How Health Information Spreads in Twitter: The Whos and Whats of

Philippine TB-data

Erika Y. Chan, Myles Russel C. Chan, Shyrene Julianne S. Ching, Stanley Lawrence Sie,

Angelyn R. Lao, Jan Michael Alexandre C. Bernadas and Charibeth K. Cheng

De La Salle University, Manila, Philippines

Keywords:

Twitter, Health Information, Social Network Analysis, Sentiment Identiﬁcation, Topic Modeling.

Abstract:

Twitter is a popular platform for disseminating health information. Unfortunately, there is no clear way to

monitor how information reaches the intended audiences. This research examined how health information

spreads on Twitter and identiﬁed factors that affect the spreading within the Philippines. We created a process

whose goal is to generate results that experts can deeply analyze to reveal insights into information spread.

The process consists of crawling Twitter data, transforming the data and applying sentiment identiﬁcation and

topic modeling, and performing Social Network Analysis (SNA). The SNA graphs allow for the study of the

interactions between Twitter users and tweets while giving insights on inﬂuential users and topics discussed

across clusters. The study explored and utilized tuberculosis-related tweets. Though the algorithms were

meant to process tweets written in Filipino, the process is mostly language-agnostic and can be applied to

Twitter data. The results also help in identifying strategies that can improve health information spread on

Twitter in the Philippines.

1 INTRODUCTION

1.1 Background of the Study

Social networking sites (SNS) have emerged as pri-

mary forms of communication within the online com-

munity, with around 46.03% of the world popula-

tion using SNS

. Furthermore, they have also become

the main sources of information online, with almost

64.5% of Internet users receiving breaking news from

online platforms, such as Facebook, Twitter, and In-

stagram, instead of traditional media. Due to their

wide use, other domains have also taken advantage of

the spread of information found within these sites as

mediums to promote and provide awareness

One such domain is that of the medical ﬁeld as

SNS have become popular avenues for publicly shar-

ing information. In fact, Twitter has been called

the most popular healthcare communication platform

(Pershad et al., 2018). Furthermore, many studies

have been conducted to better understand the behav-

ior of information spread within Twitter and its effects

https://ourworldindata.org/rise-of-social-media

https://conductscience.com/the-impact-of-social-

media-on-knowledge-dissemination/

on users (Kudchadkar and Carroll, 2020; Liang et al.,

2019; Tambuscio et al., 2015).

We found an opportunity for a localized study that

focuses on the behavior of information spread in the

Philippine context. We focused on health informa-

tion revolving around tuberculosis (TB), since TB has

been characterized as a global health threat

, with

around 9 million people acquiring TB in 2017

. It is

also more prevalent in the global South and the Philip-

pines

, which has been classiﬁed as high in both TB

and drug-resistant TB

. Moreover, in 2017, around

581,000 were diagnosed with and 27,000 died of TB

in the Philippines. One of the methods to effective

prevention and management of TB is relevant infor-

mation (Wieland et al., 2013; Brashers et al., 2004).

Communicating social support to patients is also im-

portant for continuing medication and treatment suc-

cess (Skiles et al., 2018). Thus, this study would also

like to explore how health information regarding tu-

berculosis is spread on Twitter and determine factors

that may be used to better disseminate pertinent and

https://www.tballiance.org/why-new-tb-drugs/global-

pandemic

https://www.who.int/news-room/fact-sheets/detail/

tuberculosis

https://www.who.int/tb/publications/global\ report/en/

Chan, E., Chan, M., Ching, S., Sie, S., Lao, A., Bernadas, J. and Cheng, C.

How Health Information Spreads in Twitter: The Whos and Whats of Philippine TB-data.

DOI: 10.5220/0010818000003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 5: HEALTHINF, pages 421-429

ISBN: 978-989-758-552-4; ISSN: 2184-4305

421

important information regarding tuberculosis.

There is a signiﬁcant amount of Filipinos using

social networking platforms as a way of getting news

and information (Chua, 2020). However, there is also

a lack of dissemination of reliable health information

(Sbafﬁ and Rowley, 2017). While there are alterna-

tives in disseminating information in traditional me-

dia, such as televisions and radios, these are proven

ineffective because fewer and fewer people are utiliz-

ing them (Chua, 2020).

