 756 tweets Topic modelling Although the family frame covers a wider portion of
Bolognesi90 topics, among the figurative frames, war (a highly
conventional one) was the frame used most frequently;
yet, this frame does not seem to be appropriate to
elaborate the discourse around some aspects that are
involved in the situation
(Table continues on next page)

COVID-19.99 12 articles studied infodemics that were increased, so did its infodemic.42 Gallotti and colleagues
related to COVID-19 that were circulating on social analysed over 100 million tweets and identified that, even
media platforms. Rovetta and Bhagavathula42 analysed before the onset of the COVID-19 pandemic, infodemics
over 2 million queries from Google Trends and Instagram threatened public health, although not to the same
between Feb 20 and May 6, 2020. Their findings showed extent.37 Pulido and colleagues sampled and analysed
that as global interest for COVID-19 information 942 tweets, which revealed that although false

e184 www.thelancet.com/digital-health Vol 3 March 2021

 Review

Publication Origin Social media Study population and sample size Methods Key findings
month
(Continued from previous page)
Xi et al91 September China Sina Weibo 188 unique topics, their views, Thematic analysis; temporal Six themes were identified: the most prominent theme was
and comments from Jan 20 to analysis older people contributing to the community
April 28, 2020 (46 [24%] of 188) followed by older patients (defined by
keywords—eg, “older people”, “old-aged people”,
“grandmother”, “grandfather”, “old grandmother”, “old
grandfather”, “old woman”, and “old man”) in hospitals
(43 [23%]); the theme of contributing to the community was
the most dominant in the first phase (Jan 20–Feb 20, 2020;
period of COVID-19 outbreak in China); the theme of older
patients in hospitals was most dominant in the second
(Feb 21–March 17, 2020; turnover period) and third phase
(March 18–April 28, 2020; post-peak period in China)
Xie et al92 August China Baidu Search Number of cases by Feb 29, 2020: Kendall’s Tb rank test Both the Baidu Search Index and Google Trends indices
Index and 79 968 cumulative confirmed cases, showed a similar trend in a slightly different way; daily
Google Trends 41 675 cured cases, 2873 dead cases Google Trends were correlated to seven indicators,
whereas daily Baidu Search Index was correlated to only
three indicators; these indexes and rumours are
statistically related to disease-related indicators;
information symmetry was also noted
Xue et al93 November Canada Twitter 1 015 874 tweets from April 12 to LDA Nine themes about family violence were identified
July 16, 2020
Yigitcanlar October Australia Twitter 96 666 tweets from Australia in Descriptive analysis; content Social media analytics is an efficient approach to capture
et al94 Jan 1 to May 4, 2020 analysis; sentiment analysis; attitudes and perceptions of the public during a
spatial analysis pandemic; crowdsourced social media data can guide
interventions and decisions of the authorities during a
pandemic; effective use of government social media
channels can help the public to follow the introduced
measures and restrictions
Yu et al95 July Spain Twitter 22 223 tweets Topic modelling; network Identified eight news frames for each newspaper’s Twitter
analysis account; the entire pandemic development process is
divided into three periods: precrisis, lockdown, and recovery
period; understanding of how Spanish news media cover
public health crises on social media platforms
Zhao et al96 May China Sina Weibo and 4056 topics from Dec 31, 2019, Word segmentation; word The trend of public attention could be divided into three
microblog hot to Feb 20, 2020 frequency; sentiment stages; the hot topic keywords of public attention at each
search list analysis stage were slightly different; the emotional tendency of
the public towards the COVID-19 pandemic-related hot
topics changed from negative to neutral between January
and February, 2020, with negative emotions weakening
and positive emotions increasing overall; COVID-19 topics
with the most public concern were divided into five
categories: the situation of the new cases of COVID-19
and its effects, front-line reporting of the pandemic and
the measures of prevention and control, expert
interpretation and discussion on the source of infection,
medical services on the front line of the pandemic, and
focus on the pandemic and the search for suspected cases
Zhu et al97 July China Sina Weibo 1 858 288 microblog data LDA A so-called double peaks feature appeared in the search
curve for epidemic topics; the topic changed over time,
the fluctuation of topic discussion rate gradually
decreased; political and economic centres attracted high
attention on social media; the existence of the subject of
rumours enabled people to have more communication
and discussion
All studies were published in 2020. LDA=latent Dirichlet allocation. NLP=natural language processing.

Table: Summary of chosen articles

information had a higher number of tweets, it also had 673 English tweets. Their results showed that misin­
less retweets and lower engagement than did tweets formation accounted for 24·8% (153 of 617) of all serious
comprising scientific evidence or factual statements.41 tweets (ie, not humour-related posts). Health-care or
Kouzy and colleagues39 investigated the extent to which public health accounts had the lowest amount of
misinformation or unverifiable information about the misinformation; yet still 12·3% (7 of 57) of their tweets
COVID-19 pandemic was spread on Twitter by analysing included unverifiable information. Moscadelli and

www.thelancet.com/digital-health Vol 3 March 2021 e185

 Review

the USA and 15 million tweets from the Philippines from

March 5 to March 19, 2020, and both countries showed a
s surveillance mo
edia a nito positive relation between bot activities and rate of hate
ial m rin
Soc g
speech in communities that are denser and more isolated
than others.43 Brennen and colleagues qualitatively
Mental health analysed 96 samples of visuals (ie, image or video) from
January to March, 2020, and categorised misinfor­mation
Public Detection or into six trends, noting that, fortunately, there has been no
attitudes prediction
involvement of artificial intelligence deepfake techniques
COVID-19 (ie, techniques used to make synthetic videos that closely
resemble real videos) so far.34
Government
c o nta

