Skip to main content
Download PDF

Analysis of the hikikomori phenomenon – an international infodemiology study of Twitter data in Portuguese

BMC Public Health volume 24, Article number: 518 (2024) Cite this article

Abstract

Background

Hikikomori refers to the extreme isolation of individuals in their own homes, lasting at least six months. In recent years social isolation has become an important clinical, social, and public health problem, with increased awareness of hikikomori around the globe. Portuguese is one of the six most spoken languages in the world, but no studies have analysed the content regarding this phenomenon expressed in Portuguese.

Objective

To explore the hikikomori phenomenon on Twitter in Portuguese, utilising a mixed-methods approach encompassing content analysis, emotional analysis, and correlation analysis.

Methods

A mixed methods analysis of all publicly available tweets in the Portuguese language using a specific keyword (hikikomori) between 1st January 2008 and 19th October 2022. The content analysis involved categorising tweets based on tone, content, and user types, while correlation analysis was used to investigate user engagement and geographical distribution. Statistical analysis and artificial intelligence were employed to classify and interpret the tweet data.

Results

Among the total of 13,915 tweets generated, in terms of tone 10,731 were classified as “negative”, and 3184 as “positive”. Regarding content, “curiosities” was the most posted, as well as the most retweeted and liked topic. Worldwide, most of the hikikomori related tweets in Portuguese were posted in Europe, while “individuals with hikikomori” were the users most active posting. Regarding emotion analysis, the majority of tweets were “neutral”.

Conclusions

These findings show the global prevalence of the discourse on hikikomori phenomenon among Portuguese speakers. It also indicates an increase in the number of tweets on this topic in certain continents over the years. These findings can contribute to developing specific interventions, support networks, and awareness-raising campaigns for affected individuals.

Peer Review reports

Background

Hikikomori is a Japanese concept that emerged in the last decades of the twentieth century and refers to extreme withdrawal by some people in their own homes, lasting at least six months [1]. This isolation usually begins in adolescence or early adulthood and affects a higher proportion of males [2]. Hikikomori is characterised by a lack of participation in education, work, and other daily life activities, resulting in profound suffering or functional impairment [3,4,5]. Hikikomori impacts not only individuals who suffered from this condition, but also their families who desperately try to help them with no success [6, 7].

Whilst the internet is increasingly present in our lives, and therefore, people are more digitally connected, this may sometimes lead to isolation and increased loneliness [8]. There may be different situations why people isolate themselves on the internet, from online gaming to social networking, as well as family problems and difficulties of inclusion in the school of young people [9, 10]. The COVID-19 pandemic has also led to an increased awareness of social isolation. Studies in different countries, such as Brazil, Portugal, England, China, Iran, Malaysia, Pakistan, Philippines, Thailand, and Vietnam, showed that the pandemic affected mental health, leading to anxiety, depression, and stress, which resulted in social isolation [11,12,13,14]. The pandemic brought many social changes, from mandatory physical distancing to quarantines, which may have reinforced hikikomori-prone behaviour in people with mental health conditions [15, 16].

Infodemiology is the science that includes the collection, analysis and interpretation of health-related information in electronic media (internet and other digital sources) with the aim of informing public health and public policy [17, 18]. The internet is essential for sharing opinions, knowledge, and problems, and online social networks allow new forms of (online) communication and information sharing [19]. Twitter stands out as a prominent social network, valued for its short messages with a limit of 280 characters (currently up to 25,000 for subscribers of the premium version), with the possibility of anonymity [19,20,21,22]. Users can also repost content from other users (retweet), which makes them reach a larger population, sometimes marking their tweets with hashtags to identify a theme and allow other users to see related tweets [23, 24]. Recent research has embraced the analysis of tweets to explore, identify, and reach individuals with particular characteristics of interest to seek their perspectives, which otherwise would be marginalised or excluded. Twitter is a helpful tool to investigate the hikikomori phenomenon since those affected often use social media as an online ‘refuge’ [1, 25]. Previous studies of hikikomori using infodemiology explored contents and perceptions related to hikikomori on Twitter in Japanese and some Western languages, finding personal stories as the most posted content, and several mentions of hikikomori in non-Japanese Western languages [7, 8]. Understanding the content of tweets related to the hikikomori phenomenon provides valuable information on how this social issue is perceived and discussed among different people.

Although hikikomori was considered a typical phenomenon related to Japanese culture [2, 26, 27], over the years different studies have been conducted to investigate this phenomenon in other countries [1, 7, 28, 29]. While Portuguese is the sixth most spoken language in the world, to this day, no studies have analysed this phenomenon in this language.

