Texas Research Center for Social Dynamics

Executive Summary

Background. Social contagion, the process by which behaviors, attitudes, and norms propagate through social ties, has been extensively documented in observational cohort studies. Obesity, tobacco use, happiness, and even loneliness have been shown to cluster within social networks, suggesting that health behaviors are not solely individual choices but are shaped by the actions of peers. However, causal evidence from prospective, randomized clinical trials that explicitly test whether an intervention can activate social contagion to improve health outcomes is extremely limited. Without such evidence, health systems and policymakers cannot confidently invest in network-based strategies.

Objective. The Network-ACTIVE trial was a parallel-group, cluster-randomized clinical trial designed to evaluate whether a peer-leader intervention embedded in naturally occurring social networks produces significantly greater increases in objectively measured physical activity compared to standard community health education. A second, equally important objective was to test for direct evidence of behavioral contagion: whether individuals who were not directly trained but were socially connected to a trained peer leader would increase their physical activity levels, independent of formal programming.

Methods. Twenty community clusters (faith-based congregations, neighborhood associations, and workplace wellness groups) in the greater Austin metropolitan area were randomized 1:1 to either a 12-month network-based Health Champion intervention (n=10 clusters, 604 participants) or a standard 12-week community health education program (n=10 clusters, 596 participants). In the intervention arm, sociometric surveys identified the top 15% of individuals nominated by peers as influential; these “Health Champions” received a two-day training in motivational interviewing, goal setting, and social support strategies and were tasked with promoting physical activity organically within their existing social circles. No formal classes were offered to the wider intervention clusters. Physical activity was measured objectively using waist-worn accelerometers at baseline and 12 months. The primary outcome was change in average weekly minutes of moderate-to-vigorous physical activity (MVPA). Social contagion was assessed by analyzing activity change among “secondary contacts”—participants who were not champions but had a direct social tie to one.

Results. Retention at 12 months was 89%. The network intervention group increased MVPA by 58 minutes/week (95% CI: 45 to 71), compared to a 13 minutes/week increase (95% CI: 2 to 24) in the control group, yielding a net between-group difference of 45 minutes/week (95% CI: 30 to 60, p<0.001). The proportion meeting national physical activity guidelines (≥150 min/week MVPA) rose from 22% to 47% in the intervention arm versus 24% to 31% in controls (odds ratio 2.1, p<0.001). Critically, secondary contacts in the intervention arm increased MVPA by 32 minutes/week (95% CI: 20 to 44) compared to matched controls without a trained peer, after adjustment for confounding. Further analyses demonstrated that this contagion effect persisted even among those who reported no participation in champion-organized group walks, suggesting that mere exposure to a peer’s changed behavior and the resulting normative shift was sufficient to alter personal activity levels. Qualitative data confirmed that Health Champions normalized physical activity conversations, making an active lifestyle a shared group value.

Conclusion. This trial provides the first rigorous causal evidence that intentionally activating social contagion within community networks can produce substantial, sustained increases in physical activity. The findings demonstrate a clear multiplier effect: the benefits of training a small number of influential individuals diffuse to their untrained social contacts. These results support the integration of network science methods into public health intervention design and policy.

1. Introduction

The role of social dynamics in shaping health behaviors has undergone a paradigm shift over the past two decades. Once viewed as a matter of individual motivation and knowledge, behaviors such as physical activity, dietary habits, and substance use are now recognized as being profoundly embedded in social networks. The concept of social contagion—whereby attitudes, norms, and behaviors spread through a population via interpersonal ties—offers a powerful explanatory framework. Seminal longitudinal studies by Christakis and Fowler demonstrated that obesity, smoking cessation, and happiness exhibit network clustering that cannot be explained solely by homophily or shared environment. For instance, an individual’s risk of becoming obese increased by 57% if a friend became obese over the same period, even after accounting for shared characteristics and geographic proximity. Similarly, smoking cessation occurred in clusters, with spouses, siblings, and friends mutually influencing one another’s quitting behavior.

These observational findings have been bolstered by experimental work on the spread of health behaviors online, where social influence on exercise and dietary choices has been repeatedly demonstrated. However, most of these experiments were conducted in artificial digital settings with weak ties and relatively short follow-up periods. What has been missing is a definitive clinical trial conducted in real-world communities, using objective outcome measures and rigorous randomization, that demonstrates that a network-based approach can outperform standard individual-focused interventions—and, critically, that the network effects are causal rather than merely associative.

