Can Twitter be used to predict county excessive alcohol consumption rates?

Robust language-based mental health assessments in time and space through social media

In the most comprehensive population surveys, mental health is only broadly captured through questionnaires asking about "mentally unhealthy days" or feelings of "sadness." Further, population mental health estimates are predominantly consolidated to yearly estimates at the state level, which is considerably coarser than the best estimates of physical health. Through the large-scale analysis of social media, robust estimation of population mental health is feasible at finer resolutions. In this study, we created a pipeline that used ~1 billion Tweets from 2 million geo-located users to estimate mental health levels and changes for depression and anxiety, the two leading mental health conditions. Language-based mental health assessments (LBMHAs) had substantially higher levels of reliability across space and time than available survey measures. This work presents reliable assessments of depression and anxiety down to the county-weeks level. Where surveys were available, we found moderate to strong associations between the LBMHAs and survey scores for multiple levels of granularity, from the national level down to weekly county measurements (fixed effects β = 0.34 to 1.82; p < 0.001). LBMHAs demonstrated temporal validity, showing clear absolute increases after a list of major societal events (+23% absolute change for depression assessments). LBMHAs showed improved external validity, evidenced by stronger correlations with measures of health and socioeconomic status than population surveys. This study shows that the careful aggregation of social media data yields spatiotemporal estimates of population mental health that exceed the granularity achievable by existing population surveys, and does so with generally greater reliability and validity.

Social Acceptability of Health Behavior Posts on Social Media: An Experiment

INTRODUCTION

Social media sites like Twitter (now X) are increasingly used to create health behavior metrics for public health surveillance. Yet little is known about social norms that may bias the content of posts about health behaviors. Social norms for posts about four health behaviors (smoking tobacco, drinking alcohol, physical activity, eating food) on Twitter/X were evaluated.

METHODS

This was a randomized experiment delivered via web-based survey to adult, English-speaking Twitter/X users in three Michigan, USA, counties from 2020 to 2022 (n=559). Each participant viewed 24 posts presenting experimental manipulations regarding four health behaviors and answered questions about each post's social acceptability. Principal component analysis was used to combine survey responses into one perceived social acceptability measure. Linear mixed models with the Benjamini-Hochberg correction were implemented to test seven study hypotheses in 2023.

RESULTS

Supporting six hypotheses, posts presenting healthier (CI: 0.028, 0.454), less stigmatized behaviors (CI: 0.552, 0.157) were more socially acceptable than posts regarding unhealthier, stigmatized behaviors. Unhealthy (CI: -0.268, -0.109) and stigmatized behavior (CI: -0.261, -0.103) posts were less acceptable for more educated participants. Posts about collocated activities (CI: 0.410, 0.573) and accompanied by expressions of liking (CI: 0.906, 1.11) were more acceptable than activities undertaken alone or disliked. Contrary to one hypothesis, posts reporting unusual activities were less acceptable than usual ones (CI: -0.472, 0.312).

CONCLUSIONS

Perceived social acceptability may be associated with the frequency and content of health behavior posts. Users of Twitter/X and other social media platform posts to estimate health behavior prevalence should account for potential estimation biases from perceived social acceptability of posts.

Areas of interest and sentiment analysis towards second generation antipsychotics, lithium and mood stabilizing anticonvulsants: Unsupervised analysis using Twitter

BACKGROUND

Severe mental disorders like Schizophrenia and related psychotic disorders (SRD) or Bipolar Disorder (BD) require pharmacological treatment for relapse prevention and quality of life improvement. Yet, treatment adherence is a challenge, partly due to patients' attitudes and beliefs towards their medication. Social media listening offers insights into patient experiences and preferences, particularly in severe mental disorders.

METHODS

All tweets posted between 2008 and 2022 mentioning the names of the main drugs used in SRD and BD were analyzed using advanced artificial intelligence techniques such as machine learning, and deep learning, along with natural language processing.

RESULTS

In this 15-year study analyzing 893,289 tweets, second generation antipsychotics received more mentions in English tweets, whereas mood stabilizers received more tweets in Spanish. English tweets about economic and legal aspects displayed negative emotions, while Spanish tweets seeking advice showed surprise. Moreover, a recurring theme in Spanish tweets was the shortage of medications, evoking feelings of anger among users.

LIMITATIONS

This study's analysis of Twitter data, while insightful, may not fully capture the nuances of discussions due to the platform's brevity. Additionally, the wide therapeutic use of the studied drugs, complicates the isolation of disorder-specific discourse. Only English and Spanish tweets were examined, limiting the cultural breadth of the findings.

CONCLUSION

This study emphasizes the importance of social media research in understanding user perceptions of SRD and BD treatments. The results provide valuable insights for clinicians when considering how patients and the general public view and communicate about these treatments in the digital environment.

Beyond rating scales: With targeted evaluation, large language models are poised for psychological assessment

In this narrative review, we survey recent empirical evaluations of AI-based language assessments and present a case for the technology of large language models to be poised for changing standardized psychological assessment. Artificial intelligence has been undergoing a purported "paradigm shift" initiated by new machine learning models, large language models (e.g., BERT, LAMMA, and that behind ChatGPT). These models have led to unprecedented accuracy over most computerized language processing tasks, from web searches to automatic machine translation and question answering, while their dialogue-based forms, like ChatGPT have captured the interest of over a million users. The success of the large language model is mostly attributed to its capability to numerically represent words in their context, long a weakness of previous attempts to automate psychological assessment from language. While potential applications for automated therapy are beginning to be studied on the heels of chatGPT's success, here we present evidence that suggests, with thorough validation of targeted deployment scenarios, that AI's newest technology can move mental health assessment away from rating scales and to instead use how people naturally communicate, in language.

