ArticlePDF Available

Public Perception of SARS-CoV-2 Vaccinations on Social Media: Questionnaire and Sentiment Analysis

Authors:

Abstract and Figures

Vaccine hesitancy is an ongoing concern, presenting a major threat to global health. SARS-CoV-2 COVID-19 vaccinations are no exception as misinformation began to circulate on social media early in their development. Twitter’s Application Programming Interface (API) for Python was used to collect 137,781 tweets between 1 July 2021 and 21 July 2021 using 43 search terms relating to COVID-19 vaccines. Tweets were analysed for sentiment using Microsoft Azure (a machine learning approach) and the VADER sentiment analysis model (a lexicon-based approach), where the Natural Language Processing Toolkit (NLTK) assessed whether tweets represented positive, negative or neutral opinions. The majority of tweets were found to be negative in sentiment (53,899), followed by positive (53,071) and neutral (30,811). The negative tweets displayed a higher intensity of sentiment than positive tweets. A questionnaire was distributed and analysis found that individuals with full vaccination histories were less concerned about receiving and were more likely to accept the vaccine. Overall, we determined that this sentiment-based approach is useful to establish levels of vaccine hesitancy in the general public and, alongside the questionnaire, suggests strategies to combat specific concerns and misinformation.
Content may be subject to copyright.
International Journal of
Environmental Research
and Public Health
Article
Public Perception of SARS-CoV-2 Vaccinations on Social Media:
Questionnaire and Sentiment Analysis
Charlotte Roe 1, Madison Lowe 1, Benjamin Williams 2and Clare Miller 1,*


Citation: Roe, C.; Lowe, M.;
Williams, B.; Miller, C. Public
Perception of SARS-CoV-2
Vaccinations on Social Media:
Questionnaire and Sentiment
Analysis. Int. J. Environ. Res. Public
Health 2021,18, 13028. https://
doi.org/10.3390/ijerph182413028
Academic Editors: Paolo Roma,
Merylin Monaro, Cristina Mazza and
Anthony R. Mawson
Received: 24 September 2021
Accepted: 7 December 2021
Published: 10 December 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
School of Life Sciences, University of Lincoln, Lincoln LN6 7TS, UK; 15568465@students.lincoln.ac.uk (C.R.);
16655259@students.lincoln.ac.uk (M.L.)
2School of Computer Science, University of Lincoln, Lincoln LN6 7TS, UK; bWilliams@lincoln.ac.uk
*Correspondence: cmiller@lincoln.ac.uk; Tel.: +44-(0)1522-837364
Abstract:
Vaccine hesitancy is an ongoing concern, presenting a major threat to global health. SARS-
CoV-2 COVID-19 vaccinations are no exception as misinformation began to circulate on social media
early in their development. Twitter’s Application Programming Interface (API) for Python was
used to collect 137,781 tweets between 1 July 2021 and 21 July 2021 using 43 search terms relating to
COVID-19 vaccines. Tweets were analysed for sentiment using Microsoft Azure (a machine learning
approach) and the VADER sentiment analysis model (a lexicon-based approach), where the Natural
Language Processing Toolkit (NLTK) assessed whether tweets represented positive, negative or
neutral opinions. The majority of tweets were found to be negative in sentiment (53,899), followed by
positive (53,071) and neutral (30,811). The negative tweets displayed a higher intensity of sentiment
than positive tweets. A questionnaire was distributed and analysis found that individuals with
full vaccination histories were less concerned about receiving and were more likely to accept the
vaccine. Overall, we determined that this sentiment-based approach is useful to establish levels
of vaccine hesitancy in the general public and, alongside the questionnaire, suggests strategies to
combat specific concerns and misinformation.
Keywords:
SARS-CoV-2; COVID-19; vaccinations; sentiment analysis; Twitter; anti-vax; vaccine
hesitancy; Python; VADER; NLTK
1. Introduction
1.1. Coronavirus Disease 2019 in the UK and Vaccination Uptake
Coronavirus disease 2019 (COVID-19), caused by novel severe acute respiratory
Coronavirus 2 (SARS-CoV-2), was first reported in Wuhan, China in December 2019. As
has already been well reported, COVID-19spread rapidly across the globe and was declared
a pandemic by the World Health Organisation (WHO) in March 2020. In late January 2020,
the first case was reported in the United Kingdom (UK) and by the end of March 2020,
6650 cases had been recorded in the UK and a nationwide lockdown had begun [1].
On 8 December 2020, the UK became the first country to rollout a COVID-19 vaccina-
tion programme; and by 15 August 2021, an estimated 87.1% of the adult population in the
UK had received one dose of either the Oxford/AstraZeneca, Moderna or Pfizer-BioNTech
vaccine and 74.9% were fully vaccinated with two doses [
2
]. Even before the first dose was
administered, false rumours and misinformation had begun to circulate on social media,
at times fuelled by the idea that emergency regulatory approval of these vaccines was
linked to unreliability or safety concerns, threatening to diminish public confidence in the
vaccination programme [
3
]. By 15 August 2021, the cumulative total of deaths in the UK
where the death certificate mentioned COVID-19 as one of the causes was 157,361. The
cumulative total number of doses of vaccinations administered in the UK on the same date
was 88,037,283 [2].
Int. J. Environ. Res. Public Health 2021,18, 13028. https://doi.org/10.3390/ijerph182413028 https://www.mdpi.com/journal/ijerph
Int. J. Environ. Res. Public Health 2021,18, 13028 2 of 21
1.2. Anti-Vaccination Movement
Since their introduction, vaccinations have revolutionised health care whilst at the
same time persistently facing opposition [
4
,
5
] from hesitant individuals who perceive
them as unnecessary or dangerous [
6
]. ‘Anti-vaccinators’ or ‘anti-vaxxers’ may reject
vaccinations in the belief that they contain toxins and cause serious adverse effects [
7
].
More extreme conspiracy theories accuse pharmaceutical companies of producing fake
vaccine data, concealing harmful vaccine side effects and exaggerating vaccine efficacy
statistics [8].
Hesitancy is typically associated with a lack of trust in the health-care system [
9
] and
unfamiliarity with vaccine-preventable diseases [
10
]. For example, in 1974, it was reported
that an antigen in the pertussis vaccine was responsible for 36 neurological complications
including convulsions and intellectual developmental disorders in previously healthy
children. Despite the study concluding that these complications were extremely rare and the
risks of immunisation outweighed the risks of disease [
9
], many parents in Britain refused
to vaccinate their children against pertussis throughout the 1970s and 1980s. Between 1971
and 1974, vaccination rates dropped significantly from 78.5% to 37% [
11
], leading to severe
strain on the NHS [12,13].
The measles, mumps, and rubella (MMR) controversy was the result of a now discred-
ited paper linking the MMR vaccine to autism in children [
8
,
14
], which led to a reduction in
MMR uptake after its publication in 1998 and the debate still rumbles on. Although MMR
vaccination uptake has improved since 2004, according to the WHO, it is still under the 95%
threshold to ensure herd immunity; and in 2017, an estimated 142,000 people died from
measles unnecessarily [
6
,
15
,
16
], leading the WHO to declare vaccine hesitancy as an official
threat to global health in 2019 [
17
] and highlighting the need for medical professionals to
address vaccine safety concerns to encourage uptake.
1.3. Social Media and Vaccine Hesitancy
Web 2.0 has made discovering and sharing information online more convenient than
ever with the move from passive consumption to active generation of content, leading
to Health 2.0, where social media users share advice and experiences relating to health
care [
18
]. However, despite social media being readily utilised to promote public health,
and increasing numbers of people using social media to research vaccinations [
17
,
19
],
health-care professionals remain a key source of vaccine information [
20
]. Media and
celebrity opinion on social media is known to contribute to anti-vaccine beliefs [
21
] and the
way in which research is interpreted by the media can have a profound effect on influencing
public perception [
22
,
23
]. Scientists regularly challenge inaccurate information on social
media and one high-profile example of this occurred in September 2021, when Professor
Chris Whitty, the Chief Medical Officer for England and Chief Medical Advisor for the UK
Government, was asked at a televised press conference about a tweet by rapper Nicki Minaj
which claimed that her cousin’s friend was rendered impotent after taking a Coronavirus
vaccine which caused swelling in his testicles. Prof Whitty said that these “myths
. . .
untrue
. . .
designed to scare
. . .
they should be ashamed”, leading to a conversation which
continued afterwards in the media, including on social media. Despite progress being made
to combat false reporting of science [
23
], understanding reasons behind vaccine hesitation
will allow insight into how these beliefs may be counteracted effectively. Analysis of tweets
during a 2013 measles outbreak [
24
] noted users informing each other about the importance
of vaccination in light of the outbreak, illustrating a positive application of social media to
educate others regarding the importance of vaccines to prevent outbreaks of disease.
However, the echo-chamber effect described by Piedrahita-Valdés et al. (2021), ex-
plains how users with differing beliefs consume homogeneously polarised content re-
garding vaccines and form opposing groups who rarely communicate with one another
positively [
25
]. Hence, debate regarding vaccines may have little positive outcome, as prior
personal beliefs are only reinforced in this environment. Efforts by health professionals to
promote vaccination through social media have not always received a positive response;
Int. J. Environ. Res. Public Health 2021,18, 13028 3 of 21
and in extreme cases, health-care professionals have been threatened after posting videos
online encouraging vaccination [26].
During the UK national lockdowns in 2020 and 2021, much of the conversation
regarding COVID-19 took place on social media platforms including Twitter, which has
approximately 300 million monthly users [
27
,
28
]. Social media has become a common
platform for individuals to voice their concern and share their thoughts with others during
times of crisis [
29
]; but whilst these platforms allow the rapid dissemination of information,
there is no guarantee that the information is correct, reliable or accurate [
30
] and the majority
of anti-vaccination communication and conversation takes place over the internet [
31
].
Google search interest for the term ‘vaccine’ has greatly increased since March 2020, peaking
in March 2021 [32].
In a July 2020 UK survey, 16% of participants stated that they would be unlikely to
accept a COVID-19 vaccine [33]; and between September and October 2020, 12% and 17%
of individuals were strongly hesitant or very unsure, respectively [
34
]. The likelihood of
refusal of the COVID-19 vaccine was also found to be higher among young adults who are
indifferent about COVID-19 and lack trust in scientists [33].
1.4. Sentiment Analysis and Data Mining
Natural language processing (NLP) research topics rely heavily on the use of sentiment
analysis and opinion mining, where sentiment analysis is the study of opinions, feelings
and attitudes towards a product, organisation or event [
35
37
]. Opinion—or text—mining
involves extracting knowledge and information from online text, usually focusing on a
certain topic and categorising it as positive, negative or neutral [38,39].
Python is a versatile computer programming language which can manage large
datasets, making it ideal for use in complex projects [
40
42
]. It can be used to retrieve
tweets that contain chosen search terms and store them via a designated database engine,
such as SQLite. Valance Aware Dictionary and sEntiment Reasoner (VADER) is one of
many tools found within the popular Natural Language Toolkit (NLTK), with an excess
of 9000 lexicon features and the ability to analyse sentiments extracted from social media
sources. It produces a gold-standard sentiment lexicon by combining quantitative and
qualitative methods [
43
]. Sentiment lexicons contain lists with initial lexical capabilities
(words) categorised to a semantic orientation (i.e., positive or negative) [
38
,
44
]. The VADER
lexicon is a collection of predefined words with an associated polarity score—analysing the
positive and negative aspects of text and determining overall polarity. Typically, neutral
sentiments have a polarity score of 0 due to unidentifiable sentiment in the text. Neg-
ative and positive sentiments are assigned polarity scores of less than and greater than
0, respectively [
45
]. According to Satter et al. (2021), it is one of the easiest approaches
to sentiment classification [
28
] with VADER based on a gold-standard sentiment lexicon
with an ability to process acronyms and slang words [
46
], making it highly sensitive to
sentiment expressions when applied to social media contexts. Hutto and Gilbert (2014)
determined that VADER analysis performed better in comparison to eleven other highly re-
garded sentiment models and interestingly the accuracy of VADER has been determined to
outperform individual human analysers at correctly classifying the sentiment of tweets [
47
].
In the majority of machine learning approaches to sentiment classification, for example,
Microsoft Azure’s Text Analytics suite, a labelled dataset is required, whereby the polarity
of text is predefined. Whilst Azure’s graphical interface can be utilised by individuals with
little to no formal computer programming experience, making it an ideal software to use
for novices, VADER, on the other hand, requires domain-specific knowledge of computing
to use.
1.5. Sentiment Analysis of Vaccine Hesitance
Vaccine hesitancy is a fluid and ever-changing phenomenon [
47
]. Previous studies
have typically focused on vaccine hesitance in general rather than being directed at specific
vaccines and have revealed different trends across time [
25
,
48
]. Rahim et al. (2020) analysed
Int. J. Environ. Res. Public Health 2021,18, 13028 4 of 21
approximately 100,000 tweets about vaccinations between October 2019 and March 2020
and determined that the majority (41%) were positive in sentiment, closely followed by
neutral sentiment (39%) and 20% were negative [
48
]. COVID-19-specific vaccine hesitancy
has also been investigated: in May 2020, vaccine hesitancy rates were low (20–25%) in
American and Canadian adults [
49
], whereas, in Italy, the rates of COVID-19 vaccine
hesitancy were 41% [50] and 26% in France [51].
1.6. Research Involving Questionnaires
Before the explosion of online sentiment mining, researchers solely used qualitative
data collection methods in the form of surveys and particularly questionnaires [
52
]. Online
questionnaires have many advantages, including increased collection of data, decreased
cost and time to collect data and readily exportable formats for analytical simplicity [
53
,
54
].
To establish trends, attitudes and patterns, questionnaires are usually incorporated into
mixed-method research and often yield information that computer-based programs may
not identify. For example, questionnaires can extract demographic information and include
questions exploring the reasoning behind opinions [54].
1.7. Aims and Objectives
The overall aim of this study was to determine the sentiment of public opinion
regarding COVID-19 vaccinations. This was carried out via sentiment analysis of English
language tweets on Twitter and followed up with a questionnaire which was distributed
from the UK. The goal of the questionnaire was to explore attitudes to the expression
of any particular sentiment, rather than to find any specific correlation between the two.
Specifically, we aimed to determine the following:
1. Whether negative opinion regarding COVID-19 vaccines exists on Twitter.
2.
Whether lexicon-based (PYTHON/VADER) and machine learning (Microsoft Azure)
approaches to sentiment classification yield different sentiment results.
3.
Whether low levels of concern about COVID-19 vaccines lead to high acceptance of
the vaccine.
4.
Whether public opinion towards COVID-19 vaccinations becomes more positive
over time.
2. Materials and Methods
2.1. Data Collection
In order to share information on Twitter as widely as possible, Twitter provides broad
access to public Twitter data via their own Application Programming Interface (API). In
this study, Twitter’s official API was used to collect tweets in real time between 1 July 2021
and 21 July 2021. The language filter arguments “EN” and “RT” were applied to only select
English tweets and filter out re-tweets. Tweet scraping was conducted using 43 search
terms relating to COVID-19 vaccinations (Table 1) on Twitter’s asymmetric cryptography
(OAuth2) process and saved into an SQLite database. Following a small pilot study to
establish which key words would be most useful to investigate, key words were selected
based on the COVID-19 vaccines available in the UK at the time of data collection and also
to avoid collecting a large number of tweets that would have discussed vaccines in general
rather than being specifically related to COVID.
A total of 137,781 tweets were collected and stored in a database. Data collected
included the user’s display name, twitter handle, tweet text and date/time the tweet
was published.
Int. J. Environ. Res. Public Health 2021,18, 13028 5 of 21
Table 1. Text mining parameter details.