The main aim of this study has been to explore contents related to the hikikomori phenomenon in the Portuguese language on Twitter. In particular, this study addressed the following research questions: (1) How do Twitter users express their views on hikikomori and what emotions do they associate with it? (2) Has the frequency of tweets about the hikikomori phenomenon changed over time, including during the COVID-19 pandemic? (3) What type of content related to hikikomori generates more interest on Twitter? (4) Are there any geographical differences in Portuguese-speaking countries regarding the hikikomori phenomenon tweets?

Methods

Research strategy

The research strategy focused on the collection and content analysis of Portuguese-language tweets about hikikomori. We included tweets that met the following criteria: (1) public (not private/protected) tweets; (2) use the keyword hikikomori; (3) tweets posted between 1st January 2008, and 19th October 2022; and (4) text in Portuguese language. The exclusion criteria were: (1) the majority of the text in the tweet in a language other than Portuguese; (2) tweets with only a link or image without any text. The tool used for collecting tweets was Tweet Binder, which has been widely used in previous research and provides access to 100% of public tweets [30, 31]. This tool provides the tweet text, count of retweets and likes for each tweet, as well as the date of publication, a link to the tweet in its context, user description, and geolocation data (obtained from the biography [bio] of the account that published the tweet). Regarding geolocation, we have gathered data at the country level. To enable comparison of the number of tweets posted in each country, we have grouped them into five continents: Americas (North, Central and South America), Europe, Africa, Asia and Oceania. The number of retweets and likes generated by each tweet were analysed as an indicator of user interest in a given topic.

Content analysis process

All tweets in our database were analysed using a content analysis procedure that consisted of creating codes and categories. We created a codebook to characterise each tweet, where each tweet was analysed according to the categories including: i) whether content was from a positive (tweets that express solidarity, self-disclosure, encouragement, gratitude, enthusiasm and pride) or negative perspective (tweets that express blame, stigma, and negative opinions); ii) whether it was about information relating to the hikikomori phenomenon; iii) whether it was an account of personal stories, or whether it was about curiosities of the phenomenon. Table 1 provides a detailed characterisation of the categories and examples of tweets that fall into these categories. The Twitter users were classified into three users types: “individuals with hikikomori” (people who describe themselves as hikikomori), “family and friends” (including family members and close friends or acquaintances), and “others” (unspecified as not fitting the previous types).

Table 1 Detailed characterisation of the categories and examples of tweets that fall into these categories
Full size table

Application of artificial intelligence to evaluate tweets

Recent technological advancements have led to the development of artificial intelligence (AI), including machine learning (ML) [32] and deep learning (DL) [33]. Neural networks, inspired by human brain neurons, are extensively used in various applications, such as natural language processing (NLP) [34], weather prediction [35], coronavirus detection [36], and image object detection [37]. In this study, a pre-trained neural network called BERTWEET [38], trained on 850 million English tweets, was employed to classify hikikomori-related tweets into different categories.

Before applying the BERTWEET network, the tweet database underwent preprocessing steps. Non-English tweets were translated into English using Google Translator since the network is trained only on English tweets. Previous studies have demonstrated that employing Google Translator to translate text into English and subsequently using a model trained on English tweets can enhance the performance of machine learning models [39,40,41]. The tweets were then normalised by removing special characters, separating negative tenses, and eliminating repeated characters. As BERTWEET was not originally trained to classify the desired categories, a process called fine-tuning was conducted. The manually classified tweets were randomly split into training and testing subsets, with 75% used for fine-tuning the network and 25% for validation. We validated the correct performance of the model in the validation set by computing the weighted F1 score, consistently achieving a score above 0.8 across all categories. This methodology, previously employed with positive outcomes [42], was employed to ensure the fine-tuned BERTWEET model performed well on the database. The fine-tuned BERTWEET model was then used to categorise the remaining tweets that had not been manually classified.

Additionally, the emotions expressed in the tweets were analysed using a pre-trained neural network called emotion-english-distilroberta-base [43]. This network, capable of detecting Ekman's six basic emotions plus neutral, was applied to the translated and normalised dataset of 13,915 tweets [44]. The emotion-english-distilroberta-base model, previously used in other research studies, does not require additional fine-tuning as it serves the same purpose for which it was originally trained [45, 46]. The number of likes and retweets each tweet generated, the date and time of each tweet, a permanent link to the tweet, and a description of each user's profile were collected. The nature of users who posted tweets was determined according to the available information (tweet content, description of the user's profile, or Twitter identifier).

Statistical analysis

Descriptive statistics were used to summarise tweets, likes, and retweets of users related to the content topics, tone, user type, emotions, and location. Correlation coefficients were determined to measure the strength and direction of their association. We also investigated the number of tweets and retweets generated by Twitter users by the days of the week grouped by user type. Additionally, we conducted an emotional analysis to graphically represent Twitter users' emotions by the type of user; time trends were used to describe the number of all tweets posted by continents between 2008–2022 as well as to illustrate the number of tweets posted before and after the COVID-19 pandemic; lastly, the proportions of Tweets posted by type of users across continents were graphically represented. All analyses were performed with STATA version 15 (StataCorp LP).

