DIGITAL SOCIAL LISTENING IN COVID-19 PANDEMIC FOR INFORMED INTER-VENTIONS IN THE REPUBLIC OF MOLDOVA: INTEGRATED DATA

,


INTRODUCTION
The COVID-19 pandemic has created a global public health crisis, exacerbated by the lack of prompt scientific information at its onset regarding the causes of the disease, mode of transmission, prevention measures, necessary treatment, etc., and the simultaneous falsehoods occurrence, which rapidly were distributed (with or without malicious intent) in communities through various sources of information. The proliferation of media sources, particularly online (social media), has catalysed the phenomenon of the general increase in the volume of information, but in the COVID-19 pandemic, it has facilitated the spread of false information to unprecedented levels. The phenomenon of an overabundance of information (both correct and incorrect) from online and offline sources that accompanies an epidemic or another health crisis has conceptually consolidated in the term "infodemic" (1).
The consequences of infodemic are many and alarming. At the community level, it leads to the polarization of opinion, distortion of the interpretation of scientific evidence, promotion of fear and panic, increased mental and physical exhaustion of the population, increase in conspiratorial beliefs, decreased trust in governments and public health systems, as well as in the accuracy of official health messages. Within the healthcare system, infodemics can lead to the misallocation of resources and increased stress among healthcare providers, decreased access to medical care, conspiratorial beliefs, delayed delivery of high-quality care and proper treatment to patients, and more (2,3).
Particularly, the abundance of information causes confusion and uncertainty regarding the risks of the disease and the benefits of protection/prevention measures, and subsequently, negative attitudes/perceptions and hesitant behaviors towards the interventions initiated by authorities to reduce the impact of the pandemic. The important role of risk perception and the benefits of health interventions and the adoption of healthy behaviors in communities have been demonstrated by multiple studies, including those related to communication and behavioral factors in vaccine hesitancy (4,5,6).
Based on the above, monitoring the infodemic for the benefit of public health becomes an imperative and a challenge of our time. The World Health Organization, together with other international organizations such as UNICEF, Gavi, the Vaccine Alliance, etc., have made efforts to strengthen community capacities globally in the fight against the infodemic (7,8,9), including by popularizing and promoting social listening tools conducive to achieving this objective (10).
Digital "social listening" techniques make it possible to quickly process and analyses a large amount of information regarding the questions, concerns, feelings, narratives (opinions), perceptual errors (level of misinformation) of the population, as well as information gaps and outdated or distorted information on social media platforms. Based on these analyses, reports can be created to provide timely and relevant data for targeted, appropriate interventions to shape healthy behaviors, promote the resilience of these behaviors, and develop community engagement with public health strengthening actions.
In this context, our goal is the digital monitoring of public perceptions regarding the COVID-19 pandemic, the evaluation over temporal sequences of the content of messages on social media in order to understand and inform decisionmakers regarding targeted actions on needs/problems that need to be implemented.

MATERIAL AND METHODS
Digital monitoring of public perceptions regarding the COVID-19 pandemic was carried out using the Talkwalker software. This tool, primarily designed for marketing research, was adapted to the study's objective by creating syntaxes for searching on COVID-19 topics. Thus, using Boolean search operators (AND, OR, and NOT), key words and phrases (such as virus, COVID, vaccination, childhood vaccination, misinformation, etc.) were combined to establish and limit the search to a well-defined research's area of interest. In establishing the search syntax for COVID-19 information, the taxonomy for COVID-19 conversations in social listening developed and validated by Purnat et al. (11) was taken into account: Cause (virus cause, stigma surrounding the spread, spread and immunity); Disease (confirmed symptoms, other discussed symptoms, asymptomatic transmission, pre-symptomatic transmission, modes of transmission, protection against transmission, risks, vulnerable individuals/communi-2020 ties, restrictions, myths); Treatment (current treatment, research and development, unproven treatment, context: nutrition, myths); Vaccination (vaccine types, efficacy, side effects/safety, number of doses, contraindications); Misinformation (mischaracterization of the disease or protective measures, false treatments, conspiracy theories, sources and influencers, most engaging information, statistics and data); Interventions (testing, supportive care, personal measures, measures in public spaces, travel).
The software was also adjusted to the local specificities of the country by setting the health information channels (official pages of important health institutions, such as the Ministry of Health, the National Public Health Agency, Nicolae Testemitanu State University of Medicine and Pharmacy, the National Health Insurance Company, UNICEF, etc.) and information sources (types of mass media and social media alike).
Between January 2022 and November 2022, using the adjusted software, data was collected and analyzed from social platforms and forums. Monthly reports were compiled highlighting the most interesting trends and useful information regarding the number of posts related to COVID-19 (according to the established syntax), the en-gagement for each post (the number of likes, shares, and comments on the post), the potential reach -estimated number of people who can access the post/information, public sentiment regarding the information, influencers, the profile of internet users, the messages conveyed by internet users, and other related factors. Based on the analyzed data, recommendations were developed for actions aimed at adjusting the population's perceptions, attitudes, and behaviors. This article presents the integrated (synthesized) results of the study obtained over the entire duration of the research.