Parameters Details
Search terms
Vaccineforall, Vaccine, Antivaccine, Vaccinationcovid, Covid19, AstraZeneca, Astrazenecavaccine,
Pfizer, Pfizervaccine, UKvaccinerollout, Covidvaccine, Covidvaccination, Covid19vaccine,
Covid19vaccination, Modernavaccine, Oxfordvaccine, UKvaccine, AZvaccine, vaccinesideeffects,
Antivax, Antivaxxer, Antivaxxers, OxfordAZvaccine, Moderna, Modernasideffects,
Astrazenecasideffects, Pfizersideffects, Oxfordsideffects, seconddose, firstdose, Vaccineconspiracy,
UKfightscorona, Covid19UK, Covidenier, vaccinehesitancy, AZvax, modernavax, anti-vaccination,
anti-vax, anti-vaxxers, pro-vax, covid19jab
2.2. Sentiment Data Analysis—Machine Learning Approach (MLP)
Primary sentiment analysis was conducted on the dataset using Azure on Microsoft
Excel. The software yielded the results as ‘positive’, ‘negative’ or ‘neutral’ and scored the
confidence of the analysis, with a score of 1 being most confident with the analysis and
0 being least confident.
2.3. Sentiment Data Analysis—Lexicon-Based Approach
A Python-based API for Twitter was used to collect live tweets, which were recorded
into a relational database using SQLite. Sentiment analysis was performed post-collection
using the VADER algorithm, as part of the NLTK Python package. It is worth noting that
Python version 3.9.0 was used throughout this process. Custom-made software built with
Python 3.9.0 was used to perform the word frequency analysis. NLTK was used in the
pre-processing of tweets—to remove stop words—prior to the word frequency analysis.
The provided sentiment compound—or sentiment score—calculated from the sum of
lexicon ratings, was normalised between
1 (extreme negative) and +1 (extreme positive).
This technique determined the polarity—or positivity and negativity—and the intensity of
the expressed emotion. The intensity of emotion of each tweet is divided into the quantity
of positive, negative and neutral elements the tweet contained—adding to a total value
of 1. Each tweet was classified as positive, negative or neutral according to its compound
score. Compound scores less than 0.05 were considered negative, scores between
0.05
and 0.05 were considered neutral and scores above 0.05 were classified as positive [41,55].
2.4. Statistical Analysis
Descriptive statistics analysed differences between the program outputs and to test
for significance between approaches, sentiment frequency, and sentiment against time.
Questionnaire results were analysed on JISC (www.jisc.ac.uk, (accessed on 12 August
2021)) [
56
] automatically. Chi-square tests, two-way ANOVA and descriptive statistics
were performed on Microsoft Excel and Statistics Kingdom (www.Statskingdom.com,
(accessed on 14 August 2021)) [
57
] and Welch’s and two-sample t-tests were performed
using Python 3.9.0 and MATLAB.
2.5. Questionnaire
Using the JISC software to design, distribute and record the results, the questionnaire
(Table A1)—composed of 22 questions—was distributed to anonymous adult participants
(n= 182). The questionnaire was designed to investigate attitudes towards COVID-19
disease and COVID-19 vaccinations with the aim to determine personal knowledge and
opinion of vaccinations as well as identifying factors that may influence vaccine hesitancy.
Demographic data including age (18–29, 30–39, 40–49, 50–59, 60–69, and 70+) and parent-
hood status were recorded by the respondents. Questions including whether participants
have previously received vaccinations for themselves or their children and whether they
have accepted or will accept a COVID-19 vaccination were posed. Free-text opportunities
to elaborate on the reasons for declining vaccinations for themselves or their children
were provided. The participants were also asked agree/disagree-style questions relating
to COVID-19 vaccinations and their general knowledge surrounding vaccinations. The
Int. J. Environ. Res. Public Health 2021,18, 13028 6 of 21
questionnaire was distributed via email and social media platforms including Twitter and
Facebook. Incomplete responses were excluded from this study.
3. Results
3.1. Python Sentiment Analysis
3.1.1. Tweet Sentiment Scores
The VADER algorithm is the gold standard used among sentiment researchers [
47
].
Due to its wider term coverage [
58
], quick application [
41
] and high classification accu-
racy [
59
], we opted to use the results from this approach for the rest of this study. Between
1 July 2021 and 21 July 2021, Python scraped a total of 137,781 tweets relating to the chosen
search terms. The compound scores were plotted against time (Figure 1). There was no
obvious trend from the graphical representation, and therefore sentiment groups were
investigated individually.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 6 of 23
Demographic data including age (1829, 3039, 4049, 5059, 6069, and 70+) and
parenthood status were recorded by the respondents. Questions including whether par-
ticipants have previously received vaccinations for themselves or their children and
whether they have accepted or will accept a COVID-19 vaccination were posed. Free-text
opportunities to elaborate on the reasons for declining vaccinations for themselves or their
children were provided. The participants were also asked agree/disagree-style questions
relating to COVID-19 vaccinations and their general knowledge surrounding vaccina-
tions. The questionnaire was distributed via email and social media platforms including
Twitter and Facebook. Incomplete responses were excluded from this study.
3. Results
3.1. Python Sentiment Analysis
3.1.1. Tweet Sentiment Scores
The VADER algorithm is the gold standard used among sentiment researchers [47].
Due to its wider term coverage [58], quick application [41] and high classification accuracy
[59], we opted to use the results from this approach for the rest of this study. Between 1
July 2021 and 21 July 2021, Python scraped a total of 137,781 tweets relating to the chosen
search terms. The compound scores were plotted against time (Figure 1). There was no
obvious trend from the graphical representation, and therefore sentiment groups were
investigated individually.
Figure 1. VADER sentiment scores for each tweet. Values greater than 0.05 are displayed as positive, values between 0.05
and 0.05 are neutral and values less than 0.05 are negative tweets. The lengths of the peaks represent the intensity of
negativity or positivity. Values represent the tweet number. The horizontal axis shows the tweets in order, ranging from
1 July 2021 (left of graph) to 21 July 2021 (right of graph).
3.1.2. Word Frequency
The word count (Figure 2) shows the most frequently identified term was clearly
‘#covid19′ with other terms such as ‘people’, ‘get’ and ‘vaccine’ also frequently used. There
was no mention of specific groups such as ‘children’ or ‘parents’, only the collective term
‘people’.
A word cloud (Figure 3a) displays the most frequently used words in size descending
order. The larger-sized words depict a higher frequency of the word. To further under-
stand the relationship between words and their frequency, analysis into the most preva-
lent words was conducted from the separate positive, negative and neutral groups.
Figure 1.
VADER sentiment scores for each tweet. Values greater than 0.05 are displayed as positive, values between
0.05
and 0.05 are neutral and values less than 0.05 are negative tweets. The lengths of the peaks represent the intensity of
negativity or positivity. Values represent the tweet number. The horizontal axis shows the tweets in order, ranging from 1
July 2021 (left of graph) to 21 July 2021 (right of graph).
3.1.2. Word Frequency
The word count (Figure 2) shows the most frequently identified term was clearly
‘#covid19
0
with other terms such as ‘people’, ‘get’ and ‘vaccine’ also frequently used.
There was no mention of specific groups such as ‘children’ or ‘parents’, only the collective
term ‘people’.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 7 of 23
Figure 2. Top 50 frequently recurring words.
(a)
(b)
(c)
Figure 2. Top 50 frequently recurring words.
Int. J. Environ. Res. Public Health 2021,18, 13028 7 of 21
A word cloud (Figure 3a) displays the most frequently used words in size descending
order. The larger-sized words depict a higher frequency of the word. To further understand
the relationship between words and their frequency, analysis into the most prevalent words
was conducted from the separate positive, negative and neutral groups.
In the positive category (Figure 3b), the most commonly recurring words were
‘#covid19
0
(29,661), ‘people’ (5313) and ‘please’ (4455). In the neutral category
(Figure 3c)
,
the most commonly used words were ‘#covid19
0
(14,399), ‘people’ (2469) and ‘#vaccine’
(2322). In the negative category (Figure 3d), the most commonly used words were
‘#covid19
0
(31,725), ‘people’ (7925) and ‘get’ (4282). Noticeable words in this category
include ‘don’t’, ‘get’, ‘vaccinated’ and ‘death’, which could suggest that users are advising
others not to receive the vaccinations.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 7 of 23
Figure 2. Top 50 frequently recurring words.
(a)
(b)
(c)
Figure 3. Cont.
Int. J. Environ. Res. Public Health 2021,18, 13028 8 of 21
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 8 of 23
(d)
Figure 3. (a) Word cloud of the top fifty repeated words (https://wordart.com/, (accessed on 15 Au-
gust 2021)); (b) word cloud of the top twenty-five most repeated words in the positive category; (c)
word cloud of the top twenty-five most repeated words in the neutral category; (d). word cloud of
the top twenty-five most repeated words in the negative category.
In the positive category (Figure 3b), the most commonly recurring words were
‘#covid19′ (29,661), ‘people’ (5313) and ‘please’ (4455). In the neutral category (Figure 3c),
the most commonly used words were ‘#covid19′ (14,399), ‘people’ (2469) and ‘#vaccine’
(2322). In the negative category (Figure 3d), the most commonly used words were
‘#covid19′ (31,725), ‘people’ (7925) and ‘get’ (4282). Noticeable words in this category in-
clude ‘don’t’, ‘get’, ‘vaccinated’ and ‘death’, which could suggest that users are advising
others not to receive the vaccinations.
The frequency and percentage (Table 2) of the sentiment of tweets in each week were
determined to establish whether there was a trend across time between the groups.
During week 1, positive tweets were the most frequent (14,305; 39.0%) compared to
negative (13,900; 37.9%) and neutral (8398; 22.9%). By week 2 and week 3, negative tweets
(19,691; 39.0% and 20,308; 40.0%, respectively) were most frequent compared to positive
(19,394; 38.4% and 19,372; 38.1%) and neutral (11,352; 22.5% and 11,061; 21.7%) (Table 2,
Figure 4).
Figure 4. Frequency of negative, positive and neutral tweets over a 3 week period. The frequency of all sentiment groups
increased in week 2 compared to week 1. The frequency of negative tweets continued to increase into week 3, whereas
positive and neutral tweets slightly decreased.
Figure 3.
(
a
) Word cloud of the top fifty repeated words (https://wordart.com/, (accessed on 15
August 2021)); (
b
) word cloud of the top twenty-five most repeated words in the positive category;
(
c
) word cloud of the top twenty-five most repeated words in the neutral category; (
d
). word cloud of
the top twenty-five most repeated words in the negative category.
The frequency and percentage (Table 2) of the sentiment of tweets in each week were
determined to establish whether there was a trend across time between the groups.
Table 2. Frequency and percentages of tweets collected for each week.
Week Negative Tweets Positive Tweets Neutral Tweets Total
Frequency
Frequency Percentage (%) Frequency Percentage (%) Frequency Percentage (%)
1 13,900 37.9 14,305 39.0 8398 22.9 36,603
2 19,691 39.0 19,394 38.4 11,352 22.5 50,437
3 20,308 40.0 19,372 38.1 11,061 21.7 50,741
Total 53,899 53,071 30,811
During week 1, positive tweets were the most frequent (14,305; 39.0%) compared to
negative (13,900; 37.9%) and neutral (8398; 22.9%). By week 2 and week 3, negative tweets
(19,691; 39.0% and 20,308; 40.0%, respectively) were most frequent compared to positive
(19,394; 38.4% and 19,372; 38.1%) and neutral (11,352; 22.5% and 11,061; 21.7%) (Table 2,
Figure 4).
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 8 of 23
Figure 3. (a) Word cloud of the top fifty repeated words (https://wordart.com/, (accessed on 15 Au-
gust 2021)); (b) word cloud of the top twenty-five most repeated words in the positive category; (c)
word cloud of the top twenty-five most repeated words in the neutral category; (d). word cloud of
the top twenty-five most repeated words in the negative category.
In the positive category (Figure 3b), the most commonly recurring words were
‘#covid19′ (29,661), ‘people’ (5313) and ‘please’ (4455). In the neutral category (Figure 3c),
the most commonly used words were ‘#covid19′ (14,399), ‘people’ (2469) and ‘#vaccine’
(2322). In the negative category (Figure 3d), the most commonly used words were
‘#covid19′ (31,725), ‘people’ (7925) and ‘get’ (4282). Noticeable words in this category in-
clude ‘don’t’, ‘get’, ‘vaccinated’ and ‘death’, which could suggest that users are advising
others not to receive the vaccinations.
The frequency and percentage (Table 2) of the sentiment of tweets in each week were
determined to establish whether there was a trend across time between the groups.
During week 1, positive tweets were the most frequent (14,305; 39.0%) compared to
negative (13,900; 37.9%) and neutral (8398; 22.9%). By week 2 and week 3, negative tweets
(19,691; 39.0% and 20,308; 40.0%, respectively) were most frequent compared to positive
(19,394; 38.4% and 19,372; 38.1%) and neutral (11,352; 22.5% and 11,061; 21.7%) (Table 2,
Figure 4).
Figure 4. Frequency of negative, positive and neutral tweets over a 3 week period. The frequency of all sentiment groups
increased in week 2 compared to week 1. The frequency of negative tweets continued to increase into week 3, whereas
positive and neutral tweets slightly decreased.
Figure 4.
Frequency of negative, positive and neutral tweets over a 3 week period. The frequency of all sentiment groups
increased in week 2 compared to week 1. The frequency of negative tweets continued to increase into week 3, whereas
positive and neutral tweets slightly decreased.
To determine whether there was a significant difference between the frequency of
positive, negative and neutral scores, mean values were established for each week of data
collection (Figure 5).
Int. J. Environ. Res. Public Health 2021,18, 13028 9 of 21
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 9 of 23
Table 2. Frequency and percentages of tweets collected for each week.
Week
Negative Tweets
Positive Tweets
Neutral Tweets
Total Frequency
Frequency
Percentage (%)
Frequency
Percentage (%)
Frequency
Percentage (%)
1
13,900
37.9
14,305
39.0
8398
22.9
36,603
2
19,691
39.0
19,394
38.4
11,352
22.5
50,437
3
20,308
40.0
19,372
38.1
11,061
21.7
50,741
Total
53,899
53,071
30,811
To determine whether there was a significant difference between the frequency of
positive, negative and neutral scores, mean values were established for each week of data
collection (Figure 5).
Figure 5. Average values of negative, positive and neutral scores displayed over time. During week 2, the mean values for
neutral tweets are lower (>0.01) than the previous and following week.
A two-sample t-test with equal standard deviation was performed between the first
and final week of each sentiment group to investigate difference over time. The positive
average (0.508; SD = 0.511) during week 1 was found to be equal to the positive average
in week 3 (0.498; p = 0.110). The Test statistic (t = 1.597) was found in the 95% critical value
accepted range. The negative average (0.554; SD = 0.511) values during week 1 were
found to be equal to the negative average in week 3 (−0.553; p = 0.858). The Test statistic (t
= −0.177) was in the 95% critical value accepted range. The neutral average (0.00019; SD =
0.511) values during week 1 were found to be equal to the negative average in week 3
(0.00017; p = 0.997). The Test statistic (t = 0.003) was in the 95% critical value accepted
range.
3.1.3. Intensity of Sentiment
Week 1 (−0.345, 0.508, 0.00019) and week 3 (−0.358, 0.499, 0.00017) displayed similar
trends of negative, positive and neutral tweets, respectively (Figure 5). During week 2,
neutral tweets displayed more negativity than positivity (−1.322).