The following are the contributions of this re-

search:

• Identiﬁed factors information spread will allow

Philippine health organizations to take advantage

of the popularity of Twitter to spread important

and relevant information, with an emphasis on

lesser known diseases or programs that do not re-

ceive the necessary attention, effectively and ef-

ﬁciently so that the information reaches a wider

audience.

• The formulated process in tracing health informa-

tion spread is reproducible in other contexts with

most of the modules being language agnostic.

2 RELATED WORKS

2.1 Social Network Analysis

Social network analysis is consistently used to learn

how information spreads in social networks by char-

acterizing network structures to ﬁnd unique patterns

on how information spreads and determine the po-

tential factors that may affect the spread. A previ-

ous study by (Himelboim et al., 2017) has shown that

social network analysis can also be used to classify

topic-networks on Twitter by using features such as

mentions, retweets, and replies. Twitter networks are

also directed and have edges ﬂowing towards a certain

direction. In social network analysis, these networks

are usually represented as graphs, with different met-

rics applied to measure the spread of information in

the network. One of the most commonly used metrics

is centrality.

Centrality aims to measure the inﬂuence of a spe-

ciﬁc node in a network and is commonly used to iden-

tify the source of a spread (Grandjean, 2016). In its

application to the analysis of information spread in

SNS, the out-degree centrality can be used to mea-

sure the amount of outgoing information from a cer-

tain user and represents the reach in the community

in a directed network. The out-degree centrality value

also represents how often a node interacts with other

nodes in the network (Hansen et al., 2020). Eigen-

vector centrality measures the importance of a node

based on the importance of its neighbors (Bihari and

Pandia, 2015), where its neighbors are also connected

to other nodes. Furthermore, eigenvector centrality

can be used to determine the most inﬂuential node in

a network (Maharani et al., 2014), as a high eigen-

vector centrality value indicates greater connectivity

compared to other nodes, resulting to a wider spread

of information ﬂow in the network. Another central-

ity measure, betweenness centrality, is used to mea-

sure how frequently a node acts as a bridge along the

shortest path between two nodes (Xu et al., 2015).

2.2 Non-textual Factors

As Twitter is being continuously used as a medium for

information spread, the factors that affect its spread

vary on what metrics are used to compare. A previ-

ous study has shown that different types of users play

different roles in the dissemination of information on

Twitter (Cha et al., 2012). Mass media accounts, or-

dinary accounts, and even inﬂuential accounts such

as world leaders, politicians, and celebrities affect the

ﬂow of information spread signiﬁcantly (Cha et al.,

2012).

Other features such as retweets and followers are

also effective in playing a big role in the retransmis-

sion of information. Through the retweet feature, in-

formation is spread efﬁciently, as users who retweeted

only act as a middleware in the spreading process

(Zhang et al., 2017). This implies that readers on

Twitter will be only focused on the original message

without taking into account the users who retweeted

the messages. Meanwhile, hashtags, commonly used

as a metric to measure the popularity, were proven

to ineffective in the process of information spread on

Twitter (Skaza and Blais, 2017).

Furthermore, these non-textual factors are affect-

ing each other’s process, such as when inﬂuential ac-

counts’ posts trigger the retweet actions from their

followers resulting in the spread of the information to

accounts connected to the accounts’ followers. This

effect quickly became a continuous chain of actions,

making the information spread faster from one point

to the next.

2.3 Tweet Sentiment

There have been some researches done on analyzing

the sentiments of different messages and information

and how these sentiments affect the spread of this

piece of information on SNS (Tsugawa and Ohsaki,

2015; Hansen et al., 2011; Brady et al., 2017).

