Assessment of Antipsychotic Medications on Social Media: Machine Learning Study

Assessment of beliefs and attitudes towards benzodiazepines using machine learning based on social media posts: an observational study

BACKGROUND

Benzodiazepines are frequently prescribed drugs; however, their prolonged use can lead to tolerance, dependence, and other adverse effects. Despite these risks, long-term use remains common, presenting a public health concern. This study aims to explore the beliefs and opinions held by the public regarding benzodiazepines, as understanding these perspectives may provide insights into their usage patterns.

METHODS

We collected public tweets published in English between January 1, 2019, and October 31, 2020, that mentioned benzodiazepines. The content of each tweet and the characteristics of the users were analyzed using a mixed-method approach, including manual analysis and semi-supervised machine learning.

RESULTS

Over half of the Twitter users highlighted the efficacy of benzodiazepines, with minimal discussion of their side effects. The most active participants in these conversations were patients and their families, with health professionals and institutions being notably absent. Additionally, the drugs most frequently mentioned corresponded with those most commonly prescribed by healthcare professionals.

CONCLUSIONS

Social media platforms offer valuable insights into users' experiences and opinions regarding medications. Notably, the sentiment towards benzodiazepines is predominantly positive, with users viewing them as effective while rarely mentioning side effects. This analysis underscores the need to educate physicians, patients, and their families about the potential risks associated with benzodiazepine use and to promote clinical guidelines that support the proper management of these medications.

CLINICAL TRIAL NUMBER

Not applicable.

Machine learning and natural language processing to assess the emotional impact of influencers' mental health content on Instagram

Background

This study aims to examine, through artificial intelligence, specifically machine learning, the emotional impact generated by disclosures about mental health on social media. In contrast to previous research, which primarily focused on identifying psychopathologies, our study investigates the emotional response to mental health-related content on Instagram, particularly content created by influencers/celebrities. This platform, especially favored by the youth, is the stage where these influencers exert significant social impact, and where their analysis holds strong relevance. Analyzing mental health with machine learning techniques on Instagram is unprecedented, as all existing research has primarily focused on Twitter.

Methods

This research involves creating a new corpus labelled with responses to mental health posts made by influencers/celebrities on Instagram, categorized by emotions such as love/admiration, anger/contempt/mockery, gratitude, identification/empathy, and sadness. The study is complemented by modelling a set of machine learning algorithms to efficiently detect the emotions arising when faced with these mental health disclosures on Instagram, using the previous corpus.

Results

Results have shown that machine learning algorithms can effectively detect such emotional responses. Traditional techniques, such as Random Forest, showed decent performance with low computational loads (around 50%), while deep learning and Bidirectional Encoder Representation from Transformers (BERT) algorithms achieved very good results. In particular, the BERT models reached accuracy levels between 86-90%, and the deep learning model achieved 72% accuracy. These results are satisfactory, considering that predicting emotions, especially in social networks, is challenging due to factors such as the subjectivity of emotion interpretation, the variability of emotions between individuals, and the interpretation of emotions in different cultures and communities.

Discussion

This cross-cutting research between mental health and artificial intelligence allows us to understand the emotional impact generated by mental health content on social networks, especially content generated by influential celebrities among young people. The application of machine learning allows us to understand the emotional reactions of society to messages related to mental health, which is highly innovative and socially relevant given the importance of the phenomenon in societies. In fact, the proposed algorithms' high accuracy (86-90%) in social contexts like mental health, where detecting negative emotions is crucial, presents a promising research avenue. Achieving such levels of accuracy is highly valuable due to the significant implications of false positives or false negatives in this social context.