The means of tweets were subjected to a two-way ANOVA (Table 3). The difference
between weeks is not statistically significant (p = 0.1951), which is indicative of no signifi-
cant change in mean values between weeks. The difference between averages of the sen-
timent results (i.e., negative mean value against positive mean value against neutral mean
value) is statistically significant (p < 0.0001).
Figure 5.
Average values of negative, positive and neutral scores displayed over time. During week 2, the mean values for
neutral tweets are lower (>0.01) than the previous and following week.
A two-sample t-test with equal standard deviation was performed between the first
and final week of each sentiment group to investigate difference over time. The positive
average (0.508; SD = 0.511) during week 1 was found to be equal to the positive average
in week 3 (0.498; p= 0.110). The Test statistic (t= 1.597) was found in the 95% critical
value accepted range. The negative average (
0.554; SD = 0.511) values during week 1
were found to be equal to the negative average in week 3 (
0.553; p= 0.858). The Test
statistic (t=
0.177) was in the 95% critical value accepted range. The neutral average
(0.00019; SD = 0.511) values during week 1 were found to be equal to the negative average
in week 3 (0.00017; p= 0.997). The Test statistic (t= 0.003) was in the 95% critical value
accepted range.
3.1.3. Intensity of Sentiment
Week 1 (
0.345, 0.508, 0.00019) and week 3 (
0.358, 0.499, 0.00017) displayed similar
trends of negative, positive and neutral tweets, respectively (Figure 5). During week 2,
neutral tweets displayed more negativity than positivity (1.322).
The means of tweets were subjected to a two-way ANOVA (Table 3). The difference
between weeks is not statistically significant (p= 0.1951), which is indicative of no sig-
nificant change in mean values between weeks. The difference between averages of the
sentiment results (i.e., negative mean value against positive mean value against neutral
mean value) is statistically significant (p< 0.0001).
Table 3. Descriptive statistics of two-way ANOVA of the mean values of sentiment groups.
Source DF Sum of Square (SS) Mean Square (MS) F Statistic (df1df2)p-Value
Week 2 0.0001162 0.00005809 2.528 (2,4) 0.1951
Sentiment Groups 2 1.6833 0.8416 36,625.9271 (2,4) <0.001
Error 4 0.00009192 0.00002298
Total 8 1.6835 0.2104
Negative tweets had a higher mean value (0.52706) than positive (0.48196) and neutral
(0.50119) tweets (Table 4). To compare the means between the groups, Welch’s t-test (two-
sample t-test) was performed (due to unequal variance and differing n) using MATLAB.
Firstly, the values were normalised by mapping to the range of 0–1, where 0 is the “least”
and 1 is the “most”, i.e., negative tweets were mapped from [
1,
0.05] to [0, 1], where 0
is least negative (
0.05) and 1 is most negative (
1). This was achieved using an inverse
interpolation function (t
a)/(b
a), where t is the value, a is the lower bound and b is the
upper bound.
Int. J. Environ. Res. Public Health 2021,18, 13028 10 of 21
Table 4. Descriptive statistics of collected data, post-normalisation.
Category n1Mean Std. dev 2
Positive 53,071 0.48196 0.246031
Negative 53,899 0.52706 0.258930
Neutral 30,812 0.50119 0.066879
1Sample size; 2standard deviation.
Welch’s t-test demonstrated that positive vs. negative (p< 0.001), positive vs. neutral
(p< 0.001) and negative vs. neutral (p< 0.001) groups show statistical significance between
the means. This suggests that sentiment across our dataset displays a larger intensity of
negative sentiment compared to positive or neutral., i.e., the negative tweets are “more”
negative than the positivity in positive tweets.
3.2. Machine Learning vs. Lexicon Based: A Comparison of Negative, Positive and Neutral Tweets
The Natural Language Toolkit (or NLTK) (https://www.nltk.org/, (accessed on 21
July 2021)) [
60
] was used for the VADER sentiment analysis and scored 53,899 tweets as
negative, 53,071 as positive and 30,811 as neutral, whereas Azure determined the frequency
of the categories as 67,538, 45,282 and 24,961, respectively. They reveal similar trends
whereby most tweets were negative, followed by positive and neutral tweets being least
prevalent (Table 5).
Table 5.
Comparison between Python-based VADER and Microsoft Azure sentiment analysis approaches.
Parameters VADER Azure
Positive 53,071 45,282
Negative 53,899 67,538
Neutral 30,811 24,961
Median 0 0.459178
Mean 0.01978 0.445796
Variance 0.262321 0.071255
Skewness 0.04129 0.00218
SD 10.512173 0.266937
Total 137,781 137,781
1Standard deviation.
The lexicon-based (VADER) and machine learning (Microsoft Azure) approaches to
classify sentiment were compared (Table 5, Figure 6). A total of 39.11% of tweets were
scored as negative by VADER and 49.01% were scored as negative by Azure. The percentage
of tweets scored by VADER and Azure as positive were 38.51% and 32.86%, respectively. A
total of 22.36% and 18.11% were considered neutral.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 11 of 23
The lexicon-based (VADER) and machine learning (Microsoft Azure) approaches to
classify sentiment were compared (Table 5, Figure 6). A total of 39.11% of tweets were
scored as negative by VADER and 49.01% were scored as negative by Azure. The percent-
age of tweets scored by VADER and Azure as positive were 38.51% and 32.86%, respec-
tively. A total of 22.36% and 18.11% were considered neutral.
Figure 6. Total number of negative, positive and neutral tweets as determined by Microsoft Azure
and VADER.
3.3. Questionnaire
The questionnaire collected a total of 188 responses. A total of 6 responses were ex-
cluded due to the participants not meeting the requirements for this study or not agreeing
to their data being shared and so we used the complete 182 responses in the analysis (Ta-
ble A1).
A total of 31.9% of participants were between 18 and 29 years (the largest age group
of participants), with 90.1% stating they had previously searched for information regard-
ing COVID-19 online (e.g., Google). The most common length of time spent on social me-
dia was recorded as ‘daily’ (64.3%). Most of the participants (85.7%) had previously ac-
cepted all vaccines they had been offered), 73.8% were not concerned about receiving a
COVID-19 vaccination, 17.1% were slightly concerned, 4.3% were very concerned and
4.3% stated that they were impartial.
We asked whether participants had acceptedor will accepta COVID-19 vaccine.
Of the 182 participants, 8.2% have not/will not accept the vaccine, 1.6% said they did not
know, and the majority (90.1%) stated that they had already or would accept a vaccine.
The most likely reason (40.2%) for accepting a COVID-19 vaccine was ‘I want the world
to go back to how it used to be before the COVID-19 pandemic’, whereas the most com-
mon reason for not accepting the COVID-19 vaccine was ‘I have done my own research
and do not believe them to be safe’ (52.9%).
In response to whether the participants would allow their child under the age of 18
to have a COVID-19 vaccination if they were offered them in the future, 26.8% would not
vaccinate and 5.4% probably would not vaccinate their children against COVID-19. A total
of 17.9% were unsure whether they would vaccinate their children, 8.9% probably would
and 41.1% said yes, they would vaccinate their children. Participants with adult children
(18 or older) or without children automatically skipped this question. We compared level
of concern to vaccination acceptance or rejection (Figure 7). Out of 52 participants showing
some level of concern, 15 of these participants rejected the vaccine.
Figure 6.
Total number of negative, positive and neutral tweets as determined by Microsoft Azure
and VADER.
Int. J. Environ. Res. Public Health 2021,18, 13028 11 of 21
3.3. Questionnaire
The questionnaire collected a total of 188 responses. A total of 6 responses were
excluded due to the participants not meeting the requirements for this study or not agreeing
to their data being shared and so we used the complete 182 responses in the analysis
(Table A1).
A total of 31.9% of participants were between 18 and 29 years (the largest age group of
participants), with 90.1% stating they had previously searched for information regarding
COVID-19 online (e.g., Google). The most common length of time spent on social media
was recorded as ‘daily’ (64.3%). Most of the participants (85.7%) had previously accepted
all vaccines they had been offered), 73.8% were not concerned about receiving a COVID-19
vaccination, 17.1% were slightly concerned, 4.3% were very concerned and 4.3% stated that
they were impartial.
We asked whether participants had accepted—or will accept—a COVID-19 vaccine.
Of the 182 participants, 8.2% have not/will not accept the vaccine, 1.6% said they did not
know, and the majority (90.1%) stated that they had already or would accept a vaccine.
The most likely reason (40.2%) for accepting a COVID-19 vaccine was ‘I want the world to
go back to how it used to be before the COVID-19 pandemic’, whereas the most common
reason for not accepting the COVID-19 vaccine was ‘I have done my own research and do
not believe them to be safe’ (52.9%).
In response to whether the participants would allow their child under the age of 18
to have a COVID-19 vaccination if they were offered them in the future, 26.8% would not
vaccinate and 5.4% probably would not vaccinate their children against COVID-19. A total
of 17.9% were unsure whether they would vaccinate their children, 8.9% probably would
and 41.1% said yes, they would vaccinate their children. Participants with adult children
(18 or older) or without children automatically skipped this question. We compared level
of concern to vaccination acceptance or rejection (Figure 7). Out of 52 participants showing
some level of concern, 15 of these participants rejected the vaccine.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 12 of 23
Figure 7. The relationship between level of concern and acceptance and rejection of a COVID-19 vaccine.
We asked how the participants would consider their current depth of knowledge re-
garding vaccinations generally. Knowledge scores ranged from 0 (no knowledge) to 5
(deep/thorough knowledge). Overall, 2.2% stated that they had no understanding, 74.2%
felt they had some understanding, and 23.6% had a deep understanding.
Several chi-square tests (significance level, alpha, of 0.05) were performed to deter-
mine whether there was an association between certain vaccine refusal prediction factors
(Table 6). The results show that the uptake of COVID-19 vaccines was dependent on pre-
vious vaccine history (p < 0.001) and an individuals’ level of concern (p < 0.001). However,
vaccination understanding (p = 0.949491), age (p = 0.057899) and time spent on social me-
dia (p = 0.925771) did not influence the acceptance of COVID-19 vaccinations. Chi-square
analysis was also performed between responses of the statement ‘Vaccine safety and ef-
fectiveness data are often false’ and intensity of concern and found a significant relation-
ship (p < 0.001) (Table 6). The majority of respondents who were not concerned about
receiving a COVID-19 vaccine ‘strongly disagreed’ with the statement (52.89%), whereas
those who were most concerned stated that they ‘don’t know’ (42.86%).
Table 6. Chi-square statistical analysis to determine a dependent association between accepting a COVID-19 and the var-
iables in the table. Vaccine safety (far right column) was analysed against how concerned the participant was.
Parameters
Vaccine
Knowledge
Age
Time on Social Media
Vaccine His-
tory
Level of Concern
Vaccine Safety
Chi-Square
(Observed value)
2.14521
14.25356
3.421087
56.18451
116.8076
54.87902
Chi-Square
(Critical value)
9.487729
18.30704
15.50731
9.487729
12.59159
9.487729
DF
6
10
8
4
6
15
p-value
0.905871
0.161737
0.905227
<0.001
<0.001
<0.001
4. Discussion
4.1. Machine Learning vs. Lexicon-Based Approaches
Sentiment analysis research has become popular over the past two decades [40,61,62];
as more efficient sentiment classification models are devised [63] and studies have com-
pared automated analysis of conversations on social media with manual approaches [64].
Prior studies have compared machine learning methods of text analysis (i.e., SVM)
with lexicon-based approaches [28,65,66] and often conclude the machine learning meth-
ods are more effective. For example, Sattar et al. (2021) concluded that VADER was less
Figure 7. The relationship between level of concern and acceptance and rejection of a COVID-19 vaccine.
We asked how the participants would consider their current depth of knowledge
regarding vaccinations generally. Knowledge scores ranged from 0 (no knowledge) to 5
(deep/thorough knowledge). Overall, 2.2% stated that they had no understanding, 74.2%
felt they had some understanding, and 23.6% had a deep understanding.
Several chi-square tests (significance level, alpha, of 0.05) were performed to deter-
mine whether there was an association between certain vaccine refusal prediction factors
(Table 6)
. The results show that the uptake of COVID-19 vaccines was dependent on previ-
ous vaccine history (p< 0.001) and an individuals’ level of concern (p< 0.001). However,
vaccination understanding (p= 0.949491), age (p= 0.057899) and time spent on social media
(p= 0.925771) did not influence the acceptance of COVID-19 vaccinations. Chi-square
analysis was also performed between responses of the statement ‘Vaccine safety and effec-
Int. J. Environ. Res. Public Health 2021,18, 13028 12 of 21
tiveness data are often false’ and intensity of concern and found a significant relationship
(p< 0.001) (Table 6). The majority of respondents who were not concerned about receiving
a COVID-19 vaccine ‘strongly disagreed’ with the statement (52.89%), whereas those who
were most concerned stated that they ‘don’t know’ (42.86%).
Table 6.
Chi-square statistical analysis to determine a dependent association between accepting a COVID-19 and the
variables in the table. Vaccine safety (far right column) was analysed against how concerned the participant was.
Parameters Vaccine Knowledge Age Time on Social Media Vaccine History Level of Concern Vaccine Safety
Chi-Square
(Observed value) 2.14521 14.25356 3.421087 56.18451 116.8076 54.87902
Chi-Square
(Critical value) 9.487729 18.30704 15.50731 9.487729 12.59159 9.487729
DF 6 10 8 4 6 15
p-value 0.905871 0.161737 0.905227 <0.001 <0.001 <0.001
4. Discussion
4.1. Machine Learning vs. Lexicon-Based Approaches
Sentiment analysis research has become popular over the past two decades [
40
,
61
,
62
];
as more efficient sentiment classification models are devised [
63
] and studies have com-
pared automated analysis of conversations on social media with manual approaches [64].
Prior studies have compared machine learning methods of text analysis (i.e., SVM)
with lexicon-based approaches [
28
,
65
,
66
] and often conclude the machine learning methods
are more effective. For example, Sattar et al. (2021) concluded that VADER was less accurate
than machine learning applications and used TextBlob in their study [
28
]. However,
Dhaoui et al. (2015) determined that both approaches performed similarly when analysing
Facebook reviews for both positive and negative classification [
67
]. Much of the literature
on this is contradictory and highlights the need for continued research in this area of
comparing the accuracy and precision of the machine and lexicon methods. For example,
Nguyen et al. (2018) stated that SVM displayed 89% accuracy and 90% precision in
comparison to VADER (83% and 90%, respectively) [
68
], whereas in a different study,
SVM’s accuracy and precision were different (71.8% and 66.8% and, respectively) as were
that of lexicon-based approaches (71.1% and 65.1% and, respectively) [
69
]. Despite much
of the literature claiming the inferiority of lexicon-based approaches, our research required
classification of how positive and negative online sentiment was: one advantage of the
VADER model [41].
In other studies, Microsoft Azure has been found to yield better results when com-
pared to other analyser tools such as Stanford NLP [
64
], IBM Watson Natural Language
Understanding, OpinionFinder 2.0 and Sentistrength [
70
]. However, as Azure only identi-
fies polarity, it is a less accurate method of measuring an individual’s opinion towards a
topic compared to other approaches such as VADER [
71
] and so part of this study compared
the sentiment analysis approaches of Microsoft Azure and VADER.
Previous studies have explored sentiment surrounding COVID-19 vaccinations on
Twitter [
72
,
73
]. Xue et al. (2020) used Latent Dirichlet Allocation (LDA)—a machine
learning approach—and collected four million tweets on COVID-19 using 25 search words.