HEALTHINF 2022 - 15th International Conference on Health Informatics

422

A study by (Tsugawa and Ohsaki, 2015) inves-

tigated the relationship between the sentiment of a

message on SNS and its virality, which is deﬁned as

the volume and speed of message diffusion. Through

extensive analyses, the research found that messages

which were perceived to have a negative polarity are

more likely to get re-posted signiﬁcantly, and more

quickly and frequently compared to messages with ei-

ther positive or neutral polarity. This is also supported

by the research done by (Hansen et al., 2011), which

sought to ﬁnd out what kind of sentiment would af-

fect the virality of a message on Twitter. The research

found that although negative sentiments enhanced the

virality of the message in the context of news, this was

not the case for the non-news tweets.

2.4 Tweet Topic

Understanding the topic of a tweet is crucial to iden-

tifying the information in the text. Previous studies

have used topic modeling to identify the topic a tweet

falls under.

Studies have made use of topic modeling to try

and categorize tweets into different models. A study

done by (Pirri et al., 2020) used topic modeling to

explore the nature of and extract topics posted by

users and organizations on Twitter during World Lu-

pus Day. Meanwhile, (Abd-Alrazaq et al., 2020) also

used topic modeling to identify main topics from the

tweets related to the COVID-19 pandemic. For the

ﬁrst study, twelve topics were discovered. From the

results, tweets that shares additional information is

more prevalent compared to awareness messages and

informative content. It was also found that the gen-

eral public was more interested in tweets that made

the reader understand the illness and its manifesta-

tions. The second study categorized the tweets into

twelve categories, which were grouped into four main

themes: the origin of the virus, its sources, its impact

on people, countries, and the economy, and ways of

mitigating the risk of infection. In particular, users

were focused on the impact of the virus on people and

countries, which consist of death count and emotional

and psychological impact of the virus, particularly the

fear and stress about COVID-19 and the lack of vac-

cine treatments to prevent it.

3 METHODOLOGY

Figure 1 shows the process in analyzing the Twitter

data.

Figure 1: Collection and Analysis Process.

3.1 Data Gathering

The data was gathered from Twitter with Tweepy,

a Python library that accesses the Twitter API,

from December 4, 2020, to June 4, 2021. The

data gathering was conducted with snowball sam-

pling method using custom-made crawlers to extracts

tweets every day, ﬁltering the tweets by the geolo-

cation, Philippines, and by a set of health-related

keywords — “tb”, “tuberculosis”, “WorldTBDay”,

and “TBFreePh”. The top ten Philippine-geolocated

tweets was sorted according to the number of retweets

(retweet count) and likes (favorite count). Due

to the inﬂuence of the ofﬁcial Twitter accounts of

the Department of Health (DOH), @DOHgovph, and

the World Health Organization (WHO) Philippines,

@WHOPhilippines, as sources of health information,

these accounts’ tweets were also collected.

3.2 Data Transformation

The data gathered were then converted into a CSV

format that is acceptable in Gephi along with other

details as the additional attributes for attribute-

categorization graphs. The raw Twitter data is con-

verted into graphs where the users are the nodes and

another graph where the tweets are the nodes. The

edges of both graphs are directed edges representing

directed interactions that exist between the users or

tweets.

There are two kinds of graph being generated for

social network analysis. Graphs with users as the

nodes are used to identify the inﬂuential users in the

network through their interactions with other users in

the network, and how they affect the spread of infor-

mation. During the process, all edges that have the

same source and target nodes will be skipped, to direct

How Health Information Spreads in Twitter: The Whos and Whats of Philippine TB-data

423

the focus to the interaction between different users in

the network. Meanwhile, graphs with tweets as the

nodes are used to see how the attributes of tweets such

as keyword, hashtag, and sentiment of the tweet can

affect the spread of information.

3.3 Data Cleaning

To remove the ambiguity of the keywords used to ex-

tract the Twitter data such as “tb”, the data were fur-

ther cleaned by generating the graph with tweets as

the nodes and importing the generated data in Gephi

to see the clusters of tweets built. Each central node

of each cluster was manually checked and validated

to see if the tweet was related to tuberculosis health

information. All clusters with a central node that did

not have any relation to health information were re-

moved. Tweets or nodes that did not have any in-

teraction coming out from them were also removed.

Finally, the cleaned data was exported and used to ﬁl-

ter out the other data used for sentiment identiﬁcation

and topic modeling.

