Conference Agenda

Background: Online Health Communities (OHCs) enable people with long-term conditions to exchange peer self-management experiential information, advice and support. Highly active “superusers” are essential in fostering community interaction and effective information exchange. This study examines the sentiment distribution and dynamics in posts from two UK respiratory OHCs, focusing on interactions between regular users and superusers.

Methods: Sentiment analysis was conducted with a fine-tuned BioBERT model on anonymized data from Asthma UK (AUK) and the British Lung Foundation (BLF). BioBERT was fine-tuned using the COVID-19 Twitter Dataset to categorize sentiment as positive, neutral, or negative. Superusers were defined as the top 1% most active users and via VoteRank (users with the greatest spreading ability). The sentiment of regular users’ and superusers’ aggregated posts was then calculated and analysed.

Results: The fine-tuned model achieved 96% accuracy. Posts were predominantly positive, with a trend toward increasing positivity over time. Superusers generally wrote shorter, more positive posts and superusers defined by posting activity or VoteRank largely overlapped, showing that users who posted the most were also spreaders. Threads initiated by superusers typically encouraged regular users to reply with positive sentiment. When replying to threads started by regular users with different sentiment, superusers tended to be significantly and consistently more positive than regular users.

Conclusions: Network and Sentiment Analysis highlighted the essential role of superusers in respiratory OHCs. They not only generate consistently positive posts but also stimulate similarly positive responses from regular users, thereby sustaining a supportive online environment.

In rapid-onset disasters, emergent networks become vital lifelines for individuals isolated from established relief sources. Actors improvise their collaboration, relying on altered roles and structures which are not pre-planned. While literature on disaster relief has emphasized emergent efforts (e.g., David, 2006; Wachtendorf, 2003), how they fill gaps within the broader relief network as well as the differential capacity of individuals to mobilize such support remains underexamined. This study examines the patterns of individuals’ tangible, emotional, and information support ties during post-hurricane long-term recovery. We focus specifically on how residents’ sociodemographic characteristics, along with broader community contexts, influenced the composition of social support ties involving established and emergent actors.

Survey data capturing long-term recovery experiences were collected in 2022 from five coastal counties in Texas affected by Hurricane Harvey in 2017. The analysis draws on data from 776 individuals regarding household, neighborhood, and community recovery. Respondents identified people, groups, institutions, or programs that offered support across eight different time points, ranging from three days post-landfall to 48 months later. In addition to established institutions including federal and local governments, national nonprofits, and emergency services (e.g., fire and police departments), respondents named a range of sources such as religious organizations, mutual aid groups, local schools, restaurants, and food pantries. Multi-level personal network analyses are used to examine the key predictors of support network composition and density. Further, Twitter data from the first month following Hurricane Harvey will be used to examine the emergent networks of improvised relief during the early recovery phase.

Food insecurity and pulmonary diseases, such as tuberculosis, COPD, and asthma, remain critical public health concerns in Indonesia. While food insecurity exacerbates respiratory disease risk, social networks may serve as a protective factor by influencing access to resources, health behaviors, and healthcare utilization. This study investigates how social network diversity moderates the relationship between food insecurity and pulmonary diseases, with a focus on gender differences. Utilizing data from 27,288 adults in the Indonesian Family Life Survey, we assessed food insecurity via the Food Consumption Score and identified pulmonary diseases through self-reported physician diagnoses. A composite social network diversity measure captured household size and active ties across six social relationships, including spouse, parents, children, siblings, neighbors, and groups with and without religious affiliation. Gender-stratified multivariable logistic regression models, with interaction terms between food insecurity and social network diversity, examined moderating effects while adjusting for sociodemographic and health-related covariates. Results indicate food insecurity significantly increases pulmonary disease risk. A significant interaction effect was observed between food insecurity and social network diversity among women, suggesting that higher social network diversity attenuates the adverse impact of food insecurity on pulmonary health. The moderating effect plateaus at moderate levels of social network diversity, while highly diverse networks show diminished benefits. Findings suggest that women benefit more from social support networks in mitigating the health risks associated with food insecurity. This study highlights the gendered role of social networks in health disparities, emphasizing the need for targeted, network-based interventions to improve health resilience in food-insecure populations.

Depression has historically been the most common psychiatric illness worldwide and a significant contributor to the global burden of disease. An ongoing and steady increase in its prevalence (approximately 60% in the last three decades; Liu, 2024) has now positioned this condition as the top cause of disability globally (Friedrich, 2017; GBD Mental Disorders Collaborators, 2022; WHO, 2017). Depression raises risk of a wide range of physical and psychological illnesses (Harerimana et al., 2022; Monroe & Harkness, 2022; Netsi et al., 2018; Lawrence et al., 2010; Scott et al., 2016), including cardiovascular disease and cancer (Harshfield et al., 2020; Rajan et al., 2020; Scott et. al., 2016), as well as suicide (Mann et al., 2005; Miller & Campo, 2021; Ribeiro, Huang, Fox, & Franklin, 2018). Prior work has indicated the association of depression and the composition and mental health status of face-to-face community ties (e.g., friends, household members) (Bearman & Moody, 2004; Fowler et al., 2008; Rosenquist et al., 2011; Perkins et al, 2016), but this releationship is less explores in non-WEIRD settings. Our study uses a novel dataset (Airoldi & Christakis, 2024) that incorporates a large sample of 27,274 adults living in 176 villages in rural Honduras to model the associations of signed (friends and adversaries) network features on symptoms of depression and also specifically postpartum depression, controlling for demographic and socioeconomic factors. Our study includes a subsample of prospectively observed postpartum parents and negative ties, an unusual feature of social network studies addressing depression. We estimate several models by gender and explore uncommon structural traits such as triadic network structures (e.g., the positive-negative balance of triads). We find that women in more intransitive friendships were more likely to be depressed (in keeping with past work), while the same association was not found for men or postpartum parents. We observed in both genders a higher prevalence of depression among individuals whose friends had adversarial relationships (“negative triads”), as compared with those whose friends were friends (“balanced triads”) or those whose friends had no relation (“incomplete triads”). Our findings reinforce the importance of social network structure and psychological health beyond dyadic associations, especially in high-risk settings (LMIC, rural villages).

Extensive sociological research illustrates the value of personal networks for individual outcomes, highlighting that discussions are crucial for managing personal issues. However, our understanding of what occurs within discussion networks remains limited. To address this gap, I developed a novel method that captures problem-specific networks and measures their overlap. Utilizing originally collected data on the personal networks of emerging adults, I assess discussion patterns and their links to depression. There are three principal findings. First, problem networks form around primary appraisals rather than content domains. Second, the extent of overlap among discussants varies according to primary appraisals. Challenges (potential for gain) consist of specialized and segmented discussants, whereas threats (potential for loss) and harms (actual harm) are found in overlapping networks. Third, larger social responses to problems—with more discussants—are correlated with less severe depression. These findings indicate that discussion practices that match rather than provide access to, resources are an important mechanism that links discussing problems with mental health. The benefits of personal networks for mental health in emerging adults may primarily operate through a matching mechanism driven by cognitive appraisals with specialized and segmented discussants. Theoretically, network structure arises from social interactions; consequently, an instrument based on these interactions is essential for evaluating structure. By overlooking social interactions, our capacity to characterize mechanisms is restricted, especially as digital technologies, namely social media sites, afford new ways for young adults to organize their personal networks.