Their aim was to identify popular themes, sentiment, bigrams and unigrams. The NRC
Emotion Lexicon classified sentiments into several emotions including anger, fear, surprise,
sadness, disgust, joy, trust and anticipation and revealed that Twitter users display ‘fear’
when discussing new cases of COVID-19, as opposed to ‘trust’ [
74
]. Bhagat et al. (2020)
used TextBlob to perform sentiment analysis and scraped 154 articles from blogging and
news websites. Over 90% of the articles were positive and blogs were found to be more
positive than newspaper articles [
75
]. Sattar et al. (2021) adopted a similar approach to
the present study, analysing COVID-19 vaccine sentiment using a large number of tweets
(n= ~1.2 million) using a lexicon-based classifier, namely VADER and TextBlob. They
also defined their neutral sentiments between
0.05 and 0.05 and determined that public
sentiment was more positive than negative.
Int. J. Environ. Res. Public Health 2021,18, 13028 13 of 21
4.2. Word Identification and Word Frequency
The results confirm that negativity towards the COVID-19 vaccines is present on Twit-
ter alongside tweets that are positive and neutral in sentiment. Similar studies corroborate
these results [10,49,76], with suggestions that development speed and safety concerns are
some of the reasons why hesitancy is expressed [
77
]. Chandrasekaran et al. examined the
trends of sentiment of several topics associated with COVID-19 between January 2020 and
May 2020 and found that although Twitter users expressed negativity about the spread
and symptoms of COVID-19, they determined that positive feelings were expressed when
sharing information on drugs and new therapies [
55
]. In the present study, the commonly
used term ‘people’ suggests that concerns do not specifically relate to children, elderly
or any other specific group. Although the hashtag ‘#covid19
0
was the most frequently
occurring word in all three sentiment groups, analysis found that a higher number of
negative tweets contained the hashtag (31,725) in comparison to positive (29,661) and
neutral (14,399) tweets.
A study on the sentiment surrounding human papillomavirus vaccines found different
keywords associated within their word clusters. The authors suggested that ‘HPV’ was
associated with personal words including ‘I’ and ‘me’ and ‘#HPV’ was associated with
words such as ‘learn’ and ‘prevent’. The authors considered these ‘awareness-raising
words’ [
78
]. Our findings show similar results; ‘people’, ‘don’t’, ‘health’, ‘vaccines’, and
‘death’ were noticeable in the negative groups. This could also be indicative of concerns
about the risks of accepting the vaccine [
79
]. Words including ‘people’, ‘please’, ‘help’
‘vaccine’ ‘first’ and ‘need’ were found to be frequently occurring in the positive group. These
terms suggest that discourse leans towards promotion and encouragement of vaccinating,
with similar key words found in previous studies [
79
]. The only similarities of the word
frequencies performed by Sattar et al. (2021) and this study were ‘death’ and ‘people’ in
the negative category, ‘vaccine’ in the positive category and ‘help’ and ‘first’ in both the
positive and neutral categories. They also identified words that were not found in our
study including ‘party,’ ‘happy’ and ‘thank’ [28].
Previous research suggests that social media users tend to interact with others who
share common beliefs and ignore or argue with individuals who have opposite views
[80,81]
,
creating an echo chamber. Due to this, it has been suggested that public health interventions
could reinforce vaccine hesitancy [
81
83
] and identifying keywords or hashtags that hesi-
tant individuals commonly use would be a more effective strategy [
84
] to countering the
problem. This study has identified several keywords and hashtags to assist in this process.
4.3. Relative Frequency of Tweets
We observed the frequency and relative frequency of tweets in each week of this study.
Despite most of the tweets in the dataset being negative, positive tweets (14,305; 39.0%)
were the most predominant during the first week of data collection between 1 July 2021
and 7 July 2021 whereas, in the final two weeks, between 8 July 2021 and 21 July 2021,
negative tweets (19,691; 39.0% and 20,308; 40%) were most common. Neutral tweets were
significantly lower than both negative and positive tweets throughout the entire time of
collection (22.9%, 22.5% and 21.7%). Piedrahita-Valdes et al. (2021) performed sentiment
analysis on vaccine-hesitant tweets between June 2011 and April 2019 and found neutral
tweets were predominant throughout the study, in contrast to the present study. They
also found that negative tweets peaked at times and noted that at least one of these peaks
coincided with a documentary linking autism to vaccines. Similarly, they identified positive-
related peaks occurring in April which coincided with World Immunisation week [
25
].
Furthermore, a noticeable increase in anti-vaccine discourse was experienced on Twitter
in 2015, coinciding with a measles outbreak (2014–2015), a newly released film “Vaxxed”
and the publication of the book “Vaccine Whistleblower” [
17
], supporting the idea that
conversations relating to vaccine hesitancy fluctuate over time.
The mean of neutral tweets displayed a negative sentiment compound (
0.00000132)
during week 2 of the investigation, whereas, in weeks 1 and 3, neutral tweets were positive
Int. J. Environ. Res. Public Health 2021,18, 13028 14 of 21
(0.000199 and 0.000177, respectively). This is suggestive of concurrent events that the
general public are exposed to [
17
] such as case numbers, the reporting of daily hospitalisa-
tion and death figures, the pace of the UK vaccination programme and the expansion of
testing capability in addition to wider political factors including legislated social distancing,
lockdowns, working from home mandates and face mask wearing. For example, on 5
July 2021 plans to remove the mandated wearing of facemasks from 19 July 2021 were
announced in England. This announcement could have been a key factor in the high
positive sentiment we detected in this study in week 1. By 7 July 2021, however, the UK’s
weekly COVID-19 cases had doubled in comparison to the week prior; and between 8 and
14 July (corresponding to week 2 in this study), cases continued to rise in the UK, with over
50,000 new cases reported on 17 July 2021 [
85
]. As these events unfolded, 1200 scientists
formally challenged the easing of lockdown restrictions in England [
86
], a discussion that
is likely to have added to the negative sentiment at the time. Public opinion remained
polarised and by week 3 of our study, we found the highest frequency of tweets which
reflected negative sentiment at the same time as the number of tweets that were positive in
sentiment increased from week 2 (38.4%) to week 3 (47.6%). Whilst previous research has
identified vaccine hesitancy fluctuating over time [
17
], it would be interesting to compare
the dates of specific announcements and wider discussions with daily sentiment analysis
to determine whether there is a relationship between the two.
4.4. Questionnaire: Vaccine Hesitancy towards COVID-19 Vaccinations
Our study is the only one to date to incorporate a questionnaire alongside the explo-
ration of sentiment analysis on Twitter towards COVID-19 vaccinations. Most respondents
(90.1%) had or would accept a COVID-19 vaccine, a view that is in line with conclu-
sions drawn by other studies [
87
,
88
] whilst others have reported less public support for
COVID-19 vaccinations [89].
The identification of factors that might predict hesitancy towards COVID-19 vaccines
was investigated. A positive correlation between intensity of concern regarding vaccines
and their uptake was established, suggesting that participants with higher levels of (or
more intense) concern are less likely to accept the vaccine, whereas those with low levels
(less intense) or no concern are more likely to accept the COVID-19 vaccine.
Additional predictors of vaccine hesitancy were explored by considering whether
age, vaccine history, level of vaccine understanding and usage of social media were likely
to influence an individual’s decision to take a COVID-19 vaccination. No association
was established between vaccine refusal and age, despite the Pew Research Group (2017)
finding younger adults (<30 years) were less likely to consider beneficial aspects of the
MMR vaccine outweighed the risks, compared to older age groups [
90
]. The same study
found individuals with higher levels of understanding considered the risk of vaccine side
effects as low, whereas there was no association found between vaccination understanding
and vaccination uptake in our study. Survey research on COVID-19 vaccine hesitancy
corroborated our results by also finding no association between age and vaccine refusal [
91
]
although Bendau et al. (2021) did establish an association between vaccine hesitance and
concern [
92
]. Interestingly, 17.2% of respondents in the present study somewhat or strongly
agreed that “vaccine safety and effectiveness data are often false”, suggesting a significant
proportion of the general public have concerns trusting this information as evidenced
previously [
9
]. Anecdotal evidence from the questionnaire suggests that participants are
more likely to write negative comments. This view is supported by the literature where it is
understood that negative emotions (such as anger, frustration, sadness and disappointment)
motivate individuals to articulate their views [93,94].
Reports suggest that the acceptance of vaccines in emergency situations (such as a
pandemic) differs to that of routinely administered vaccines in non-crisis situations [
87
].
However, contrastingly, public concerns surrounding safety are higher with the uncertain-
ties that come with novel vaccines and new emerging infectious diseases [
87
,
95
97
]. For
example, in the UK, France, Greece, America and Australia, only 17% to 67% of the general
Int. J. Environ. Res. Public Health 2021,18, 13028 15 of 21
public was willing to accept the vaccine for the H1N1 pandemic in 2009 [
95
102
], highlight-
ing public concern in this area and also likely variable uptake figures.
Chaudhri et al. (2021)
established the public had a weakly positive sentiment towards receiving a COVID-19
vaccine [
73
]. Vaccination history has previously been identified as a major predictor of
vaccine uptake [
95
,
98
,
101
,
103
], a view also identified in the present study which established
an association between vaccine history and acceptance. Individuals with full previous
vaccination history were more likely to accept a COVID-19 vaccine, further confirming the
idea of the echo chamber effect.
The present study has confirmed the idea that vaccine compliance remains inconsis-
tent with negative opinions and hesitancy still widespread [
91
,
92
] and the inclusion of
a questionnaire provided a greater picture of overall sentiment towards vaccines. The
questionnaire revealed generally positive sentiment, whereas more negative sentiment
was found online, alongside positive and neutral views. The questionnaire revealed that
concerns about vaccines typically centred around trust in safety and effectiveness.
4.5. Limitations and Further Work
As part of the pilot work for the present study, we manually categorised the sources
(Twitter accounts) as ‘personal’, ‘accredited medical’, ‘news’ or ‘government/public health’.
It would have been helpful if we could have extended this into the main study to facilitate
a better understanding of the most common sources of misinformation. However, with the
large dataset in the main study, this was unrealistic, and we seek an automated approach
to this for future studies.
The data were collected over a short period in July 2021 and so it would be interesting
to extend this study to look at historical and future tweets to further understand whether
public opinion regarding COVID-19 vaccinations changed during the course of the pan-
demic. It would also be interesting to compare the dates of specific events in the media
with daily sentiment analysis to determine whether they are closely related.
The questionnaire was distributed via social media and so responses were limited to
people with access and were typically in the authors’ extended networks. Future studies
should endeavour to distribute the questionnaire more widely and in particular to reach
public without access to social media. Concern exists in the UK that certain groups are
more susceptible to vaccine misinformation and we would like to reach those communities
with future research. This is also the case with the sentiment analysis which only collected
tweets in English and therefore had the potential to miss the view of non-English speaking
groups in the UK.
A simplified interface would benefit this research as the low accuracy of Microsoft
Azure and the complexity of using data mining and analysis tools such as Python requires
specific computing expertise. Thus, a simplified graphical interface is in development that
would benefit future projects seeking to collect datasets for analysis without a need for an
understanding of Python or the VADER algorithm.
Sentiment analysis is a popular and rapidly developing area. An interesting avenue
for further research would be to compare our approach using VADER to other language-
encoder-based approaches (such as using Bert or GPT), in particular exploring whether
these could be useful developments that would work with NLTK.
5. Conclusions
This study established that machine learning and lexicon-based sentiment analysis
methods yielded different frequencies of sentiment results. Negative sentiment was found
to be most frequent online, with a higher intensity of negativity within the neutral tweets.
There was no significant change in sentiment towards COVID-19 across the three-week
data collection period. Positive correlations were established between COVID-19 vaccine
acceptance with full vaccination history and low levels of concern.
Sentiment analysis provides evidence to assess public perception about various top-
ics [
104
], allowing officials in charge of managing the impact of COVID-19 and health
Int. J. Environ. Res. Public Health 2021,18, 13028 16 of 21
policy makers insight into how the public feel about vaccination safety and efficacy so they
can identify areas and misconceptions that need to be addressed [93,94].
The identification of frequently occurring negative terms and of predictors that influ-
ence vaccine hesitancy can be utilised to deploy effective strategies such as educational
campaigns to increase public confidence in the COVID-19 vaccines and improve vaccine
uptake. To ensure vaccination uptake targets are met, this requires continued attention.
Author Contributions:
Conceptualisation, C.M., M.L. and C.R.; methodology, C.M., M.L., C.R. and
B.W.; analysis, C.M., C.R. and B.W.; writing—original draft preparation, M.L. and C.R.;
writing—review
and editing, C.M., C.R. and B.W.; supervision, C.M. All authors have read and agreed to the published
version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement:
The study was conducted according to the guidelines of the
Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of
the School of Life Sciences, University of Lincoln. Reference: BGY9013M15568465, 17 June 2021.
Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.
Data Availability Statement:
Data available on request due to ethical restrictions. The data presented
in this study are available on request from the corresponding author. The data are not publicly
available due to conditions of ethical approval.
Acknowledgments: The authors would like to thank Jonathan Roe and Jamie Smith.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A
Table A1.
Summary of the raw data from participants’ answers (n= 182). Due to the different nature of written response
options to certain questions, these have been distinguished with quotation marks.
Question Responses (%)
1What is your age? 18–29
(31.9) 30–39
(17.6) 40–49
(12.1) 50–59
(20.9) 60–69
(13.2) 70+
(4.4)
2
Have you used a search engine
(e.g., Google) since January 2020 to
search for information about
Coronavirus or COVID-19?
Yes
(90.1) No
(9.4) Don’t know
(0.6)
3
How often do you use social media
(e.g., Twitter, Instagram, Facebook
and Snapchat)
Never
(2.7) Rarely (2.2) Monthly (0.0) Weekly
(3.8)
Daily
(64.3)
More
frequently
than daily
(26.9)
4Do you believe that information on
social media is reliable?
Always
reliable
(1.1)
Sometimes
reliable (70.9)
Rarely reliable
(24.2)
Never reliable
(2.7) Don’t know
(1.1)
5Have you ever tested positive for
COVID-19?
Yes
(7.7) No
(92.3) Don’t know
(0.0)
6
As far as you are aware, have you
accepted all of the vaccinations you
have been invited to (excluding
COVID-19) since the age of 18?
Yes I have had
all vaccin-
ations I have
been invited to
(85.7)
I have had some
of my
vaccinations
(8.2)
I have not had
any of my
vaccinations
(2.7)
I have not had
vaccinations
due to an
underlying
cause
(0.5)
I have
decided to
opt out of
vaccinations
(2.7)
Don’t know
(0.0)
7Have you already or are you going to
accept a vaccine against COVID-19?
Yes
(90.1) No
(8.2) Don’t know
(1.6)
7a If you selected don’t know, please
specify: (optional)
Response 1: “Too early to be sure of safety.”
Response 2: “Not sure if I will have my second vaccine.”
Response 3: “I would like to know more long term side effects before committing to being vaccinated.”
8Have you received a vaccination to
protect you against COVID-19
Yes
(98.2) No
(1.8) Don’t know
(0.0)
9Which vaccine did you receive? Pfizer
(49.1)
Oxford Astra
Zeneca
(48.4)
Modern
(1.9)
Janssen
(Johnson &
Johnson)
(0.0)
Don’t know
(0.6) Other
(0.0)
10
Are you concerned about accepting the
COVID-19 vaccine/did you have
concerns before receiving the vaccine?
I am not/was
not concerned
(73.8)
I feel/felt
impartial
(4.3)
I am/was
slightly
concerned
(17.1)
I am/was very
concerned
(4.3)
Other
(0.6)
10a If you selected other, please specify:
(optional)
Response 1: “I’m informed about side effects and don’t believe what you see in the news without looking at
the actual data. So initially concerned but not after looking into the clotting issue.”
Int. J. Environ. Res. Public Health 2021,18, 13028 17 of 21
Table A1. Cont.