3.3.1 Text Preprocessing

To determine the sentiment of the tweets, tweets that

are in Filipino or English were further processed. The

full

text attribute of the tweet is used to deter-

mine the sentiment of the tweet. The preprocessing

done to the full text consists of lowercasing and

removing handles and hyperlinks. For topic mod-

eling, additional preprocessing consisted of remov-

ing single quotes and newlines. Before tokenizing,

English tweets underwent lemmatization and Part-

Of-Speech (POS) tagging, preserving only “nouns”,

“proper nouns”, and “adjectives”. The tweets were

then combined and underwent the following prepro-

cessing: removing stopwords for English and Taga-

log, converting text to bigrams, creating dictionaries,

and converting to Bag-of-Words (BOW).

The social network analysis provides a visualiza-

tion of the network and analyzes the relationships of

the nodes to each other while measuring their central-

ities. Two types of networks are generated, one where

the user is the node and the other where the tweet is

the node. Moreover, it is to explore the different fac-

tors that might affect the spread of information in the

network.

3.3.2 Visualize using Gephi

In the visualization, the username of the Twitter user

is displayed. This is done to identify inﬂuential users

that affect the spread of Tuberculosis information in

the Philippines. Furthermore, data privacy is not a

concern, since Section 1.2 Public Information in Twit-

ter Privacy Policy states that most activities on Twit-

ter and all users’ information including their proﬁle,

tweets, and interactions (replies, retweets, likes, and

quote tweets) are considered public data.

3.3.3 Measuring the Centrality

After the network is built, Gephi is used to visualize

the network in a form of a graph. Gephi also has built-

in features to compute the different centrality mea-

sures of the network (Grandjean, 2015).

3.3.4 Sentiment Identiﬁcation

The sentiment of a tweet is naively determined by

checking if the tweet contains a negative keyword. If

it contained predeﬁned negative words, emojis, and

emoticons. If the words, emojis, or emoticons exist,

the tweet is then labeled negative, otherwise it is la-

beled positive. After the categorization, each node in

the network is labeled by its sentiment. Then the sen-

timent is graphed together with the other attributes

of the tweet to see if the sentiment is a factor in the

spread of the tb data.

3.3.5 Topic Modeling

To identify the topic of the tweet, a network with

tweets as the nodes was generated and plugged into

Gephi. The visualized tweets were grouped together

according to their modularity class using the com-

pute modularity feature of Gephi to distinguish the

different communities that exist within the network,

and then were exported to undergo topic modeling.

The full text of every tweet from the exported data

underwent text preprocessing and were plugged into

the unsupervised topic modeling algorithm, Latent

Dirichlet Allocation (LDA). The keywords per topic

that were generated by the LDA were then used to la-

bel each modularity class with its respective topic by

a domain expert following a speciﬁc framework, with

an example shown in Table 1. Afterwards, the labeled

community were once again plugged into Gephi to vi-

sualize the change in topics across the network.

3.4 Result Interpretation

The different graphs from the SNA were analyzed

to see the factors affecting the information spread.

Each centrality measure was observed to identify if

the nodes with high centrality values were one of

the most inﬂuential nodes in the network. Inﬂuen-

tial nodes were determined by being the top source

of health information and nodes that cause the most

HEALTHINF 2022 - 15th International Conference on Health Informatics

424

Table 1: Sample results for topic modeling.

Comm ID Keywords Theme

1 trafﬁc, car, road,

drive, stop, passen-

ger, park, wheel

Vehicle

2 sick, fever, symp-

toms, recovery,

cough, temperature

Illness

3 shot, efﬁcacy, vac-

cine, antibody, im-

munity, dose

Vaccine

propagation of information. Different factors that can

affect the spread of information are analyzed by de-

termining which non-textual factors are recurring the

most in the inﬂuential nodes.

4 RESULT AND DISCUSSION

Different graphs were generated with Gephi, produc-

ing different insights regarding the relationships be-

tween the users and the non-textual and textual fac-

tors of tweets. The aforementioned centrality mea-

sures were also explored in Gephi, producing graphs

with the respective insights discussed. Analyses of

the results were also given by domain experts knowl-

edgeable in SNA and health communication.