Disease, Scapegoating, and Social Contexts: Examining Social Contexts of the Support for Racist Naming of COVID-19 on Twitter

In early 2020, when COVID-19 began to spread in the United States, many Twitter users called it the "Chinese virus," blaming racial outgroups for the pandemic. I collected tweets containing the "Chinese virus" derivatives posted from March to August 2020 by users within the United States and created a data set with 141,290 tweets published by 50,695 users. I calculated the ratio of users who supported the racist naming of COVID-19 per county and merged Twitter data with the county-level census. Multilevel regression models show that counties with higher COVID-19 mortality or infection rates have more support for the racist naming. Second, the mortality and infection rates effects are stronger in counties with faster minority growth. Moreover, it is mainly in poor counties that minority growth enlarges the effects of infection and mortality rates. These findings relate to the theories on disease-induced xenophobia and the debate between conflict and contact theories.

A linguistic analysis of dehumanization toward substance use across three decades of news articles

Introduction

Substances and the people who use them have been dehumanized for decades. As a result, lawmakers and healthcare providers have implemented policies that subjected millions to criminalization, incarceration, and inadequate resources to support health and wellbeing. While there have been recent shifts in public opinion on issues such as legalization, in the case of marijuana in the U.S., or addiction as a disease, dehumanization and stigma are still leading barriers for individuals seeking treatment. Integral to the narrative of "substance users" as thoughtless zombies or violent criminals is their portrayal in popular media, such as films and news.

Methods

This study attempts to quantify the dehumanization of people who use substances (PWUS) across time using a large corpus of over 3 million news articles. We apply a computational linguistic framework for measuring dehumanization across three decades of New York Times articles.

Results

We show that (1) levels of dehumanization remain high and (2) while marijuana has become less dehumanized over time, attitudes toward other substances such as heroin and cocaine remain stable.

Discussion

This work highlights the importance of a holistic view of substance use that places all substances within the context of addiction as a disease, prioritizes the humanization of PWUS, and centers around harm reduction.

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.

Filling in the white space: Spatial interpolation with Gaussian processes and social media data

Full national coverage below the state level is difficult to attain through survey-based data collection. Even the largest survey-based data collections, such as the CDC's Behavioral Risk Factor Surveillance System or the Gallup-Healthways Well-being Index (both with more than 300,000 responses p.a.) only allow for the estimation of annual averages for about 260 out of roughly U.S. 3,000 counties when a threshold of 300 responses per county is used. Using a relatively high threshold of 300 responses gives substantially higher convergent validity-higher correlations with health variables-than lower thresholds but covers a reduced and biased sample of the population. We present principled methods to interpolate spatial estimates and show that including large-scale geotagged social media data can increase interpolation accuracy. In this work, we focus on Gallup-reported life satisfaction, a widely-used measure of subjective well-being. We use Gaussian Processes (GP), a formal Bayesian model, to interpolate life satisfaction, which we optimally combine with estimates from low-count data. We interpolate over several spaces (geographic and socioeconomic) and extend these evaluations to the space created by variables encoding language frequencies of approximately 6 million geotagged Twitter users. We find that Twitter language use can serve as a rough aggregate measure of socioeconomic and cultural similarity, and improves upon estimates derived from a wide variety of socioeconomic, demographic, and geographic similarity measures. We show that applying Gaussian Processes to the limited Gallup data allows us to generate estimates for a much larger number of counties while maintaining the same level of convergent validity with external criteria (i.e., N = 1,133 vs. 2,954 counties). This work suggests that spatial coverage of psychological variables can be reliably extended through Bayesian techniques while maintaining out-of-sample prediction accuracy and that Twitter language adds important information about cultural similarity over and above traditional socio-demographic and geographic similarity measures. Finally, to facilitate the adoption of these methods, we have also open-sourced an online tool that researchers can freely use to interpolate their data across geographies.

AI-based analysis of social media language predicts addiction treatment dropout at 90 days

The reoccurrence of use (relapse) and treatment dropout is frequently observed in substance use disorder (SUD) treatment. In the current paper, we evaluated the predictive capability of an AI-based digital phenotype using the social media language of patients receiving treatment for substance use disorders (N = 269). We found that language phenotypes outperformed a standard intake psychometric assessment scale when predicting patients' 90-day treatment outcomes. We also use a modern deep learning-based AI model, Bidirectional Encoder Representations from Transformers (BERT) to generate risk scores using pre-treatment digital phenotype and intake clinic data to predict dropout probabilities. Nearly all individuals labeled as low-risk remained in treatment while those identified as high-risk dropped out (risk score for dropout AUC = 0.81; p < 0.001). The current study suggests the possibility of utilizing social media digital phenotypes as a new tool for intake risk assessment to identify individuals most at risk of treatment dropout and relapse.