Question Responses (%)
11
Why did (or why will) you
accept the COVID-19 vaccine?
(Please select the most
likely reason)
I have done
my own
research and I
believe them
to be safe
(20.7)
I want the world to
go back to how it
used to be before
the COVID-19
pandemic
(40.2)
I know of or have
lost someone to
COVID-19 who did
not receive the
vaccination in time
(5.5)
For protection
for myself
(27.4)
Other
(6.1)
11a If you selected other, please
specify: (optional)
Response 1: “Mainly to protect others.”
Response 2: “For protection of the weak and vulnerable as well as myself.”
Response 3: “Family member I care for is vulnerable otherwise I may have declined.”
Response 4: “NHS worker.”
Response 5: “Protection for my high risk family (mother and father).”
12
Why did (or why will) you not
accept the COVID-19 vaccine?
(tick all that apply)
I worry I
might get
COVID019
(0.0)
I have done my
own research and I
do not believe
them to be safe
(52.9)
I worry about the
adverse reactions
(23.5)
I do not believe
the trials have
been long
enough to ensure
accurate results
(64.7)
Other
(23.5)
12a If you selected other, please
specify: (optional)
Response 1: “I have had both vaccine doses.”
Response 2: “I have an immune system. The majority of people do not need a vaccine for covid 19 . . . . In my
opinion. My mother also had a severe adverse reaction to the Astra Zeneca jab and is now suffering high
blood pressure.”
Response 3: “I’ve had the flu jab—that’s all I needed!”
Response 4: “I keep myself fit and healthy, I do not have any medical conditions, I ensure I eat a balanced diet and
maintain a normal BMI, I exercise frequently and take my general health very seriously thus I did not feel it
necessary to have the vaccine. I felt that pressure from colleagues, family and social media made me feel like I didn’t
have a choice. I work in an nhs hospital.”
13
If you have children, what age
are they? (If you have multiple
children, please select the age
of the youngest)
0–4 years
(16.3)
5–10 years
(7.6)
11–15 years
(4.1)
16–17 years
(1.2)
18 years +
(32.6)
I do not
have
children
(38.4)
14
As of 1 July 2021 in the UK,
children under the age of 18
are not routinely offered a
COVID-19 vaccine. If this
changed and children were
offered the vaccine, would you
give permission for your
child/children to have
the vaccine?
Yes
(41.1)
Probably
(8.9)
Don’t know
(17.9)
Probably not
(5.4)
No
(26.8)
15
If you selected no/probably
not to the previous question,
please tick the most
relevant box
They have an
underlying
disorder that
prevents them
from having
vaccinations
(0.0)
I do not trust what
is in the vaccine
(22.2)
I do not believe
that they work
(0.0)
I do not want
them to suffer
possible long
term adverse
reactions
(50.0)
Other
(27.8)
15a If you selected other, please
specify: (optional)
Response 1: “Given that the effects on children of the virus is known and proven to be low on children on balance I
don’t think any benefits outweigh the negatives as the vaccine has not been out for long.”
Response 2: “Children were never in the at risk group. I believe this experimental poison that’s only approved for
EMERGENCY use (e.g., not approved like measles/chicken pox/meningitis) will cause life changing side effects or
even death. How many dead children from this vaccine are acceptable? 1? 10? 100? We are vaccinating a population
over a disease with a 99.7% survival rate-oh and it’s not even 100% effective!”
Response 3: “Covid 19 does not affect children . . . why would anyone vaccinate a child against something that
wouldn’t cause them any harm in the first place?”
Response 4: “I would like to see more long term data on infants receiving a vaccine before making my mind.”
16
Have/would you use Twitter
to find out information about
COVID-19 or Coronavirus?
Yes
(11.5) No
(83.5) Don’t know
(4.9)
17
I would describe my attitude
towards receiving a COVID-19
vaccine as:
Very
interested
(52.7)
Interested
(19.2) Neutral
(12.1)
Uneasy
(8.8)
Against it
(7.1)
Don’t know
(0.0)
18
If friends or family were
offered a COVID-19 vaccine
I would:
Strongly
encourage
them
(61.0)
Encourage them
(19.8)
Not say anything
(12.1)
Discourage them
(1.6)
Strongly
discourage
them
(3.3)
Don’t know
(2.2)
19 Taking a COVID-19
vaccination is:
Extremely
important
(64.6)
Important
(21.5)
Neither important
nor unimportant
(6.1)
Unimportant
(2.2)
Extremely
unimpor-
tant
(2.8)
Don’t know
(2.8)
20
Do you consider the COVID-19
vaccine more dangerous than
the COVID-19 disease?
Strongly agree
(6.6)
Somewhat agree
(6.6)
Neither agree nor
disagree
(7.7)
Somewhat
disagree
(12.1)
Strongly
disagree
(64.3)
Don’t know
(2.7)
21
Vaccine safety and
effectiveness data are
often false
Strongly agree
(5.0)
Somewhat agree
(12.2)
Neither agree nor
disagree
(16.0)
Somewhat
disagree
(20.4)
Strongly
disagree
(40.3)
Don’t know
(6.1)
22
How would you describe your
general knowledge of
vaccinations?
Deep/thorough
understanding
(23.6)
Some
understanding
(74.2)
No understanding
(2.2)
Don’t know
(0.0)
Int. J. Environ. Res. Public Health 2021,18, 13028 18 of 21
References
1.
Chadeau-Hyam, M.; Bodinier, B.; Elliot, J.; Whitaker, M.D.; Tzoulaki, I.; Vermeulen, R.; Kelly-Irving, M.; Delpierre, C.; Elliot, P.
Risk factors for positive and negative COVID-19 tests: A cautious and in-depth analysis of UK biobank data. Int. J. Epidemiol.
2020,49, 1454–1467. [CrossRef]
2.
Public Health England, Coronavirus (COVID-19) in the UK. 2021. Available online: https://coronavirus.data.gov.uk (accessed on
8 July 2021).
3.
Puri, N.; Coomes, E.A.; Haghbayan, H.; Gunaratne, K. Social media and vaccine hesitancy: New updates for the era of COVID-19
and globalized infectious diseases. Hum. Vaccine Immunother. 2020,16, 2586–2593. [CrossRef] [PubMed]
4.
Kata, A. Anti-vaccine activists, Web 2.0, and the postmodern paradigm—An overview of tactics and tropes used online by the
anti-vaccination movement. Vaccine 2012,30, 3778–3789. [CrossRef] [PubMed]
5. Jacobson, R.M.; St Sauver, J.L.; Finney Rutten, L.J. Vaccine Hesitancy. Mayo Clin. Proc. 2015,90, 1562–1568. [CrossRef]
6.
Cacciatore, M.A.; Nowak, G.J.; Evans, N.J. It’s complicated: The 2014–2015 US measles outbreak and parents’ vaccination beliefs,
confidence, and intentions. Risk Anal. 2018,38, 2178–2192. [CrossRef] [PubMed]
7. Smith, N.; Graham, T. Mapping the anti-vaccine movement on Facebook. Inf. Commun. Soc. 2017,22, 1310–1327. [CrossRef]
8.
Taylor, B.; Miller, E.; Farrington, C.P.; Petropoulos, M.C.; Favot-Mayaud, I.; Li, J.; Waight, P.A. Autism and measles, mumps, and
rubella vaccine: No epidemiological evidence for a causal association. Lancet 1999,353, 2026–2029. [CrossRef]
9.
Kulenkampff, M.; Schwartzman, J.S.; Wilson, J. Neurological complications of pertussis inoculation. Arch. Dis. Child.
1974
,49,
46–49. [CrossRef]
10.
Salmon, D.A.; Dudley, M.Z.; Glanz, J.M.; Omer, S.B. Vaccine hesitancy: Causes, consequences, and a call to action. Vaccine
2015
,
33, 66–71. [CrossRef]
11.
Swansea Research Unit of the Royal College of General Practitioners. Effect of a low pertussis vaccination take-up on a large
community. Br. Med. J. (Clin. Res. Ed.) 1981,282, 23–26. [CrossRef]
12. Begg, N.; White, J.; Bozoky, Z. Media dents confidence in MMR vaccine. Br. Med. J. 1998,316, 561. [CrossRef]
13.
Gangarosa, E.; Galazka, A.M.; Wolfe, C.R.; Philips, L.M.; Gangarosa, R.E.; Miller, E.; Chen, R.T. Impact of anti-vaccine movements
on pertussis control: The untold story. Lancet 1998,351, 356–361. [CrossRef]
14.
Hansen, N.D.; Mølbak, K.; Cox, I.J.; Lioma, C. Relationship between Media Coverage and Measles-Mumps-Rubella (MMR)
vaccination Uptake in Denmark: Retrospective Study. JMIR Pub. Health Surveill. 2019,5, e9544. [CrossRef] [PubMed]
15.
World Health Organisation. News Updates. 2021. Available online: https://www.who.int/news-room/news-updates (accessed
on 14 August 2021).
16. Kupferschmidt, K. Study pushes emergence of measles back to antiquity. Science 2020,367, 11–12. [CrossRef] [PubMed]
17.
Gunaratne, K.; Coomes, E.A.; Haghbayan, H. Temporal trends in anti-vaccine discourse on Twitter. Vaccine
2019
,37, 4867–4871.
[CrossRef]
18.
Scanfeld, D.; Scanfeld, V.; Larson, E.L. Dissemination of health information through social networks: Twitter and antibiotics. Am.
J. Infect. Control 2010,38, 182–188. [CrossRef]
19.
Betsch, C.; Brewer, N.T.; Brocard, P.; Davies, P.; Gaissmaier, W.; Haase, N.; Leask, J.; Renkewitz, F.; Renner, B.; Reyna, V.F.; et al.
Opportunities and challenges of Web 2.0 for vaccine decisions. Vaccine 2012,30, 3727–3733. [CrossRef] [PubMed]
20.
Ortiz-Sánchez, E.; Velando-Soriano, A.; Pradas-Hernández, L.; Vargas-Román, K.; Gómez-Urquiza, J.L.; Cañdas-De la Fuente, G.;
Albendín-García, L. Analysis of the Anti-Vaccine Movement in Social Networks: A Systematic Review. Int. J. Environ. Res. Public
Health 2020,17, 5394. [CrossRef] [PubMed]
21. Anderson, P. Another media scare about MMR vaccine hits Britain. Br. Med. J. 1999,318, 1578. [CrossRef]
22.
Diresta, R.; Lotan, G. Anti-Vaxxers Using Twitter to Manipulate a Vaccine Bill. 2015. Available online: https://cran.rproject.org/
web/packages/SocialMediaLab/SocialMediaLab.pdf (accessed on 19 July 2021).
23. Sense about Science. 2021. Available online: https://senseaboutscience.org (accessed on 11 August 2021).
24.
Mollema, L.; Harmsen, I.A.; Broekhuizen, E.; Clijnk, R.; Melker, H.D.; Paulussen, T.; Kok, G.; Ruiter, R.; Das, E. Disease Detection
or Public Opinion Reflection? Content Analysis of Tweets, Other Social Media, and Online Newspapers During the Measles
Outbreak in the Netherlands in 2013. J. Med. Internet Res. 2015,17, e128. [CrossRef]
25.
Piedrahita-Valdes, H.; Piedrahita-Castillo, D.; Bermejo-Higuera, J.; Guillem-Saiz, P.; Bermejo-Higuera, J.R.; Guillem-Saiz, J.;
Sicilia-Montalvo, J.A.; Machio-Regidor, F. Vaccine hesitancy on social media: Sentiment analysis from June 2011 to April 2019.
Vaccines 2021,9, 28. [CrossRef] [PubMed]
26.
Kullar, R.; Goff, D.A.; Gauthier, T.P.; Smith, T.C. To Tweet or Not to Tweet—A Review of the Viral Power of Twitter for Infectious
Diseases. Technol. Infect. Dis. 2020,22, 14. [CrossRef]
27.
Twitter Revenue and Usage Statistics. 2021. Available online: https://www.businessofapps.com/data/twitter-statistics/
(accessed on 14 August 2021).
28.
Sattar, N.S.; Arifuzzaman, S. COVID-19 Vaccination awareness and aftermath: Public sentiment analysis on Twitter data and
vaccinated population prediction in the USA. Appl. Sci. 2021,11, 6128. [CrossRef]
29.
Neubaum, G.; Rosner, L.; Rosenthal-von der Putten, A.; Kramer, N.C. Psychosocial functions of social media usage in a disaster
situation: A multi-methodological approach. Comput. Hum. Behav. 2014,34, 28–38. [CrossRef]
30.
Skinner, G. Who’s least likely to say they’ll get a COVID-19 vaccine? Ipsos Mori.
2020
. Available online: https://www.ipsos.com/
en-uk/whos-least-likely- say-they\T1\textquoterightll-get- covid-19-vaccine (accessed on 17 August 2021).
Int. J. Environ. Res. Public Health 2021,18, 13028 19 of 21
31.
Freeman, D.; Loe, B.S.; Chadwick, A.; Vaccari, C.; Waite, F.; Rosebrock, L.; Jenner, L.; Petit, A.; Lewandowsky, S.;
Vanderslott, S.; et al.
COVID-19 vaccine hesitancy in the UK: The Oxford coronavirus explanations, attitudes, and narra-
tives survey (Oceans) II. Psychol. Med. 2020, 1–15. [CrossRef]
32. Google Trends. 2021. Available online: www.trends.google.com (accessed on 19 July 2021).
33. Kaur, C.; Sharma, A. Twitter Sentiment Analysis on Coronavirus using Textblob. EasyChair Prepr. 2020,2974, 1–10.
34. Tholana, V.S.S.A. Literature Review on Sentiment Analysis. Int. J. Adv. Res. Ideas Innov. Technol. 2017,3, 5.
35.
Li, N.; Wu, D.D. Using text mining and sentiment analysis for online forums hotspot detection and forecast. Decis. Support Syst.
2010,48, 354–368. [CrossRef]
36.
Bari, M.D.; Sharoff, S.; Thomas, M. Sentiml: Functional annotation for multilingual sentiment analysis. ICPS Proc.
2013
,15, 1–7.
[CrossRef]
37.
Ghassi, M.; Skinner, J.; Zimbra, D. Twitter brand sentiment analysis: A hybrid system using n-gram analysis and dynamic
artificial neural network. Expert Syst. Appl. 2013,40, 6266–6282. [CrossRef]
38.
Tane, J.; Schmitz, C.; Stumme, G. Semantic resource management for the web: An e-learning application. In Proceedings of the
13th International World Wide Web conference on Alternate Track Papers and Posters, New York, NY, USA, 19–21 May 2004;
pp. 1–10. [CrossRef]
39.
Raghupathi, V.; Ren, J.; Raghupathi, W. Studying public perception about vaccination: A sentiment analysis of Tweets. Int. J.
Environ. Res. Public Health 2020,17, 3464. [CrossRef] [PubMed]
40.
Cieliebak, M.; Durr, O.; Uzdilli, F. Potential and Limitations of Commercial Sentiment Detection Tools; Zurich University of Applied
Sciences: Winterthur, Switzerland, 2013.
41.
Bonta, V.; Janardhan, N.K.N. A Comprehensive Study on Lexicon Based Approaches for Sentiment Analysis. Asian J. Comput. Sci.
Technol. 2019,8, 1–6. [CrossRef]
42.
Shelar, A.; Huang, C.-Y. Sentiment Analysis of Twitter Data. In Proceedings of the International Conference on Computational
Science and Computational Intelligence, Las Vegas, NV, USA, 12–14 December 2018; pp. 1301–1302. [CrossRef]
43.