Social networks use in the context of Schizophrenia: a review of the literature

Schizophrenia is a persistent mental health condition that, while presenting challenges, underscores the dynamic nature of cognitive functions and encourages a unique perspective on how individuals engage with their surroundings. Social networks, as a means of communication of great importance at the present time, are for this type of people a way of interacting with their environment with a high level of security. The aim is to find out how schizophrenia is dealt with in different social networks and to differentiate between different types of articles dealing with the use of Facebook, X (former Twitter), YouTube, TikTok, Instagram, and Weibo. A total of 45 articles to i) Social networks used, ii) Country of analyzed users, iii) age of the users analyzed, iv) focus of the analyzed manuscript (mental health literacy, stigmatization, detection of patterns associated with schizophrenia, and Harmful substance use). It was observed that 45.45% of the studies analyzed were conducted in the USA population, followed by UK and China (13.64%). The most analyzed social networks were those based on audiovisual communication (60%). Furthermore, the two main foci addressed in these articles were: stigmatization of schizophrenia with 16 articles (35.55%), following by the prediction of schizophrenia-detecting patterns with 15 articles (33.33%) and the use of social networks to stigmatize people with schizophrenia (38%) and only 14 articles (31.11%) were focused on mental health literacy. Likewise, it was found that there is great potential in the use of the analysis of the content generated, as possible predictors of the presence of this disease, which would allow rapid detection and intervention for psychosis and schizophrenia.

Insights from the Twittersphere: a cross-sectional study of public perceptions, usage patterns, and geographical differences of tweets discussing cocaine

Introduction

Cocaine abuse represents a major public health concern. The social perception of cocaine has been changing over the decades, a phenomenon closely tied to its patterns of use and abuse. Twitter is a valuable tool to understand the status of drug use and abuse globally. However, no specific studies discussing cocaine have been conducted on this platform.

Methods

111,508 English and Spanish tweets containing "cocaine" from 2018 to 2022 were analyzed. 550 were manually studied, and the largest subset underwent automated classification. Then, tweets related to cocaine were analyzed to examine their content, types of Twitter users, usage patterns, health effects, and personal experiences. Geolocation data was also considered to understand regional differences.

Results

A total of 71,844 classifiable tweets were obtained. Among these, 15.95% of users discussed the harm of cocaine consumption to health. Media outlets had the highest number of tweets (35.11%) and the most frequent theme was social/political denunciation (67.88%). Regarding the experience related to consumption, there are more tweets with a negative sentiment. The 9.03% of tweets explicitly mention frequent use of the drug. The continent with the highest number of tweets was America (55.44% of the total).

Discussion

The findings underscore the significance of cocaine as a current social and political issue, with a predominant focus on political and social denunciation in the majority of tweets. Notably, the study reveals a concentration of tweets from the United States and South American countries, reflecting the high prevalence of cocaine-related disorders and overdose cases in these regions. Alarmingly, the study highlights the trivialization of cocaine consumption on Twitter, accompanied by a misleading promotion of its health benefits, emphasizing the urgent need for targeted interventions and antidrug content on social media platforms. Finally, the unexpected advocacy for cocaine by healthcare professionals raises concerns about potential drug abuse within this demographic, warranting further investigation.

Exploring Perceptions About Paracetamol, Tramadol, and Codeine on Twitter Using Machine Learning: Quantitative and Qualitative Observational Study

BACKGROUND

Paracetamol, codeine, and tramadol are commonly used to manage mild pain, and their availability without prescription or medical consultation raises concerns about potential opioid addiction.

OBJECTIVE

This study aims to explore the perceptions and experiences of Twitter users concerning these drugs.

METHODS

We analyzed the tweets in English or Spanish mentioning paracetamol, tramadol, or codeine posted between January 2019 and December 2020. Out of 152,056 tweets collected, 49,462 were excluded. The content was categorized using a codebook, distinguishing user types (patients, health care professionals, and institutions), and classifying medical content based on efficacy and adverse effects. Scientific accuracy and nonmedical content themes (commercial, economic, solidarity, and trivialization) were also assessed. A total of 1000 tweets for each drug were manually classified to train, test, and validate machine learning classifiers.

RESULTS

Of classifiable tweets, 42,840 mentioned paracetamol and 42,131 mentioned weak opioids (tramadol or codeine). Patients accounted for 73.10% (60,771/83,129) of the tweets, while health care professionals and institutions received the highest like-tweet and tweet-retweet ratios. Medical content distribution significantly differed for each drug (P<.001). Nonmedical content dominated opioid tweets (23,871/32,307, 73.9%), while paracetamol tweets had a higher prevalence of medical content (33,943/50,822, 66.8%). Among medical content tweets, 80.8% (41,080/50,822) mentioned drug efficacy, with only 6.9% (3501/50,822) describing good or sufficient efficacy. Nonmedical content distribution also varied significantly among the different drugs (P<.001).