The promise of zero-shot learning for alcohol image detection: comparison with a task-specific deep learning algorithm

Exposure to alcohol content in media increases alcohol consumption and related harm. With exponential growth of media content, it is important to use algorithms to automatically detect and quantify alcohol exposure. Foundation models such as Contrastive Language-Image Pretraining (CLIP) can detect alcohol exposure through Zero-Shot Learning (ZSL) without any additional training. In this paper, we evaluated the ZSL performance of CLIP against a supervised algorithm called Alcoholic Beverage Identification Deep Learning Algorithm Version-2 (ABIDLA2), which is specifically trained to recognise alcoholic beverages in images, across three tasks. We found ZSL achieved similar performance compared to ABIDLA2 in two out of three tasks. However, ABIDLA2 outperformed ZSL in a fine-grained classification task in which determining subtle differences among alcoholic beverages (including containers) are essential. We also found that phrase engineering is essential for improving the performance of ZSL. To conclude, like ABIDLA2, ZSL with little phrase engineering can achieve promising performance in identifying alcohol exposure in images. This makes it easier for researchers, with little or no programming background, to implement ZSL effectively to obtain insightful analytics from digital media. Such analytics can assist researchers and policy makers to propose regulations that can prevent alcohol exposure and eventually prevent alcohol consumption.

Linguistic Methodologies to Surveil the Leading Causes of Mortality: Scoping Review of Twitter for Public Health Data

BACKGROUND

Twitter has become a dominant source of public health data and a widely used method to investigate and understand public health-related issues internationally. By leveraging big data methodologies to mine Twitter for health-related data at the individual and community levels, scientists can use the data as a rapid and less expensive source for both epidemiological surveillance and studies on human behavior. However, limited reviews have focused on novel applications of language analyses that examine human health and behavior and the surveillance of several emerging diseases, chronic conditions, and risky behaviors.

OBJECTIVE

The primary focus of this scoping review was to provide a comprehensive overview of relevant studies that have used Twitter as a data source in public health research to analyze users' tweets to identify and understand physical and mental health conditions and remotely monitor the leading causes of mortality related to emerging disease epidemics, chronic diseases, and risk behaviors.

METHODS

A literature search strategy following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) extended guidelines for scoping reviews was used to search specific keywords on Twitter and public health on 5 databases: Web of Science, PubMed, CINAHL, PsycINFO, and Google Scholar. We reviewed the literature comprising peer-reviewed empirical research articles that included original research published in English-language journals between 2008 and 2021. Key information on Twitter data being leveraged for analyzing user language to study physical and mental health and public health surveillance was extracted.

RESULTS

A total of 38 articles that focused primarily on Twitter as a data source met the inclusion criteria for review. In total, two themes emerged from the literature: (1) language analysis to identify health threats and physical and mental health understandings about people and societies and (2) public health surveillance related to leading causes of mortality, primarily representing 3 categories (ie, respiratory infections, cardiovascular disease, and COVID-19). The findings suggest that Twitter language data can be mined to detect mental health conditions, disease surveillance, and death rates; identify heart-related content; show how health-related information is shared and discussed; and provide access to users' opinions and feelings.

CONCLUSIONS

Twitter analysis shows promise in the field of public health communication and surveillance. It may be essential to use Twitter to supplement more conventional public health surveillance approaches. Twitter can potentially fortify researchers' ability to collect data in a timely way and improve the early identification of potential health threats. Twitter can also help identify subtle signals in language for understanding physical and mental health conditions.

Predicting U.S. county opioid poisoning mortality from multi-modal social media and psychological self-report data

Opioid poisoning mortality is a substantial public health crisis in the United States, with opioids involved in approximately 75% of the nearly 1 million drug related deaths since 1999. Research suggests that the epidemic is driven by both over-prescribing and social and psychological determinants such as economic stability, hopelessness, and isolation. Hindering this research is a lack of measurements of these social and psychological constructs at fine-grained spatial and temporal resolutions. To address this issue, we use a multi-modal data set consisting of natural language from Twitter, psychometric self-reports of depression and well-being, and traditional area-based measures of socio-demographics and health-related risk factors. Unlike previous work using social media data, we do not rely on opioid or substance related keywords to track community poisonings. Instead, we leverage a large, open vocabulary of thousands of words in order to fully characterize communities suffering from opioid poisoning, using a sample of 1.5 billion tweets from 6 million U.S. county mapped Twitter users. Results show that Twitter language predicted opioid poisoning mortality better than factors relating to socio-demographics, access to healthcare, physical pain, and psychological well-being. Additionally, risk factors revealed by the Twitter language analysis included negative emotions, discussions of long work hours, and boredom, whereas protective factors included resilience, travel/leisure, and positive emotions, dovetailing with results from the psychometric self-report data. The results show that natural language from public social media can be used as a surveillance tool for both predicting community opioid poisonings and understanding the dynamic social and psychological nature of the epidemic.

Opioid death projections with AI-based forecasts using social media language

Targeting of location-specific aid for the U.S. opioid epidemic is difficult due to our inability to accurately predict changes in opioid mortality across heterogeneous communities. AI-based language analyses, having recently shown promise in cross-sectional (between-community) well-being assessments, may offer a way to more accurately longitudinally predict community-level overdose mortality. Here, we develop and evaluate, TROP (Transformer for Opiod Prediction), a model for community-specific trend projection that uses community-specific social media language along with past opioid-related mortality data to predict future changes in opioid-related deaths. TOP builds on recent advances in sequence modeling, namely transformer networks, to use changes in yearly language on Twitter and past mortality to project the following year's mortality rates by county. Trained over five years and evaluated over the next two years TROP demonstrated state-of-the-art accuracy in predicting future county-specific opioid trends. A model built using linear auto-regression and traditional socioeconomic data gave 7% error (MAPE) or within 2.93 deaths per 100,000 people on average; our proposed architecture was able to forecast yearly death rates with less than half that error: 3% MAPE and within 1.15 per 100,000 people.