4.1 Twitter Interaction Graphs

Figure 2 shows the user interactions, with Figure 2a

showing all user interactions while Figure 2b to Fig-

ure 2d isolate the retweet, reply, and quote tweet in-

teractions respectively. Notable accounts with high

number of users interacting with them consist of

WHOPhilippines, DOHgovph, and fan accounts of

Alden Richards, a local celebrity who spoke at a

World TB Day event. There is also signiﬁcant dis-

parity between the number of retweet interactions that

make up most of the interactions with the number of

reply and quote tweet interactions, which may suggest

that replying and quote tweeting are not as strong as

retweets in spreading information due to them requir-

ing users to add more information than what is writ-

ten in the original tweet, rather than retweeting which

simply forwards the original message.

The experts also noted that two organizational ac-

tors are driving the conversation, both WHOPhilip-

pines and DOHgovph, while the presence of fan ac-

counts could denote the power of fan bases in driv-

ing the conversation as well. Another insight noted

is that WHOPhilippines and DOHgovph did not in-

teract with each other’s tweets, instead having some

users interact with both their tweets. Another sig-

niﬁcant ﬁnding by the experts is that the same set

of actors, namely WHOPhilippines and DOHgovph,

are still prominent regardless of interaction, and there

seems to be a lack of prominent medical accounts or

media accounts that drive the conversation regarding

tuberculosis. One possible explanation for the lack

of media accounts is that since tuberculosis is not as

relevant to the public, media accounts would instead

focus on other news which garners more viewership.

4.2 Attribute-categorized Graphs

In the Attribute-Categorized Graphs, we checked how

the following four (4) attributes affect the spread of

information as shown in the interactions of the tweets:

1. keywords used in the tweet

2. hashtags used in the tweet

3. media (e.g. video or image) attached in the tweet

4. sentiment of the tweet

4.2.1 Keyword

The network in Figure 4 shows the interaction of in-

divudal tweets to each other, with the colors of the

nodes denoting the use of speciﬁc keyword (i.e. tb, tu-

berculosis). The sizes of the nodes are based on their

out-degree centrality. As seen in the graph, the key-

words “tb” is the dominant keyword. The two biggest

(a) All Interactions (b) Retweets (c) Replies (d) Quote Tweets

Figure 2: Twitter Interactions Graphs.

How Health Information Spreads in Twitter: The Whos and Whats of Philippine TB-data

425

(a) tb (b) tuberculosis (c) TBFreePh (d) WorldTBDay

Figure 3: Keyword Categorized (Users as Nodes).

nodes belong to one cluster, where the root tweet is

made by WHOPhilippines and the biggest node is a

reply WHOPhilippines made to its own tweet. The

use of certain keywords (in this case, “tb”) is essential

in propagating the information within the network as

more users interact with speciﬁc keywords compared

to the others.

Figure 4: Keyword Categorized (Tweet as Nodes).

To further analyze the effects of keywords in in-

formation spread, each tuberculosis-related keyword

was plotted wherein the user is the node of the net-

work as seen in Figure 3. The size of the node is

the out-degree of the users, which is the number of

the users that interacted with them. The ”tb” keyword

graph has the most number of nodes and edges. More-

over, the difference from the other graphs are drastic.

This means that most users used “tb” keyword when

they are talking about tuberculosis.

Overall, the use of certain keywords, speciﬁ-

cally “tb”, is essential in propagating the informa-

tion within the network, as the majority of the tweets

used “tb” when they are referring to tuberculosis. Ad-

ditionally, more users interacted with this keyword

compared to the others.