Amin, A.; Hossain, I.; Akther, A.; Alam, K.M. Bengali VADER: A sentiment analysis approach using modified VADER. In
Proceedings of the 2019 International Conference on Electrical, Computer and Communication Engineering (ECCE), Cox’sBazar,
Bangladesh, 7–9 February 2019; pp. 1–6. [CrossRef]
44.
Hatzivassiloglou, V.; McKeown, K.R. Predicting the semantic orientation of adjectives. In Proceedings of the 35th Annual
Meeting of the Association for Computational Linguistics and Eighth Conference of the European Chapter of the Association for
Computational Linguistics, Madrid, Spain, 7–12 July 1997; pp. 174–181.
45.
Sarker, D.; Bali, R.; Sahrma, T. Analyzing movie reviews sentiment. In Practical Machine Learning with Python: A Problem-Solver’s
Guide to Building Real-World Intelligent Systems; Apress: Berkley, CA, USA, 2018; pp. 246–342.
46.
Ao, S.; Chan, A.S.; Castillo, O.; Kim, H.; Katagiri, H. Transactions on Engineering Sciences: Special Issue for the International Association
of Engineers Conferences 2019, IAENG; World Scientific Publishing Pte: Singapore, 2020; pp. 72–73.
47.
Hutto, C.; Gilbert, E. VADER: A Parsimonious Rule-Based Model for Sentiment Analysis of Social Media Text [Scholarly Project]; Comp.
Social Lab: Atlanta, GA, USA, 2014; Available online: http://comp.social.gatech.edu/papers/icwsm14.vader.pdf (accessed on 1
August 2021).
48.
Rahim, N.H.A.; Rafie, S.H.M. Sentiment analysis of social media data in vaccination. Int. J. Emerg. Trends Eng. Res.
2020
,8,
5259–5264. [CrossRef]
49.
Taylor, S.; Landry, C.A.; Paluszek, M.M.; Groenewoud, R.; Rachoe, G.S.; Asmundson, G.J.G. A proactive Approach for Managing
COVID-19: The importance of Understanding the Motivational Roots of Vaccination Hesitancy for SARS-CoV-2. Front. Psychol.
2020,11, 575950. [CrossRef]
50.
Graffigna, G.; Palamenghi, L.; Boccia, S.; Barello, S. Relationship between Citizens’ Health Engagement and Intention to Take the
COVID-19 Vaccine in Italy: A Mediation Analysis. Vaccines 2020,8, 576. [CrossRef]
51.
Group, C. A future vaccination campaign against COVID-19 at risk of vaccine hesitancy and politicization. Lancet Infect. Dis.
2020,20, 769–770.
52.
Zucco, C.; Paglia, C.; Graziano, S.; Bella, S.; Cannataro, M. Sentiment analysis and text mining of questionnaires to support
telemonitoring programs. Information 2020,11, 550. [CrossRef]
53. Allery, L.A. Design and use questionnaires for research in medical education. Educ. Prim. Care 2016,27, 234–238. [CrossRef]
54.
McGuirk, P.M.; O’Neill, P. Using Questionnaires in Qualitative Human Geography. Faculty of Social Sciences—Papers. 2016,
Volume 2518, pp. 246–273. Available online: https://ro.uow.edu.au/sspapers/2518 (accessed on 16 August 2021).
55.
Chandrasekaran, R.; Mehta, V.; Valkunde, T.; Moustakas, E. Topics, Trends, and Sentiments of Tweets About the COVID-19
Pandemic: Temporal Infoveillance Study. J. Med. Internet Res. 2020,22, e22624. [CrossRef]
56. JISC Online Surveys. 2021. Available online: www.JISC.ac.uk (accessed on 1 July 2021).
57. StatsKingdom. 2021. Available online: www.statskingdom.com (accessed on 13 August 2021).
58.
D’Andrea, A.; Ferri, F.; Grifoni, P.; Guzzo, T. Approaches, Tools and Applications for Sentiment Analysis Implementation. Int. J.
Comput. Appl. 2015,125, 0975–8887. [CrossRef]
59.
Hutto, C.J.; Gilbert, E. VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Test. Assoc. Adv. Artif.
Intell. 2015. Available online: https://www.researchgate.net/publication/275828927_VADER_A_Parsimonious_Rule- based_
Model_for_Sentiment_Analysis_of_Social_Media_Text (accessed on 12 August 2021).
Int. J. Environ. Res. Public Health 2021,18, 13028 20 of 21
60. NLTK. 2021. Available online: https://www.nltk.org/ (accessed on 1 July 2021).
61.
Karn, A.; Shrestha, A.; Pudasaini, A.; Mahara, B.; Jaiswal, A. Statistic based Sentiment Analysis of Social Media Data. Int. Res. J.
Innov. Eng. Technol. 2018,2, 28–32.
62.
Mantyla, M.V.; Graziotin, D.; Kuutila, M. The evolution of Sentiment Analysis—A review of research topics, venues and top cited
papers. Comput. Sci. Rev. 2018,27, 16–32. [CrossRef]
63.
Abdelwahab, O.; Bahgat, M.; Lowrance, C.J.; Elmaghraby, A. Effect of training set size on SVM and Naïve Bayes for Twitter
sentiment analysis. In Proceedings of the 2015 IEEE International Symposium on Signal Processing and Information Technology
(ISSPIT), Abu Dhabi, United Arab Emirates, 7–10 December 2015; pp. 46–51.
64.
Canhoto, A.I.; Padmanabhan, Y. ‘We (don’t) know how you feel’—A comparative study of automated vs. manual analysis of
social media conversations. J. Mark. Manag. 2015,13, 1141–1157. [CrossRef]
65.
Chaovalit, P.; Zhou, L. Movie review mining: A comparison between supervised and unsupervised classification approaches. In
System Sciences, Proceedings of the 38th Annual Hawaii International Conference HICSS’05, Big Island, HI, USA, 3–6 January 2005; IEEE:
Piscataway, NJ, USA, 2005; pp. 1–9.
66.
Kolchyna, O.; Souza, T.T.P.; Treleaven, P.C.; Aste, T. Twitter sentiment analysis: Lexicon method, machine learning method and
their combination. arXiv 2015, arXiv:1507.00955.
67.
Dhaoui, C.; Webster, C.M.; Tan, L.P. Social media sentiment analysis: Lexicon versus machine learning. J. Consum. Mark.
2017
,34,
480–488. [CrossRef]
68.
Nguyen, H.; Veluchamy, A.; Diop, M.; Iqbal, R. Comparative study of sentiment analysis with product reviews using machine
learning and lexicon based approaches. Data Sci. Rev. 2018,1, 4–7.
69.
Lu, B.; Tsou, B.K. Combining a large sentiment lexicon and machine learning for subjectivity classification. In Proceedings of the
2010 International Conference on Machine Learning and Cybernetics, Qingdao, China, 11–14 July 2010; pp. 3311–3316. [CrossRef]
70.
Dolianiti, F.S.; Iakovakis, D.; Dias, S.B.; Hadjileontiadou, S.J.; Diniz, J.A.; Natsiou, G.; Tsitoiuridou, M.; Bamidis, P.D.;
Hadjileontiadis, L.J.
Sentiment analysis on educational datasets: A comparative evaluation of commercial tools. Educ. J. Univ.
Patras UNESCO Chair 2019,6, 262–273.
71.
Sweigart, A. Invent Your Own Computer Games with Python. 2nd Edition. 2012. Available online: https://inventwithpython.
com/ (accessed on 19 July 2021).
72.
Villavicencio, C.; Macrohon, J.J.; Inbaraj, X.A.; Jeng, J.H.; Hsieh, J.G. Twitter Sentiment Analysis towards COVID-19 Vaccines in
the Philippines Using Naïve Bayes. Information 2021,12, 204. [CrossRef]
73.
Chaudhri, A.A.; Saranya, S.; Dubey, S. Implementation Paper on Analyzing COVID-18 Vaccines on Twitter Dataset Using Tweepy
and Text Blob. Ann. Rom. Soc. Cell Biol. 2021,25, 8393–8396.
74.
Xue, J.; Chen, J.; Hu, R.; Chen, C.; Zheng, C.; Liu, X.; Zhu, T. Twitter discussions and emotions and COVID-19 pandemic: A
machine learning approach (2020). arXiv 2020, arXiv:2005.12830.
75.
Baghat, K.K.; Mishra, S.; Dixit, A.; Chang, C.Y. Public Opinions about online learning during COVID-19: A sentiment analysis
approach. Sustainability 2021,13, 3364. [CrossRef]
76.
Del Vicario, M.; Bessi, A.; Zollo, F.; Petroni, F.; Scala, A.; Caldarelli, G.; Quattrociocchi, W. The spreading of misinformation online.
Proc. Natl. Acad. Sci. USA 2016,113, 554–559. [CrossRef] [PubMed]
77.
Dredze, M.; Broniatowski, D.A.; Hilyard, K.M. Zika vaccine misconception: A social media analysis. Vaccine
2016
,34, 3441.
[CrossRef] [PubMed]
78.
Massey, P.M.; Leader, A.; Yom-Tov, E.; Budena, A.; Fisher, K.; Klasse, A.C. Applying multiple data collection tools to quantify
Human Papillomavirus vaccine communication on Twitter. J. Med. Internet Res. 2016,18, 12. [CrossRef]
79.
Mahajan, R.; Romine, W.; Miller, M.; Banerjee, T. Analyzing public outlook towards vaccination using Twitter. In Proceedings
of the 2019 IEEE International Conference on Big Data (Big Data), Los Angeles, CA, USA, 9–12 December 2019; pp. 2763–2772.
[CrossRef]
80.
Yuan, X.; Schuchard, R.J.; Crooks, A.T. Examining emergent communities and social bots within the polarized online vaccination
debate in Twitter. Soc. Media+ Soc. 2019,5, 20563051198546. [CrossRef]
81.
Meyer, S.B.; Violette, R.; Aggarwal, R.; Simeoni, M.; MacDoughall, H.; Waite, N. Vaccine hesitancy and Web 2.0: Exploring how
attitudes and beliefs about influenza vaccination are exchanged in online threaded user comments. Vaccine
2019
,37, 1769–1774.
[CrossRef]
82.
Nyhan, B.; Reifler, J.; Richey, S.; Freed, G.L. Effective messages in vaccine promotion: A randomized trial. Pediatrics
2014
,133,
e835–e842. [CrossRef]
83.
Nyhan, B.; Reifler, J. Does correcting myths about the flu vaccine work? An experimental evaluation of the effects of corrective
information. Vaccine 2015,33, 459–464. [CrossRef]
84.
Massey, P.M.; Budenz, A.; Leader, A.; Fisher, K.; Klassen, A.C.; Yom-Tov, E. What drives heath professionals to tweet about # HPV
vaccine? Identifying strategies for effective communication. Prev. Chronic Dis. 2018,15, 15. [CrossRef]
85.
Medicines & Healthcare Products Regulatory Agency, MHRA Guidance on Coronavirus (COVID-19). 2021. Available online:
https://www.gov.uk/government/collections/mhra- guidance-on-coronavirus-covid-19 (accessed on 13 July 2021).
86.
British Foreign Policy Group. COVID-19 Timeline. 2021. Available online: https://bfpg.co.uk/2020/04/covid-19-timeline/
(accessed on 14 August 2021).
Int. J. Environ. Res. Public Health 2021,18, 13028 21 of 21
87.
Huang, C.; Wang, Y.; Li, X.; Ren, L.; Zhao, J.; Hu, Y.; Zhang, L.; Fan, G.; Xu, J.; Gu, X.; et al. Clinical features of patients infected
with 2019 novel coronavirus in Wuhan, China. Lancet 2020,395, 497–506. [CrossRef]
88.
Dodd, R.H.; Pickles, K.; Nickel, B.; Cvejic, E.; Ayre, J.; Batcup, C.; Bonner, C.; Copp, T.; Cornell, S.; Dakin, T.; et al. Concerns and
motivations about COVID-19 vaccination. Lancet Infect. Dis. 2021,21, 163–191. [CrossRef]
89.
Chowdhury, A.A.; Das, A.; Saha, S.K.; Rahman, M.; Hasan, K.T. Sentiment Analysis of COVID-19 Vaccination from Survey
Responses in Bangladesh. Researchsquare Preprint. 2021. Available online: https://assets.researchsquare.com/files/rs-482293
/v1_stamped.pdf?c=1620421587 (accessed on 16 August 2021).
90.
Pew Research Centre. Parents of Young Children Are More ‘Vaccine Hesitant’; Pew Research Centre: Washington, DC, USA, 2017;
Available online: https://www.pewresearch.org/fact-tank/2017/02/06/parents-of-young- children-are-more-vaccine-hesitant/
(accessed on 13 August 2021).
91.
Dror, A.A.; Eisenbach, N.; Taiber, S.; Morozov, N.G.; Mizrachi, M.; Zigron, A.; Srouji, S.; Sela, E. Vaccine hesitancy: The next
challenge in the fight against COVID-19. Eur. J. Epidemiol. 2020,35, 775–779. [CrossRef]
92. Bendau, A.; Plag, J.; Petzold, M.B.; Strohle, A. COVID-19 vaccine hesitancy and related fears and anxiety. Int. Immunopharmacol.
2021,97, 107724. [CrossRef]
93.
Alicke, M.D.; Braun, J.C.; Glor, J.E.; Klotz, K.L.; Magee, J.; Sederholm, H.; Siegel, R. Complaining Behaviour in social interactions.
Personal. Soc. Psychol. Bull. 1992,18, 286–295. [CrossRef]
94. You, K.H.; Gretzel, U. What motivates consumers to write online travel reviews? Fac. Commer. 2008,1, 282–296. [CrossRef]
95.
Eastwood, K.; Durrheim, D.N.; Jones, A.; Butler, M. Acceptance of pandemic (H1N1) 2009 influenza vaccination by the Australian
public. Med. J. Aust. 2010,192, 33–36. [CrossRef]
96.
Raude, J.; Caille-Brillet, A.-L.; Setbon, M. The 2009 pandemic H1N1 influenza vaccination in France: Who accepted to receive the
vaccine and why? PLoS Curr. 2010,2. [CrossRef] [PubMed]
97.
Maurer, J.; Harris, K.M.; Parker, A.M.; Lurie, N. Does receipt of seasonal influenza vaccine predict intention to receive novel
H1N1 vaccine: Evidence from a nationally representative survey of U.S. adults. Vaccine 2009,27, 5732–5734. [CrossRef]
98.
Maurer, J.; Uscher-Pines, L.; Harris, K.M. Perceived seriousness of seasonal and A(H1N1) influenzas, attitudes toward vaccination,
and vaccine uptake among U.S. adults: Does the source of information matter? Prev. Med.
2010
,51, 185–187. [CrossRef] [PubMed]
99.
Schwarzinger, M.; Flicoteaux, R.; Cortarenoda, S.; Obadia, Y.; Moatti, J.-P. Low Acceptability of A/H1N1 Pandemic Vaccination in
French Adult Population: Did Public Health Policy Fuel Public Dissonance? PLoS ONE 2010,5, 4. [CrossRef] [PubMed]
100.
Sypsa, V.; Livanios, T.; Psichogiou, M.; Malliori, M.; Tsiodras, S.; Nikolakopoulos, I.; Hatzakis, A. Public perceptions in relation to
intention to receive pandemic influenza vaccination in a random population sample: Evidence from a cross-sectional telephone
survey. Eurosurveillance 2009,14, 19437. [CrossRef] [PubMed]
101.
Rubin, G.J.; Potts, H.; Michie, S. The impact of communications about swine flu (influenza A H1N1v) on public responses to the
outbreak: Results from 36 national telephone surveys in the UK. Health Technol. Assess. 2010,14, 183–266. [CrossRef] [PubMed]
102.