Physical inactivity is a leading modifiable risk factor for noncommunicable diseases globally, responsible for an estimated 5.3 million premature deaths annually. Current public health guidelines recommend at least 150 minutes of moderate-to-vigorous physical activity per week, yet adherence rates in the United States hover around 25%. Traditional interventions such as supervised exercise classes, one-on-one counseling, and mass media campaigns have proven effective in the short term but often suffer from high attrition and fail to produce durable behavior change. Furthermore, these interventions rarely generate the secondary spread needed to reach population-level impact.

The Network-ACTIVE trial was designed to address these gaps. It conceptualizes the community not as a collection of independent individuals but as a structured web of relationships. By identifying and training a minority of highly connected and respected individuals—"Health Champions"—the intervention sought to seed physical activity behaviors and norms that would then propagate through existing social ties. This approach builds on threshold models of social contagion, which posit that behaviors become self-sustaining once a critical mass of adopters is reached, as well as on diffusion of innovations theory, which highlights the role of opinion leaders in accelerating adoption.

The primary objective of the trial was to determine whether the network-based intervention results in a greater increase in objectively measured MVPA at 12 months compared to a standard community health education program. The secondary objective was to provide direct, causally rigorous evidence for social contagion by measuring behavior change among individuals who were never directly exposed to the intervention but were connected to a trained peer.

2. Methods

2.1 Trial Design and Oversight

This was a parallel-group, cluster-randomized controlled superiority trial with two arms and a 12-month follow-up period. Clusters were randomized at the group level to avoid contamination. The study protocol, including the statistical analysis plan, was published prior to enrollment completion and registered at ClinicalTrials.gov (NCT-FICT-2847). The study was approved by the Institutional Review Board of the Texas Research Center for Social Dynamics and overseen by an independent Data and Safety Monitoring Board. All participants provided written informed consent.

2.2 Participants and Clusters

Twenty community-based groups were recruited from the greater Austin metropolitan area. Eligible groups had a membership roster of at least 50 adults, met at least twice monthly for social or organizational purposes, and had not recently participated in a formal physical activity program. A total of 1,200 individual participants were enrolled across the 20 clusters, with cluster sizes ranging from 52 to 78. Individual inclusion criteria were: age 25–65 years, insufficiently active at baseline (self-reported <150 min/week MVPA on the International Physical Activity Questionnaire, confirmed by accelerometry with <150 min/week MVPA during screening), willing to wear an accelerometer, and able to participate in moderate-intensity physical activity without medical contraindication.

2.3 Randomization and Blinding

Clusters were randomly assigned in a 1:1 ratio using a computer-generated sequence stratified by cluster size (small: 50–64 members; large: 65–78 members) and baseline mean cluster MVPA (below vs. above the median). Allocation was performed by a central statistician not involved in participant recruitment. Outcome assessors responsible for accelerometer data processing were blinded to group assignment. Participants and Health Champions could not be blinded due to the nature of the intervention.

2.4 Interventions

Network-Based Health Champion Intervention. In clusters assigned to the intervention arm, a three-step process was implemented. First, a sociometric survey was administered to all consenting cluster members. Participants were asked to nominate up to 10 individuals in the group whom they went to for advice, whom they considered a friend, and with whom they discussed personal matters. Using aggregated nomination data, the top 15% of individuals by indegree centrality (the number of nominations received) were identified as influential.

Second, these individuals were invited to a two-day Health Champion training workshop. The training covered: the principles of motivational interviewing, evidence-based physical activity guidelines, goal-setting techniques, how to model active behaviors, how to provide effective social support, and strategies for initiating informal walking groups or activity-based social gatherings. Champions were explicitly instructed not to deliver formal educational sessions to the group; instead, they were to integrate physical activity promotion naturally into their everyday interactions. They were encouraged to share their own activity goals, invite friends to join them for walks, and positively reinforce peers’ efforts.

Third, over the subsequent 12 months, champions received brief monthly telephone check-ins from a study coordinator for support and problem solving but were otherwise autonomous. The wider cluster members received no direct study intervention.

Standard Health Education Control. Clusters in the control arm received a 12-week community health education program consisting of weekly 60-minute group classes delivered by an external certified health educator. The curriculum covered physical activity guidelines, basic nutrition, stress management, and general wellness. This represents a typical, resource-feasible community intervention.

2.5 Outcomes

The primary outcome was change from baseline to 12 months in average weekly minutes of MVPA measured by ActiGraph GT3X+ accelerometers. Participants wore the monitor on the waist during waking hours for seven consecutive days at each assessment. Data were considered valid with at least four days of ≥10 hours of wear time. MVPA was defined using established cut points (≥1952 counts per minute).

Secondary outcomes included:

• Proportion meeting physical activity guidelines (≥150 min/week MVPA).

