Using Machine Learning of Online Expression to Explain Recovery Trajectories: Content Analytic Approach to Studying a Substance Use Disorder Forum.

BACKGROUND: Smartphone-based apps are increasingly used to prevent relapse among those with substance use disorders (SUDs). These systems collect a wealth of data from participants, including the content of messages exchanged in peer-to-peer support forums. How individuals self-disclose and exchange social support in these forums may provide insight into their recovery course, but a manual review of a large corpus of text by human coders is inefficient. OBJECTIVE: The study sought to evaluate the feasibility of applying supervised machine learning (ML) to perform large-scale content analysis of an online peer-to-peer discussion forum. Machine-coded data were also used to understand how communication styles relate to writers' substance use and well-being outcomes. METHODS: Data were collected from a smartphone app that connects patients with SUDs to online peer support via a discussion forum. Overall, 268 adult patients with SUD diagnoses were recruited from 3 federally qualified health centers in the United States beginning in 2014. Two waves of survey data were collected to measure demographic characteristics and study outcomes: at baseline (before accessing the app) and after 6 months of using the app. Messages were downloaded from the peer-to-peer forum and subjected to manual content analysis. These data were used to train supervised ML algorithms using features extracted from the Linguistic Inquiry and Word Count (LIWC) system to automatically identify the types of expression relevant to peer-to-peer support. Regression analyses examined how each expression type was associated with recovery outcomes. RESULTS: Our manual content analysis identified 7 expression types relevant to the recovery process (emotional support, informational support, negative affect, change talk, insightful disclosure, gratitude, and universality disclosure). Over 6 months of app use, 86.2% (231/268) of participants posted on the app's support forum. Of these participants, 93.5% (216/231) posted at least 1 message in the content categories of interest, generating 10,503 messages. Supervised ML algorithms were trained on the hand-coded data, achieving F1-scores ranging from 0.57 to 0.85. Regression analyses revealed that a greater proportion of the messages giving emotional support to peers was related to reduced substance use. For self-disclosure, a greater proportion of the messages expressing universality was related to improved quality of life, whereas a greater proportion of the negative affect expressions was negatively related to quality of life and mood. CONCLUSIONS: This study highlights a method of natural language processing with potential to provide real-time insights into peer-to-peer communication dynamics. First, we found that our ML approach allowed for large-scale content coding while retaining moderate-to-high levels of accuracy. Second, individuals' expression styles were associated with recovery outcomes. The expression types of emotional support, universality disclosure, and negative affect were significantly related to recovery outcomes, and attending to these dynamics may be important for appropriate intervention.

Effect of an eHealth intervention on older adults' quality of life and health-related outcomes: a randomized clinical trial.

BACKGROUND: By 2030, the number of US adults age ≥65 will exceed 70 million. Their quality of life has been declared a national priority by the US government. OBJECTIVE: Assess effects of an eHealth intervention for older adults on quality of life, independence, and related outcomes. DESIGN: Multi-site, 2-arm (1:1), non-blinded randomized clinical trial. Recruitment November 2013 to May 2015; data collection through November 2016. SETTING: Three Wisconsin communities (urban, suburban, and rural). PARTICIPANTS: Purposive community-based sample, 390 adults age ≥65 with health challenges. EXCLUSIONS: long-term care, inability to get out of bed/chair unassisted. INTERVENTION: Access (vs. no access) to interactive website (ElderTree) designed to improve quality of life, social connection, and independence. MEASURES: Primary outcome: quality of life (PROMIS Global Health). Secondary: independence (Instrumental Activities of Daily Living); social support (MOS Social Support); depression (Patient Health Questionnaire-8); falls prevention (Falls Behavioral Scale). Moderation: healthcare use (Medical Services Utilization). Both groups completed all measures at baseline, 6, and 12 months. RESULTS: Three hundred ten participants (79%) completed the 12-month survey. There were no main effects of ElderTree over time. Moderation analyses indicated that among participants with high primary care use, ElderTree (vs. control) led to better trajectories for mental quality of life (OR=0.32, 95% CI 0.10-0.54, P=0.005), social support received (OR=0.17, 95% CI 0.05-0.29, P=0.007), social support provided (OR=0.29, 95% CI 0.13-0.45, P<0.001), and depression (OR= -0.20, 95% CI -0.39 to -0.01, P=0.034). Supplemental analyses suggested ElderTree may be more effective among people with multiple (vs. 0 or 1) chronic conditions. LIMITATIONS: Once randomized, participants were not blind to the condition; self-reports may be subject to memory bias. CONCLUSION: Interventions like ET may help improve quality of life and socio-emotional outcomes among older adults with more illness burden. Our next study focuses on this population. TRIAL REGISTRATION: ClinicalTrials.gov ; registration ID number: NCT02128789.