Understanding Alcohol Use Discourse and Stigma Patterns in Perinatal Care on Twitter

(1) Background: perinatal alcohol use generates a variety of health risks. Social media platforms discuss fetal alcohol spectrum disorder (FASD) and other widespread outcomes, providing personalized user-generated content about the perceptions and behaviors related to alcohol use during pregnancy. Data collected from Twitter underscores various narrative structures and sentiments in tweets that reflect large-scale discourses and foster societal stigmas; (2) Methods: We extracted alcohol-related tweets from May 2019 to October 2021 using an official Twitter search API based on a set of keywords provided by our clinical team. Our exploratory study utilized thematic content analysis and inductive qualitative coding methods to analyze user content. Iterative line-by-line coding categorized dynamic descriptive themes from a random sample of 500 tweets; (3) Results: qualitative methods from content analysis revealed underlying patterns among inter-user engagements, outlining individual, interpersonal and population-level stigmas about perinatal alcohol use and negative sentiment towards drinking mothers. As a result, the overall silence surrounding personal experiences with alcohol use during pregnancy suggests an unwillingness and sense of reluctancy from pregnant adults to leverage the platform for support and assistance due to societal stigmas; (4) Conclusions: identifying these discursive factors will facilitate more effective public health programs that take into account specific challenges related to social media networks and develop prevention strategies to help Twitter users struggling with perinatal alcohol use.

Tweeting about alcohol: Exploring differences in Twitter sentiment during the onset of the COVID-19 pandemic

This study explores pandemic-related changes in Twitter communication by examining differences in emotional, psychological and social sentiment between alcohol-related tweets and a random sample of non-alcohol tweets during the onset of the COVID-19 pandemic. Two equivalent size sets of English-language, COVID-specific tweets posted between February 1^st and April 20^th, 2020 are examined. The first set includes 1.5 million tweets containing alcohol-related keywords, while the second set does not contain such references. LIWC software analyzed the tweets for sentiment factors. ANCOVAs were used to determine whether language use significantly differed between the sets, considering differences in the pandemic period (before or after the pandemic declaration) while controlling for the number of tweets. The study found that tweets in the 40 days after March 11, 2020 contained more authentic language, more affiliation-oriented language, and exhibited more positive emotion than tweets in the 40 days pre-declaration. Alcohol-related status was a significant factor only when tweets contained personal concerns, regardless of pandemic period. Authenticity levels increased significantly in alcohol-related tweets post-declaration. The findings suggest alcohol may play a lesser role in the expression of psychological, social, and emotional sentiment than the pandemic period, but interaction between authentic language and alcohol references may reflect an increased use of alcohol for coping.

Substance use and sentiment and topical tendencies: a study using social media conversations of youth experiencing homelessness

Purpose

This study investigates associations between Facebook (FB) conversations and self-reports of substance use among youth experiencing homelessness (YEH). YEH engage in high rates of substance use and are often difficult to reach, for both research and interventions. Social media sites provide rich digital trace data for observing the social context of YEH's health behaviors. The authors aim to investigate the feasibility of using these big data and text mining techniques as a supplement to self-report surveys in detecting and understanding YEH attitudes and engagement in substance use.

Design/methodology/approach

Participants took a self-report survey in addition to providing consent for researchers to download their Facebook feed data retrospectively. The authors collected survey responses from 92 participants and retrieved 33,204 textual Facebook conversations. The authors performed text mining analysis and statistical analysis including ANOVA and logistic regression to examine the relationship between YEH's Facebook conversations and their substance use.

Findings

Facebook posts of YEH have a moderately positive sentiment. YEH substance users and non-users differed in their Facebook posts regarding: (1) overall sentiment and (2) topics discussed. Logistic regressions show that more positive sentiment in a respondent's FB conversation suggests a lower likelihood of marijuana usage. On the other hand, discussing money-related topics in the conversation increases YEH's likelihood of marijuana use.

Originality/value

Digital trace data on social media sites represent a vast source of ecological data. This study demonstrates the feasibility of using such data from a hard-to-reach population to gain unique insights into YEH's health behaviors. The authors provide a text-mining-based toolkit for analyzing social media data for interpretation by experts from a variety of domains.

Correcting Sociodemographic Selection Biases for Population Prediction from Social Media

Social media is increasingly used for large-scale population predictions, such as estimating community health statistics. However, social media users are not typically a representative sample of the intended population - a "selection bias". Within the social sciences, such a bias is typically addressed with restratification techniques, where observations are reweighted according to how under- or over-sampled their socio-demographic groups are. Yet, restratifaction is rarely evaluated for improving prediction. In this two-part study, we first evaluate standard, "out-of-the-box" restratification techniques, finding they provide no improvement and often even degraded prediction accuracies across four tasks of esimating U.S. county population health statistics from Twitter. The core reasons for degraded performance seem to be tied to their reliance on either sparse or shrunken estimates of each population's socio-demographics. In the second part of our study, we develop and evaluate Robust Poststratification, which consists of three methods to address these problems: (1) estimator redistribution to account for shrinking, as well as (2) adaptive binning and (3) informed smoothing to handle sparse socio-demographic estimates. We show that each of these methods leads to significant improvement in prediction accuracies over the standard restratification approaches. Taken together, Robust Poststratification enables state-of-the-art prediction accuracies, yielding a 53.0% increase in variance explained (R ²) in the case of surveyed life satisfaction, and a 17.8% average increase across all tasks.

Using Facebook language to predict and describe excessive alcohol use

BACKGROUND

Assessing risk for excessive alcohol use is important for applications ranging from recruitment into research studies to targeted public health messaging. Social media language provides an ecologically embedded source of information for assessing individuals who may be at risk for harmful drinking.