Ethical considerations

This study used publicly available tweets. This study received approval from the Ethics Committee of the Biomedical Sciences Institute Abel Salazar at the University of Porto (Ref 2023/CE/P03/(P401/CETI/ICBAS)).

Results

Overall Tweets, Likes, and Retweets

Among a total of 13,915 tweets posted in Portuguese about hikikomori, in terms of tone 3,184 (22.9%) were classified as “positive”, and 10,731 (77.1%) as “negative”. The content of the tweets was classified as “curiosities” (n = 9,818, 70.6%), “personal stories” (n = 3,675, 26.4%) or “concept information” (n = 422, 3.0%) (Table 2).

Table 2 Number of tweets, retweets, and likes related to the content topics, tone, user type, emotions, and location
Full size table

Within the user type, these tweets were classified as “individuals with hikikomori” (n = 5,571, 40.0%), “family and friends” (n = 4,653, 33.4%), or “others” (n = 3,691, 26.5%).

In regard to the emotion analysis, the higher proportion of tweets were “neutral” (n = 4,991, 35.9%). The emotion “joy” was present in 2,453 (17.6%) tweets, “surprise” in 1,943 (14.0%) tweets, “sadness” in 1,822 (13.1%) tweets, “fear” in 1,319 (9.5%) tweets, “anger” in 1,114 (8.0%) tweets, and “disgust” in 273 (2.0%) tweets.

Regarding the world regions, the majority of tweets (n = 12,400; 89.1%) did not report their location (“Missing”). Among the tweets with available location, the greater proportion of tweets was posted in “Europe” (n = 553, 4.0%). From the rest of the tweets, 487 (3.5%) tweets were posted in the “Americas”, 377 (2.7%) tweets were posted in “Asia”, 91 (0.7%) tweets were posted in “Africa” and 7 (0.1%) tweets were posted in “Oceania”.

There was a total of 21,307 likes and 5,359 retweets of hikikomori-related phenomenon in Portuguese. Regarding content, “curiosities” showed the highest proportion of the number of likes and retweets. Tweets posted in a positive way have more likes and retweets than tweets posted in a negative way. The "others" type of user presented a higher proportion of number of likes and retweets and, and “disgust” feelings was the sentiment with the most likes and retweets. “Europe” was the continent with the highest number of likes and retweets of hikikomori-related tweets in Portuguese-language.

The full details of the frequency of these tweets are reported in Table 2.

Content topics

The information regarding content topics and tweets is illustrated in Table 3, categorised by user type, emotions, continents and tone.

Table 3 Content topics related to tweets classified according to the type of user, emotions, continents, and tone
Full size table

Concept information: In terms of tone, 397 (94.1%) tweets were classified as “negative”, and 25 (5.9%) as “positive”. Within the user type, these were made by “others” in 398 (94.3%), “family and friends” in 15 (3.6%) tweets and “individuals with hikikomori” in 9 (2.1%) tweets. With respect to emotion analysis, the majority of tweets were “neutral” (n = 204, 48.3%). The emotion “fear” was present in 75 (17.8%) tweets, “sadness” in 53 (12.6%) tweets, “surprise” in 37 (8.8%) tweets, “joy” in 27 (6.4%) tweets, “anger” in 22 (5.2%) tweets and “disgust” in 4 (1.0%) tweets. Across the world, a total of 330 (78.2%) tweets did not report their location (“Missing”), 47 (11.1%) tweets were posted in “Americas”, 21 (5.0%) tweets were posted in “Asia”, 17 (4.0%) tweets were posted in “Europe”, 7 (1.7%) tweets were posted in “Africa”, and 0 (0%) tweets were posted in “Oceania”.

Personal stories: In terms of tone, 2,665 (72.5%) tweets were classified as “negative”, and 1,010 (27.5%) tweets as “positive”. Within the user type, these were made by “individuals with hikikomori” in 3,536 (96.2%) tweets, “family and friends” in 112 (3.1%), and “others” in 27 (0.7%) tweets. With respect to emotion analysis, most tweets were “neutral” (n = 1,058, 28.8%). The emotion “sadness” was present in 876 (23.8%) tweets, “joy” in 519 (14.1%) tweets, “surprise” in 479 (13.0%) tweets, “fear” in 391 (10.6%) tweets, “anger” in 258 (7.0%) tweets and “disgust” in 94 (2.6%) tweets. Worldwide, of 3,299 (89.8%) tweets we have no report of their location (“Missing”), 137 (3.7%) tweets were posted in “Europe”, 107 (2.9%) tweets were posted in “Asia”, 106 (2.9%) tweets were posted in “Americas”, 22 (0.6%) tweets were posted in “Africa” and 4 (0.1%) tweets were posted in “Oceania”.