Infodemics
S oci ion a

responses Social media for surveillance and monitoring

al m n d
g

Information Three themes emerged under this category: public

e d ve

attitudes, mental health, and detection or prediction of

ia ct

quality
as

or l
tr o
n COVID-19 cases. Public attitudes and mental health are
co
as e
s di
se reflections regarding the public perceptions and mental
dia a
S o ci a l m e health effects of the pandemic; detection or prediction
of COVID-19 cases includes typical surveillance studies
aiming to propose ways to detect or predict COVID-19
Figure 2: Modified Social Media and Public Health Epidemic and Response cases.
framework
48 selected articles gauged the attitudes and emotions
that were expressed by social media users regarding the
colleagues40 collected and reviewed 2102 news articles COVID-19 pandemic, mainly by use of content and
that were circulated on the internet. Their analysis sentiment analysis. Twitter accounted for 33 articles
showed that fake news was shared over 2 million times, and Sina Weibo accounted for 8 articles. Public attitude
which accounted for 23·1% (2 352 585 of 10 184 351) of can be further divided into the following sub-themes:
total shares between Dec 31, 2019, and April 30, 2020.40 public sentiment towards the COVID-19 pandemic and
Similarly, another quantitative study by Galhardi and interventions, stigma and racism, and ageism.
colleagues comparing the proportion of fake news shared To learn about the public sentiment towards the overall
on WhatsApp, Instagram, and Facebook in Brazil showed COVID-19 pandemic and its interventions, Abd-Alrazaq
that fake news was mainly shared on WhatsApp.36 A UK and colleagues7 analysed 167 073 unique English tweets
study by Ahmed and colleagues analysed 22 785 tweets that were divided into four categories: origin, source,
posted by 11 333 Twitter users with #FilmYourHospital regional and global effects on people and society, and
to identify and evaluate the source of the conspiracy methods to reduce transmission of SARS-CoV-2. Tweets
theory on Twitter.32 Their work uncovered that ordinary regarding economic loss had the highest mean number
people were the major driver behind the spread of of likes, whereas travel bans and warnings had the lowest
conspiracy theories.32 Another study investigated the 5G number of likes.7 Kwon and colleagues investigated
and COVID-19 conspiracy theory that was circulating on 259 529 English tweets in the USA, using trending and
Twitter with a random subsample of 233 tweets. The spatiotemporal analyses, and noted that tweets about
content analysis showed that 34·8% (81) of tweets linked social disruptiveness had the highest number of retweets,
5G and COVID-19 and 32·2% (75) condemned such whereas tweets about COVID-19 interventions had the
theory.33 Similar research by Bruns and colleagues highest number of likes.73 A content analysis of 522 Reddit
investigated 89 664 distinct Facebook posts in Australia comments showed that the topic of symptoms accounted
that were related to this conspiracy from Jan 1 to for 27% (141) of all comments, followed by the topic of
April 12, 2020, by use of time series and network prevention (25% [131]).74 Likewise, another content
analysis.35 The results showed that this conspiracy went analysis of 155 353 unique English tweets showed that
viral after March 19, 2020, with unusual coalition among the most mentioned topic was “peril of COVID-19”.76
various groups on Facebook. Islam and colleagues Additionally, a study that examined 126 049 English
analysed 2311 infodemic reports that were related to tweets by use of sentiment analysis and latent
COVID-19 from Dec 31, 2019, to April 5, 2020, and Dirichlet analysis for topic modelling showed that the
showed that misinformation was mainly driven by most common emotion that was mentioned was fear,
rumours, stigma, and conspiracy theories that were and the most common topic that was mentioned was the
circulating on various social media and other online economic and political effects.79 Al-Rawi and colleagues
platforms.38 Associations between infodemic and bot studied emojis in over 50 million tweets and identified
activities on social media are another important research five primary subjects: morbidity fears, health concerns,
direction. One study analysed 12 million tweets from employment and financial issues, praise for front-line

e186 www.thelancet.com/digital-health Vol 3 March 2021

 Review

workers, and unique gendered emoji use.51 Samuel fear to anger and from sadness to gratefulness.77 Chang
and colleagues investigated 293 597 tweets with senti­ and colleagues examined over 1·07 million Chinese texts
ment analysis and noted more positive emotions than from various online sources in Taiwan using deductive
negative emotions towards the US economy reopening.86 analysis and identified that negative sentiments mainly
Analysing 2 558 474 English tweets by use of clustering came from online news with stigmatising language
and network analyses, Odlum and colleagues identified linked with the COVID-19 pandemic.56 In India, one
that African Americans shared positive sentiments study investigated 410 643 tweets via sentiment analysis
and encouraged virtual discussions and prevention and latent Dirichlet analysis and showed that positive
behaviours.82 A study investigated gender differences in emotions were overall substantially higher than negative
terms of topics by analysing 3 038 026 English tweets.88 sentiments, but this observation diminished at indivi­
The results showed that tweets from women were more dual levels.61 Another study analysed 29 554 tweets
likely to be about family, physical distancing, and health from the second lockdown (ie, April 15–May 3, 2020)
care, whereas tweets from men were more likely to and 47 672 tweets from the third lockdown (ie,
be about sports cancellations, pandemic severity, and May 4–May 17, 2020) via sentiment analysis uncovered
politics. In Canada, Xue and colleagues analysed positive attitudes towards the second lockdown but
1 015 874 tweets via latent Dirichlet analysis to identify negative attitudes towards the third lockdown in India.57
nine themes about family violence.93 In Australia, One study analysed 868 posts from Reddit and noted
Yigitcanlar and colleagues analysed 96 666 tweets and sentiments to be 50% (434) neutral, 22% (191) positive,
identified that the public’s attitude could be captured and 28% (243) negative in India.63 A study in South Korea
efficiently through social media analytics.94 One qualita­ examined 43 832 unique users and their relations on
tive content analysis of 30 profiles from Instagram, Twitter by use of content and network analyses and
Twitter, and TikTok in Brunei identified five types of showed that tweets including medical news were more
attitudes towards physical distancing: fear, responsibility, popular than tweets containing non-medical news.83 A
annoyance, fun, and resistance.80 In Turkey, to show the study from Ireland analysed 203 756 tweets through topic
effects of social media on human psychology and modelling and identified that war was the most frequently
behaviour, Arpaci and colleagues52 used evolutionary used frame for the pandemic.90 In the USA, Damiano
clustering analysis on 43 million tweets between March 22 and colleagues qualitatively analysed 600 English tweets
and March 30, 2020. The study suggested that high- and showed neutral sentiment across most tweets.59
frequency word clusters, such as death, test, spread, and Politics also had an essential role in shaping people’s
lockdown denoted the public’s underlying fear of infection opinion.59 A study of 19 803 tweets from Democrats and
and death from the virus, whereas terms such as stay 11 084 tweets from Republicans by use of random forest
home and social distancing corresponded to behavioural in the USA showed that Democrats put more emphasis
shifts.52 A study in Luzon, Philippines,84 in which on public health and direct aid to US workers, whereas
sentiment analysis was done by use of natural language Republicans put more emphasis on national unity,
processing, showed that most Filipino Twitter users China, and businesses.67 Results of a study involving
expressed negative emotions towards COVID-19, and the various online data sources from Italy, the UK, the USA,
negative mood grew stronger over time in lockdown.84 and Canada showed that media was the major driver of
Sentiment analysis of 107 990 English tweets uncovered the public’s attention, but attention decreased with
that a negative feeling towards the COVID-19 pandemic saturation of the media with news about COVID-19.66
dominated, and topic modelling showed three major Compared with other users, Reddit users focused more
themes in people’s concerns: the COVID-19 pandemic on health, data related to new disease, and preventative
emergency, how to control COVID-19, and reports on interventions. Researchers in Spain studied 22 223 tweets
COVID-19.54 Another study analysed 373 908 Belgian by use of topic modelling and network analysis.95 They
tweets and retweets, which showed that the public relied identified eight frames and noted that the entire pan­-
on the EU coalition to tackle the pandemic.72 De Santis demic could be divided into three periods: precrisis,
and colleagues analysed 1 044 645 tweets to identify lockdown, and recovery periods. Using 563 079 English
daily hot topics in Italy that were related to the Reddit posts that were related to COVID-19, Jelodar and
COVID-19 pandemic and developed a framework for colleagues proposed a novel method to detect meaningful
prospective research.62 One thematic analysis study of latent topics and sentiment–comment classification.69
1 920 593 Arabic tweets in Egypt showed that negative Samuel and colleagues examined over 900 000 tweets to
emotions and sadness were high in tweets showing study the accuracy of tweet classifications among logistic
affective discussions, and the dominant themes regression and Naive Bayes methods.85 They identified
included the outbreak of the pandemic, metaphysics that Naive Bayes had 91% of accuracy compared with
responses, signs and symptoms in confirmed cases, 74% from the logistic regression model.85
and conspiracism.64 In Singapore, Lwin and colleagues Han and colleagues analysed 1 413 297 Sina Weibo posts
examined 20 325 929 tweets using sentiment analysis and observed that the public paid attention to information
and showed that public emotions shifted over time: from regarding the epidemic, especially in metro areas.68