METHODS

Using data collected on 3664 respondents from the general population, we examine how accurately language used on social media classifies individuals as at-risk for alcohol problems based on Alcohol Use Disorder Identification Test-Consumption score benchmarks.

RESULTS

We find that social media language is moderately accurate (area under the curve = 0.75) at identifying individuals at risk for alcohol problems (i.e., hazardous drinking/alcohol use disorders) when used with models based on contextual word embeddings. High-risk alcohol use was predicted by individuals' usage of words related to alcohol, partying, informal expressions, swearing, and anger. Low-risk alcohol use was predicted by individuals' usage of social, affiliative, and faith-based words.

CONCLUSIONS

The use of social media data to study drinking behavior in the general public is promising and could eventually support primary and secondary prevention efforts among Americans whose at-risk drinking may have otherwise gone "under the radar."

Machine learning of language use on Twitter reveals weak and non-specific predictions

Depressed individuals use language differently than healthy controls and it has been proposed that social media posts can be used to identify depression. Much of the evidence behind this claim relies on indirect measures of mental health and few studies have tested if these language features are specific to depression versus other aspects of mental health. We analysed the Tweets of 1006 participants who completed questionnaires assessing symptoms of depression and 8 other mental health conditions. Daily Tweets were subjected to textual analysis and the resulting linguistic features were used to train an Elastic Net model on depression severity, using nested cross-validation. We then tested performance in a held-out test set (30%), comparing predictions of depression versus 8 other aspects of mental health. The depression trained model had modest out-of-sample predictive performance, explaining 2.5% of variance in depression symptoms (R² = 0.025, r = 0.16). The performance of this model was as-good or superior when used to identify other aspects of mental health: schizotypy, social anxiety, eating disorders, generalised anxiety, above chance for obsessive-compulsive disorder, apathy, but not significant for alcohol abuse or impulsivity. Machine learning analysis of social media data, when trained on well-validated clinical instruments, could not make meaningful individualised predictions regarding users’ mental health. Furthermore, language use associated with depression was non-specific, having similar performance in predicting other mental health problems.

Natural language analyzed with AI-based transformers predict traditional subjective well-being measures approaching the theoretical upper limits in accuracy

We show that using a recent break-through in artificial intelligence –transformers–, psychological assessments from text-responses can approach theoretical upper limits in accuracy, converging with standard psychological rating scales. Text-responses use people's primary form of communication –natural language– and have been suggested as a more ecologically-valid response format than closed-ended rating scales that dominate social science. However, previous language analysis techniques left a gap between how accurately they converged with standard rating scales and how well ratings scales converge with themselves – a theoretical upper-limit in accuracy. Most recently, AI-based language analysis has gone through a transformation as nearly all of its applications, from Web search to personalized assistants (e.g., Alexa and Siri), have shown unprecedented improvement by using transformers. We evaluate transformers for estimating psychological well-being from questionnaire text- and descriptive word-responses, and find accuracies converging with rating scales that approach the theoretical upper limits (Pearson r = 0.85, p < 0.001, N = 608; in line with most metrics of rating scale reliability). These findings suggest an avenue for modernizing the ubiquitous questionnaire and ultimately opening doors to a greater understanding of the human condition.

When are alcohol-related blackout Tweets written in the United States?

BACKGROUND

Alcohol use varies throughout the year and often peaks on weekends or during celebrations (e.g., New Year's). There is not a perfect correlation between alcohol use and negative consequences, and the extent to which one particularly risky consequence-an alcohol-related blackout-is more common during certain times of the year is unknown. Identifying when blackouts occur may help identify which periods are associated with more risk and be critical in designing public health campaigns. Thus, we examined Twitter data to ascertain whether alcohol-related blackouts occur more during certain holidays/celebrations than typical weekends and whether they differed in timing from general alcohol-related Tweets.

METHODS

We used a Twitter-sponsored platform to access unique Tweets written in the United States referencing blackouts (e.g., "blackout") and alcohol generally (e.g., "drunk").

RESULTS

The final dataset included 3.5 million blackout Tweets and 591 million alcohol Tweets (written between 2009 and 2020). Both blackout and alcohol Tweets were written in the late evening, on weekends, and during certain holidays (New Years, St. Patrick's). However, relative to typical weekends, only blackout Tweets were more common during Thanksgiving and only general alcohol-related Tweets were more common during Cinco de Mayo.

CONCLUSION

While blackout and alcohol-related Tweets were similar in time of day (peaking in the evening) and day of week (peaking on weekends), they differed during certain celebrations/holidays, suggesting that while alcohol use may be more common during some celebrations, others are more associated with serious harms.

Lessons for science and technology policy? Probing the Linkedin network of an RDI organisation

In this paper, we seek to examine the network of the Greek National Documentation Centre (EKT) as formed by its LinkedIn followers. By applying specific data collection and processing techniques, we explore the network of all the individuals that follow EKT’s LinkedIn page. Significant manual and automatic approaches have been implemented with regard to data extraction, data curation and data homogenization. The aim is to identify the network’s advancement over time, the institutions involved and the countries. The timeframe of the study spans from when the relevant LinkedIn page was constructed in 2015 to 2020. Findings indicate that there is a steady increase in the number of new followers, peaking in 2020. On an international scale, the evolution of the network of followers is imprinted and distributed in worldwide maps. In total, 68 countries have followed EKT over the examined time period. Also, in terms of followers’ institutional sector the Business Sector (BES) stands out (46.5%). Higher Education (HES) and Government Sector (GOV) are associated with 26.4 and 22.2% of the followers, respectively. Lastly, this paper provides a first institutional and country-level mapping of who constitutes the organisation’s interlocutors in the national and global RDI ecosystem.