Curiosities: In terms of tone, 7,669 (78.1%) tweets were classified as “negative”, and 2,149 (21.9%) as “positive”. Within the user type, these were made by “family and friends” in 4,526 (46.1%), “individuals with hikikomori” in 2,026 (20.6%) tweets, and “others” in 3,266 (33.3%) tweets. With respect to emotion analysis, the majority of tweets were “neutral” (n = 3,729, 38.0%). The emotion “joy” was present in 1,907 (19.4%) tweets, “surprise” in 1,427 (14.5%) tweets, “sadness” in 893 (9.1%) tweets, “fear” in 853 (8.7%) tweets, “anger” in 834 (8.5%) tweets and “disgust” in 175 (1.8%) tweets. Worldwide, we had no report of the location of 8,771 (89.3%) tweets (“Missing”), 399 (4.1%) tweets were posted in “Europe”, 334 (3.4%) tweets were posted in “Americas”, 249 (2.5%) tweets were posted in “Asia”, 62 (0.6%) tweets were posted in “Africa” and 3 (0%) tweets were posted in “Oceania”.

Emotion analysis

Figure 1 displays the Twitter users' emotions by type of user and reports that in all user types, the majority of tweets were “neutral”, whereas “disgust” emotion was the least frequent. Among the “individuals with hikikomori” users, the second most frequent emotion was “sadness”, whilst in the “family and friends” the second most frequent emotion was “surprise”, and in “others” types of users the second most prevalent emotion was “joy”.

Fig. 1
figure 1

Proportion of type of user by considered emotions of tweets

Full size image

Number of tweets over time (2008–2022)

Over the years (Fig. 2), there was a progressive increase in the number of tweets about hikikomori in Portuguese on all continents except in Oceania, where tweets were only retrieved since 2018 and where there are fewer tweets. Particularly in the Americas, Europe, and Asia, there was a higher increase in the number of tweets as the years passed.

Fig. 2
figure 2

Temporal evolution of tweets published from 2008 to 2022 by continents

Full size image

Regarding the COVID-19 pandemic period, a difference in the number of tweets posted before and after the pandemic can graphically be observed. During and after the pandemic the number of tweets was higher (Fig. 3).

Fig. 3
figure 3

Time trend of all tweets posted 2 years before and 2 years after the beginning of the COVID-19 pandemic (1st March 2020). In this figure, the line represents the frequency of tweets posted by users

Full size image

Continents

In Africa, Asia and Oceania, "individuals with hikikomori" were the type of user posting the most tweets. In America, there was an equal distribution of tweets posted by “individuals with hikikomori” and "others". In Europe, the majority of tweets were posted by “family and friends” (Fig. 4).

Fig. 4
figure 4

Proportion of tweets published by each user type (Y-axis) according to the continent (X-axis)

Full size image

Discussion

Key findings

The number of likes and retweets retrieved reflects an interest in the hikikomori phenomenon in Portuguese-speaking Twitter users. Regarding the content, tweets about curiosities were the most frequent, with the majority reporting hikikomori as something negative. Concerning the user type, “individuals with hikikomori” posted more tweets about the story of their day (personal stories), while for “family and friends” of people with hikikomori, curiosities was the topic that got more attention, and “others” posted more tweets about the concept of hikikomori itself. Among all the types of users, “individuals with hikikomori” were the most frequent. In terms of emotions, most tweets were neutral, with the least frequent emotion being disgust.

The knowledge of the hikikomori phenomenon in Portuguese language is globally spread, with users in Europe posting the most tweets and users in Oceania the least. The distribution of the different user types was relatively homogeneous, with most continents having more tweets posted by “individuals with hikikomori”. Many people seemed to join social networks to occupy their time during the COVID-19 period [15], increasing the number of tweets posted during the pandemic compared to before.

Comparison with the other literature

A study conducted with psychiatrists from several countries, supported the notion that the hikikomori phenomenon is global [27, 47]. Consistent with our findings, a previous study [7] identified Twitter contents suggesting the existence of hikikomori in Western countries based on the languages analysed (Catalan, English, French, Italian, and Spanish). The number of tweets posted from different continents has increased significantly over the years. This phenomenon may be explained, on the one hand, as people are using Twitter more to express their thoughts and feelings, and on the other hand, because the hikikomori phenomenon is becoming worldwide a more known and trending topic [1, 25, 47]. A study exploring the general population’s attitudes toward prison volunteering, demonstrated that users posted more tweets in Portuguese about volunteering in prisons in America (Ferrão Nunes et al.: Public discourse towards volunteering in prisons: An infodemiology study of Twitter data, submitted), whereas our study revealed that users from Europe posted the most. This may be related to the fact that people from Europe are more familiar with the term hikikomori than people from other continents.