www.thelancet.com/digital-health Vol 3 March 2021 e187

 Review

Zhao and colleagues studied 4056 topics from the emotions led to increased citizen engagement through
Sina Microblog hot search list and noted that the public government social media.58 Yin and colleagues65 proposed
emotions shifted from negative to neutral to positive over a new multiple-information susceptible-discussing-
time and that five major public concerns existed: the immune model to analyse the public opinion propagation
situation of the new cases of COVID-19 and its effects, of COVID-19 from Sina Weibo posts that were collected
front-line reporting of the pandemic and the measures of from Dec 31, 2019, to Feb 27, 2020. The researchers
prevention and control, expert interpretation and reported that the reprod­ uction rate of this proposed
discussion on the source of infection, medical services model reached 1·78 in the early stage of COVID-19 but
on the front line of the pandemic, and focus on the decreased to around 0·97 and was maintained at this
pandemic and the search for suspected cases.96 Li and level. Such a result showed that the information on
colleagues75 did an observational infoveillance study with COVID-19 would continue to increase slowly in the future
a linear regression model by analysing 115 299 Sina Weibo until it stabilises. However, this stability would depend on
posts. The results showed that the number of Sina Weibo how much information is received on COVID-19. Wang
posts positively correlated with the number of reported and colleagues89 analysed 999 978 randomly selected
cases of COVID-19 in Wuhan. Additionally, the qualita­ Sina Weibo posts that were related to COVID-19 through
tive analysis classified the topics into the following an unsupervised Bidirectional Encoder Representations
four overarching themes: cause of the virus, epidemio­ from Trans­ formers model for sentiments and a term
logical characteristics of COVID-19, public responses, frequency-inverse document frequency model for topic
and others.75 Chen and colleagues examined relationships modelling. The authors identified four public concerns:
between citizen engagement through government social the virus origin, symptom, production activity, and public
media and media richness, dialogic loop, content type, health control in China.89 Xi and colleagues examined
and emotion valence.58 Citizen engagement through 241 topics with their views and comments via thematic
government social media refers to sum of shares, likes, and temporal analysis and noted that older adults
and comments in this study, so the higher the sum, the contributing to the community was the most frequent
greater the citizen engagement through government theme in the first phase of COVID-19 in China
social media. Media richness quantifies how much (ie, Jan 20–Feb 20, 2020).91 The theme of older patients
information that a sender transfers to a receiver via a in hospitals was most frequent in the second (ie,
medium and is based on the media richness theory (ie, Feb 21–March 17, 2020) and third phase (ie,
“the potential information load of communication March 18–April 28, 2020). Using Wilcoxon tests, Su and
media, emhasising the abilities of promoting shared colleagues examined posts from 850 Sina Weibo users
meaning”).101 Dialogic loop, or dialogic communication and 14 269 tweets from Italy.87 The findings showed that
theory, is defined as an approach that promotes a Italian people paid more attention to leisure, whereas
dialogue between a speaker and audience. According to Chinese people paid more attention to the community,
the American Psychological Association, emotion religion, and emotions after lockdowns. Analysing the
valence refers to “the value associated with a stimulus, top 200 accounts from WeChat via regressions and
expressed on a continuum from pleasant to unpleasant content analysis, Ma and colleagues showed that both
or from attractive to aversive”.100 For instance, happiness non-medical and medical reports had positive effects on
is typically considered to be pleasant valence. Chen and people’s behaviours.78 Using Kendall’s Tau-B rank test,
colleagues analysed 1411 posts that were related to Xie and colleagues investigated relations among the
COVID-19 from Healthy China, an official account of the Baidu Attention Index, daily Google Trends, and numbers
National Health Commission of China on Sina Weibo. of COVID-19 cases and deaths.92 Daily Google Trends
Findings showed an inverse association between media were correlated to seven indicators, whereas daily Baidu
richness and citizen engagement through government Search Index was correlated only to three indicators.92
social media, indi­cating that posts with plain texts had Zhu and colleagues analysed 1 858 288 Sina Weibo posts
higher citizen engagement through government social and noted that topics changed over time but political and
media than did posts with pictures or videos. A positive economic posts attracted greater attention than did other
association between dialogic loop and citizen engage­ topics.97
ment through government social media was noted, as Regarding stigma and racism, Kim71 analysed
evidenced by 96% (1355 of 1411) of responses to these 27 849 individual tweets in South Korea by use of a binary
posts having hashtags and 25% (353 of 1411) containing logistic regression to gauge network size and semantic
questions. In terms of media richness, when posts had network analysis to capture contextual and subjective
both a high media richness and positive emotion, citizen factors. The results indicated that size of personal social
engagement through government social media increased, network was inversely correlated with impolite language
whereas when posts had a high media richness and use. Namely, users with larger social networks were less
negative emotion, citizen engagement decreased. likely to post uncivil messages on Twitter than were users
Regarding content type, when posts were related to the with smaller social networks. This study suggested that
latest news about the pandemic, stronger negative the size of the social network influenced the language

e188 www.thelancet.com/digital-health Vol 3 March 2021

 Review

choice of social media users in their postings.71 Research indignation, after the declaration and decreased positive
compared public stigma before and after the introduction sentiments expressed in the Oxford happiness score.
of the terms Chinese virus or China virus in 16 535 English Additionally, cognitive indicators showed increased
tweets from before introduction and 177 327 tweets sensitivity to social risks but decreased life satisfaction
from after introduction.55 The results showed an almost after the declaration.45
10 times increase, nationwide and statewide and in Six of 81 studies investigated the detection or prediction
the USA, from 0·38 tweets posted per 10 000 people of COVID-19 outbreaks with social media data. Qin and
referencing the two terms before introduction to colleagues22 attempted to predict the number of newly
4·08 tweets posted per 10 000 after introduction. A similar suspected or confirmed COVID-19 cases by collecting
study examined 339 063 tweets from non-Asian respon­ social media search indexes for symptoms (eg, dry cough,
dents via local polynomial regression and interrupted fever, and chest distress), coronavirus, and pneumonia.
time-series analysis.60 The findings showed that, when The data were analysed by use of subset selection,
stigmatising terms, such as Chinese virus, were used by forward selection, lasso regression, ridge regression, and
media (starting from March 8, 2020), the bias index (ie, elastic net. Results showed that the optimal model was
Implicit Americanness Bias) began to increase, and such constructed via the subset selection. The lagged social
bias was more profound in conservatives than in members media search indexes were a predictor of new suspected
of any other political subgroup. Nguyen and colleagues COVID-19 cases and could be detected 6–9 days before
analysed 3 377 295 tweets that were related to race in confirmation of new cases.22 To evaluate the possibility of
the USA using sentiment analysis and uncovered a early prediction of COVID-19 cases via internet searches
68·4% increase in negative tweets referring to Asian and social media data, Li and colleagues17 used the
people, whereas tweets referring to other races remained keywords coronavirus and pneumonia to retrieve corres­
stable.81 ponding trend data from Google Trends, Baidu Search
Regarding ageism, a study70 investigating Twitter Index, and Sina Weibo Index. By use of the lag correlation,
content that was related to both COVID-19 and older the results showed that the correlation between trend
adults analysed a random sample of 351 English tweets. data with the keyword coronavirus and number of
21·1% (74) of the tweets implied diminished regard for laboratory-confirmed cases was highest 8–12 days before
older adults by downplaying or dismissing concerns increase in confirmed COVID-19 cases in the three
over the high fatality of COVID-19 in this population.70 platforms. Similarly, the correlation between trend data
Similar research examined 188 tweets via thematic for the keyword coronavirus and new suspected
analysis and showed that 90% (169) of tweets opposed COVID-19 cases was highest 6–8 days before increase in
ageism, whereas 5% (9) of tweets favoured ageism, and new suspected cases. The correlation between trend data
5% (10) of tweets were neutral.53 for the keyword pneumonia and new suspected cases
Two of 81 reviewed studies, both based in China, was highest 8–10 days before increase in new suspected
focused on assessing the mental health of social COVID-19 cases across the three platforms.17 Peng and
media users.44,45 A cross-sectional study44 investigated the colleagues studied 1200 Sina Weibo records using
relationship between anxiety and social media exposure, spatiotemporal analysis, kernel density analysis, and
which is theoretically defined as “the extent to which ordinary least square regression and noted that scattered
audience members have encountered specific infection, community spread, and full-scale outbreak
messages”.102 The researchers distributed an online were three phases of early COVID-19 transmission in
survey based on the Chinese version of WHO-Five Well- Wuhan, China.21 Older people are at high risk of severe
Being Index for depression and the Chinese version of COVID-19 and accounted for over 50% of help seeking
Generalized Anxiety Disorder Scale for anxiety. on Sina Weibo. To identify COVID-19 patients with poor
Respondents included 4872 Chinese citizens aged prognosis, Liu and colleagues analysed Sina Weibo
18 years and older from 31 provinces and autonomous messages from 599 patients along with telephone follow-
regions in China. After controlling for all covariates ups.18 The findings suggested risk factors involving older
through a multivariable logistic regression, the study age, diffuse distribution of pneumonia, and hypoxaemia.
showed that frequent social media exposure increased A regression study analysed Google Trends searches,
the odds ratio of anxiety, showing that frequent social Wikipedia page views, and tweets and showed that
media exposure is potentially contributing to mental current Wikipedia page views, tweets from a week before,
health problems during the COVID-19 outbreak.44 To and Google Trends searches from two weeks before can
explore how people’s mental health was influenced by be used to model the number of COVID-19 cases. To
COVID-19, Li and colleagues45 analysed posts from model the number of deaths, all three variables should
17 865 active Sina Weibo users to compare sentiments be one week earlier than for cases.19
before and after declaration of COVID-19 outbreak
by the National Health Commission in China on Social media as disease control
Jan 20, 2020. The researchers identified increased To inoculate the public against misinformation, public
negative sentiments, including anxiety, depression, and health organisations and governments should create and