• Change in average daily step count.

• Social contagion metric: change in MVPA among “secondary contacts,” defined as intervention-arm participants who were not Health Champions but who had at least one direct, reciprocated friendship or advice tie to a Champion as identified in the baseline network survey. The comparator group was constructed from control-arm participants who lacked a tie to any trained peer, matched using propensity scores on baseline MVPA, age, sex, BMI, and cluster size.

• Self-reported physical activity enjoyment, social support for exercise, and descriptive norms (perception of how many peers are active), measured via validated questionnaires.

2.6 Statistical Analysis

The target sample size of 20 clusters with an average of 60 participants each provided >90% power to detect a 30-minute/week between-group difference in MVPA change, assuming an intracluster correlation coefficient (ICC) of 0.04 and a standard deviation of change of 70 minutes/week, with a two-sided alpha of 0.05.

All primary analyses were conducted according to the intention-to-treat principle. We used linear mixed-effects models with random intercepts for clusters to account for within-cluster correlation. The model included fixed effects for treatment arm, time point (baseline vs. 12 months), their interaction, and covariates for age, sex, BMI, and cluster size. The treatment effect was estimated as the difference in adjusted mean change from baseline between arms.

For the social contagion analysis, we implemented propensity score weighting to balance the secondary contact group with a comparable subset of control participants. The propensity model included baseline MVPA, age, sex, education, and health status. We then used a weighted linear mixed model with cluster as a random effect to estimate the contagion effect. Sensitivity analyses excluded secondary contacts who self-reported any participation in group walks or scheduled activities with a Champion to isolate “passive” contagion.

3. Results

3.1 Participant Flow and Baseline Characteristics

Between September 2024 and February 2025, 1,200 participants were enrolled and underwent baseline assessment. Randomization resulted in balanced groups: mean age was 44.3 years (SD 11.2), 62% were women, mean BMI was 29.4 kg/m² (SD 5.8), and mean baseline MVPA was 98 minutes/week (SD 42). Racial and ethnic composition was representative of the Austin area (55% non-Hispanic White, 30% Hispanic, 10% Black, 5% Asian/Other). The 20 clusters ranged from 52 to 78 members. A total of 112 Health Champions were identified and trained across the 10 intervention clusters. At 12 months, 1,068 participants (89%) completed follow-up accelerometry, with no differential attrition between arms (intervention 88.7%, control 89.3%).

3.2 Primary Outcome

The network intervention arm showed a mean increase in MVPA of 58 minutes/week (95% CI: 45 to 71) from baseline, while the control arm increased by 13 minutes/week (95% CI: 2 to 24). The adjusted between-group difference was 45 minutes/week (95% CI: 30 to 60, p<0.001). This represents a clinically meaningful improvement that, if sustained, is associated with reduced cardiovascular and metabolic disease risk.

3.3 Secondary Outcomes

The proportion meeting physical activity guidelines rose from 22% to 47% in the intervention arm and from 24% to 31% in the control arm (adjusted odds ratio 2.1, 95% CI: 1.5–2.9, p<0.001). Mean daily step counts increased by 1,250 steps (SD 980) in the intervention group versus 380 steps (SD 1,120) in the control group (p<0.001). Intervention participants also reported significantly greater increases in social support for exercise and in the perception that peers were physically active (descriptive norms), suggesting a shift in the social environment consistent with contagion mechanisms.

3.4 Evidence of Social Contagion

Among the 487 secondary contacts identified in the intervention arm, mean MVPA increased by 32 minutes/week (95% CI: 20 to 44) relative to the matched control group without a trained peer, after propensity score adjustment and accounting for clustering. Importantly, sensitivity analyses restricted to the 312 secondary contacts who explicitly reported no participation in any Champion-organized walks or group exercise events still showed a significant increase of 22 minutes/week (95% CI: 8 to 36, p=0.002). This strongly suggests that the effect was not merely due to formal group activities but reflects a genuine contagion process operating through normative influence, behavioral modeling, and casual social encouragement.

Further post hoc network analysis revealed that the magnitude of contagion increased with the strength of the tie: secondary contacts with reciprocated friendship ties to a Champion increased MVPA by 41 minutes/week, compared to 19 minutes/week among those with unidirectional advice ties only. This dose-response relationship aligns with theories emphasizing the importance of strong ties in transmitting complex behavioral norms.

3.5 Adverse Events

No serious adverse events were attributed to the intervention. Minor musculoskeletal discomfort was reported by 8% of intervention and 7% of control participants.