While a previous study [8] of tweets about hikikomori in Japanese showed that personal stories were the most posted content, in our study, curiosities about hikikomori were predominant, suggesting that the hikikomori phenomenon has generated more interest among people since the users who posted the tweets know some curiosities about this phenomenon.

The interest of users in a specific topic can be assessed by analysing the number of retweets and likes that each tweet generates [48]. A study [8] that explored this phenomenon by analysing tweets had more likes than retweets, similar to our study. This could be attributed to the fact that people are more comfortable to show their interest in the subject, but less inclined to actively share the tweets for wider dissemination.

Two previous studies [7, 8] have additionally shown that the majority of tweets have reported hikikomori as a problem, which corroborates in the findings of our study, suggesting widespread awareness of the seriousness of the phenomenon.

Implication of the findings for future policies and research

Twitter is a helpful tool to study the hikikomori phenomenon since the affected individuals often use social networks as a refuge [1, 25]. Internet platforms allow us to reach socially withdrawn youth, and in this way, warn of risks and provide support.

This study suggests, similarly to other studies, that the hikikomori phenomenon is not only restricted to Japan [7, 8, 27, 47]; it specifically shows that the phenomenon is also known and spoken about among Portuguese speakers all over the world.

This study provides valuable insights into the hikikomori phenomenon on Twitter in Portuguese, offering implications for future research and policies aimed at addressing this complex social issue. These findings are a call to action to further investigate and develop methods to assist individuals with hikikomori who have been identified through the Twitter platform. The identification of individuals with hikikomori tendencies in this study underscores the need to provide them with the necessary support and resources to help these individuals seek help. Future research could focus on effective intervention and support strategies tailored to the needs of individuals affected by hikikomori. Exploring the potential of social media platforms like Twitter to provide psychosocial support and establish a good support network for individuals affected by hikikomori, as well as their families and friends, encouraging professional help as well, is another area for future investigation. Understanding how these platforms can be utilised to promote mental health and well-being in the context of hikikomori is crucial for developing effective support systems. It is also essential to promote education on the subject, alerting the public to its risks and consequences, and raising awareness of ways to prevent its development.

Future studies could consider conducting comparative analyses by examining data from Twitter and other social media platforms in other languages and countries where the hikikomori phenomenon has not yet been researched. This approach would provide a more comprehensive understanding of the global impact of hikikomori and its manifestations in different cultural contexts.

Strengths and limitations

This is the first study to investigate tweets related to hikikomori in the Portuguese language and to investigate its geographical differences. However, this study has some limitations. First, using a specific keyword may have limited our collection of tweets, since users may have used another term to refer to the hikikomori phenomenon, and thus tweets that refer to it in a different way may not have been collected. Furthermore, the potential deletion of tweets over time, whether by the social network, users, or through account deletion or protection, may have resulted in the exclusion of valuable data from the analysis, potentially impacting the comprehensiveness of the study findings. Additionally, the increase in tweets observed after the pandemic may not be related to users' interest in the topic. Therefore, it may not be conclusive that the pandemic has significantly impacted the level of interest in the hikikomori phenomenon compared to other topics. Another limitation of the study is that a significant portion of the tweets lacked geographical location data, limiting the ability to gain insights into regional variations of the hikikomori phenomenon among Portuguese speakers. Finally, whilst we analysed the number of retweets and likes generated by each tweet as an indicator of user interest in a given topic, these engagement metrics can be influenced by other factors and thus may not truly match user interest.

Conclusions

These findings demonstrate that the hikikomori phenomenon is actively discussed and appears to be prevalent among Portuguese speakers on all continents of the world. This supports and enhances our understanding of the globalisation of this phenomenon. The majority of tweets seemed to originate from individuals apparently affected by hikikomori. They reported curiosities about the phenomenon and discussed it in a negative light, suggesting that most people perceive it as a problem that needs to be addressed. Tweets about hikikomori in the Americas, Asia, and Europe have been steadily increasing in recent years, without a noticeable surge during the COVID-19 pandemic. This suggests that the prevalence of hikikomori is gradually growing in these continents. This study underscores the need to provide individuals with hikikomori tendencies with the necessary support and resources to improve their quality of life. It also offers implications for future research and policies addressing this complex social issue on Twitter.

Availability of data and materials

The datasets analysed available from the corresponding author upon reasonable request.

References

  1. Tateno M, et al. Internet Addiction, Smartphone Addiction, and Hikikomori Trait in Japanese Young Adult: Social Isolation and Social Network. Front Psychiatry. 2019;10:455.