Intraindividual, Dyadic, and Network Communication in a Digital Health Intervention: Distinguishing Message Exposure from Message Production.

Communicating within digital health interventions involves a range of behaviors that may contribute to the management of chronic illnesses in different ways. This study examines whether communication within a smartphone-based application for addiction recovery produces distinct effects depending on 1) the "level" of communication, defined as intraindividual communication (e.g., journal entries to oneself); dyadic communication (e.g., private messaging to other individuals); or network communication (e.g., discussion forum posts to all group members), and 2) whether individuals produce or are exposed to messages. We operationalize these communication levels and behaviors based on system use logs as the number of clicks dedicated to each activity and assess how each category of system use relates to changes in group bonding and substance use after 6 months with the mobile intervention. Our findings show that (1) intraindividual exposure to one's own past posts marginally predicts decreased drug use; (2) dyadic production predicts greater perceived bonding; while dyadic exposure marginally predicts reduced drug use; (3) network production predicts decreased risky drinking. Implications for digital health interventions are discussed.

Patient-provider communication while using a clinical decision support tool: explaining satisfaction with shared decision making for mammography screening.

BACKGROUND: Clinical decision aids may support shared decision-making for screening mammography. To inform shared decision-making between patients and their providers, this study examines how patterns of using an EHR-integrated decision aid and accompanying verbal patient-provider communication predict decision-making satisfaction. METHODS: For 51 patient visits during which a mammography decision aid was used, linguistic characteristics of patient-provider verbal communication were extracted from transcribed audio recordings and system logs automatically captured uses of the decision aid. Surveys assessed patients' post-visit decisional satisfaction and its subcomponents. Linear mixed effects models assessed how patients' satisfaction with decision making was related to patterns of verbal communication and navigation of the decision aid. RESULTS: The results indicate that providers' use of quantitative language during the encounter was positively associated with patients' overall satisfaction, feeling informed, and values clarity. Patients' question-asking was negatively associated with overall satisfaction, values clarity, and certainty perception. Where system use data indicated the dyad had cycled through the decision-making process more than once ("looping" back through pages of the decision aid), patients reported improved satisfaction with shared decision making and all subcomponents. Overall satisfaction, perceived support, certainty, and perceived effectiveness of decision-making were lowest when a high number of navigating clicks occurred absent "looping." CONCLUSIONS: Linguistic features of patient-provider communication and system use data of a decision aid predict patients' satisfaction with shared decision making. Our findings have implications for the design of decision aid tools and clinician training to support more effective shared decision-making for screening mammography.

Characterizing computer-mediated communication, friendship, and social participation in adults with traumatic brain injury.

BACKGROUND: Adults with traumatic brain injury (TBI) report loss of friendship and reduced social participation after injury, but there is limited information regarding quantity of friends and methods of communication. Our objective was to characterize friendship networks, social participation, and methods of communication, including computer-mediated communication (CMC), used by adults with TBI compared to uninjured adults. METHODS: Participants were 25 adults with TBI and 26 uninjured healthy comparisons (HC) adults, who completed the Participation Assessment with Recombined Tools-Objective (PART-O) and the Social Network Questionnaire (SNQ). RESULTS: Adults with TBI had significantly fewer total friends and significantly lower levels of productivity and overall social participation. Face-to-face interaction was the preferred method of contact for both groups. Adults with TBI were significantly less likely to use texting as a primary method of communication than their uninjured peers, but used other methods of communication at similar rates. CONCLUSION: Our study supports prior findings of reduced friendships and reduced social participation after TBI and adds new information about similarities and differences in communication methods between adults with and without TBI.