Regional personality assessment through social media language

OBJECTIVE

We explore the personality of counties as assessed through linguistic patterns on social media. Such studies were previously limited by the cost and feasibility of large-scale surveys; however, language-based computational models applied to large social media datasets now allow for large-scale personality assessment.

METHOD

We applied a language-based assessment of the five factor model of personality to 6,064,267 U.S. Twitter users. We aggregated the Twitter-based personality scores to 2,041 counties and compared to political, economic, social, and health outcomes measured through surveys and by government agencies.

RESULTS

There was significant personality variation across counties. Openness to experience was higher on the coasts, conscientiousness was uniformly spread, extraversion was higher in southern states, agreeableness was higher in western states, and emotional stability was highest in the south. Across 13 outcomes, language-based personality estimates replicated patterns that have been observed in individual-level and geographic studies. This includes higher Republican vote share in less agreeable counties and increased life satisfaction in more conscientious counties.

CONCLUSIONS

Results suggest that regions vary in their personality and that these differences can be studied through computational linguistic analysis of social media. Furthermore, these methods may be used to explore other psychological constructs across geographies.

Deep Learning for Identification of Alcohol-Related Content on Social Media (Reddit and Twitter): Exploratory Analysis of Alcohol-Related Outcomes

BACKGROUND

Many social media studies have explored the ability of thematic structures, such as hashtags and subreddits, to identify information related to a wide variety of mental health disorders. However, studies and models trained on specific themed communities are often difficult to apply to different social media platforms and related outcomes. A deep learning framework using thematic structures from Reddit and Twitter can have distinct advantages for studying alcohol abuse, particularly among the youth in the United States.

OBJECTIVE

This study proposes a new deep learning pipeline that uses thematic structures to identify alcohol-related content across different platforms. We apply our method on Twitter to determine the association of the prevalence of alcohol-related tweets with alcohol-related outcomes reported from the National Institute of Alcoholism and Alcohol Abuse, Centers for Disease Control Behavioral Risk Factor Surveillance System, county health rankings, and the National Industry Classification System.

METHODS

The Bidirectional Encoder Representations From Transformers neural network learned to classify 1,302,524 Reddit posts as either alcohol-related or control subreddits. The trained model identified 24 alcohol-related hashtags from an unlabeled data set of 843,769 random tweets. Querying alcohol-related hashtags identified 25,558,846 alcohol-related tweets, including 790,544 location-specific (geotagged) tweets. We calculated the correlation between the prevalence of alcohol-related tweets and alcohol-related outcomes, controlling for confounding effects of age, sex, income, education, and self-reported race, as recorded by the 2013-2018 American Community Survey.

RESULTS

Significant associations were observed: between alcohol-hashtagged tweets and alcohol consumption (P=.01) and heavy drinking (P=.005) but not binge drinking (P=.37), self-reported at the metropolitan-micropolitan statistical area level; between alcohol-hashtagged tweets and self-reported excessive drinking behavior (P=.03) but not motor vehicle fatalities involving alcohol (P=.21); between alcohol-hashtagged tweets and the number of breweries (P<.001), wineries (P<.001), and beer, wine, and liquor stores (P<.001) but not drinking places (P=.23), per capita at the US county and county-equivalent level; and between alcohol-hashtagged tweets and all gallons of ethanol consumed (P<.001), as well as ethanol consumed from wine (P<.001) and liquor (P=.01) sources but not beer (P=.63), at the US state level.

CONCLUSIONS

Here, we present a novel natural language processing pipeline developed using Reddit's alcohol-related subreddits that identify highly specific alcohol-related Twitter hashtags. The prevalence of identified hashtags contains interpretable information about alcohol consumption at both coarse (eg, US state) and fine-grained (eg, metropolitan-micropolitan statistical area level and county) geographical designations. This approach can expand research and deep learning interventions on alcohol abuse and other behavioral health outcomes.

What on-line searches tell us about public interest and potential impact on behaviour in response to minimum unit pricing of alcohol in Scotland

AIMS

To investigate whether the introduction of minimum unit pricing (MUP) in Scotland on 1 May 2018 was reflected in changes in the likelihood of alcohol-related queries submitted to an internet search engine, and in particular whether there was any evidence of increased interest in purchasing of alcohol from outside Scotland.

DESIGN

Observational study in which individual queries to the internet Bing search engine for 2018 in Scotland and England were captured and analysed. Fluctuations over time in the likelihood of specific topic searches were examined. The patterns seen in Scotland were contrasted with those in England.

SETTING

Scotland and England.

PARTICIPANTS

People who used the Bing search engine during 2018.

MEASUREMENTS

Numbers of daily queries submitted to Bing in 2018 on eight alcohol-related topics expressed as a proportion of queries on that day on any topic. These daily likelihoods were smoothed using a 14-day moving average for Scotland and England separately.

FINDINGS

There were substantial peaks in queries about MUP itself, cheap sources of alcohol and online alcohol outlets at the time of introduction of MUP in May 2018 in Scotland, but not England. These were relatively short-lived. Queries related to intoxication and alcohol problems did not show a MUP peak, but were appreciably higher in Scotland than in England throughout 2018.

CONCLUSIONS

Analysis of internet search engine queries appears to show that a fraction of people in Scotland may have considered circumventing minimum unit pricing in 2018 by looking for on-line alcohol retailers. The overall higher levels of queries related to alcohol problems in Scotland compared with England mirrors the corresponding differences in alcohol consumption and harms between the countries.