    Article  PubMed  PubMed Central  Google Scholar 

  2. Amendola S, et al. Psychometric Properties of the Italian Version of the 25-Item Hikikomori Questionnaire. Int J Environ Res Public Health. 2022;19(20):13552.

  3. Mota AM, Lázaro M, Gaspar E, Rosa D, Farias AC. Case report of Hikikomori syndrome in Portugal: a primary syndrome or expression of another mental disorder? Braz J Psychiat. 2019;68(3):177–80.

  4. Yuen JWM, et al. A Physical Health Profile of Youths Living with a “Hikikomori” Lifestyle. Int J Environ Res Public Health. 2018;15(2):315.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Teo AR, et al. Development and validation of the 25-item Hikikomori Questionnaire (HQ-25). Psychiatry Clin Neurosci. 2018;72(10):780–8.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Teo AR. A new form of social withdrawal in Japan: a review of hikikomori. Int J Soc Psychiatry. 2010;56(2):178–85.

    Article  PubMed  Google Scholar 

  7. Pereira-Sanchez V, et al. Exploring the Extent of the Hikikomori Phenomenon on Twitter: Mixed Methods Study of Western Language Tweets. J Med Internet Res. 2019;21(5):e14167.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Pereira-Sanchez V, et al. Examining Tweet Content and Engagement of Users With Tweets About Hikikomori in Japanese: Mixed Methods Study of Social Withdrawal. J Med Internet Res. 2022;24(1):e31175.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Frankova I. Similar but Different: Psychological and Psychopathological Features of Primary and Secondary Hikikomori. Front Psychiatry. 2019;10:558.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Abbasi-Perez A, et al. Analysis of Tweets Containing Information Related to Rheumatological Diseases on Twitter. Int J Environ Res Public Health. 2021;18(17):9094.

  11. Duarte MDQ, Santos MAD, Lima C, Giordani JP, Trentini CM. Covid-19 and the impacts on mental health: a sample from Rio Grande do Sul. Brazil Ciência & Saúde Coletiva. 2020;25:3401–11.

    Article  Google Scholar 

  12. P., A., The Impact of the COVID-19 Pandemic on Mental Health. Acta Médica Periódica. 2020;33:356–7.

  13. Wang C, et al. The impact of COVID-19 pandemic on physical and mental health of Asians: A study of seven middle-income countries in Asia. PLoS ONE. 2021;16(2):e0246824.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Fancourt D, Steptoe A, Bu F. Trajectories of anxiety and depressive symptoms during enforced isolation due to COVID-19 in England: a longitudinal observational study. The lancet Psychiatry. 2021;8(2):141–9.

    Article  PubMed  Google Scholar 

  15. Herold M, et al. Hikikomori: a possible mental health consequence of the COVID-19 epidemic. Orv Hetil. 2021;162(41):1637–42.

    Article  PubMed  Google Scholar 

  16. Kubo H, et al. Risk factors of hikikomori among office workers during the COVID-19 pandemic: A prospective online survey. Curr Psychol, 2022;42(27):1–19.

  17. Eysenbach G. Infodemiology and infoveillance: framework for an emerging set of public health informatics methods to analyze search, communication and publication behavior on the Internet. J Med Internet Res. 2009;11(1):e11.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Mavragani A. Infodemiology and Infoveillance: Scoping Review. J Med Internet Res. 2020;22(4):e16206.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Alvarez-Mon MA, et al. Increasing Interest of Mass Communication Media and the General Public in the Distribution of Tweets About Mental Disorders: Observational Study. J Med Internet Res. 2018;20(5):e205.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Robillard JM, Johnson TW, Hennessey C, Beattie BL, Illes J. Aging 20: Health Information about Dementia on Twitter. PloS one. 2013;8(7):e69861.

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  21. Joseph AJ, et al. #Schizophrenia: Use and misuse on Twitter. Schizophr Res. 2015;165(2–3):111–5.

    Article  PubMed  ADS  Google Scholar 

  22. Prieto VM, et al. Twitter: a good place to detect health conditions. PLoS ONE. 2014;9(1):e86191.

    Article  PubMed  PubMed Central  ADS  Google Scholar 

  23. Reavley NJ, Pilkington PD. Use of Twitter to monitor attitudes toward depression and schizophrenia: an exploratory study. PeerJ. 2014;2:e647.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Lienemann BA, et al. Methods for Coding Tobacco-Related Twitter Data: A Systematic Review. J Med Internet Res. 2017;19(3):e91.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Liu LL, et al. Harnessing Social Media to Explore Youth Social Withdrawal in Three Major Cities in China: Cross-Sectional Web Survey. JMIR Ment Health. 2018;5(2):e34.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Kato TA, Kanba S. Boundless syndromes in modern society: An interconnected world producing novel psychopathology in the 21st century. Psychiatry Clin Neurosci. 2016;70(1):1–2.