www.thelancet.com/digital-health Vol 3 March 2021 e189

 Review

spread accurate information on social media because COVID-19 via temporal and network analyses.31 They
social media has had an increasingly important role categorised 16 types of messages and identified incon­
in policy announcement and health education. Six of sistent and incongruent messages expressed in four
81 articles were categorised as government responses crucial pre­vention topics: mask wearing, risk assessments,
because they examined how government messages and stay at home order, and disinfectants or sanitisers.
health education material were generated and consumed Eight chosen studies investigated the quality (ie, the
on social media platforms. Two studies analysed data number of recommended prevention behaviours that
from Sina Weibo,23,27 and the other four studies analysed were covered in the videos—eg, wearing a facemask,
data from Twitter.28–31 washing hands, physical distancing, etc) of YouTube
Zhu and colleagues23 measured the attention of Chinese videos with COVID-19 prevention information. Basch
netizens—ie, citizen of the net—to COVID-19 by and colleagues24 did a cross-sectional study and retrieved
analysing 1101 Sina Weibo posts. They noted that Chinese the top 100 YouTube videos with the most views that
netizens paid little attention to the disease until the were uploaded in January, 2020, with the keyword of
Chinese Government acknowledged and declared the coronavirus in English, with English subtitles, or in
COVID-19 outbreak on Jan 20, 2020. Since then, high Spanish. These 100 videos generated over 125 million
levels of social media traffic occurred when Wuhan, views in total. However, fewer than 33 videos included
China, began its quarantine (Jan 23–Jan 24, 2020), during any of the seven key prevention behaviours that are
a Red Cross Society of China scandal (Feb 1, 2020), and recommended by the US Centers for Disease Control
following the death of Li Wenliang (Feb 6–Feb 7, 2020).23 and Prevention.24 A follow-up study with the same criteria
Li and colleagues27 collected 36 746 Sina Weibo posts to and a successive sampling design gathered the top
identify and categorise the situational information using 100 YouTube videos that were most viewed in January and
support vector machines, Naive Bayes, and random forest March, 2020.25 Findings showed that, in total, the January
as well as features in predicting the number of reports sample generated over 125 million views, and the March
using linear regression. Except for posts that were sample had over 355 million views. Yet, fewer than
categorised as counter rumours (ie, used to oppose 50 videos in either sample contained any of the preven­
rumours), they identified that the higher the word count, tion behaviours that are recommended by the US Centers
the more reposts there were. Likewise, posts from for Disease Control and Prevention.25 Additionally, a
unverified users had more reposts for all categories than study investigated the top 100 YouTube videos about do it
did posts from verified users, excluding the counter yourself hand sanitiser with the most views and showed
rumours. For counter rumours, reposts increased with that the average number of daily calls about paediatric
the number of followers and if the followers were from poisoning increased substantially in March, 2020,
urban areas.27 A qualitative content analysis was done to compared with the previous 2 years.46
investigate how G7 leaders used Twitter for matters To analyse the information quality on YouTube about
concerning the COVID-19 pandemic by collecting the COVID-19 pandemic and to compare the contents in
203 tweets.29 The findings showed that 166 of 203 tweets English and Chinese Mandarin videos, Khatri and
were informative, 48 tweets were linked to official colleagues26 collected 150 videos with the keywords
government resources, 19 (9·4%) tweets were morale- 2019 novel coronavirus and Wuhan virus in English and
boosting, and 14 (6·9%) tweets were political.29 To assess Mandarin. The DISCERN score and the medical
the political partisan polarisation in Canada regarding information and content index were calculated as a
COVID-19, Merkley and colleagues28 randomly sampled reliable way to measure the quality of health information.
1260 tweets from the social media of 292 federal members The mean DISCERN score for reliability was low:
of parliament and collected 87 Google Trends for the 3·12 of 5·00 for English videos and 3·25 for Mandarin
search term coronavirus. 2499 Canadian respondents videos. The mean cumulative medical information and
aged 18 years and above were also surveyed. The results content index score of useful videos was also undesirable:
showed that, regardless of party affiliation, members of 6·71 of 25·00 for English videos and 6·28 for Mandarin
parliament emphasised the importance of measures for videos.26 In Spain, a similar study of 129 videos in Spanish
physical distancing and proper hand-hygiene practices to identified that information in videos about preventing
cope with the COVID-19 pandemic, without tweets exag­ COVID-19 was usually incomplete and differed according
gerating concerns or misinformation about COVID-19. to the type of authorship (ie, mass media, health
Search interest in COVID-19 among munici­palities was professionals, individual users, and others).47 Likewise,
strongly determined by socioeconomic and urban factors one study in South Korea noted that misleading videos
rather than Conservative Party vote share.28 Sutton and accounted for 37·14% (39 of 105) of most-viewed videos
colleagues studied 149 335 tweets from public health, and had more likes, fewer comments, and longer viewing
emergency management, and elected officials and times than did useful videos.48 Two studies in Turkey
observed that the underlying emotion of messages investigated the quality of YouTube videos regarding
changed positively and negatively over time.30 Wang and COVID-19 information in dentistry.49,50 One of these
colleagues investigated 13 598 tweets that were related to studies analysed the top 116 English videos with at least