Fluctuation of interest on COVID-19 topic
During 01 January -30 November 2022, 18 thousand posts (results) regarding COVID-19 were shared, structured in descending order according to the engagement rate per post. Interest in this topic during the studied period shows a fluctuating curve, with several moments of increased interest (  In the period of 21 March 2022 -27 March 2022 (the second peak) 1000 results were recorded. Several news items presented high interest. The first news item posted by the WHO was about the meeting between the WHO Regional Director for Europe and the President of the Republic of Moldova, during which, support was provided by the WHO on the COVAX platform for the prioritized distribution of COVID-19 vaccine in the Republic 0 300 600 900 1200 No of results of Moldova. The message had a potential audience of 14,900 people, generated 139 interactions (engagement: likes -127, shares -10, comments -2), and registered a positive emotional reaction. The second news item referred to a message issued by Jurnal.md, stating that Moderna is set to seek emergency authorization for vaccinating children under the age of six for protection against the coronavirus. The potential audience of the post was 180600 people, and the engagement rate was 53 (9 shares, 8 likes, and 36 comments). Although the software qualifies the reaction to the message as neutral, an analysis of the comments reveals an evident negative feeling. The message "Hands off children, you scoundrels" received the most likes (22 in total), while "Take your hands off children" received 20 likes. During the same period, TV8.md reported that mobile vaccination points are no longer as popular as they used to be, and doctors say that the flow of people wishing to be immunized has decreased significantly. The message had a potential target audience of 178600 people and 20 interactions (engagement: 14 likes, 5 comments, and 1 share). The software qualifies the sentiment to the message as neutral, but the comments remained negative as related to the interventions of the authorities: "It was a stupid idea from the beginning. They want to control the population". In the period of 1 August 2022 -7 January 2022 (the fourth peak), 610 results were recorded. The post with the highest engagement was launched by the WHO and refers to promoting a proactive attitude towards vaccination by telling the story and opinion of a nurse who, being responsible for administering the second dose of the COVID-19 vaccine booster, encourages the public to get vaccinated. A total of 167 people showed engagement towards the post (including 15 shares, 128 likes, and 24 positive comments about the character of the post, most of which apparently came from the person's close circle). The potential reach for the message was 15000 people, and the sentiment towards the message was qualified as neutral.

Engagement or reactions to the topic
Engagement is a term that refers to taking a stance (involvement) or interaction/reaction of internet users to posted messages. In the analysed period, the engagement towards posts regarding COVID-19 amounts to 38500 shares, likes, and comments. The intensity of the recorded engagement is relatively uniform and low with only one significant increase throughout the year ( fig. 2).

2020
The increase in engagement intensity is recorded in the period of 28.02.2022-06.03.2022 and is related to the publication made by Unica.md regarding the decision of the French authorities to cease the enforcement of vaccine passports and protective masks. About 17500 people showed interest in this news, which represents almost half of all reactions/interactions recorded in a year. The news was shared 2400 times, received likes from 15100 internet users, and was commented on 521 times. The message with the most likes (183 likes) was: "2 years of theatre have ended, they got bored and lifted COVID restrictions. Now we have a war on stage and played very, let's see what's next. Only one thing is certain, we are simple people who lose our lives, our loved ones, our homes and are forced to leave the country. Cursed be you, LEAD-ERS OF COUNTRIES, who play dirty games and kill innocent people and children for your own goals".