CONCLUSIONS

Patients seeking relief from pain are highly interested in the effectiveness of drugs rather than potential side effects. Alarming trends include a significant number of tweets trivializing drug use and recreational purposes, along with a lack of awareness regarding side effects. Monitoring conversations related to analgesics on social media is essential due to common illegal web-based sales and purchases without prescriptions.

Understanding Public Perceptions and Discussions on Opioids Through Twitter: Cross-Sectional Infodemiology Study

BACKGROUND

Opioids are used for the treatment of refractory pain, but their inappropriate use has detrimental consequences for health. Understanding the current experiences and perceptions of patients in a spontaneous and colloquial environment regarding the key drugs involved in the opioid crisis is of utmost significance.

OBJECTIVE

The study aims to analyze Twitter content related to opioids, with objectives including characterizing users participating in these conversations, identifying prevalent topics and gauging public perception, assessing opinions on drug efficacy and tolerability, and detecting discussions related to drug dispensing, prescription, or acquisition.

METHODS

In this cross-sectional study, we gathered public tweets concerning major opioids posted in English or Spanish between January 1, 2019, and December 31, 2020. A total of 256,218 tweets were collected. Approximately 27% (69,222/256,218) were excluded. Subsequently, 7000 tweets were subjected to manual analysis based on a codebook developed by the researchers. The remaining databases underwent analysis using machine learning classifiers. In the codebook, the type of user was the initial classification domain. We differentiated between patients, family members and friends, health care professionals, and institutions. Next, a distinction was made between medical and nonmedical content. If it was medical in nature, we classified it according to whether it referred to the drug's efficacy or adverse effects. In nonmedical content tweets, we analyzed whether the content referred to management issues (eg, pharmacy dispensation, medical appointment prescriptions, commercial advertisements, or legal aspects) or the trivialization of the drug.

RESULTS

Among the entire array of scrutinized pharmaceuticals, fentanyl emerged as the predominant subject, featuring in 27% (39,997/148,335 posts) of the tweets. Concerning user categorization, roughly 70% (101,259/148,335) were classified as patients. Nevertheless, tweets posted by health care professionals obtained the highest number of retweets (37/16,956, 0.2% of their posts received over 100 retweets). We found statistically significant differences in the distribution concerning efficacy and side effects among distinct drug categories (P<.001). Nearly 60% (84,401/148,335) of the posts were devoted to nonmedical subjects. Within this category, legal facets and recreational use surfaced as the most prevalent themes, while in the medical discourse, efficacy constituted the most frequent topic, with over 90% (45,621/48,777) of instances characterizing it as poor or null. The opioid with the greatest proportion of tweets concerning legal considerations was fentanyl. Furthermore, fentanyl was the drug most frequently offered for sale on Twitter, while methadone generated the most tweets about pharmacy delivery.

CONCLUSIONS

The opioid crisis is present on social media, where tweets discuss legal and recreational use. Opioid users are the most active participants, prioritizing medication efficacy over side effects. Surprisingly, health care professionals generate the most engagement, indicating their positive reception. Authorities must monitor web-based opioid discussions to detect illicit acquisitions and recreational use.

Assessment of beliefs and attitudes about electroconvulsive therapy posted on Twitter: An observational study

BACKGROUND

Electroconvulsive therapy (ECT) is an effective and safe medical procedure that mainly indicated for depression, but is also indicated for patients with other conditions. However, ECT is among the most stigmatized and controversial treatments in medicine. Our objective was to examine social media contents on Twitter related to ECT to identify and evaluate public views on the matter.