What Do You Say Before You Relapse? How Language Use in a Peer-to-peer Online Discussion Forum Predicts Risky Drinking among Those in Recovery.

Increasingly, individuals with alcohol use disorder (AUD) seek and provide support for relapse prevention in text-based online environments such as discussion forums. This paper investigates whether language use within a peer-to-peer discussion forum can predict future relapse among individuals treated for AUD. A total of 104 AUD sufferers who had completed residential treatment participated in a mobile phone-based relapse-prevention program, where they communicated via an online forum over the course of a year. We extracted patterns of language use on the forum within the first four months on study using Linguistic Inquiry and Word Count (LIWC), a dictionary-based text analysis program. Participants reported their incidence of risky drinking via a survey at 4, 8, and 12 months. A logistic regression model was built to predict the likelihood that individuals would engage in risky drinking within a year based on their language use, while controlling for baseline characteristics and rates of utilizing the mobile system. Results show that all baseline characteristics and system use factors explained just 13% of the variance in relapse, whereas a small number of linguistic cues, including swearing and cognitive mechanism words, accounted for an additional 32% of the total 45% of variance in relapse explained by the model. Effective models for predicting relapse are needed. Messages exchanged on AUD forums could provide an unobtrusive and cost-effective window into the future health outcomes of AUD sufferers, and their psychological underpinnings. As online communication expands, models that leverage user-submitted text toward predicting relapse will be increasingly scalable and actionable.

Detecting Recovery Problems Just in Time: Application of Automated Linguistic Analysis and Supervised Machine Learning to an Online Substance Abuse Forum.

BACKGROUND: Online discussion forums allow those in addiction recovery to seek help through text-based messages, including when facing triggers to drink or use drugs. Trained staff (or "moderators") may participate within these forums to offer guidance and support when participants are struggling but must expend considerable effort to continually review new content. Demands on moderators limit the scalability of evidence-based digital health interventions. OBJECTIVE: Automated identification of recovery problems could allow moderators to engage in more timely and efficient ways with participants who are struggling. This paper aimed to investigate whether computational linguistics and supervised machine learning can be applied to successfully flag, in real time, those discussion forum messages that moderators find most concerning. METHODS: Training data came from a trial of a mobile phone-based health intervention for individuals in recovery from alcohol use disorder, with human coders labeling discussion forum messages according to whether or not authors mentioned problems in their recovery process. Linguistic features of these messages were extracted via several computational techniques: (1) a Bag-of-Words approach, (2) the dictionary-based Linguistic Inquiry and Word Count program, and (3) a hybrid approach combining the most important features from both Bag-of-Words and Linguistic Inquiry and Word Count. These features were applied within binary classifiers leveraging several methods of supervised machine learning: support vector machines, decision trees, and boosted decision trees. Classifiers were evaluated in data from a later deployment of the recovery support intervention. RESULTS: To distinguish recovery problem disclosures, the Bag-of-Words approach relied on domain-specific language, including words explicitly linked to substance use and mental health ("drink," "relapse," "depression," and so on), whereas the Linguistic Inquiry and Word Count approach relied on language characteristics such as tone, affect, insight, and presence of quantifiers and time references, as well as pronouns. A boosted decision tree classifier, utilizing features from both Bag-of-Words and Linguistic Inquiry and Word Count performed best in identifying problems disclosed within the discussion forum, achieving 88% sensitivity and 82% specificity in a separate cohort of patients in recovery. CONCLUSIONS: Differences in language use can distinguish messages disclosing recovery problems from other message types. Incorporating machine learning models based on language use allows real-time flagging of concerning content such that trained staff may engage more efficiently and focus their attention on time-sensitive issues.

Implementing a Mobile Health System to Integrate the Treatment of Addiction Into Primary Care: A Hybrid Implementation-Effectiveness Study.