e190 www.thelancet.com/digital-health Vol 3 March 2021

 Review

300 views and showed moderate quality and useful that are based on misinformation. Bot posts are another
information from these videos.49 The other study, topic to be addressed and studies evaluating effective
however, showed poor quality for 24 of 55 (43·6%) counter-infodemic interventions are also needed.
English videos, whereas good quality accounted for only Articles regarding public attitudes towards the
2 (3·6%) videos.50 COVID-19 pandemic have shown sentiments that shifted
over time. Yet, this theme can be a useful indicator when
Discussion evaluating interventions, such as physical distancing and
Studies on social media data showed our attitudes and wearing masks, that aim to reduce the risk of COVID-19
mental state to some extent during the COVID-19 crisis. infection. However, public sentiments had not been
These studies also showed how we generated, consumed, incorporated into many intervention studies by the time
and propagated information on social media platforms that we did this Review. When a disease, such as
when facing the rapid spread of the SARS-CoV-2 and COVID-19, starts spreading and causing negative senti­
extraordinary measures for the containment. In our ments, timely, proper, and effective risk communication is
Review, public attitudes accounted for nearly 59% needed to help ease people’s anxiety or negative attitudes
(48 of 81) of the reviewed articles. In terms of social regarding the COVID-19 pandemic, especially through
media platforms, 56% (45 of 81) of the chosen articles social media.
used data from Twitter, followed by Sina Weibo (20% Mental health is another issue that requires further
[16 of 81]). Machine learning analyses, such as latent investigation. Our chosen studies did not address mental
Dirichlet analysis and random forest, were applied in health issues on the basis of age, as symptoms and inter­
research that studied public attitudes. ventions tend to vary with age. Public health measures,
We identified six themes on the basis of our modified such as physical distancing, that were implemented in the
SPHERE framework, including infodemics, public atti­ COVID-19 pandemic exacerbated risk factors and adverse
tudes, mental health, detection or prediction of COVID-19 health behaviours at the individual and population levels.
cases, government responses to the pandemic, and quality Studies showed that social media data were useful to detect
of prevention education videos. However, a common mental health issues at the population level. Due to the
limitation in all chosen studies on social media data is the early outbreak of COVID-19 and the prevalence of social
comparison of data due to differences in quality, such as media use (eg, Sina Weibo and WeChat) in China, two
formats, metrics, or even the definition of common studies reported increased issues of mental health among
variables (eg, the amount of time required for a post to be the Chinese population.44,45 A similar trend of deteriorating
on an individuals screen to be counted as a view). For mental health could happen in other regions. At the time
instance, the definition of a view on one social media of writing, British Columbia has recorded the highest
platform is likely to be different from another. Besides, number of overdose deaths in Canada (May, 2020).103
not every social media platform offers accessible data, like In terms of the surveillance of the COVID-19 pandemic,
Twitter and Sina Weibo. To address these challenges, the six chosen studies showed methods to detect or predict
selected studies have controlled for many factors, the number of COVID-19 cases by use of social media
including social media platforms, languages, locations, data. Accoridng to our Review, unlike other infectious
time, misspellings, keywords, or hashtags. However, such diseases, such as influenza and malaria, COVID-19 has
search strategies resulted in many study limitations, such not had real-time monitoring surveillance developed with
as non-representative sample sizes, selection bias, cross- social media data. It is possible that the pandemic has
sectional study design, or retrospective study design. We evolved so rapidly that finding COVID-19 vaccinations or
also observed that, given the large amount of available therapies has been prioritised over real-time monitoring
data, most studies across all domains sampled small data surveillance with social media. Besides, scarcity of
size for analyses, except for four studies under the theme accurate and reliable data sources might discourage the
of public attitudes that analysed over one million posts development of the COVID-19 real-time surveillance.
via machine learning methods. Additionally, data from Moreover, whether COVID-19 is a one-time event or
Twitter and Sina Weibo accounted for over 70% (59 of 81) will become seasonal, like influenza, is unknown. If
of our selected studies. Research examining other social COVID-19 becomes seasonal, then it might be meaningful
media platforms, including Facebook, Instagram, TikTok, and useful to establish a real-time model to monitor the
Snapchat, and WhatsApp, is scarce due to barriers of data disease by use of social media data.
availability and accessibility. We also identified future Government responses that were distributed via social
research topics that are needed for each category during media have been increasingly crucial in combating
the COVID-19 pandemic. From an infodemics perspec­ infodemics and promoting accurate and reliable
tive, additional research is needed to investigate how information for the public. However, little has been
misinformation, rumours, and fake news (eg, anti-mask studied about how efficient and effective these official
wearing reports) undermine preventions and compromise responses are at leading to public belief or behavioural
public health, although social media companies, such as changes. It also remained unknown whether govern­
Twitter and Facebook, have started to remove accounts ment posts would reach greater numbers of social

