Human Behavior, Not Just Biology, Shapes Epidemic Outcomes

Epidemics unfold through biology and behavior together, and a new editorial in Frontiers in Epidemiology argues that the second half of that equation has been neglected for far too long. Romulus Breban of the Institut Pasteur in Paris and Raffaele Vardavas of Causal Paths Analytics in Los Angeles surveyed a research collection of twelve studies that model how human decisions—from vaccinating a child to imposing a lockdown—reshape the course of infectious disease outbreaks. Their assessment, published as an editorial introducing the collection, traces how the field of behavioral epidemiology grew over more than twenty-five years and why its influence on public health practice still lags behind its intellectual output.

The starting point is a simple observation with profound consequences. The trajectory of an epidemic is not determined solely by the biology of host-pathogen interactions; it depends heavily on the behavior of the host. Decisions to vaccinate, to wear a mask in public, to self-quarantine at home, and the political calculus of when and how to impose a lockdown are all interwoven into every aspect of an emerging epidemic. Yet for most of the twentieth century, mathematical epidemiology treated behavior with fixed parameters—a constant transmission rate here, a fixed vaccination coverage there—or simply assumed it was implicitly bundled into other quantities. The COVID-19 crisis exposed the limits of that simplification with unprecedented clarity.

Populations with broadly similar demographics and healthcare capacity sometimes experienced markedly different outcomes. Voluntary compliance with protective measures, vaccine acceptance, and the political environment contributed to those differences alongside age structure, socioeconomic disparities, testing capacity, prior immunity, viral variants, and access to healthcare. The non-pharmaceutical interventions implemented by public health authorities were themselves highly consequential behavioral interventions. According to the editorial authors, this experience shed new light on the fundamental role of human behavior in epidemic prevention and control.

The intellectual roots of the field reach back further than many modelers might assume. A landmark step, crystallized around the turn of the millennium, was the realization that private vaccine markets are unlikely to achieve socially optimal vaccination because of positive externalities. In a 1997 paper, Geoffard and Philipson showed that pricing vaccines for the free market within a game-theoretic framework is actually misaligned with social goals, because the benefit that accrues to non-vaccinators when their neighbors immunize systematically undercuts the individual incentive to purchase. The implication is not merely academic: state governments need to intervene to subsidize vaccine markets, and the extent to which they do is critical to epidemic outcomes.

Building on that foundation, researchers including Bauch, Earn, d’Onofrio, and Manfredi developed early game-theoretic models of vaccination decision-making, in which individuals weigh the risk of infection against the risk and cost of vaccination while updating their strategies in response to the overall epidemic. This line of work established a foundational insight known as the vaccinee’s dilemma: once coverage rises enough to prevent notable outbreaks, the perceived benefit of vaccination falls and coverage erodes, potentially allowing the epidemic to re-emerge. Another key modeling result showed that voluntary vaccination with a universal, long-lasting vaccine may not prevent severe influenza epidemics—a counterintuitive illustration of how behavioral feedbacks can subvert expectations. The literature then expanded rapidly, adopting evolutionary game theory, complex networks, and stochastic agent-based models, and accumulating findings on when voluntary vaccination achieves herd immunity, how information and perceived risk shape behavior, and how behavioral feedbacks can drive epidemic oscillations.

Even the most celebrated quantity in epidemiology carries a behavioral imprint. The basic reproduction number, R0, has sometimes been interpreted—including in COVID-19 press coverage—as a parameter determined entirely by host-pathogen biology. More precisely, the editorial notes, R0 is defined for a susceptible population under specified baseline contact and mixing patterns, before behavioral adaptation to an emerging epidemic. Its estimated value therefore reflects both biology and baseline community conditions, including contact structure, cultural norms, and routine hygiene practices. Once individuals become aware of an epidemic and respond through self-isolation, reduced mobility, mask use, or avoidance of crowded settings, transmission is better summarized by the effective reproduction number or its time-varying counterpart, Rt, which reflect changes in population immunity due to interventions or behavioral changes. Classical models provide the foundation for these quantities; behavioral models extend it by representing how adaptive behavior reshapes transmission over time.

Despite this productivity, the editorial is candid about the field’s limitations. Behavioral epidemic models have had limited impact on public health practice, and the authors examine the gap honestly. One structural reason is disciplinarity: much of the literature is written primarily for a technical readership, and the mathematical language of game theory, evolutionary dynamics, and agent-based models is unfamiliar to most epidemiologists trained in observational methods, let alone field practitioners. A second gap lies in the behavioral sciences themselves, where psychologists, sociologists, and behavioral economists publish in different journals and speak a different technical language from mathematical epidemiologists. A third tension concerns the models: one parsimonious enough to be interpretable is unlikely to capture all the heterogeneity and multiscale dynamics of a realistic epidemic, while one rich enough for numerical accuracy risks becoming a black box. Modern methods—global sensitivity analysis, uncertainty quantification, transparent calibration, and interpretable machine learning—can mitigate that risk, and the authors argue such models are often most valuable as tools for structured thinking and scenario comparison, with carefully calibrated versions potentially useful for forecasting within a clearly defined setting and prediction horizon.