BACKGROUND: Despite the near ubiquity of mobile phones, little research has been conducted on the implementation of mobile health (mHealth) apps to treat patients in primary care. Although primary care clinicians routinely treat chronic conditions such as asthma and diabetes, they rarely treat addiction, a common chronic condition. Instead, addiction is most often treated in the US health care system, if it is treated at all, in a separate behavioral health system. mHealth could help integrate addiction treatment in primary care. OBJECTIVE: The objective of this paper was to report the effects of implementing an mHealth system for addiction in primary care on both patients and clinicians. METHODS: In this implementation research trial, an evidence-based mHealth system named Seva was introduced sequentially over 36 months to a maximum of 100 patients with substance use disorders (SUDs) in each of three federally qualified health centers (FQHCs; primary care clinics that serve patients regardless of their ability to pay). This paper reports on patient and clinician outcomes organized according to the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework. RESULTS: The outcomes according to the RE-AIM framework are as follows: Reach-Seva reached 8.31% (268/3226) of appropriate patients. Reach was limited by our ability to pay for phones and data plans for a maximum of 100 patients per clinic. Effectiveness-Patients who were given Seva had significant improvements in their risky drinking days (44% reduction, (0.7-1.25)/1.25, P=.04), illicit drug-use days (34% reduction, (2.14-3.22)/3.22, P=.01), quality of life, human immunodeficiency virus screening rates, and number of hospitalizations. Through Seva, patients also provided peer support to one another in ways that are novel in primary care settings. Adoption-Patients sustained high levels of Seva use-between 53% and 60% of the patients at the 3 sites accessed Seva during the last week of the 12-month implementation period. Among clinicians, use of the technology was less robust than use by patients, with only a handful of clinicians using Seva in each clinic and behavioral health providers making most referrals to Seva in 2 of the 3 clinics. Implementation-At 2 sites, implementation plans were realized successfully; they were delayed in the third. Maintenance-Use of Seva dropped when grant funding stopped paying for the mobile phones and data plans. Two of the 3 clinics wanted to maintain the use of Seva, but they struggled to find funding to support this. CONCLUSIONS: Implementing an mHealth system can improve care among primary care patients with SUDs, and patients using the system can support one another in their recovery. Among clinicians, however, implementation requires figuring out how information from the mHealth system will be used and making mHealth data available in the electronic health (eHealth) record. In addition, paying for an mHealth system remains a challenge.

Effects of Televised Direct-to-Consumer Advertising for Varenicline on Prescription Dispensing in the United States, 2006-2009.

INTRODUCTION: Televised direct-to-consumer advertising (DTCA) for prescription drugs is controversial, especially for tobacco cessation products such as varenicline, given safety concerns that arose only after its market approval. We aim to quantify the extent to which DTCA influenced varenicline use. METHODS: We linked monthly DTCA television ratings with monthly prescription data from IMS Health's National Prescription Audit across top 75 media markets in 2006-2009. We used Poisson models with Generalized Estimating Equations to analyze effects of exposures to DTCA for both varenicline and nicotine replacement therapies on rate of dispensed varenicline prescriptions among smokers, controlling for population characteristics and varenicline-related events. RESULTS: Varenicline prescriptions increased dramatically following DTCA launch and declined sharply after safety risks were publicized and US Food and Drug Administration (FDA) issued an advisory. DTCA had significant impact on new prescription dispensing in the subsequent month: before the FDA advisory, one additional exposure to varenicline DTCA was associated with a 1.8% (rate ratio [RR] = 1.018 [1.015-1.021]) higher rate of new prescriptions; no effect was observed after the advisory (RR = 1.000 [0.997-1.003]). Prior to the advisory, cross-product effects of nicotine replacement therapy advertising on varenicline prescribing were negligible (RR = 1.002 [0.999-1.004]); after the advisory, effects were positive (RR = 1.015 [1.012-1.019]). CONCLUSIONS: DTCA for varenicline had a significant impact on varenicline prescribing when the drug's safety profile was not well characterized, supporting arguments to limit DTCA for newly approved products whose real-world safety is unclear. IMPLICATIONS: We examined the fluctuations in varenicline use in association with DTCA for varenicline and other tobacco cessation aids. To our knowledge this is the first study to quantify the effects of televised DTCA for varenicline and other tobacco cessation aids on varenicline prescription dispensing. We believe that understanding these relationships is critical for formulating effective public health policy and interventions.