Monitoring the sentiment of cannabis-related tweets in the lead up to New Zealand's cannabis referendum

INTRODUCTION AND AIMS

In October 2020, New Zealanders will vote on whether cannabis should be legalised for recreational use. With this in mind, the aim of the present study is to gauge the views and opinions of the New Zealand population on cannabis via tweets. To achieve this, we conducted a sentiment analysis of all historic cannabis-related tweets and referendum-specific tweets written in New Zealand.

DESIGN AND METHODS

We used a Twitter-sponsored commercial platform to access all historic cannabis-related tweets written in New Zealand and used search terms to remove non-cannabis-related terms. Next, we used the platform's machine learning function to code the sentiment of tweets (i.e. positive/pro-cannabis, negative/anti-cannabis or neutral).

RESULTS

Between July 2009 and August 2020, 304 760 cannabis-related tweets were written in New Zealand. Overall, the tweets were predominantly positive (62.0%) and there was a higher proportion of positive tweets written in 2020 (65.3%) compared to negative or neutral tweets. Similarly, for referendum-specific tweets, the 2020 data reveal a generally positive view of cannabis (53.5%).

DISCUSSION AND CONCLUSIONS

Both cannabis-related, and referendum-specific tweets, suggest that Twitter users in New Zealand have a generally positive view of cannabis. Given the nature of Twitter, the current method will allow us to study whether views toward cannabis change as the referendum nears and capture any late swings in pro- or anti-cannabis sentiment (abcd-lab.shinyapps.io/cannabis_sentiment/).

Cultural Differences in Tweeting about Drinking Across the US

Excessive alcohol use in the US contributes to over 88,000 deaths per year and costs over $250 billion annually. While previous studies have shown that excessive alcohol use can be detected from general patterns of social media engagement, we characterized how drinking-specific language varies across regions and cultures in the US. From a database of 38 billion public tweets, we selected those mentioning “drunk”, found the words and phrases distinctive of drinking posts, and then clustered these into topics and sets of semantically related words. We identified geolocated “drunk” tweets and correlated their language with the prevalence of self-reported excessive alcohol consumption (Behavioral Risk Factor Surveillance System; BRFSS). We then identified linguistic markers associated with excessive drinking in different regions and cultural communities as identified by the American Community Project. “Drunk” tweet frequency (of the 3.3 million geolocated “drunk” tweets) correlated with excessive alcohol consumption at both the county and state levels (r = 0.26 and 0.45, respectively, p < 0.01). Topic analyses revealed that excessive alcohol consumption was most correlated with references to drinking with friends (r = 0.20), family (r = 0.15), and driving under the influence (r = 0.14). Using the American Community Project classification, we found a number of cultural markers of drinking: religious communities had a high frequency of anti-drunk driving tweets, Hispanic centers discussed family members drinking, and college towns discussed sexual behavior. This study shows that Twitter can be used to explore the specific sociocultural contexts in which excessive alcohol use occurs within particular regions and communities. These findings can inform more targeted public health messaging and help to better understand cultural determinants of substance abuse.

Exploring the association between problem drinking and language use on Facebook in young adults

Recent literature suggests that variations in both formal and content aspects of texts shared on social media tend to reflect user-level differences in demographic, psychosocial, and behavioral characteristics. In the present study, we examined associations between language use on Facebook and problematic alcohol use. We collected texts shared on Facebook by a sample of 296 adult social media users (66.9% females; mean age = 28.44 years (SD = 7.38)). Texts were mined using the closed-vocabulary approach based on the Linguistic Inquiry Word Count (LIWC) semantic dictionary, and an open-vocabulary approach performed via Latent Dirichlet Allocation (LDA). Then, we examined associations between emerging textual features and alcohol-drinking scores as assessed using the AUDIT-C questionnaire. As a final aim, we employed the Random Forest machine-learning algorithm to determine and compare the predictive accuracy of closed- and open-vocabulary features over users' AUDIT-C scores. We found use of words about family, school, and positive feelings and emotions to be negatively associated with alcohol use and problematic drinking, while words suggesting interest in sport events, politics and economics, nightlife, and use of coarse language were more frequent among problematic drinkers. Results coming from LIWC and LDA analyses were quite similar, but LDA added information that could not be retrieved only with LIWC analysis. Furthermore, open-vocabulary features outperformed closed-vocabulary features in terms of predictive power over participants’ AUDIT-C scores (r = .46 vs. r = .28, respectively). Emerging relationships between text features and offline behaviors may have important implications for alcohol screening purposes in the online environment.

Recent Advances in Using Natural Language Processing to Address Public Health Research Questions Using Social Media and ConsumerGenerated Data

OBJECTIVE

We present a narrative review of recent work on the utilisation of Natural Language Processing (NLP) for the analysis of social media (including online health communities) specifically for public health applications.

METHODS

We conducted a literature review of NLP research that utilised social media or online consumer-generated text for public health applications, focussing on the years 2016 to 2018. Papers were identified in several ways, including PubMed searches and the inspection of recent conference proceedings from the Association of Computational Linguistics (ACL), the Conference on Human Factors in Computing Systems (CHI), and the International AAAI (Association for the Advancement of Artificial Intelligence) Conference on Web and Social Media (ICWSM). Popular data sources included Twitter, Reddit, various online health communities, and Facebook.