www.thelancet.com/digital-health Vol 3 March 2021 e191

 Review

media users or have greater effects on them than would machine learning methods, whereas most studies used
infodemics. traditional statistical methods. Unlike influenza, we were
YouTube has served as one of the major platforms to not able to find studies documenting real-time surveillance
spread information concerning the control of COVID-19. that was developed with social media data on COVID-19.
Nonetheless, our chosen studies showed that most Our Review also identified studies that were related to
YouTube videos were of undesirable quality because they COVID-19 on infodemics, mental health, and prediction.
contained few recommended preventions from govern­ For COVID-19, accurate and reliable information through
ments or public health organisations. The undesirable social media platforms can have a crucial role in tackling
quality is a worrisome observation if accurate and reliable infodemics, misinformation, and rumours. Additionally,
videos and other types of information are not created and real-time surveillance from social media about COVID-19
disseminated in a timely manner. Therefore, videos, can be an important tool in the armamentarium of
especially from public health authorities, should include interventions by public health agencies and organisations.
accurate and reliable medical and scientific information Contributors
and use relevant hashtags to reach a large audience, HC and ZAB conceived the Review. S-FT searched for articles and
generate a high number of views, and increase responses. screened them, analysed data, and wrote the first draft of the
manuscript. TT refined the search strategy and searched for articles.
Moreover, our selected studies were limited to YouTube YY refined the search strategy and created the tables. LL assisted in
videos only. Additionally, a substantial proportion of the screening. ZAB and HC supervised the review process and prepared the
studies were done using Sina Weibo, which, although used final draft for submission. All authors contributed to the interpretation
by many people, is exclusive to China and might lead to an of results, manuscript preparation, and revisions. All authors read and
approved the final manuscript.
over-representation of a single country in this Review.
In summary, although our Review has limitations that Declaration of interests
We declare no competing interests.
are embedded from the chosen studies, we recognised
six themes that have been studied so far and identified References
1 Johns Hopkins University and Medicine. Coronavirus resource
future research directions. Our adopted framework can center: world map. https://coronavirus.jhu.edu/map.html (accessed
serve as a fundamental and flexible guideline when Jan 18, 2021).
studying social media and epidemiology. 2 Pérez-Escoda A, Jiménez-Narros C, Perlado-Lamo-de-Espinosa M,
Pedrero-Esteban LM. Social networks’ engagement during the
COVID-19 pandemic in Spain: health media vs. healthcare
Conclusion professionals. Int J Environ Res Public Health 2020; 17: 5261.
Our Review identified various topics, themes, and 3 Jordan SE, Hovet SE, Fung IC, Liang H, Fu K, Tse ZTH. Using
Twitter for public health surveillance from monitoring and
methodological approaches in studies on social media prediction to public response. Data (Basel) 2019; 4: 6.
and COVID-19. Among the six identified themes, public 4 Shah Z, Surian D, Dyda A, Coiera E, Mandl KD, Dunn AG.
attitudes comprised most of the articles. Among the Automatically appraising the credibility of vaccine-related web
pages shared on social media: a Twitter surveillance study.
selected studies, Twitter was the leading social media J Med Internet Res 2019; 21: e14007.
platform, followed by Sina Weibo. Few studies included 5 Sinnenberg L, Buttenheim AM, Padrez K, Mancheno C, Ungar L,
Merchant RM. Twitter as a tool for health research: a systematic
review. Am J Public Health 2017; 107: e1–8.
6 Steffens MS, Dunn AG, Wiley KE, Leask J. How organisations
Search strategy and selection criteria promoting vaccination respond to misinformation on social media:
a qualitative investigation. BMC Public Health 2019; 19: 1348.
The search strategies used index terms, where applicable, and free-text terms to capture
7 Abd-Alrazaq A, Alhuwail D, Househ M, Hamdi M, Shah Z.
two concepts: social media, including both general terms and specific platform names Top concerns of tweeters during the COVID-19 pandemic:
(eg, Twitter, Facebook, Sina Weibo, and YouTube); and COVID-19. For each database, both infoveillance study. J Med Internet Res 2020; 22: e19016.
indexed terms (ie, MeSH and Emtree) and natural language keywords were used with 8 Freberg K, Palenchar MJ, Veil SR. Managing and sharing
H1N1 crisis information using social media bookmarking services.
Boolean operators (ie, AND, OR, and NOT) and truncations (panel). Since each database Public Relat Rev 2013; 39: 178–84.
has distinctive search functionality, individually tailored search statements were developed 9 Giustini D, Ali SM, Fraser M, Kamel Boulos MN. Effective uses of
with appropriate search filters for each database. Final search statements, along with a list social media in public health and medicine: a systematic review of
systematic reviews. Online J Public Health Inform 2018; 10: e215.
of search results, were downloaded from each database. Articles were included if they 10 Al-Dmour H, Masa’deh R, Salman A, Abuhashesh M, Al-Dmour R.
discussed the use of social media for COVID-19 research and if they were original, empirical Influence of social media platforms on public health protection
studies. Only peer-reviewed articles, including peer-reviewed preprints, in English or against the COVID-19 pandemic via the mediating effects of public
health awareness and behavioral changes: integrated model.
Mandarin, were included. A decision to include Chinese publications was based on the fact J Med Internet Res 2020; 22: e19996.
that COVID-19 cases were first reported in Wuhan, China, and many initial and relevant 11 Bridgman A, Merkley E, Loewen PJ, et al. The causes and
studies were published in Mandarin; therefore, we wanted to capture most studies consequences of COVID-19 misperceptions: understanding the role
of news and social media. June 18, 2020. https://misinforeview.hks.
regarding the use of social media for COVID-19 research. All articles published between harvard.edu/article/the-causes-and-consequences-of-covid-19-
Nov 1, 2019, and Nov 4, 2020, were included. Publications such as reviews, opinion pieces, misperceptions-understanding-the-role-of-news-and-social-media
books, book chapters, articles and preprints that were not peer-reviewed, and articles that (accessed Sept 15, 2020).
12 Tang L, Bie B, Park SE, Zhi D. Social media and outbreaks of
were written in languages other than English or Mandarin were automatically excluded. emerging infectious diseases: a systematic review of literature.
The final reference list was generated on the basis of originality and relevance to the broad Am J Infect Control 2018; 46: 962–72.
scope of this review. 13 Arksey H, O’Malley L. Scoping studies: towards a methodological
framework. Int J Soc Res Methodol 2005; 8: 19–32.