Price-related promotions for tobacco products on Twitter.

OBJECTIVES: This cross-sectional study examined price-related promotions for tobacco products on Twitter. METHODS: Through the Twitter Firehose, we obtained access to all public tweets posted between 6 December 2012 and 20 June 2013 that contained a keyword suggesting a tobacco-related product or behaviour (eg, cigarette, vaping) in addition to a keyword suggesting a price promotion (eg, coupon, discount). From this data set of 155 249 tweets, we constructed a stratified sampling frame based on the price-related keywords and randomly sampled 5000 tweets (3.2%). Tweets were coded for product type and promotion type. Non-English tweets and tweets unrelated to a tobacco or cessation price promotion were excluded, leaving an analytic sample of 2847 tweets. RESULTS: The majority of tweets (97.0%) mentioned tobacco products while 3% mentioned tobacco cessation products. E-cigarettes were the most frequently mentioned product (90.1%), followed by cigarettes (5.4%). The most common type of price promotion mentioned across all products was a discount. About a third of all e-cigarette-related tweets included a discount code. Banned or restricted price promotions comprised about 3% of cigarette-related tweets. CONCLUSIONS: This study demonstrates that the vast majority of tweets offering price promotions focus on e-cigarettes. Future studies should examine the extent to which Twitter users, particularly youth, notice or engage with these price promotion tweets.

Effect of Direct-to-Consumer Advertising on Asthma Medication Sales and Healthcare Use.

RATIONALE: The United States is one of only two countries that permit direct-to-consumer advertising (DTCA) of prescription drugs, and many questions remain regarding its effects. OBJECTIVES: To quantify the association between asthma-related DTCA, pharmacy sales, and healthcare use. METHODS: This was an ecological study from 2005 through 2009 using linked data from Nielsen (DTCA television ratings), the IMS Health National Prescription Audit (pharmacy sales), and the MarketScan Commercial Claims data (healthcare use) for 75 designated market areas in the United States. We used multilevel Poisson regression to model the relationship between DTCA and rates of prescriptions and use within and across designated market areas. Main outcome measures include (1) volume of total, new, and refilled prescriptions for advertised products based on pharmacy sales; (2) prescription claims for asthma medications; and asthma-related (3) emergency department use, (4) hospitalizations, and (5) outpatient encounters among the commercially insured. MEASUREMENTS AND MAIN RESULTS: Four Food and Drug Administration-approved asthma medicines were advertised during the period examined: (1) fluticasone/salmeterol (Advair), (2) mometasone furoate (Asmanex), (3) montelukast (Singulair), and (4) budesonide/formoterol (Symbicort). After adjustment, each additional televised advertisement was associated with 2% (incident rate ratio, 1.02; 95% confidence interval, 1.01-1.03) higher pharmacy sales rate from 2005 through 2009, although this effect varied across the three consistently advertised therapies examined. Among the commercially insured, DTCA was positively and significantly associated with emergency room visits related to asthma (incident rate ratio, 1.02; 95% confidence interval, 1.01-1.04), but there was no relationship with hospitalizations or outpatient encounters. CONCLUSIONS: Among this population, DTCA was associated with higher prescription sales and asthma-related emergency department use.

100 Million Views of Electronic Cigarette YouTube Videos and Counting: Quantification, Content Evaluation, and Engagement Levels of Videos.