Seale, H.; Heywood, A.E.; McLaws, M.; Ward, K.F.; Lowbridge, C.P.; Van, D.; MacIntyre, C.R. Why do I need it? I am not at risk!
Public perceptions towards the pandemic (H1N1) 2009 vaccine. BMC Infect. Dis. 2010,10, 99. [CrossRef]
103.
Nguyen, T.; Henningsen, K.H.; Brehaut, J.C.; Hoe, E.; Wilson, K. Acceptance of a pandemic influenza vaccine: A systematic
review of surveys of the general public. Infect. Drug Resist. 2011,4, 197–207. [PubMed]
104.
Soni, R.; Mathai, J. Improved Twitter Sentiment Prediction through ‘Cluster-then-Predict Model. Int. J. Comput. Sci. Netw.
2015
,4,
2277–5420.
... It is mainly driven by the antivaccine misinformation and conspiracy theories that have been circulating on social media since the release of the first vaccine [6]. Social media platforms, such as Twitter, have become a common platform for individuals to exchange opinions and concerns about the effectiveness and safety of the COVID-19 vaccines [7]. To many people, the rapid development of numerous COVID-19 vaccinations is questionable. ...
... Examples include VADER [21] that utilizes a sentiment lexicon of predefined words with an associated polarity score. VADER can process acronyms and slang words making it highly suitable to be applied to social media contexts [7]. Syuzhet package [22], LIWC, ANEW and GI [10] are similar lexiconbased SA tools. ...
... Their results show that nearly 40% of the population in both countries have a negative attitude toward COVID-19 vaccines and that celebrities could lead the opinion shift on social media in vaccination progress. Similarly, Roe et al. [7] collected 137,781 tweet datasets using 43 search terms relating to COVID-19 vaccines and analyzed the sentiment of the tweets using VADER and Microsoft Azure (a machine learning approach) to assess whether tweets represented positive, negative, or neutral opinions. They found that Most tweets are negative in sentiment, followed by positive and neutral. ...
Full-text available
Conference Paper
Despite the evidence that shows the benefits and safety of immunizations, the widespread vaccine-related misinformation and conspiracy theories online have fueled a general vaccine hesitancy, and coronavirus disease (COVID-19) vaccinations are no exception. COVID-19 vaccine hesitancy is considered a global threat to public health that undermines the efforts to control the COVID-19 pandemic. Twitter and other social media platforms allow people to exchange information and express concerns and emotions on COVID-19-related issues. This research aims to understand people’s sentiment on COVID-19 vaccines from data collected from Twitter. Analyzing the public attitude toward the vaccines helps the authorities to make better decisions and reach the intended herd immunity. In this paper, we utilize the state-of-the-art transformer-based classification models, RoBERTa and BERT, along with multiple task-specific versions, to classify people’s opinions about COVID-19 vaccinations into positive, negative, and neutral. A Twitter dataset that consists of people’s opinions about vaccines is used to train and evaluate the presented models. Two ensemble learning techniques that aggregate the individual classifiers are presented for further performance improvement: majority voting and stacking with Support Vector Machine (SVM) as meta-learner. The results also show that applying ensemble learning significantly outperforms the individual classifiers using all evaluation measures. We also found that ensembling with stacking has an advantage over simple majority voting.
... Furthermore, they found that age, marital status, and education are the significant features for the willingness of the SARS-CoV-2 vaccine. However, another study has used the combined approach, that is, the use of questionnaire and twitter data for exploring the SARS-CoV-2 vaccination hesitancy [21]. Conversely, they found that the number of negative tweets was higher, followed by the positive tweets and then the neutral tweets. ...
... Furthermore, samples were collected from individuals with at least postgraduate qualification. Conversely, Roe et al. [21] analysed the data collected via questionnaires and tweets and found a higher number of negative tweets when compared with the positive ones. However, the data collection was made in July 2021 and during that time vaccine has already been introduced and most of the countries have already imposed the SARS-CoV-2 vaccination. ...
Full-text available
Article
Countries around the world are facing so many challenges to slow down the spread of the current SARS-CoV-2 virus. Vaccination is an effective way to combat this virus and prevent its spreading among individuals. Currently, there are more than 50 SARS-CoV-2 vaccine candidates in trials; only a few of them are already in use. The primary objective of this study is to analyse the public awareness and opinion toward the vaccination process and to develop a model that predicts the awareness and acceptability of SARS-CoV-2 vaccines in Saudi Arabia by analysing a dataset of Arabic tweets related to vaccination. Therefore, several machine learning models such as Support Vector Machine (SVM), Naïve Bayes (NB), and Logistic Regression (LR), sideways with the N-gram and Term Frequency-Inverse Document Frequency (TF-IDF) techniques for feature extraction and Long Short-Term Memory (LSTM) model used with word embedding. LR with unigram feature extraction has achieved the best accuracy, recall, and F1 score with scores of 0.76, 0.69, and 0.72, respectively. However, the best precision value of 0.80 was achieved using SVM with unigram and NB with bigram TF-IDF. However, the Long Short-Term Memory (LSTM) model outperformed the other models with an accuracy of 0.95, a precision of 0.96, a recall of 0.95, and an F1 score of 0.95. This model will help in gaining a complete idea of how receptive people are to the vaccine. Thus, the government will be able to find new ways and run more campaigns to raise awareness of the importance of the vaccine.
... This phase of emotional tension can be concretely illustrated by the context of the occurrence of COVID-19 we are still experiencing. Most of us have witnessed the global emotional impact of this disruptive event, provoking surprise, fear, sadness, or anger in many countries Roe et al., , 2021;Xue et al., 2020). As mentioned earlier, emotions are markers of the subjective experience of reality. ...
Full-text available
Article
One of the main focuses of the theory of Social Representations (SRs) consists of examining its sensitivity to context immediacy. Arising from everyday life, emotions could be particularly relevant to this aim as they could constitute modalities through which SRs can emerge, be reinforced, or be transformed. The inherently unstable nature of reality requires a signaling system of this variability. The general assumption is that emotions provide, at an individual level, a signaling function of the relevance of SRs in the social integration of reality. Triggering this signal function, the role of tension lies at the heart of the process. With the emotion-driven tension, we approach SRs as cognitive-emotional processes of construction, conservation, and transformation of social knowledge. We, therefore, situate the study of the relationships between SRs and emotions in a conceptual approach to the dynamics of stability and change. Considering SRs as dynamic objects of social change, this article proposes to conceptualize SRs as cognitive-emotional processes by promoting an integrative model grounded on a socio-constructivist as well as a discursive perspective. In this model, emotions are addressed as individual dispositions at the service of the sociogenesis of SRs. Occurring from individual experience, they contribute through their sharing to the construction of social knowledge. Implications of this conceptual proposal for SR theory are discussed.
... For a better insight, the authors also presented the sentiment polarities across a multitude of countries. Roe et al. [9] studied the COVID-19 vaccine hesitancy among people across the globe through their Tweets. The authors found that the majority of Tweets were negative in sentiment, while the positive and neutral Tweets appeared at second and third places respectively. ...
Full-text available
Chapter
In the realm of contemporary soft computing practices, analysis of public perceptions and opinion mining (OM) have received considerable attention due to the easy availability of colossal data in the form of unstructured text generated by social media, e-commerce portals, blogs, and other similar web resources. The year 2020 witnessed the gravest epidemic in the history of mankind, and in the present year, we stand amidst a global, massive and exhaustive vaccination movement. Since the inception of the COVID-19 vaccines and their applications, people across the globe, from the ordinary public to celebrities and VIPs have been expressing their fears, doubts, experiences, expectations, dilemmas and perceptions about the current COVID-19 vaccination program. Being very popular among a large class of modern human society, the Twitter platform has been chosen in this research to study public perceptions about this global vaccination drive. More than 112 thousand Tweets from users of different countries around the globe are extracted based on hashtags related to the affairs of the COVID-19 vaccine. A three-tier framework is being proposed in which raw Tweets are extracted and cleaned first, visualized and converted into numerical vectors through word embedding and N-gram models next, and finally analyzed through a few machine learning classifiers with the standard performance metrics, accuracy, precision, recall, and F1-measure. The Logistic Regression (LR) and Adaptive Boosting (AdaBoost) classifiers attended the highest accuracies of 87% and 89% with the Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF) word embedding models respectively. Overall, the BoW model achieved slightly better average classification accuracy (78.33%) than that of the TF-IDF model (77.89%). Moreover, the experimental results show that most of the people have a neutral attitude towards the current COVID-19 vaccination drive and people favoring the COVID-19 vaccination program are greater in number than those who doubt it and its consequences.KeywordsVaccineCOVID-19 vaccination programSentiments analysis Twitter Machine learningN-gram
... Antivaccine advocates have utilized Facebook and Twitter to disseminate exaggerated claims [12,46]. In a study on vaccine hesitancy on Twitter, it was found that most of the negative tweets on COVID-19 contained a hashtag as opposed to positive and neutral tweets [69]. Similarly, a study showed that antivaccine claims on Twitter in 2018 relied on the use of hashtags [46]. ...
Preprint
BACKGROUND Vaccines serve an integral role in containing pandemics, yet vaccine hesitancy is prevalent globally. One key reason for this hesitancy is the pervasiveness of misinformation on social media. Although considerable research attention has been drawn to how exposure to misinformation is closely associated with vaccine hesitancy, little scholarly attention has been given to the investigation or robust theorizing of the various content themes pertaining to antivaccine misinformation about COVID-19 and the writing strategies in which these content themes are manifested. Virality of such content on social media exhibited in the form of comments, shares, and reactions has practical implications for COVID-19 vaccine hesitancy. OBJECTIVE We investigated whether there were differences in the content themes and writing strategies used to disseminate antivaccine misinformation about COVID-19 and their impact on virality on social media. METHODS We constructed an antivaccine misinformation database from major social media platforms during September 2019-August 2021 to examine how misinformation exhibited in the form of content themes and how these themes manifested in writing were associated with virality in terms of likes, comments, and shares. Antivaccine misinformation was retrieved from two globally leading and widely cited fake news databases, COVID Global Misinformation Dashboard and International Fact-Checking Network Corona Virus Facts Alliance Database, which aim to track and debunk COVID-19 misinformation. We primarily focused on 140 Facebook posts, since most antivaccine misinformation posts on COVID-19 were found on Facebook. We then employed quantitative content analysis to examine the content themes (ie, safety concerns, conspiracy theories, efficacy concerns) and manifestation strategies of misinformation (ie, mimicking of news and scientific reports in terms of the format and language features, use of a conversational style, use of amplification) in these posts and their association with virality of misinformation in the form of likes, comments, and shares. RESULTS Our study revealed that safety concern was the most prominent content theme and a negative predictor of likes and shares. Regarding the writing strategies manifested in content themes, a conversational style and mimicking of news and scientific reports via the format and language features were frequently employed in COVID-19 antivaccine misinformation, with the latter being a positive predictor of likes. CONCLUSIONS This study contributes to a richer research-informed understanding of which concerns about content theme and manifestation strategy need to be countered on antivaccine misinformation circulating on social media so that accurate information on COVID-19 vaccines can be disseminated to the public, ultimately reducing vaccine hesitancy. The liking of COVID-19 antivaccine posts that employ language features to mimic news or scientific reports is perturbing since a large audience can be reached on social media, potentially exacerbating the spread of misinformation and hampering global efforts to combat the virus.
... Antivaccine advocates have utilized Facebook and Twitter to disseminate exaggerated claims [12,46]. In a study on vaccine hesitancy on Twitter, it was found that most of the negative tweets on COVID-19 contained a hashtag as opposed to positive and neutral tweets [69]. Similarly, a study showed that antivaccine claims on Twitter in 2018 relied on the use of hashtags [46]. ...
Full-text available
Article
Background: Vaccines serve an integral role in containing pandemics, yet vaccine hesitancy is prevalent globally. One key reason for this is the pervasiveness of misinformation on social media. While considerable research attention has been drawn to how exposure to misinformation is closely associated with vaccine hesitancy, little scholarly attention has been given to the investigation or robust theorizing of the various content themes pertaining to antivaccine misinformation about COVID-19 and the writing strategies in which these content themes are manifested. Virality of such content on social media exhibited in the form of comments, shares, and reactions, has practical implications for COVID-19 vaccine hesitancy. Objective: We investigated whether there were differences in the content themes and writing strategies used to disseminate antivaccine misinformation about COVID-19 and their impact on virality on social media. Methods: This study constructed an antivaccine misinformation database from major social media platforms during September 2019-August 2021 to examine how misinformation exhibited in the form of content themes and how these themes were manifested in writing were associated with virality in terms of likes, comments, and shares. Antivaccine misinformation was retrieved from two globally leading and widely cited fake news databases - COVID Global Misinformation Dashboard and IFCN Corona Virus Facts Alliance Database, which aim to track and debunk COVID-19 misinformation. We primarily focused on 140 Facebook posts since most antivaccine misinformation posts on COVID-19 were found on Facebook. Then, we employed quantitative content analysis to examine the content themes (i.e. safety concerns; conspiracy theories; efficacy concerns) and manifestation strategies of misinformation (i.e. mimicking of news and scientific reports in terms of the format and language features; use of a conversational style; use of amplification) in these posts and their association with virality of misinformation in the form of likes, comments, and shares. Results: Our study revealed that safety concern was the most prominent content theme and a negative predictor of likes and shares. Regarding the writing strategies manifested in content themes, a conversational style and mimicking of news and scientific reports via the format and language features were frequently employed in COVID-19 antivaccine misinformation, with the latter being a positive predictor of likes. Conclusions: This study contributes to a richer research-informed understanding of which concerns about content theme and manifestation strategy need to be countered on antivaccine misinformation circulating on social media so that accurate information on COVID-19 vaccines can be disseminated to the public, ultimately reducing vaccine hesitancy. The liking of COVID-19 antivaccine posts that employed language features to mimic news or scientific reports is perturbing since a large audience can be reached on social media, potentially exacerbating the spread of misinformation and hampering global efforts to combat the virus. Clinicaltrial:
... 97 Finally, in July 2021 the UK Twitter users expressed similar rates of negative and positive sentiments towards current immunization campaign, as shown by the portion of positive and negative tweets obtained from the study period. 98 Studies reporting fluctuation trends of vaccine sentiments on social networks caused by some driver events Nine studies reported how social media vaccine sentiments varied depending on some key events ( Table 4). Events that were identified as those that triggered an increase in the positive emotions towards vaccination were: the announcement about the vaccines' effectiveness, 99−102 the arrival of the vaccines in UK hospitals, 99 the announcement of the first human vaccine trial (UK Twitter and Facebook users) and Donald Trump's announcement regarding a vaccine being ready in a few weeks (US Twitter and Facebook users), 100 as well as the decrease in number of positive COVID-19 cases in Korea. ...