4.2.2 Hashtags

Figure 5 shows a graph where the nodes are tweets

and the color of the nodes represent a group of hash-

tags the tweets use. The node size is the out-degree

Figure 5: Categorized by the Existence of Hashtags (Tweets

as Nodes).

value of the nodes. The majority of the tweets about

tuberculosis do not make use of any hashtags in their

tweets. However, it can be seen that the most preva-

lent hashtag is #ALDENRichards. While the orig-

inal tweet contains nothing about the artist Alden

Richards, since he is the ambassador of World TB

Day, his fans are using his name as a hashtag in the

tuberculosis-related tweets. Moreover, the network

shows that not the same hashtag sets are being used

in one cluster. Oftentimes, the presence of hashtags

decreases per interaction made by the users. Some-

times, there are new hashtags that pop up while the

information is cascading, like the “ALDENRichards”

hashtag.

4.2.3 Media Attachment

The graph in Figure 6 is a network, where the tweet

is the node and the color is categorized by the pres-

ence or absence of media attachment. It can be seen

that the majority of the tweets don’t interact using

media attachments. Even though the majority of the

root nodes contain media attachments in their tweets,

the node with the highest out-degree centrality does

not contain any media attachments. Additionally, the

3rd biggest node also does not contain media attach-

ments. This shows that media attachments do not

signiﬁcantly play a role in the spread of information

but rather, it is because of the users being inﬂuential,

HEALTHINF 2022 - 15th International Conference on Health Informatics

426

Figure 6: Categorized by the Existence of Media Attach-

ment (Tweets as Nodes).

which is why the clusters are big.

4.2.4 Sentiment

Figure 7: Categorized by Sentiment (Tweet as Nodes).

The graph in Figure 7 shows the tweet as a node

network, where the node colors are categorized by

sentiment and the edges are categorized by interac-

tion. It can be seen that the majority of the tweets

are negative, with a percentage of 84.7% and with

retweets inheriting the sentiment of the original tweet.

In some cases, the polarity of the tweet changes when

the sentiment is changed when replying or quote

tweeting.

The second biggest cluster showed that the ma-

jority of the children nodes with differing sentiments

were made from quote tweeting the parent node.

The cluster on the lower right show that the interac-

tions were replies to the original tweet. Upon manu-

ally checking the tweets, they contain negative words

against the disease. This could be because, in the

context of diseases, people naturally disagree or go

against it rather than support or encourage it. Overall,

the sentiment of the tweet does not affect the spread

of tuberculosis information in the network.

4.3 SNA Centrality Measures

Figure 8 shows a set of graphs with users as nodes,

where the colors of the nodes are based on the modu-

larity class or the community in the network, and the

sizes of the nodes are based on the out-degree, eigen-

vector, betweenness, and closeness centrality.

It can be seen in Figure 8a that when measur-

ing based on out-degree centrality, the top inﬂuen-

tial users are health organizational accounts and a

fan account. As noted by the experts, the differ-

ence of impact towards the spread of information be-

tween the two health organizations was mainly caused

by how often they tweeted, where DOHgovph only

tweeted once as compared to WHOPhilippines. Fur-

thermore, the impact of fan base towards the spread

of information can be seen through the network built

by urmyﬂashlight.

On the other hand, Figure 8b shows that the inﬂu-

ential users are fan accounts or individual accounts,

as this type of account were most likely to react or

response to anyone who interacted with their tweets,

compared to health organizational accounts, thus,

raising their centrality value. The experts also noted

that health organizational accounts will most likely to

have low eigenvector centrality value due to the lack

of interactions they had with their followers, making

them merely sources of information.

Furthermore, ﬁgure 8c shows that the inﬂuential

users are health professionals and individual users,

including the fan accounts, signifying how informa-

tion are being passed often through the aforemen-

tioned user types, and was forwarded to other con-

nected users in the network.

As for the closeness centrality, the experts con-

cluded that the use of closeness centrality as a metric

to measure how inﬂuential a user in this case was not

very useful, as there were too many small and discon-

nected clusters in the network as shown in Figure 8d.

Furthermore, when it comes to the tuberculosis

data, the user with the highest number of followers

is DOHgovph, and its number of followers is signiﬁ-

cantly higher compared to the rest of users in the net-

work. This shows how the number of followers of

each user does not have signiﬁcant contribution or ef-

fect towards the spread of information in the network.