e192 www.thelancet.com/digital-health Vol 3 March 2021

 Review

14 Levac D, Colquhoun H, O’Brien KK. Scoping studies: advancing the 36 Galhardi CP, Freire NP, Minayo MCS, Fagundes MCM. Fact or
methodology. Implement Sci 2010; 5: 69. fake? An analysis of disinformation regarding the COVID-19
15 Ose SO. Using Excel and Word to structure qualitative data. pandemic in Brazil. Cien Saude Colet 2020; 25 (suppl 2): 4201–10.
J Appl Soc Sci (Boulder) 2016; 10: 147–62. 37 Gallotti R, Valle F, Castaldo N, Sacco P, De Domenico M. Assessing
16 Braun V, Clarke V. Using thematic analysis in psychology. the risks of ‘infodemics’ in response to COVID-19 epidemics.
Qual Res Psychol 2006; 3: 77–101. Nat Hum Behav 2020; 4: 1285–93.
17 Li C, Chen LJ, Chen X, Zhang M, Pang CP, Chen H. Retrospective 38 Islam MS, Sarkar T, Khan SH, et al. COVID-19-related infodemic
analysis of the possibility of predicting the COVID-19 outbreak and its impact on public health: a global social media analysis.
from Internet searches and social media data, China, 2020. Am J Trop Med Hyg 2020; 103: 1621–29.
Euro Surveill 2020; 25: 2000199. 39 Kouzy R, Abi Jaoude J, Kraitem A, et al. Coronavirus goes viral:
18 Liu D, Wang Y, Wang J, et al. Characteristics and outcomes of a quantifying the COVID-19 misinformation epidemic on Twitter.
sample of patients with COVID-19 identified through social media Cureus 2020; 12: e7255.
in Wuhan, China: observational study. J Med Internet Res 2020; 40 Moscadelli A, Albora G, Biamonte MA, et al. Fake news and
22: e20108. COVID-19 in Italy: results of a quantitative observational study.
19 O’Leary DE, Storey VC. A Google–Wikipedia–Twitter model as a Int J Environ Res Public Health 2020; 17: 5850.
leading indicator of the numbers of coronavirus deaths. 41 Pulido CM, Villarejo-Carballido B, Redondo-Sama G, Gómez A.
Intell Syst Account Finance Manag 2020; 27: 151–58. COVID-19 infodemic: more retweets for science-based information
20 Worldometer. COVID-19 coronavirus pandemic. 2020. https://www. on coronavirus than for false information. Int Sociol 2020;
worldometers.info/coronavirus (accessed Sept 15, 2020). 35: 377–92.
21 Peng Z, Wang R, Liu L, Wu H. Exploring urban spatial features of 42 Rovetta A, Bhagavathula AS. Global infodemiology of COVID-19:
COVID-19 transmission in Wuhan based on social media data. analysis of Google web searches and Instagram hashtags.
ISPRS Int J Geoinf 2020; 9: 402. J Med Internet Res 2020; 22: e20673.
22 Qin L, Sun Q, Wang Y, et al. Prediction of number of cases of 43 Uyheng J, Carley KM. Bots and online hate during the COVID-19
2019 novel coronavirus (COVID-19) using social media search pandemic: case studies in the United States and the Philippines.
index. Int J Environ Res Public Health 2020; 17: 2365. J Comput Soc Sc 2020; published online Oct 20. https://doi.
23 Zhu Y, Fu KW, Grépin KA, Liang H, Fung IC. Limited early org/10.1007/s42001-020-00087-4.
warnings and public attention to coronavirus disease 2019 in China, 44 Gao J, Zheng P, Jia Y, et al. Mental health problems and social
January–February, 2020: a longitudinal cohort of randomly sampled media exposure during COVID-19 outbreak. PLoS One 2020;
Weibo users. Disaster Med Public Health Prep 2020; published online 15: e0231924.
April 3. https://doi.org/10.1017/dmp.2020.68. 45 Li S, Wang Y, Xue J, Zhao N, Zhu T. The impact of COVID-19
24 Basch CH, Hillyer GC, Meleo-Erwin ZC, Jaime C, Mohlman J, epidemic declaration on psychological consequences: a study on
Basch CE. Preventive behaviors conveyed on YouTube to mitigate active Weibo users. Int J Environ Res Public Health 2020; 17: 2032.
transmission of COVID-19: cross-sectional study. 46 Hakimi AA, Armstrong WB. Hand sanitizer in a pandemic: wrong
JMIR Public Health Surveill 2020; 6: e18807. formulations in the wrong hands. J Emerg Med 2020; 59: 668–72.
25 Basch CE, Basch CH, Hillyer GC, Jaime C. The role of YouTube and 47 Hernández-García I, Giménez-Júlvez T. Characteristics of YouTube
the entertainment industry in saving lives by educating and videos in Spanish on how to prevent COVID-19.
mobilizing the public to adopt behaviors for community mitigation Int J Environ Res Public Health 2020; 17: 4671.
of COVID-19: successive sampling design study. 48 Moon H, Lee GH. Evaluation of Korean-language COVID-19-related
JMIR Public Health Surveill 2020; 6: e19145. medical information on YouTube: cross-sectional infodemiology
26 Khatri P, Singh SR, Belani NK, et al. YouTube as a source of study. J Med Internet Res 2020; 22: e20775.
information on 2019 novel coronavirus outbreak: a cross-sectional 49 Ozdede M, Peker I. Analysis of dentistry YouTube videos related to
study of English and Mandarin content. Travel Med Infect Dis 2020; COVID-19. Braz Dent J 2020; 31: 392–98.
35: 101636. 50 Yüce MÖ, Adalı E, Kanmaz B. An analysis of YouTube videos as
27 Li L, Zhang Q, Wang X, et al. Characterizing the propagation of educational resources for dental practitioners to prevent the spread
situational information in social media during covid-19 epidemic: of COVID-19. Ir J Med Sci 2020; published online July 23.
a case study on Weibo. IEEE Trans Comput Soc Syst 2020; https://doi.org/10.1007/s11845-020-02312-5.
7: 556–62. 51 Al-Rawi A, Siddiqi M, Morgan R, Vandan N, Smith J, Wenham C.
28 Merkley E, Bridgman A, Loewen PJ, Owen T, Ruths D, Zhilin O. COVID-19 and the gendered use of emojis on Twitter:
A rare moment of cross-partisan consensus: elite and public infodemiology study. J Med Internet Res 2020; 22: e21646.
response to the COVID-19 pandemic in Canada. Can J Polit Sci 52 Arpaci I, Alshehabi S, Al-Emran M, et al. Analysis of twitter data
2020; published April 16. https://doi.org/10.1017/ using evolutionary clustering during the COVID-19 pandemic.
S0008423920000311. Comput Mater Contin 2020; 65: 193–204.
29 Rufai SR, Bunce C. World leaders’ usage of Twitter in response to 53 Barrett AE, Michael C, Padavic I. Calculated ageism: generational
the COVID-19 pandemic: a content analysis. J Public Health (Oxf) sacrifice as a response to the COVID-19 pandemic.
2020; 42: 510–16. J Gerontol Psychol Sci Soc Sci 2020; published online Aug 25.
30 Sutton J, Renshaw SL, Butts CT. COVID-19: Retransmission of https://doi.org/10.1093/geronb/gbaa132.
official communications in an emerging pandemic. PLoS One 2020; 54 Boon-Itt S, Skunkan Y. Public perception of the COVID-19
15: e0238491. pandemic on Twitter: sentiment analysis and topic modelling study.
31 Wang Y, Hao H, Platt LS. Examining risk and crisis JMIR Public Health Surveill 2020; 6: e21978.
communications of government agencies and stakeholders during 55 Budhwani H, Sun R. Creating COVID-19 stigma by referencing the
early-stages of COVID-19 on Twitter. Comput Human Behav 2021; novel coronavirus as the “Chinese virus” on Twitter: quantitative
114: 106568. analysis of social media data. J Med Internet Res 2020; 22: e19301.
32 Ahmed W, López Seguí F, Vidal-Alaball J, Katz MS. COVID-19 and 56 Chang A, Schulz PJ, Tu S, Liu MT. Blaming devices in online
the “Film Your Hospital” conspiracy theory: social network analysis communication of the COVID-19 pandemic: stigmatizing cues and
of Twitter data. J Med Internet Res 2020; 22: e22374. negative sentiment gauged with automated analytic techniques.
33 Ahmed W, Vidal-Alaball J, Downing J, López Seguí F. COVID-19 J Med Internet Res 2020; 22: e21504.
and the 5G conspiracy theory: social network analysis of Twitter 57 Chehal D, Gupta P, Gulati P. COVID-19 pandemic lockdown:
data. J Med Internet Res 2020; 22: e19458. an emotional health perspective of Indians on Twitter.
34 Brennen JS, Simon FM, Nielsen RK. Beyond (mis)representation: Int J Soc Psychiatry 2020; published online July 7. https://doi.org/
visuals in COVID-19 misinformation. Int J Press/Polit 2020; 10.1177/0020764020940741.
published online Oct 10. https://doi.org/10.1177/1940161220964780. 58 Chen Q, Min C, Zhang W, Wang G, Ma X, Evans R. Unpacking the
35 Bruns A, Harrington S, Hurcombe E. ‘Corona? 5G? or both?’: black box: how to promote citizen engagement through government
the dynamics of COVID-19/5G conspiracy theories on Facebook. social media during the COVID-19 crisis. Comput Hum Behav 2020;
Media Int Aust 2020; 177: 12–29. 110: 106380.