Full-text available
Article
Background: Vaccine hesitancy continues to limit global efforts in combatting the COVID-19 pandemic. Emerging research demonstrates the role of social media in disseminating information and potentially influencing people's attitudes towards public health campaigns. This systematic review sought to synthesize the current evidence regarding the potential role of social media in shaping COVID-19 vaccination attitudes, and to explore its potential for shaping public health interventions to address the issue of vaccine hesitancy. Methods: We performed a systematic review of the studies published from inception to 13 of March2022 by searching PubMed, Web of Science, Embase, PsychNET, Scopus, CINAHL, and MEDLINE. Studies that reported outcomes related to coronavirus disease 2019 (COVID-19) vaccine (attitudes, opinion, etc.) gathered from the social media platforms, and those analyzing the relationship between social media use and COVID-19 hesitancy/acceptance were included. Studies that reported no outcome of interest or analyzed data from sources other than social media (websites, newspapers, etc.) will be excluded. The Newcastle Ottawa Scale (NOS) was used to assess the quality of all cross-sectional studies included in this review. This study is registered with PROSPERO (CRD42021283219). Findings: Of the 2539 records identified, a total of 156 articles fully met the inclusion criteria. Overall, the quality of the cross-sectional studies was moderate - 2 studies received 10 stars, 5 studies received 9 stars, 9 studies were evaluated with 8, 12 studies with 7,16 studies with 6, 11 studies with 5, and 6 studies with 4 stars. The included studies were categorized into four categories. Cross-sectional studies reporting the association between reliance on social media and vaccine intentions mainly observed a negative relationship. Studies that performed thematic analyses of extracted social media data, mainly observed a domination of vaccine hesitant topics. Studies that explored the degree of polarization of specific social media contents related to COVID-19 vaccines observed a similar degree of content for both positive and negative tone posted on different social media platforms. Finally, studies that explored the fluctuations of vaccination attitudes/opinions gathered from social media identified specific events as significant cofactors that affect and shape vaccination intentions of individuals. Interpretation: This thorough examination of the various roles social media can play in disseminating information to the public, as well as how individuals behave on social media in the context of public health events, articulates the potential of social media as a platform of public health intervention to address vaccine hesitancy. Funding: None.
Article
Objective Online patient reviews influence a patient’s choice of a vascular surgeon. The aim of this study is to examine underlying factors that contribute to positive and negative patient reviews by leveraging sentiment analysis and machine learning methods. Methods The Society of Vascular Surgeons (SVS) publicly accessible member directory was queried and cross-referenced with a popular patient-maintained physician review website, healthgrades.com. Sentiment analysis and machine learning methods were utilized to analyze several parameters. Demographics (gender, age, state of practice), star rating (out of 5 stars), and written reviews were obtained for corresponding vascular surgeons. A sentiment analysis model was applied to patient-written reviews and validated against the star ratings. Student’s t-test or one-way ANOVA assessed demographic relationships with reviews. Word frequency assessments and multivariable logistic regression analyses were conducted to identify common and determinative components of written reviews. Results A total of 1799 vascular surgeons had public profiles with reviews. Female gender of surgeon was associated with lower star ratings (male = 4.19, female = 3.95, p < 0.01) and average sentiment score (male = 0.50, female = 0.40, p <0.01). Younger physician age was associated `with higher star rating (p = 0.02), but not average sentiment score (p = 0.12). In the Best reviews, the most commonly used one-words were Care (N=999), Caring (N=767) and Kind (N=479) while the most commonly used two-word pairs were Saved/Life (N=189), Feel/Comfortable (N=106), and Kind/Caring (N=104). For the Worst reviews, the most commonly used one-words were Pain (N=254) and Rude (N=148) while the most commonly used two-word pairs were No/One (N=27), Waste/Time (N=25), and Severe/Pain (N=18). In multiple logistic regression, satisfactory reviews were associated with word such as Confident (OR=8.93), Pain-free (OR=4.72), Listens (OR=2.55), and Bedside Manner (OR=1.70), while unsatisfactory reviews were associated with words such as Rude (OR=0.01), Arrogant (OR=0.09), Infection (OR=0.20), and Wait (OR=0.48). Conclusions Female surgeons received significantly worse reviews, and younger surgeons tended to receive better reviews. The positivity and negativity of reviews were largely related to words associated with the patient-doctor experience and pain. Vascular surgeons should focus on these two areas to improve patient experiences and their own reviews.
Full-text available
Article
Social media, such as Twitter, is a source of exchanging information and opinion on global issues such as COVID-19 pandemic. In this study, we work with a database of around 1.2 million tweets collected across five weeks of April–May 2021 to draw conclusions about public sentiments towards the vaccination outlook when vaccinations become widely available to the population during the COVID-19 pandemic. We deploy natural language processing and sentiment analysis techniques to reveal insights about COVID-19 vaccination awareness among the public. Our results show that people have positive sentiments towards taking COVID-19 vaccines instead of some adverse effects of some of the vaccines. We also analyze people’s attitude towards the safety measures of COVID-19 after receiving the vaccines. Again, the positive sentiment is higher than that of negative in terms of maintaining safety measures against COVID-19 among the vaccinated population. We also project that around 62.44% and 48% of the US population will get at least one dose of vaccine and be fully vaccinated, respectively, by the end of July 2021 according to our forecast model. This study will help to understand public reaction and aid the policymakers to project the vaccination campaign as well as health and safety measures in the ongoing global health crisis.
Full-text available
Article
A year into the COVID-19 pandemic and one of the longest recorded lockdowns in the world, the Philippines received its first delivery of COVID-19 vaccines on 1 March 2021 through WHO’s COVAX initiative. A month into inoculation of all frontline health professionals and other priority groups, the authors of this study gathered data on the sentiment of Filipinos regarding the Philippine government’s efforts using the social networking site Twitter. Natural language processing techniques were applied to understand the general sentiment, which can help the government in analyzing their response. The sentiments were annotated and trained using the Naïve Bayes model to classify English and Filipino language tweets into positive, neutral, and negative polarities through the RapidMiner data science software. The results yielded an 81.77% accuracy, which outweighs the accuracy of recent sentiment analysis studies using Twitter data from the Philippines.
Full-text available
Article
The aim of this study was to analyze public opinion about online learning during the COVID-19 (Coronavirus Disease 2019) pandemic. A total of 154 articles from online news and blogging websites related to online learning were extracted from Google and DuckDuckGo. The articles were extracted for 45 days, starting from the day the World Health Organization (WHO) declared COVID-19 a worldwide pandemic, 11 March 2020. For this research, we applied the dictionary-based approach of the lexicon-based method to perform sentiment analysis on the articles extracted through web scraping. We calculated the polarity and subjectivity scores of the extracted article using the TextBlob library. The results showed that over 90% of the articles are positive, and the remaining were mildly negative. In general, the blogs were more positive than the newspaper articles; however, the blogs were more opinionated compared to the news articles.
Full-text available
Article
Vaccine hesitancy was one of the ten major threats to global health in 2019, according to the World Health Organisation. Nowadays, social media has an important role in the spread of information, misinformation, and disinformation about vaccines. Monitoring vaccine-related conversations on social media could help us to identify the factors that contribute to vaccine confidence in each historical period and geographical area. We used a hybrid approach to perform an opinion-mining analysis on 1,499,227 vaccine-related tweets published on Twitter from 1st June 2011 to 30th April 2019. Our algorithm classified 69.36% of the tweets as neutral, 21.78% as positive, and 8.86% as negative. The percentage of neutral tweets showed a decreasing tendency, while the proportion of positive and negative tweets increased over time. Peaks in positive tweets were observed every April. The proportion of positive tweets was significantly higher in the middle of the week and decreased during weekends. Negative tweets followed the opposite pattern. Among users with ≥2 tweets, 91.83% had a homogeneous polarised discourse. Positive tweets were more prevalent in Switzerland (71.43%). Negative tweets were most common in the Netherlands (15.53%), Canada (11.32%), Japan (10.74%), and the United States (10.49%). Opinion mining is potentially useful to monitor online vaccine-related concerns and adapt vaccine promotion strategies accordingly.
Full-text available
Article
Background Our aim was to estimate provisional willingness to receive a coronavirus 2019 (COVID-19) vaccine, identify predictive socio-demographic factors, and, principally, determine potential causes in order to guide information provision. Methods A non-probability online survey was conducted (24th September−17th October 2020) with 5,114 UK adults, quota sampled to match the population for age, gender, ethnicity, income, and region. The Oxford COVID-19 vaccine hesitancy scale assessed intent to take an approved vaccine. Structural equation modelling estimated explanatory factor relationships. Results 71.7% (n=3,667) were willing to be vaccinated, 16.6% (n=849) were very unsure, and 11.7% (n=598) were strongly hesitant. An excellent model fit (RMSEA=0.05/CFI=0.97/TLI=0.97), explaining 86% of variance in hesitancy, was provided by beliefs about the collective importance, efficacy, side-effects, and speed of development of a COVID-19 vaccine. A second model, with reasonable fit (RMSEA=0.03/CFI=0.93/TLI=0.92), explaining 32% of variance, highlighted two higher-order explanatory factors: ‘excessive mistrust’ (r=0.51), including conspiracy beliefs, negative views of doctors, and need for chaos, and ‘positive healthcare experiences’ (r=−0.48), including supportive doctor interactions and good NHS care. Hesitancy was associated with younger age, female gender, lower income, and ethnicity, but socio-demographic information explained little variance (9.8%). Hesitancy was associated with lower adherence to social distancing guidelines. Conclusions COVID-19 vaccine hesitancy is relatively evenly spread across the population. Willingness to take a vaccine is closely bound to recognition of the collective importance. Vaccine public information that highlights prosocial benefits may be especially effective. Factors such as conspiracy beliefs that foster mistrust and erode social cohesion will lower vaccine up-take.
Full-text available
Article
Background It is important to measure the public response to the COVID-19 pandemic. Twitter is an important data source for infodemiology studies involving public response monitoring. Objective The objective of this study is to examine COVID-19–related discussions, concerns, and sentiments using tweets posted by Twitter users. Methods We analyzed 4 million Twitter messages related to the COVID-19 pandemic using a list of 20 hashtags (eg, “coronavirus,” “COVID-19,” “quarantine”) from March 7 to April 21, 2020. We used a machine learning approach, Latent Dirichlet Allocation (LDA), to identify popular unigrams and bigrams, salient topics and themes, and sentiments in the collected tweets. Results Popular unigrams included “virus,” “lockdown,” and “quarantine.” Popular bigrams included “COVID-19,” “stay home,” “corona virus,” “social distancing,” and “new cases.” We identified 13 discussion topics and categorized them into 5 different themes: (1) public health measures to slow the spread of COVID-19, (2) social stigma associated with COVID-19, (3) COVID-19 news, cases, and deaths, (4) COVID-19 in the United States, and (5) COVID-19 in the rest of the world. Across all identified topics, the dominant sentiments for the spread of COVID-19 were anticipation that measures can be taken, followed by mixed feelings of trust, anger, and fear related to different topics. The public tweets revealed a significant feeling of fear when people discussed new COVID-19 cases and deaths compared to other topics. Conclusions This study showed that Twitter data and machine learning approaches can be leveraged for an infodemiology study, enabling research into evolving public discussions and sentiments during the COVID-19 pandemic. As the situation rapidly evolves, several topics are consistently dominant on Twitter, such as confirmed cases and death rates, preventive measures, health authorities and government policies, COVID-19 stigma, and negative psychological reactions (eg, fear). Real-time monitoring and assessment of Twitter discussions and concerns could provide useful data for public health emergency responses and planning. Pandemic-related fear, stigma, and mental health concerns are already evident and may continue to influence public trust when a second wave of COVID-19 occurs or there is a new surge of the current pandemic.
Full-text available
Article
Background With restrictions on movement and stay-at-home orders in place due to the COVID-19 pandemic, social media platforms such as Twitter have become an outlet for users to express their concerns, opinions, and feelings about the pandemic. Individuals, health agencies, and governments are using Twitter to communicate about COVID-19. Objective The aims of this study were to examine key themes and topics of English-language COVID-19–related tweets posted by individuals and to explore the trends and variations in how the COVID-19–related tweets, key topics, and associated sentiments changed over a period of time from before to after the disease was declared a pandemic. Methods Building on the emergent stream of studies examining COVID-19–related tweets in English, we performed a temporal assessment covering the time period from January 1 to May 9, 2020, and examined variations in tweet topics and sentiment scores to uncover key trends. Combining data from two publicly available COVID-19 tweet data sets with those obtained in our own search, we compiled a data set of 13.9 million English-language COVID-19–related tweets posted by individuals. We use guided latent Dirichlet allocation (LDA) to infer themes and topics underlying the tweets, and we used VADER (Valence Aware Dictionary and sEntiment Reasoner) sentiment analysis to compute sentiment scores and examine weekly trends for 17 weeks. Results Topic modeling yielded 26 topics, which were grouped into 10 broader themes underlying the COVID-19–related tweets. Of the 13,937,906 examined tweets, 2,858,316 (20.51%) were about the impact of COVID-19 on the economy and markets, followed by spread and growth in cases (2,154,065, 15.45%), treatment and recovery (1,831,339, 13.14%), impact on the health care sector (1,588,499, 11.40%), and governments response (1,559,591, 11.19%). Average compound sentiment scores were found to be negative throughout the examined time period for the topics of spread and growth of cases, symptoms, racism, source of the outbreak, and political impact of COVID-19. In contrast, we saw a reversal of sentiments from negative to positive for prevention, impact on the economy and markets, government response, impact on the health care industry, and treatment and recovery. Conclusions Identification of dominant themes, topics, sentiments, and changing trends in tweets about the COVID-19 pandemic can help governments, health care agencies, and policy makers frame appropriate responses to prevent and control the spread of the pandemic.
Full-text available
Article
While several studies have shown how telemedicine and, in particular, home telemonitoring programs lead to an improvement in the patient’s quality of life, a reduction in hospitalizations, and lower healthcare costs, different variables may affect telemonitoring effectiveness and purposes. In the present paper, an integrated software system, based on Sentiment Analysis and Text Mining, to deliver, collect, and analyze questionnaire responses in telemonitoring programs is presented. The system was designed to be a complement to home telemonitoring programs with the objective of investigating the paired relationship between opinions and the adherence scores of patients and their changes through time. The novel contributions of the system are: (i) the design and software prototype for the management of online questionnaires over time; and (ii) an analysis pipeline that leverages a sentiment polarity score by using it as a numerical feature for the integration and the evaluation of open-ended questions in clinical questionnaires. The software pipeline was initially validated with a case-study application to discuss the plausibility of the existence of a directed relationship between a score representing the opinion polarity of patients about telemedicine, and their adherence score, which measures how well patients follow the telehomecare program. In this case-study, 169 online surveys sent by 38 patients enrolled in a home telemonitoring program provided by the Cystic Fibrosis Unit at the “Bambino Gesù” Children’s Hospital in Rome, Italy, were collected and analyzed. The experimental results show that, under a Granger-causality perspective, a predictive relationship may exist between the considered variables. If supported, these preliminary results may have many possible implications of practical relevance, for instance the early detection of poor adherence in patients to enable the application of personalized and targeted actions.
Article
Background Vaccination is crucial to limit the pandemic spread of SARS-CoV-2/COVID-19. Therefore, besides the development and supply of vaccines, it is essential that sufficient individuals are willing to get vaccinated, but concerning proportions of populations worldwide show vaccine hesitancy. This makes it important to determine factors that are associated with vaccine acceptance. Methods 1,779 adults of a non-probability convenience sample in Germany were assessed with an online survey in a cross-sectional survey period from 1st to 11th January, 2021 (a few days after the beginning of vaccinations in Germany). Results 64.5 % of the sample stated that they absolutely would accept the vaccination, 13.8 % would rather accept it, 10.4 % were undecided, and 5.2 % would rather not and 6.0 % absolutely not get vaccinated. COVID-19-related anxiety, and fears of infection and health-related consequences correlated significantly positively with vaccine acceptance (all p<.001). In contrast, social (p=.006) and economic fears (p<.001) showed significant negative associations with vaccination willingness. The broader constructs of unspecific anxiety and depressive symptoms were not significantly associated with vaccine acceptance. Vaccine acceptance differed between users/non-users of social media and official websites to gain information about the pandemic (p<.001). Conclusions COVID-19-related anxiety and health-related fears were associated with higher vaccine acceptance, whereas the fear of social and economic consequences showed the contrary direction. These findings highlight the need to differentiate between several types of fears and anxiety to predict their influence on vaccine acceptance, and provide important information and an essential base for future studies and interventions.