The twelve contributions in the collection illustrate the range of approaches now in play. Magli and colleagues extend game-theoretic epidemic modeling from the individual to the institutional level, examining how strategic interactions among governments, political opposition, and economic lobbies affect the timing and effectiveness of epidemic response. Taking the COVID-19 first wave as their empirical reference—including stark excess mortality in parts of northern Italy—the model predicts that when political actors weigh electoral consequences against health protection, response implementation is systematically delayed, and that epidemic growth eventually forces even partially irresponsible actors toward protective measures only after significant preventable mortality has accumulated. The authors of the editorial call the case for including explicit ethical statements in pandemic preparedness plans compelling, while cautioning that conclusions drawn from a stylized strategic structure applied to one historical setting require validation in other political and epidemiological contexts.

Other studies push into cognitive science and labor economics. Vardavas and colleagues model healthcare workers as economic agents whose labor supply responds to pandemic conditions, using a dynamic equilibrium model of the nursing labor market to document how COVID-19 disrupted nurse supply through elevated workforce exit, reluctance to accept high-exposure assignments, and wage adjustments that failed to fully compensate for perceived risk. Mitsopoulos and colleagues propose an agent-based framework grounded in the ACT-R cognitive architecture and instance-based learning, applied to mask-wearing during COVID-19; when agents respond to local social cues, masking behavior clusters tightly (Pearson r = 0.76), whereas clustering nearly disappears when agents respond only to global information (r = 0.09). Walsh and colleagues apply a similar cognitive model to influenza vaccination decisions using a multi-year longitudinal survey, finding that personal vaccination history dominates, social network influences are secondary, and recent experiences carry the largest weight. Teng and colleagues introduce psychologically valid agents that predict mask-wearing and social distancing across the 50 US states during the first three COVID-19 waves, achieving a strong fit for daily mask-wearing (R² = 0.93) and a moderate fit for social distancing (R² = 0.62). Hotton and colleagues offer a methodological roadmap for incorporating structural social determinants of health—housing instability, incarceration, employment precarity—into HIV transmission models, while Gantayat and colleagues argue for multi-level approaches that engage structural and social dynamics alongside individual cognition in HIV and tuberculosis prevention. Pestow uses an integrated economic-epidemiological model calibrated to COVID-19 data to show that increased perceived disease risk reduces infection peaks and total incidence but has ambiguous effects on overall social well-being, arguing against state-orchestrated manipulation of public risk perception. Four empirical studies round out the collection: Williams and colleagues examine physical distancing and quality of life among immunocompromised adolescents; Fan and colleagues document urban-rural knowledge gaps about infectious disease prevention in China; Li and Tang quantify how government information release affects public protective behaviors; and Jung and Song test risk perception as a mediator between e-health literacy and prevention behaviors among young adults.

The editorial closes with priorities for a field it describes as in productive, if sometimes unruly, expansion. The authors call for tighter collaboration between mathematical modelers and behavioral scientists, standardized reporting of behavioral model assumptions, benchmark datasets for comparative evaluation, reproducible code and calibration workflows, and external validation before models are transferred across populations or used to inform policy. They urge modelers to design studies around the question of what decision a model informs and for whom, to communicate model scope honestly—these models are useful for structuring thinking, identifying mechanisms, and comparing policy scenarios, but are not in general reliable predictive instruments for near-term epidemic outcomes—and to preserve parsimony and interpretability as models grow in complexity. The central unresolved challenges, they conclude, are to measure adaptive behavior over time, identify causal mechanisms that transfer across settings, represent social and institutional feedbacks without sacrificing interpretability, and validate models against independent data. Addressing those priorities, the authors argue, will determine whether behavioral epidemiology can move from compelling explanation toward reliable and equitable decision support.

Subject of Research: Behavioral epidemiology and mathematical modeling of human behavior in infectious disease dynamics

Article Title: Editorial: Modeling the impact of human behavior on infectious disease epidemiology

Article References: Breban, R., & Vardavas, R. (2026). Editorial: Modeling the impact of human behavior on infectious disease epidemiology. Frontiers in Epidemiology, 6, Article 1931979. https://doi.org/10.3389/fepid.2026.1931979

Image Credits: AI Generated

DOI: 10.3389/fepid.2026.1931979

Keywords: behavioral epidemiology, mathematical modeling, infectious disease, COVID-19, vaccination decisions, game theory, agent-based models, reproduction number, public health, risk perception, health behavior, pandemic preparedness