    Article  PubMed  Google Scholar 

  27. Kato TA, et al. Are Japan’s hikikomori and depression in young people spreading abroad? Lancet. 2011;378(9796):1070.

    Article  PubMed  Google Scholar 

  28. Orsolini L, et al. Hikikomori and modern-type depression in Italy: A new phenotypical trans-cultural characterization? Int J Soc Psychiatry. 2022;68(5):1010–7.

    Article  PubMed  Google Scholar 

  29. Roza TH, Paim Kessler FH, Cavalcante Passos I. Hikikomori in Brazil: Context, clinical characteristics, and challenges. Int J Soc Psychiatry. 2023;69(4):1059–61.

    Article  PubMed  Google Scholar 

  30. Morstatter F, Pfeffer J, Liu H, Carley KM. Is the Sample Good Enough? Comparing Data from Twitter’s Streaming API with Twitter’s Firehose. Association for the Advancement of Artificial Intelligence. 2013;130. https://doi.org/10.1609/icwsm.v7i1.14401.

  31. Alvarez-Mon MA, et al. Areas of Interest and Stigmatic Attitudes of the General Public in Five Relevant Medical Conditions: Thematic and Quantitative Analysis Using Twitter. J Med Internet Res. 2019;21(5):e14110.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Helm JM, et al. Machine Learning and Artificial Intelligence: Definitions, Applications, and Future Directions. Curr Rev Musculoskelet Med. 2020;13(1):69–76.

    Article  PubMed  PubMed Central  Google Scholar 

  33. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–44.

    Article  CAS  PubMed  ADS  Google Scholar 

  34. Deng L, Liu Y. Deep Learning in Natural Language Processing. Singapore: Springer; 2018.

    Book  Google Scholar 

  35. Hongkang W, Li L, Yong W, Fanjia M, Haihua W, Sigrimis NA. Recurrent Neural Network Model for Prediction of Microclimate in Solar Greenhouse. IFAC-PapersOnLine. 2018;51(17):790–5.

    Article  Google Scholar 

  36. Islam MZ, Islam MM, Asraf A. A combined deep CNN-LSTM network for the detection of novel coronavirus (COVID-19) using X-ray images. Inform Med Unlocked. 2020;20:100412.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Kousik N, Natarajan Y, Raja RA, Kallam S, Patan R, Gandomi AH. Improved salient object detection using hybrid Convolution Recurrent Neural Network. Expert Systems with Applications, 2021;166:1–9.

  38. Nguyen DQ, Vu T, Nguyen AT. BERTweet: A pre-trained language model for English Tweets. 2020.

    Google Scholar 

  39. Chew EW, et al. Machine Translation for Accessible Multi-Language Text Analysis, in arXiv preprint arXiv:2301.08416. 2023: .

  40. Mahajan D, Chaudhary DK. Sentiment Analysis Using Rnn and Google Translator. In: 2018 8th International Conference on Cloud Computing, Data Science & Engineering (Confluence). Noida, India: IEEE; 2018. p. 798–802.

    Chapter  Google Scholar 

  41. Sagnika S, et al. A review on multi-lingual sentiment analysis by machine learning methods. Journal of Engineering Science and Technology Review. 2020;13(2):154.

    Article  Google Scholar 

  42. de Anta L, Alvarez-Mon MA, Donat-Vargas C, Lara-Abelanda FJ, Pereira-Sanchez V, Gonzalez Rodriguez C, Mora F, Ortega MA, Quintero J, Alvarez-Mon M. Assessment of beliefs and attitudes about electroconvulsive therapy posted on Twitter: An observational study. European psychiatry : the journal of the Association of European Psychiatrists. 2023;66(1): e11.

    Article  PubMed  Google Scholar 

  43. J., H. Emotion English DistilRoBERTa-base. [cited 2022; Available from: https://huggingface.co/j-hartmann/emotion-english-distilroberta-base.

  44. Ekman P. Basic emotions. In: Handbook of cognition and emotion. Hoboken, NJ, US: John Wiley & Sons Ltd; 1999. p. 45–60.

    Chapter  Google Scholar 

  45. Butt S, Sharma S, Sharma R, Sidorov G, Gelbukh A. What goes on inside rumour and non-rumour tweets and their reactions: A psycholinguistic analyses. Computers in Human Behavior, 2022;135:107345.

  46. Rozado D, Hughes R, Halberstadt J. Longitudinal analysis of sentiment and emotion in news media headlines using automated labelling with Transformer language models. PLoS ONE. 2022;17(10):e0276367.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Kato TA, et al. Does the “hikikomori” syndrome of social withdrawal exist outside Japan? A preliminary international investigation. Soc Psychiatry Psychiatr Epidemiol. 2012;47(7):1061–75.