METHODS

We collected Twitter posts in English and Spanish mentioning ECT between January 1, 2019 and October 31, 2020. Identified tweets were subject to a mixed method quantitative-qualitative content and sentiment analysis combining manual and semi-supervised natural language processing machine-learning analyses. Such analyses identified the distribution of tweets, their public interest (retweets and likes per tweet), and sentiment for the observed different categories of Twitter users and contents.

RESULTS

"Healthcare providers" users produced more tweets (25%) than "people with lived experience" and their "relatives" (including family members and close friends or acquaintances) (10% combined), and were the main publishers of "medical" content (mostly related to ECT's main indications). However, more than half of the total tweets had "joke or trivializing" contents, and such had a higher like and retweet ratio. Among those tweets manifesting personal opinions on ECT, around 75% of them had a negative sentiment.

CONCLUSIONS

Mixed method analysis of social media contents on Twitter offers a novel perspective to examine public opinion on ECT, and our results show attitudes more negative than those reflected in studies using surveys and other traditional methods.

Transferability Based on Drug Structure Similarity in Automatic Classification of Noncompliant Drug Use on Social Media: Natural Language Processing Approach (Preprint)

BACKGROUND

Medication noncompliance is a critical issue because of the increased number of drugs sold on the web. Web-based drug distribution is difficult to control, causing problems such as drug noncompliance and abuse. The existing medication compliance surveys lack completeness because it is impossible to cover patients who do not go to the hospital or provide accurate information to their doctors, so a social media-based approach is being explored to collect information about drug use. Social media data, which includes information on drug usage by users, can be used to detect drug abuse and medication compliance in patients.

OBJECTIVE

This study aimed to assess how the structural similarity of drugs affects the efficiency of machine learning models for text classification of drug noncompliance.

METHODS

This study analyzed 22,022 tweets about 20 different drugs. The tweets were labeled as either noncompliant use or mention, noncompliant sales, general use, or general mention. The study compares 2 methods for training machine learning models for text classification: single-sub-corpus transfer learning, in which a model is trained on tweets about a single drug and then tested on tweets about other drugs, and multi-sub-corpus incremental learning, in which models are trained on tweets about drugs in order of their structural similarity. The performance of a machine learning model trained on a single subcorpus (a data set of tweets about a specific category of drugs) was compared to the performance of a model trained on multiple subcorpora (data sets of tweets about multiple categories of drugs).

RESULTS

The results showed that the performance of the model trained on a single subcorpus varied depending on the specific drug used for training. The Tanimoto similarity (a measure of the structural similarity between compounds) was weakly correlated with the classification results. The model trained by transfer learning a corpus of drugs with close structural similarity performed better than the model trained by randomly adding a subcorpus when the number of subcorpora was small.

CONCLUSIONS

The results suggest that structural similarity improves the classification performance of messages about unknown drugs if the drugs in the training corpus are few. On the other hand, this indicates that there is little need to consider the influence of the Tanimoto structural similarity if a sufficient variety of drugs are ensured.

Analyzing Psychotherapy on Twitter: An 11-Year Analysis of Tweets From Major U.S. Media Outlets

BACKGROUND

The Internet has become the main source of information on health issues, and information now determines the therapeutic preferences of patients. For this reason, it is relevant to analyze online information discussing psychotherapy.

OBJECTIVE

To investigate tweets posted by 25 major US media outlets between 2009 and 2019 concerning psychotherapy.

METHODS

We investigated tweets posted by 25 major US media outlets about psychotherapy between January 2009 and December 2019 as well as the likes generated. In addition, we measured the sentiment analysis of these tweets.

RESULTS

Most of the tweets analyzed focused on Mindfulness (5,498), while a low number were related to Psychoanalysis (376) and even less to Cognitive-Behavioral Therapy (61). Surprisingly, Computer-supported therapy, Psychodynamic therapy, Systemic therapy, Acceptance and commitment therapy, and Dialectical behavior therapy did not generate any tweet. In terms of content, efficacy was the main focus of the posted tweets, receiving Cognitive-Behavioral Therapy and Mindfulness a positive appraisal.

CONCLUSIONS

US media outlets focused their interest on Mindfulness which may have contributed to the growing popularity in the past years of this therapeutic modality.