BACKGROUND: The video-sharing website, YouTube, has become an important avenue for product marketing, including tobacco products. It may also serve as an important medium for promoting electronic cigarettes, which have rapidly increased in popularity and are heavily marketed online. While a few studies have examined a limited subset of tobacco-related videos on YouTube, none has explored e-cigarette videos' overall presence on the platform. OBJECTIVE: To quantify e-cigarette-related videos on YouTube, assess their content, and characterize levels of engagement with those videos. Understanding promotion and discussion of e-cigarettes on YouTube may help clarify the platform's impact on consumer attitudes and behaviors and inform regulations. METHODS: Using an automated crawling procedure and keyword rules, e-cigarette-related videos posted on YouTube and their associated metadata were collected between July 1, 2012, and June 30, 2013. Metadata were analyzed to describe posting and viewing time trends, number of views, comments, and ratings. Metadata were content coded for mentions of health, safety, smoking cessation, promotional offers, Web addresses, product types, top-selling brands, or names of celebrity endorsers. RESULTS: As of June 30, 2013, approximately 28,000 videos related to e-cigarettes were captured. Videos were posted by approximately 10,000 unique YouTube accounts, viewed more than 100 million times, rated over 380,000 times, and commented on more than 280,000 times. More than 2200 new videos were being uploaded every month by June 2013. The top 1% of most-viewed videos accounted for 44% of total views. Text fields for the majority of videos mentioned websites (70.11%); many referenced health (13.63%), safety (10.12%), smoking cessation (9.22%), or top e-cigarette brands (33.39%). The number of e-cigarette-related YouTube videos was projected to exceed 65,000 by the end of 2014, with approximately 190 million views. CONCLUSIONS: YouTube is a major information-sharing platform for electronic cigarettes. YouTube appears to be used unevenly for promotional purposes by e-cigarette brands, and our analyses indicated a high level of user engagement with a small subset of content. There is evidence that YouTube videos promote e-cigarettes as cigarette smoking cessation tools. Presence and reach of e-cigarette videos on YouTube warrants attention from public health professionals and policymakers.

Televised obesity-prevention advertising across US media markets: exposure and content, 2010-2011.

OBJECTIVE: To examine levels of exposure and content characteristics for recent televised obesity-prevention campaigns sponsored by state and community health departments, federal agencies, non-profit organizations and television stations in the USA. DESIGN: Nielsen television ratings for obesity-prevention advertising were collected for the top seventy-five US media markets and were used to calculate household exposure levels for 2010 and 2011. Governmental advertisements were coded for content. SETTING: United States. RESULTS: Average household exposure to obesity-prevention campaigns was 2·6 advertisements per month. Exposure increased by 31 % between 2010 and 2011, largely driven by increases in federal advertisements. In 2011, the federal government accounted for 62 % of obesity-prevention exposure, non-profit organizations for 9 %, community departments for 8 %, state departments for 3 %, and television station-sponsored public-service announcements for 17 %. The greatest percentage increase between 2010 and 2011 was in community advertising, reflecting efforts funded by the Communities Putting Prevention to Work (CPPW) programme. Among thirty-four state and community campaigns, the majority advocated both healthy eating and physical activity (53 %). Campaigns typically had positive or neutral emotional valence (94 %). Obesity or overweight was mentioned in 47 % of campaigns, but only 9 % specifically advocated weight loss. CONCLUSIONS: Exposure to televised obesity-prevention advertising increased from 2010 to 2011 and was higher than previously found in 1999-2003, apart from in 2003 during the federal VERB campaign. Nevertheless, exposure remains low relative to advertising for unhealthy foods. New federal campaigns have increased exposure to obesity-prevention advertising nationally, while CPPW grants have increased exposure for targeted areas.

Earned media and public engagement with CDC's "Tips from Former Smokers" campaign: an analysis of online news and blog coverage.