    Article  PubMed  Google Scholar 

  48. Edney S, et al. Creating Engaging Health Promotion Campaigns on Social Media: Observations and Lessons From Fitbit and Garmin. J Med Internet Res. 2018;20(12):e10911.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

The authors wish to thank Mr Javier Abrego from Twitter Binder for his support in the retrieval of tweets through their search engine. The authors would also like to thank Dr Paul Wai-Ching Wong for his valuable comments on our findings.

Funding

This research benefited from funds to the Instituto de Salud Carlos III (ISCIII) through the project 'PI22/00653' co-funded by the European Union.

Author information

Authors and Affiliations

  1. Institute of Biomedical Sciences Abel Salazar, University of Porto, Porto, Portugal

    Francisca Correia Lopes & Mariana Pinto da Costa

  2. Institute of Psychiatry Psychology & Neuroscience, King´s College London, London, UK

    Mariana Pinto da Costa

  3. Department of Preventive Medicine and Public Health, School of Medicine, University of Navarra, IdiSNA, Navarra Institute for Health Research, Pamplona, Spain

    Cesar I Fernandez-Lazaro

  4. Department of Signal Theory and Communications and Telematic Systems and Computing, School of Telecommunications Engineering, Rey Juan Carlos University, 28942, Madrid, Spain

    Francisco J Lara-Abelenda

  5. Child Mind Institute, New York, NY, USA

    Victor Pereira-Sanchez

  6. Health Services Research & Development Center to Improve Veteran Involvement in Care, VA Portland Health Care System, Portland, OR, USA

    Alan R Teo

  7. Department of Psychiatry, Oregon Health & Science University, Portland, OR, USA

    Alan R Teo

  8. Department of Medicine and Medical Specialties, Faculty of Medicine and Health Sciences, University of Alcala, Alcala de Henares, Spain

    Miguel Angel Alvarez-Mon

  9. Department of Psychiatry and Mental Health, University Hospital Infanta Leonor, Madrid, Spain

    Miguel Angel Alvarez-Mon

  10. Ramón y Cajal Institute of Sanitary Research (IRYCIS), Madrid, Spain

    Miguel Angel Alvarez-Mon

Authors
  1. Francisca Correia Lopes
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Mariana Pinto da Costa
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Cesar I Fernandez-Lazaro
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Francisco J Lara-Abelenda
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Victor Pereira-Sanchez
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Alan R Teo
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Miguel Angel Alvarez-Mon
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

This article reports work for the research masters of FCL under the supervision of MPdC and MAAM. FCL, MPdC and MAAM contributed equally to writing this article. FJLA applied computational analysis. CIFL conducted the statistical analysis. FJLA, CIFL, ART and VPS each reviewed the article and provided critique and feedback to the article.

Corresponding author

Correspondence to Mariana Pinto da Costa.

Ethics declarations

The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs.

Ethics approval and consent to participate

This study obtained approval from the Ethics Committee of the Biomedical Sciences Institute Abel Salazar at the University of Porto (Ref 2023/CE/P03/(P401/CETI/ICBAS)).

Consent for publication

Not applicable.

Competing Interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Glossary

Hikikomori

Term used to describe the social phenomenon of individuals, often young adults, who withdraw from social life and isolate themselves at home for a prolonged period, avoiding social interaction.

Infodemiology

Refers to the study of how information spreads and impacts public health, especially on the internet. Analyses patterns of information disclosure to understand how it affects people's knowledge and behaviour.

Machine Learning

Subset of artificial intelligence that involves the development of algorithms and statistical models that enable computers to perform tasks without explicit programming. It relies on patterns and inference from data to improve performance over time.

Deep Learning

A specialised area of machine learning that involves using neural networks with multiple layers (deep neural networks) to model and solve complex problems. It is particularly effective in tasks such as image and speech recognition.

BERTweet

Natural language processing model based on BERT (Bidirectional Encoder Representations from Transformers) architecture, specifically tailored for processing and understanding Twitter data. It enhances language understanding in a social media context.

Emotion-English-Distilroberta-Base

Pre-trained language model based on the Distilroberta architecture, specialised in recognizing and understanding emotions in English text. It is trained to analyse and categorise emotions expressed in written content.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Correia Lopes, F., Pinto da Costa, M., Fernandez-Lazaro, C.I. et al. Analysis of the hikikomori phenomenon – an international infodemiology study of Twitter data in Portuguese. BMC Public Health 24, 518 (2024). https://doi.org/10.1186/s12889-023-17617-0

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12889-023-17617-0

Keywords

BMC Public Health

ISSN: 1471-2458

Contact us