A social networks-driven approach to understand the unique alcohol mixing patterns of tuberculosis patients: reporting methods and findings from a high TB-burden setting

The study addresses how alcohol-driven social mixing patterns among tuberculosis (TB) patients and their close social contacts influence TB transmission risk in a high-burden Indian setting. Alcohol use disorders contribute substantially to TB incidence in India and are associated with higher risk of active disease, reinfection, and poorer treatment outcomes. Prior work suggested that drinking groups and venues may amplify transmission outside households, but quantitative characterization of these social mixing patterns has been limited. This study aims to quantify the social networks of TB patients who regularly share alcohol, assess how venue-based mixing structures connect patients and contacts, and evaluate the implications for TB transmission and contact tracing strategies.

Background literature identifies alcohol use as a significant risk factor for TB incidence and adverse outcomes (WHO 2020; Lönnroth et al., 2008; Rehm et al., 2009). Prior epidemiologic and molecular studies indicate that community and extra-household contacts, including drinking groups and venues, can drive transmission (Classen et al., 1999; Diel et al., 2002). However, these approaches often lack granular relational data on social mixing patterns. Social network analysis has been proposed to quantify transmission-relevant structures and centralities in contact networks (Salathé and Jones, 2010), but quantitative network studies of alcohol-driven mixing among TB patients have been scarce, particularly in high-burden settings.

Design and setting: Exploratory cross-sectional personal social network (ego-centric) survey conducted in Chennai, India, within the catchment of 24 Designated Microscopy Centers (DMCs) of the National Tuberculosis Elimination Program (NTEP) in the northern part of the city, a high-TB burden area. Period: February 2018 to June 2019. Population: Newly diagnosed adult pulmonary drug-sensitive TB patients residing in the area for at least one year and consenting to provide extra-household social mixing information. Sampling and recruitment: Of 713 consecutively screened TB patients, 300 met eligibility criteria and were enrolled. Data collection: A validated social network questionnaire elicited first-degree social network contacts (name generator method) and social mixing venues during the 6 months prior to TB diagnosis (pre-diagnostic reference period). Definitions: Consistent social relation was defined as living/socializing/working together ≥3 days/week for 2–4 hours during the reference period. Patients with regular alcohol use (RAU) were those who shared alcohol ≥3 days/week with a first-degree contact in a neighborhood venue; contacts with RAU were first-degree contacts who shared alcohol with the patient on the same schedule. Venues were neighborhood places where RAU mixing occurred. Attributes collected included age, sex, occupation, TB diagnostic status (per NTEP), and relationship type (family, extended family, friendship, neighborhood, occupational). TB status of contacts was verified via documentation at health facilities. Network construction and analysis: Two undirected networks were created: (1) Patient-contact network (nodes: patients with RAU and their contacts with RAU; edges: alcohol-sharing ties) and (2) Patient-contact-venue network (nodes: patients with RAU, contacts with RAU, and venues; edges: regular alcohol sharing at a venue). Networks were analyzed in Gephi (v0.9.2) using ForceAtlas and Fruchterman-Reingold layouts. Metrics: degree centrality (immediate transmission potential), betweenness centrality (bridging/transmission potential), network density (observed ties over possible ties), and components (connected subgraphs). Summary metrics were compared between the two networks using mean differences and percent change. Statistical analysis: Descriptive statistics, standard errors, and exact 95% confidence intervals were computed in STATA 15.1. Ethics: Approved by ICMR-NIRT Institutional Ethics Committee (IEC No. 2017018). Written informed consent obtained.

- Among 300 enrolled TB patients, 52 (17%) were identified as patients with regular alcohol use (RAU), all male. These 52 patients regularly shared alcohol with 106 first-degree contacts (4% of 2544 total contacts). - Alcohol sharing took place at 16 overcrowded, poorly ventilated neighborhood venues. - Contacts with RAU had a higher proportion diagnosed with TB than contacts without RAU: 12.3% (13/106; 95% CI: 6.6–20.0; SE 3.1) vs. 3.5% (86/2438; 95% CI: 2.8–4.3; SE 0.3). - Relationship types among RAU ties were predominantly non-household: friends, neighbors, or occupational contacts (96.2%). By type among contacts with RAU: friendship 6/68 (5.6%), neighborhood 5/6 (4.7%), occupational 2/26 (1.8%) had TB; family and extended family RAU contacts showed 0% TB in this sample (small n). - Patient-contact network (patients + RAU contacts only): sparse and weakly connected with density ratio 0.009, 52 disconnected components, mean degree centrality 1.3, and low mean betweenness centrality 0.5, indicating limited transmission potential within isolated dyads/triads. - Patient-contact-venue network (adding venues as nodes): denser and more connected with density ratio 0.018, 10 components, mean degree centrality 3.1, and markedly higher mean betweenness centrality 154.2 (≈+198% vs. patient-contact network), indicating substantially greater bridging and potential exposure pathways when venues are considered. - A giant component linked four alcohol venues and connected 67.3% of patients with RAU, 38.4% of contacts with RAU who had TB, and 72.3% of contacts with RAU without TB, illustrating venue-driven co-location of high-risk mixing. - From a network perspective, the pooled TB transmission exposure for contacts with RAU increased roughly tenfold when venue-based connections were included.

The study demonstrates that TB patients and first-degree contacts who regularly share alcohol are embedded in social structures that elevate transmission risk, particularly through non-household ties (friends, neighbors, occupational). Contacts with RAU exhibited substantially higher TB prevalence than contacts without RAU, supporting prior evidence that community mixing is a key driver of transmission outside households. Network analysis revealed that considering venues transforms a fragmented set of patient–contact dyads into a more connected structure with higher centralities and a giant component, indicating that venues function as bridging nodes that can amplify exposure. These findings directly address the research question by quantifying how alcohol-driven social mixing and specific venues structure transmission-relevant connections. Programmatically, results suggest expanding TB contact tracing beyond households to include RAU contacts and prioritizing screening and preventive interventions at frequently shared alcohol venues to identify cases earlier and reduce transmission.

Using primary social network data and analysis, the study quantified alcohol-driven social mixing among TB patients in a high-burden Indian setting. Contacts who regularly shared alcohol with TB patients had a higher proportion of TB than contacts without such sharing. Incorporating alcohol venues into the network analysis identified them as critical nodes that increased connectivity and potential exposure, forming a giant component linking many patients and contacts. These insights support adapting TB contact tracing strategies to prioritize RAU contacts and targeted activities at alcohol-sharing venues. Future research could integrate molecular epidemiology with network data, evaluate intervention effectiveness at venues, and examine temporal dynamics of mixing and transmission.

The study did not assess TB patients’ social mixing patterns in non-alcohol venues; identified non-alcohol (casual) mixing was sparse and mostly occurred in open public spaces, limiting comparability with closed, crowded alcohol-sharing settings. Consequently, the analysis may underrepresent other mixing contexts relevant to transmission.