4.4 Topic Results

As seen in Figure 9, the topics of the interactions be-

tween users of the same clusters were not limited to

what the central nodes were talking about, as its topic

was directed towards a different focus after the inter-

actions For example, the two biggest nodes were the

How Health Information Spreads in Twitter: The Whos and Whats of Philippine TB-data

427

(a) Out-Degree (b) Eigenvector (c) Betweenness (d) Closeness

Figure 8: Centralities Graphs.

Figure 9: Topic Graph generated from results of Topic Mod-

eling.

tweets by WHOPhilippines, reminding people about

the virtual event of World TB Day. Even though both

central nodes talked about “Event”, the topics that

were being discussed by most of the tweets that in-

teracted with the central nodes were about “Ambas-

sador”, which in this context would be the fan base

of Alden Richards as he was the ambassador of the

World TB Day on last March 24, 2021. Furthermore,

aside from World TB Day, when it comes to tuber-

culosis health information, other users in the network

were also seen to be exchanging information about

tuberculosis treatments and health recommendations

for people afﬂicted with tuberculosis.

5 CONCLUSIONS

Twitter is a social media platform that can be utilized

to spread information, especially health information.

With tuberculosis being identiﬁed as a global public

health threat, this paper utilized social network anal-

ysis, topic modeling, and sentiment identiﬁcation to

analyze the spread of tuberculosis health information

on Twitter in the Philippine setting.

The results show that among the interactions

available in Twitter, retweets are proven to have more

impact in triggering more interactions and spreading

information, followed by quote tweets, and replies as

the Twitter interactions that are less likely to have any

impact when it comes to tuberculosis health informa-

tion. Furthermore, among the attributes of the tweet

only the keyword attribute played a major role in the

spread of tuberculosis information in the network.

Meanwhile, the centrality measures show that

WHOPhilippines is the most inﬂuential user in the

spread of tuberculosis health information on Twitter.

However, with a bigger dataset or a different focus on

health information, it is still possible that other health

organizations, health professionals, or even individual

users might arise as inﬂuential users in the network.

6 FUTURE WORKS

Future studies may further analyze the results and

conﬁrm whether these ﬁndings hold for other similar

datasets. The results may also be used as a foundation

for other studies which may opt to ﬁnd information re-

lated to tuberculosis health information spread in the

Philippines, such as ﬁnding efﬁcient ways to spread

health information through Twitter. Meanwhile, other

studies may also build upon the pipeline created by

this study, such as including ﬁnding topics commonly

used within different clusters, and use the pipeline in

other domains outside of tuberculosis and health in-

formation.

REFERENCES

Abd-Alrazaq, A., Alhuwail, D., Househ, M., Hamdi, M.,

and Shah, Z. (2020). Top concerns of tweeters during

the covid-19 pandemic: Infoveillance study. Journal

of Medical Internet Research, 22(4):e19016.

Bihari, A. and Pandia, M. K. (2015). Eigenvector central-

ity and its application in research professionals’ rela-

tionship network. In 2015 International Conference

on Futuristic Trends on Computational Analysis and

Knowledge Management (ABLAZE). IEEE.

Brady, W. J., Wills, J. A., Jost, J. T., Tucker, J. A.,

and Van Bavel, J. J. (2017). Emotion shapes the

HEALTHINF 2022 - 15th International Conference on Health Informatics

428

diffusion of moralized content in social networks.

Proceedings of the National Academy of Sciences,

114(28):7313–7318.

Brashers, D. E., Neidig, J. L., and Goldsmith, D. J. (2004).

Social support and the management of uncertainty for

people living with hiv or aids. Health Communication,

16(3):305–331.

Cha, M., Benevenuto, F., Haddadi, H., and Gummadi,

K. (2012). The world of connections and informa-

tion ﬂow in twitter. IEEE Transactions on Systems,

Man, and Cybernetics - Part A: Systems and Humans,

42(4):991–998.

Chua, Y. T. (2020). Interest in news high, but trust low in

the philippines—2020 digital news report.

Grandjean, M. (2015). Gephi: Introduction to network anal-

ysis and visualisation. page 12.