www.thelancet.com/digital-health Vol 3 March 2021 e193

 Review

59 Damiano AD, Allen Catellier JR. A content analysis of coronavirus 81 Nguyen TT, Criss S, Dwivedi P, et al. Exploring U.S. shifts in
tweets in the United States just prior to the pandemic declaration. anti-Asian sentiment with the emergence of COVID-19.
Cyberpsychol Behav Soc Netw 2020; published online Dec 14. Int J Environ Res Public Health 2020; 17: 7032.
https://doi.org/10.1089/cyber.2020.0425. 82 Odlum M, Cho H, Broadwell P, et al. Application of topic modeling
60 Darling-Hammond S, Michaels EK, Allen AM, et al. After “The to tweets as the foundation for health disparity research for
China Virus” went viral: racially charged coronavirus coverage and COVID-19. Stud Health Technol Inform 2020; 272: 24–27.
trends in bias against Asian Americans. Health Educ Behav 2020; 83 Park HW, Park S, Chong M. Conversations and medical news
47: 870–79. frames on twitter: infodemiological study on COVID-19 in
61 Das S, Dutta A. Characterizing public emotions and sentiments South Korea. J Med Internet Res 2020; 22: e18897.
in COVID-19 environment: a case study of India. 84 Pastor CK. Sentiment analysis of filipinos and effects of extreme
J Hum Behav Soc Environ 2020; published online July 14. community quarantine due to coronavirus (COVID-19) pandemic.
https://doi.org/10.1080/10911359.2020.1781015. SSRN 2020; published online April 13. https://papers.ssrn.com/
62 De Santis E, Martino A, Rizzi A. An infoveillance system for sol3/papers.cfm?abstract_id=3574385 (preprint).
detecting and tracking relevant topics from Italian tweets during the 85 Samuel J, Ali NGGM, Rahman MM, Esawi E, Samuel Y. COVID-19
COVID-19 event. IEEE Access 2020; 8: 132527–38. public sentiment insights and machine learning for tweets
63 Dheeraj K. Analysing COVID-19 news impact on social media classification. Information (Basel) 2020; 11: 314.
aggregation. Int J Adv Trends Comput Sci Eng 2020; 9: 2848–55. 86 Samuel J, Rahman MM, Ali GGMN, et al. Feeling positive about
64 Essam BA, Abdo MS. How do Arab tweeters perceive the COVID-19 reopening? New normal scenarios from COVID-19 US reopen
pandemic? J Psycholinguist Res 2020; published online Aug 14. sentiment analytics. IEEE Access 2020; 8: 142173–90.
https://doi.org/10.1007/s10936-020-09715-6. 87 Su Y, Xue J, Liu X, et al. Examining the impact of COVID-19
65 Yin FL, Lv JH, Zhang XJ, Xia XY, Wu JH. COVID-19 information lockdown in Wuhan and Lombardy: a psycholinguistic analysis on
propagation dynamics in the Chinese Sina-microblog. Weibo and Twitter. Int J Environ Res Public Health 2020; 17: 4552.
Math Biosci Eng 2020; 17: 2676–92. 88 Thelwall M, Thelwall S. COVID-19 tweeting in English: gender
66 Gozzi N, Tizzani M, Starnini M, et al. Collective response to media differences. Relaciones Públicas 2020; 29: e290301.
coverage of the COVID-19 pandemic on Reddit and Wikipedia: 89 Wang T, Lu K, Chow KP, Zhu Q. COVID-19 sensing: negative
mixed-methods analysis. J Med Internet Res 2020; 22: e21597. sentiment analysis on social media in China via BERT model.
67 Green J, Edgerton J, Naftel D, Shoub K, Cranmer SJ. Elusive IEEE Access 2020; 8: 138162–69.
consensus: polarization in elite communication on the COVID-19 90 Wicke P, Bolognesi MM. Framing COVID-19: how we conceptualize
pandemic. Sci Adv 2020; 6: eabc2717. and discuss the pandemic on Twitter. PLoS One 2020; 15: e0240010.
68 Han X, Wang J, Zhang M, Wang X. Using social media to mine and 91 Xi W, Xu W, Zhang X, Ayalon L. A thematic analysis of Weibo topics
analyze public opinion related to COVID-19 in China. (Chinese twitter hashtags) regarding older adults during the
Int J Environ Res Public Health 2020; 17: 2788. COVID-19 outbreak. J Gerontol Psychol Sci Soc Sci 2020; published
69 Jelodar H, Wang Y, Orji R, Huang S. Deep Sentiment classification online Sept 3. https://doi.org/10.1093/geronb/gbaa148.
and topic discovery on novel coronavirus or COVID-19 online 92 Xie T, Tan T, Li J. An extensive search trends-based analysis of
discussions: NLP using LSTM recurrent neural network approach. public attention on social media in the early outbreak of COVID-19
IEEE J Biomed Health Inform 2020; 24: 2733–42. in China. Risk Manag Healthc Policy 2020; 13: 1353–64.
70 Jimenez-Sotomayor MR, Gomez-Moreno C, Soto-Perez-de-Celis E. 93 Xue J, Chen J, Chen C, Hu R, Zhu T. The hidden pandemic of
Coronavirus, ageism, and Twitter: an evaluation of tweets about family violence during COVID-19: unsupervised learning of tweets.
older adults and COVID-19. J Am Geriatr Soc 2020; 68: 1661–65. J Med Internet Res 2020; 22: e24361.
71 Kim B. Effects of social grooming on incivility in COVID-19. 94 Yigitcanlar T, Kankanamge N, Preston A, et al. How can social
Cyberpsychol Behav Soc Netw 2020; 23: 519–25. media analytics assist authorities in pandemic-related policy
72 Kurten S, Beullens K. #Coronavirus: monitoring the Belgian Twitter decisions? Insights from Australian states and territories.
discourse on the severe acute respiratory syndrome coronavirus 2 Health Inf Sci Syst 2020; 8: 37.
pandemic. Cyberpsychol Behav Soc Netw 2020; published online 95 Yu J, Lu Y, Muñoz-Justicia J. Analyzing Spanish news frames on
Aug 27. https://doi.org/10.1089/cyber.2020.0341. Twitter during COVID-19–a network study of El País and El Mundo.
73 Kwon J, Grady C, Feliciano JT, Fodeh SJ. Defining facets of social Int J Environ Res Public Health 2020; 17: 5414.
distancing during the COVID-19 pandemic: Twitter analysis. 96 Zhao Y, Cheng S, Yu X, Xu H. Chinese public’s attention to the
J Biomed Inform 2020; 111: 103601. COVID-19 epidemic on social media: observational descriptive
74 Lai D, Wang D, Calvano J, Raja AS, He S. Addressing immediate study. J Med Internet Res 2020; 22: e18825.
public coronavirus (COVID-19) concerns through social media: 97 Zhu B, Zheng X, Liu H, Li J, Wang P. Analysis of spatiotemporal
utilizing Reddit’s AMA as a framework for public engagement with characteristics of big data on social media sentiment with
science. PLoS One 2020; 15: e0240326. COVID-19 epidemic topics. Chaos Solitons Fractals 2020; 140: 110123.
75 Li J, Xu Q, Cuomo R, Purushothaman V, Mackey T. Data mining 98 Schillinger D, Chittamuru D, Ramírez AS. From “infodemics” to
and content analysis of the Chinese social media platform Weibo health promotion: a novel framework for the role of social media in
during the early COVID-19 outbreak: retrospective observational public health. Am J Public Health 2020; 110: 1393–96.
infoveillance study. JMIR Public Health Surveill 2020; 6: e18700. 99 WHO. Novel coronavirus (2019-nCoV): situation report—13. 2020.
76 Li Y, Twersky S, Ignace K, et al. Constructing and communicating https://www.who.int/docs/default-source/coronaviruse/situation-
COVID-19 stigma on Twitter: a content analysis of tweets during the reports/20200202-sitrep-13-ncov-v3.pdf (accessed Feb 15, 2020).
early stage of the COVID-19 outbreak. 100 PsychCentral. Your emotional brain on resentment, part 2. 2020.
Int J Environ Res Public Health 2020; 17: 6847. https://pro.psychcentral.com/your-emotional-brain-on-resentment-
77 Lwin MO, Lu J, Sheldenkar A, et al. Global sentiments surrounding part-2 (accessed Sept 15, 2020).
the COVID-19 pandemic on Twitter: analysis of Twitter trends. 101 Daft RL, Lengel RH. Organizational information requirements,
JMIR Public Health Surveill 2020; 6: e19447. media richness and structural design. Manag Sci 1986; 32: 554–71.
78 Ma R, Deng Z, Wu M. Effects of health information dissemination 102 de Vreese CH, Neijens P. Measuring media exposure in a changing
on user follows and likes during COVID-19 outbreak in China: communications environment. J Commun Methods Meas 2016;
data and content analysis. Int J Environ Res Public Health 2020; 10: 69–80.
17: 5081.
103 Schmunk R. B.C. records highest number of fatal overdoses in a
79 Medford RJ, Saleh SN, Sumarsono A, Perl TM, Lehmann CU. single month, with 170 deaths. June 11, 2020. https://www.cbc.ca/
An “infodemic”: leveraging high-volume Twitter data to understand news/canada/british-columbia/overdose-deaths-bc-1.5607792
early public sentiment for the coronavirus disease 2019 outbreak. (accessed Sept 15, 2020).
Open Forum Infect Dis 2020; 7: ofaa258.
80 Mohamad SM. Creative production of ‘COVID-19 social distancing’ Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an
narratives on social media. Tijdsschr Econ Soc Geog 2020; Open Access article under the CC BY-NC-ND 4.0 license.
111: 347–59.

e194 www.thelancet.com/digital-health Vol 3 March 2021