BACKGROUND: In March 2012, the US Centers for Disease Control and Prevention (CDC) launched the first-ever paid national tobacco education campaign. At a cost of US $54 million, "Tips from Former Smokers" (Tips) ran for 3 months across multiple media, depicting the suffering experienced by smokers and their families in graphic detail. The potential impact and reach of the Tips campaign was not limited to that achieved through paid media placements. It was also potentially extended through "earned media", including news and blog coverage of the campaign. Such coverage can shape public understanding of and facilitate public engagement with key health issues. OBJECTIVE: To better understand the contribution of earned media to the public's engagement with health issues in the current news media environment, we examined the online "earned media" and public engagement generated by one national public health campaign. METHODS: We constructed a purposive sample of online media coverage of the CDC's 2012 Tips from Former Smokers television campaign, focusing on 14 influential and politically diverse US news outlets and policy-focused blogs. We identified relevant content by combining campaign and website-specific keywords for 4 months around the campaign release. Each story was coded for content, inclusion of multimedia, and measures of audience engagement. RESULTS: The search yielded 36 stories mentioning Tips, of which 27 were focused on the campaign. Story content between pieces was strikingly similar, with most stories highlighting the same points about the campaign's content, cost, and potential impact. We saw notable evidence of audience engagement; stories focused on Tips generated 9547 comments, 8891 Facebook "likes", 1027 tweets, and 505 story URL shares on Facebook. Audience engagement varied by story and site, as did the valence and relevance of associated audience comments. Comments were most oppositional on CNN and most supportive on Yahoo. Comment coding revealed approximately equal levels of opposition and support overall. We identified four common arguments among oppositional comments: government intrusion on personal behaviors, problematic allocation of governmental spending, questionable science, and challenges regarding campaign efficacy. Supportive comments tended to convey personal stories and emotions. CONCLUSIONS: The Tips campaign received limited coverage on either online news or blog sources, but the limited number of stories generated engagement among online audiences. In addition to the content and volume of blog and news coverage, audience comments and websites' mechanisms for sharing stories via social media are likely to determine the influence of online earned media. In order to facilitate meaningful evaluation of public health campaigns within the rapidly advancing media environment, there is a need for the public health community to build consensus regarding collection and assessment of engagement data.

A cross-sectional examination of marketing of electronic cigarettes on Twitter.

BACKGROUND: Rapid increases in marketing of e-cigarettes coincide with growth in e-cigarette use in recent years; however, little is known about how e-cigarettes are marketed on social media platforms. METHODS: Keywords were used to collect tweets related to e-cigarettes from the Twitter Firehose between 1 May 2012 and 30 June 2012. Tweets were coded for smoking cessation mentions, as well as health and safety mentions, and were classified as commercial or non-commercial ('organic') tweets using a combination of Naïve Bayes machine learning methods, keyword algorithms and human coding. Metadata associated with each tweet were used to examine the characteristics of accounts tweeting about e-cigarettes. RESULTS: 73,672 tweets related to e-cigarettes were captured in the study period, 90% of which were classified as commercial tweets. Accounts tweeting commercial e-cigarette content were associated with lower Klout scores, a measure of influence. Commercial tweeting was largely driven by a small group of highly active accounts, and 94% of commercial tweets included links to websites, many of which sell or promote e-cigarettes. Approximately 10% of commercial and organic tweets mentioned smoking cessation, and 34% of commercial tweets included mentions of prices or discounts for e-cigarettes. CONCLUSIONS: Twitter appears to be an important marketing platform for e-cigarettes. Tweets related to e-cigarettes were overwhelmingly commercial, and a substantial proportion mentioned smoking cessation. E-cigarette marketing on Twitter may have public health implications. Continued surveillance of e-cigarette marketing on social media platforms is needed.

Understanding the Role of Large Language Models in Personalizing and Scaffolding Strategies to Combat Academic Procrastination.

Traditional interventions for academic procrastination often fail to capture the nuanced, individual-specific factors that underlie them. Large language models (LLMs) hold immense potential for addressing this gap by permitting open-ended inputs, including the ability to customize interventions to individuals' unique needs. However, user expectations and potential limitations of LLMs in this context remain underexplored. To address this, we conducted interviews and focus group discussions with 15 university students and 6 experts, during which a technology probe for generating personalized advice for managing procrastination was presented. Our results highlight the necessity for LLMs to provide structured, deadline-oriented steps and enhanced user support mechanisms. Additionally, our results surface the need for an adaptive approach to questioning based on factors like busyness. These findings offer crucial design implications for the development of LLM-based tools for managing procrastination while cautioning the use of LLMs for therapeutic guidance.

Using Adaptive Bandit Experiments to Increase and Investigate Engagement in Mental Health.