Grandjean, M. (2016). A social network analysis of twitter:

Mapping the digital humanities community. Cogent

Arts & Humanities, 3(1).

Hansen, D. L., Shneiderman, B., Smith, M. A., and Himel-

boim, I. (2020). Twitter: Information ﬂows, inﬂu-

encers, and organic communities, page 161–178. El-

sevier.

Hansen, L. K., Arvidsson, A., Nielsen, F. A., Colleoni, E.,

and Etter, M. (2011). Good Friends, Bad News - Af-

fect and Virality in Twitter, volume 185, page 34–43.

Springer Berlin Heidelberg.

Himelboim, I., Smith, M. A., Rainie, L., Shneiderman,

B., and Espina, C. (2017). Classifying twitter topic-

networks using social network analysis. Social Media

+ Society, 3(1):205630511769154.

Kudchadkar, S. R. and Carroll, C. L. (2020). Using social

media for rapid information dissemination in a pan-

demic: #pedsicu and coronavirus disease 2019. Pedi-

atric Critical Care Medicine, Publish Ahead of Print.

Liang, H., Fung, I. C.-H., Tse, Z. T. H., Yin, J., Chan, C.-H.,

Pechta, L. E., Smith, B. J., Marquez-Lameda, R. D.,

Meltzer, M. I., Lubell, K. M., and et al. (2019). How

did ebola information spread on twitter: broadcasting

or viral spreading? BMC Public Health, 19(1):438.

Maharani, W., Adiwijaya, and Gozali, A. A. (2014). Degree

centrality and eigenvector centrality in twitter. In 2014

8th International Conference on Telecommunication

Systems Services and Applications (TSSA). IEEE.

Pershad, Y., Hangge, P., Albadawi, H., and Oklu, R. (2018).

Social medicine: Twitter in healthcare. Journal of

Clinical Medicine, 7(6):121.

Pirri, S., Lorenzoni, V., Andreozzi, G., Mosca, M., and

Turchetti, G. (2020). Topic modeling and user net-

work analysis on twitter during world lupus awareness

day. International Journal of Environmental Research

and Public Health, 17(15):5440.

Sbafﬁ, L. and Rowley, J. (2017). Trust and credibility in

web-based health information: A review and agenda

for future research. Journal of Medical Internet Re-

search, 19(6):e218.

Skaza, J. and Blais, B. (2017). Modeling the infectiousness

of twitter hashtags. Physica A: Statistical Mechanics

and its Applications, 465:289–296.

Skiles, M. P., Curtis, S. L., Angeles, G., Mullen, S., and

Senik, T. (2018). Evaluating the impact of social

support services on tuberculosis treatment default in

ukraine. PLOS ONE, 13(8):e0199513.

Tambuscio, M., Ruffo, G., Flammini, A., and Menczer,

F. (2015). Fact-checking effect on viral hoaxes: A

model of misinformation spread in social networks.

In Proceedings of the 24th International Conference

on World Wide Web - WWW ’15 Companion, page

977–982. ACM Press.

Tsugawa, S. and Ohsaki, H. (2015). Negative messages

spread rapidly and widely on social media. In Pro-

ceedings of the 2015 ACM on Conference on Online

Social Networks - COSN ’15, page 151–160. ACM

Press.

Wieland, M. L., Nelson, J., Palmer, T., O’Hara, C., Weis,

J. A., Nigon, J. A., and Sia, I. G. (2013). Evaluation of

a tuberculosis education video among immigrants and

refugees at an adult education center: A community-

based participatory approach. Journal of Health Com-

munication, 18(3):343–353.

Xu, W. W., Chiu, I.-H., Chen, Y., and Mukherjee, T.

(2015). Twitter hashtags for health: applying network

and content analyses to understand the health knowl-

edge sharing in a twitter-based community of practice.

Quality & Quantity, 49(4):1361–1380.

Zhang, X., Han, D.-D., Yang, R., and Zhang, Z.

(2017). Users’ participation and social inﬂuence dur-

ing information spreading on twitter. PLOS ONE,

12(9):e0183290.

How Health Information Spreads in Twitter: The Whos and Whats of Philippine TB-data

429