4. Discussion

The Network-ACTIVE trial provides rigorous, prospective evidence that a network-based intervention leveraging natural peer influence can produce meaningful, sustained increases in physical activity at the community level. The effect size—an additional 45 minutes of MVPA per week compared to standard health education—places this intervention among the most effective community-based physical activity programs tested to date. More importantly, the clear demonstration of behavioral spillover to untrained social contacts confirms the causal role of social contagion in health behavior change and underscores the potential for network interventions to achieve population-level reach beyond directly exposed individuals.

These results extend prior observational findings on social networks and health by providing experimental verification that social influence processes can be harnessed intentionally. The magnitude of the contagion effect, a 32-minute/week increase among secondary contacts, suggests that the intervention’s total population impact far exceeds what would be measured by focusing solely on trained individuals. When factoring in the secondary, and possibly tertiary, spread through the network, the cost-effectiveness of this approach becomes exceptionally compelling.

The mechanism underlying the contagion effect appears to involve several reinforcing pathways. Qualitative data indicated that Health Champions normalized physical activity within their social circles, making conversations about walking, step goals, and active commuting commonplace. As descriptive norms shifted—people began to perceive that “everyone is moving more”—the social cost of inactivity increased. Behavioral modeling, where individuals imitate the actions of respected peers, likely played a supplementary role. This combination of normative and informational social influence is consistent with Centola’s model of complex contagion, where behaviors that require social reinforcement spread more effectively through clustered, strong-tie networks.

The failure of the standard health education arm to produce durable change or any secondary diffusion highlights the limitations of individually focused, didactic interventions. Without embedding behavior change within the ongoing social fabric, improvements tend to decay once the structured program ends. In contrast, the network approach created a self-sustaining social dynamic that persisted long after the initial training workshop.

Limitations of this trial should be acknowledged. The study was conducted in a single metropolitan area with a predominantly middle-income, moderately diverse sample; generalizability to lower-resource or culturally distinct settings requires further investigation. Although we used objective physical activity measures, the network data relied on self-reported ties, which may contain recall biases. The 12-month follow-up, while longer than many lifestyle trials, leaves open the question of multi-year durability. Additionally, the impossibility of blinding participants introduces potential for expectancy bias, although the objective primary outcome partly mitigates this concern.

Further research should explore the application of this network-based approach to other health behaviors such as dietary change, medication adherence, and smoking cessation. Hybrid implementation-effectiveness trials in healthcare systems and underserved communities are needed to translate these findings into policy. Moreover, digital tools for efficient, scalable sociometric mapping could reduce the cost and complexity of identifying peer leaders.

5. Conclusion

Social dynamics are not an incidental backdrop to individual health choices; they are a substrate that can be strategically activated to amplify behavior change. The Network-ACTIVE trial demonstrates that by identifying and empowering influential peers within existing community networks, public health can unlock a social multiplier effect that extends far beyond the reach of conventional interventions. The Texas Research Center for Social Dynamics is committed to continuing this line of inquiry, working to integrate network science into the design of clinical and community systems to address the most pressing health challenges of our time.

6. References

1. Christakis, N.A., & Fowler, J.H. (2007). The spread of obesity in a large social network over 32 years. New England Journal of Medicine, 357(4), 370–379.

2. Rosenquist, J.N., Murabito, J., Fowler, J.H., & Christakis, N.A. (2010). The spread of alcohol consumption behavior in a large social network. Annals of Internal Medicine, 152(7), 426–433.

3. Centola, D. (2018). How Behavior Spreads: The Science of Complex Contagions. Princeton University Press.

4. Valente, T.W. (2012). Network interventions. Science, 337(6090), 49–53.

5. Lee, I.M., Shiroma, E.J., Lobelo, F., Puska, P., Blair, S.N., & Katzmarzyk, P.T. (2012). Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. The Lancet, 380(9838), 219–229.

6. Aral, S., & Nicolaides, C. (2017). Exercise contagion in a global social network. Nature Communications, 8, 14753.

7. Zhang, J., Centola, D. (2019). Social networks and health: new developments in diffusion and influence. Annual Review of Sociology, 45, 91–109.

8. Freedson, P.S., Melanson, E., & Sirard, J. (1998). Calibration of the Computer Science and Applications, Inc. accelerometer. Medicine & Science in Sports & Exercise, 30(5), 777–781.

9. Bandura, A. (2004). Health promotion by social cognitive means. Health Education & Behavior, 31(2), 143–164.

10. Latkin, C.A., & Knowlton, A.R. (2015). Social network assessments and interventions for health behavior change: a critical review. Behavioral Medicine, 41(3), 90–97.

www.trcsd.org | contact@trcsd.org

June 2026