Influential sources of information (top influencers)
Influencers are active authors or websites with a large potential reach and whose posts generate high-intensity engagement (with multiple interactions) and therefore have the potential to shape the opinions and behaviors of the population. During the studied period, the Unica.md Facebook page is qualified by the Talkwalker software as the most influential author (with an engagement of 18300), while News.yam.md is the most active author and site (having 1400 posts). The most intense interactions of internet users are recorded on the social media platform Facebook. The top influencers with posts on this site (following the sequence outlined in tab. 1, according to the level of Engagement) are: Unica.md, World Health Organization in Moldova, TV8.md, JurnalTV.md, and UNICEF Moldova. The table provides information on the following for each source: the total number of posts; the aggregated sentiments; the general and result-specific potential reach; the cumulative and per-post engagement.

Sentiment analysis on the topic of COVID-19
Social media sentiment refers to the emotion or attitude people express on social media about a particular message, topic, or post (12). Sentiment analysis in "digital social listening" is an automated process of attaching sentiments to the processed text in general terms of positive, negative or neutral polarity, but also in terms specific to the concrete type of emotion.
prise. The net sentiment score (NSS, calculated by subtracting the percentage of negative sentiment from the percentage of positive sentiment) is negative -44.6%. Figure 3. Share of sentiment (Talkwalker).

The social demographic profile of internet users
The profile of internet users is outlined through the analysis of the following independent variables:

DISCUSSIONS
In the study, it was found that online messages are critical and distrustful of pandemic health protection measures, vaccination, and authorities in general. The health interventions of authorities are perceived as inventions intended for controlling and manipulating the population and vacci-nation is seen as "a means of making money", which presents clear signs of misinformation and false attitudes typical of the infodemic. Public trust is an important factor in the effective implementation of safety measures in public health crises. Acceptance of prevention measures, including vaccination, depends on the public's trust in the safety and effectiveness of these measures, trust in the healthcare system, healthcare professionals, etc. (13,14).
Neutral sentiments, which are a priority in this study, can arise from the need for emotional detachment caused by the fatigue associated with the uncontrollable nature of the pandemic and the resulting uncertainty (15). At the same time, both negative and neutral sentiments marking the messages of Internet users are predictors/indicators of negative attitudes and hesitant behaviors. Information about threats and the fear generated thereby lead to avoidance and reduce intentions to adopt recommended protective behavior (16).
Taking into account that emotional responses to the pandemic are inevitable, and that emotions can make the population more susceptible to distorted information, it is necessary to develop and implement targeted strategies based on the needs of the population in public health crises (17).
Some recommendations for developing such strategies would be: (1) Promoting a single source for verifying information about COVID-19 (e.g. www.covidinfo.gov.md); (2) Strengthening a fact-checking platform for health information; (3) Forming teams to combat misinformation; (4) Developing partnerships between health authorities and media institutions; (5) Increasing trust in authorities, including through assertive communication strategies; (6) Promoting information that would lead to a positive attitude towards COVID-19 vaccination; (7) Using opinion leaders; (8) Diversifying accurate information offered to national minorities in different languages; (9) Building resistance to misinformation by developing internet users' competencies; (10) Using Social Listening reports to adjust communication strategies.

CONCLUSIONS
1. The population's interest in the COVID-19 topic is constantly decreasing and is characterised by a weak intensity of interactions (engagement) per post, which corresponds to a decrease in the intensity of the pandemic as a public health problem. 2020 2. The net sentiment score of the population regarding the topic of COVID-19 is negative, with a dominance of the emotion of anger.
3. Some of the posts with a high level of engagement express distrust regarding the reality of the pandemic, reluctance towards the actions taken by authorities to counter the pandemic, and/or conspiracy theories, which indicate the presence of falsehoods on social media and the need for targeted responses from authorities to increase the population's resistance to misinformation.
4. The phenomenon of infodemic in public health crises is very complex, and in order to address it at the national level, long-term communication strategies are needed.
5. In the context of the limitations observed in this study, it would be beneficial to combine future digital analyses of public opinions on public health topics with data from complementary research (such as KAP studies, offline population dialogues, observational qualitative studies, etc.).

LIMITATIONS
The present study has several limitations: The study gathered the opinions of those who are active on social media platforms, predominantly those aged 18-34, therefore not taking into account the voices of the entire population. At the same time, the opinions of people with limited access to social media (usually classified as vulnerable individuals) are not known. According to World Bank data, 76% of the country's population has access to the Internet (18).
Other limitations are related to the privacy policies of social media platforms, which restrict access to posts with private content (including closed groups), therefore only public posts were analysed in the study.
A particular suspicion is determined by the accuracy of evaluating internet users' sentiments. Some of the automated assessments made by the software did not correspond with the researchers' assessments, which is why sentiment assessments were sometimes done manually.