Digital mental health (DMH) interventions, such as text-message-based lessons and activities, offer immense potential for accessible mental health support. While these interventions can be effective, real-world experimental testing can further enhance their design and impact. Adaptive experimentation, utilizing algorithms like Thompson Sampling for (contextual) multi-armed bandit (MAB) problems, can lead to continuous improvement and personalization. However, it remains unclear when these algorithms can simultaneously increase user experience rewards and facilitate appropriate data collection for social-behavioral scientists to analyze with sufficient statistical confidence. Although a growing body of research addresses the practical and statistical aspects of MAB and other adaptive algorithms, further exploration is needed to assess their impact across diverse real-world contexts. This paper presents a software system developed over two years that allows text-messaging intervention components to be adapted using bandit and other algorithms while collecting data for side-by-side comparison with traditional uniform random non-adaptive experiments. We evaluate the system by deploying a text-message-based DMH intervention to 1100 users, recruited through a large mental health non-profit organization, and share the path forward for deploying this system at scale. This system not only enables applications in mental health but could also serve as a model testbed for adaptive experimentation algorithms in other domains.

Small Steps over time: A longitudinal usability test of an automated interactive text messaging intervention to support self-management of depression and anxiety symptoms.

BACKGROUND: Digital mental health interventions (DMHIs) offer potential solutions for addressing mental health care gaps, but often suffer from low engagement. Text messaging is one promising medium for increasing access and sustaining user engagement with DMHIs. This paper examines the Small Steps SMS program, an 8-week, automated, adaptive text message-based intervention for depression and anxiety. METHODS: We conducted an 8-week longitudinal usability test of the Small Steps SMS program, recruiting 20 participants who met criteria for major depressive disorder and/or generalized anxiety disorder. Participants used the automated intervention for 8 weeks and completed symptom severity and usability self-report surveys after 4 and 8 weeks of intervention use. Participants also completed individual interviews to provide feedback on the intervention. RESULTS: Participants responded to automated messages on 70 % of study days and with 85 % of participants sending responses to messages in the 8th week of use. Usability surpassed established cutoffs for software that is considered acceptable. Depression symptom severity decreased significantly over the usability test, but reductions in anxiety symptoms were not significant. Participants noted key areas for improvement including addressing message volume, aligning message scheduling to individuals' availability, and increasing the customizability of content. LIMITATIONS: This study does not contain a control group. CONCLUSIONS: An 8-week automated interactive text messaging intervention, Small Steps SMS, demonstrates promise with regard to being a feasible, usable, and engaging method to deliver daily mental health support to individuals with symptoms of anxiety and depression.

Late presentation of colorectal cancer in a vulnerable population.

OBJECTIVES: We examined colorectal cancer (CRC) stage at presentation and mortality in a vulnerable population compared with nationally representative data. METHODS: CRC cases were identified from San Francisco General Hospital (SFGH) and the Surveillance Epidemiology and End Results (SEER) database. RESULTS: Fifty-five percent of the SFGH cohort presented with advanced disease, compared with 44% of the SEER cohort. Increased risk of advanced stage at presentation at SFGH compared with SEER was most evident among blacks and Asians. There was weak evidence for worse survival at SFGH compared with SEER overall. This varied by race with poorer survival at SFGH among whites and possibly blacks but some evidence for better survival among Asians. Among CRC patients at SFGH, Asians and Hispanics had better survival than whites and blacks. At SFGH, 44% had a diagnosis of CRC within 1 year of establishing care there. Of those who had established care at SFGH for at least 1 year, only 22% had exposure to CRC screening tests. CONCLUSIONS: These findings allow examination of CRC presentation by ethnicity in vulnerable populations and identify areas where access and utilization of CRC screening can be improved.

Investigating the Role of Context in the Delivery of Text Messages for Supporting Psychological Wellbeing.

Without a nuanced understanding of users' perspectives and contexts, text messaging tools for supporting psychological wellbeing risk delivering interventions that are mismatched to users' dynamic needs. We investigated the contextual factors that influence young adults' day-to-day experiences when interacting with such tools. Through interviews and focus group discussions with 36 participants, we identified that people's daily schedules and affective states were dominant factors that shape their messaging preferences. We developed two messaging dialogues centered around these factors, which we deployed to 42 participants to test and extend our initial understanding of users' needs. Across both studies, participants provided diverse opinions of how they could be best supported by messages, particularly around when to engage users in more passive versus active ways. They also proposed ways of adjusting message length and content during periods of low mood. Our findings provide design implications and opportunities for context-aware mental health management systems.