RESULTS

In the recent past, communicable diseases (e.g., influenza, dengue) have been the focus of much social media-based NLP health research. However, mental health and substance use and abuse (including the use of tobacco, alcohol, marijuana, and opioids) have been the subject of an increasing volume of research in the 2016 - 2018 period. Associated with this trend, the use of lexicon-based methods remains popular given the availability of psychologically validated lexical resources suitable for mental health and substance abuse research. Finally, we found that in the period under review "modern" machine learning methods (i.e. deep neural-network-based methods), while increasing in popularity, remain less widely used than "classical" machine learning methods.

"Can't Wait to Blackout Tonight": An Analysis of the Motives to Drink to Blackout Expressed on Twitter

BACKGROUND

Alcohol-related blackouts are associated with a range of negative consequences and are common among social drinkers. Discussing alcohol use on social networking platforms (e.g., Twitter) is common and related to higher alcohol consumption levels. Due to the widespread nature of alcohol-related social networking posts and alcohol-related blackouts, we examined the content of alcohol-related blackouts posts/"Tweets" on Twitter, with a focus on intentions to blackout and specific motivations for blacking out.

METHODS

A set of Tweets containing "blackout," "blackout," "blacking out," "blacked out," or "blacks out" were collected from April 26, 2018, and April 29, 2018. Using NVivo software, we coded all preblackout Tweets (i.e., before the blackout experience) for intentions and motives to blackout.

RESULTS

Most Tweets that we collected expressed an intention to blackout and these intentions ranged in strength (i.e., will blackout vs. might blackout). With respect to specific motives for blacking out, celebration motives were identified. For example, Tweets addressed blacking out to celebrate one's birthday, someone else's birthday, a school or work accomplishment, a sports win, during a vacation, or a holiday. Another endorsed motive for blacking out was loss or coping motives. For example, the Tweets commented on blacking out to deal with stress or a bad day.

CONCLUSION

Our findings suggest that Twitter users express intentions to blackout due to celebration or coping reasons. Given the consequences associated with blackout drinking, future research should consider the link between blackout intentions, blackout motives, and alcohol-related harm.

Comparison of Smartphone Ownership, Social Media Use, and Willingness to Use Digital Interventions Between Generation Z and Millennials in the Treatment of Substance Use: Cross-Sectional Questionnaire Study

Background

Problematic substance use in adolescence and emerging adulthood is a significant public health concern in the United States due to high recurrence of use rates and unmet treatment needs coupled with increased use. Consequently, there is a need for both improved service utilization and availability of recovery supports. Given the ubiquitous use of the internet and social media via smartphones, a viable option is to design digital treatments and recovery support services to include internet and social media platforms.

Objective

Although digital treatments delivered through social media and the internet are a possibility, it is unclear how interventions using these tools should be tailored for groups with problematic substance use. There is limited research comparing consumer trends of use of social media platforms, use of platform features, and vulnerability of exposure to drug cues online. The goal of this study was to compare digital platforms used among adolescents (Generation Zs, age 13-17) and emerging adults (Millennials, age 18-35) attending outpatient substance use treatment and to examine receptiveness toward these platforms in order to support substance use treatment and recovery.

Methods

Generation Zs and Millennials enrolled in outpatient substance use treatment (n=164) completed a survey examining social media use, digital intervention acceptability, frequency of substance exposure, and substance use experiences. Generation Zs (n=53) completed the survey in July 2018. Millennials (n=111) completed the survey in May 2016.

Results

Generation Zs had an average age of 15.66 (SD 1.18) years and primarily identified as male (50.9%). Millennials had an average age of 27.66 (SD 5.12) years and also primarily identified as male (75.7%). Most participants owned a social media account (Millennials: 82.0%, Generation Zs: 94.3%) and used it daily (Millennials: 67.6%, Generation Zs: 79.2%); however, Generation Zs were more likely to use Instagram and Snapchat, whereas Millennials were more likely to use Facebook. Further, Generation Zs were more likely to use the features within social media platforms (eg, instant messaging: Millennials: 55.0%, Generation Zs: 79.2%; watching videos: Millennials: 56.8%, Generation Zs: 81.1%). Many participants observed drug cues on social media (Millennials: 67.5%, Generation Zs: 71.7%). However, fewer observed recovery information on social media (Millennials: 30.6%, Generation Zs: 34.0%). Participants felt that social media (Millennials: 55.0%, Generation Zs: 49.1%), a mobile phone app (Millennials: 36.9%, Generation Zs: 45.3%), texting (Millennials: 28.8%, Generation Zs: 45.3%), or a website (Millennials: 39.6%, Generation Zs: 32.1%) would be useful in delivering recovery support.

Conclusions

Given the high rates of exposure to drug cues on social media, disseminating recovery support within a social media platform may be the ideal just-in-time intervention needed to decrease the rates of recurrent drug use. However, our results suggest that cross-platform solutions capable of transcending generational preferences are necessary and one-size-fits-all digital interventions should be avoided.

Identifying substance use risk based on deep neural networks and Instagram social media data

Social media may provide new insight into our understanding of substance use and addiction. In this study, we developed a deep-learning method to automatically classify individuals' risk for alcohol, tobacco, and drug use based on the content from their Instagram profiles. In total, 2287 active Instagram users participated in the study. Deep convolutional neural networks for images and long short-term memory (LSTM) for text were used to extract predictive features from these data for risk assessment. The evaluation of our approach on a held-out test set of 228 individuals showed that among the substances we evaluated, our method could estimate the risk of alcohol abuse with statistical significance. These results are the first to suggest that deep-learning approaches applied to social media data can be used to identify potential substance use risk behavior, such as alcohol use. Utilization of automated estimation techniques can provide new insights for the next generation of population-level risk assessment and intervention delivery.