When Behavior Splits a Population: Two-Group SEIQR Model Reveals How Heterogeneity Reshapes Optimal COVID-19 Control

Mathematical models of infectious disease spread have long treated populations as relatively uniform collections of individuals, an assumption that simplifies the equations but often obscures the reality of how epidemics unfold. During the COVID-19 pandemic, it became strikingly clear that people do not behave the same way: some individuals scrupulously masked, distanced, and quarantined at the first sign of exposure, while others continued their routines with little regard for transmission risk. A new study published in BMC Infectious Diseases by Mo’tassem Al-Arydah of Khalifa University, Hailay Weldegiorgis Berhe of Mekelle University, and Abdessamad Tridane of United Arab Emirates University takes this behavioral heterogeneity seriously, building a two-group epidemiological model that separates the population into high-risk and low-risk categories and then asks a deceptively simple question: given that people behave differently, how should vaccination and awareness campaigns be deployed to control an outbreak most effectively?

The researchers constructed what epidemiologists call an SEIQR model, an acronym that tracks individuals through five compartments: Susceptible, Exposed, Infected, Quarantined, and Recovered. What distinguishes their framework from classical single-population models is the explicit division into two behavioral groups, each with its own transmission dynamics and its own quarantine rate. The model is written as a system of ordinary differential equations, with cross-transmission terms capturing how infection flows between the groups. A susceptible individual in the high-risk group can be infected by infectious individuals in either group, and the same holds for the low-risk group, creating a transmission matrix whose entries encode the full pattern of contact and behavioral differences across the population.

Before any control strategy could be evaluated, the team established the mathematical foundations of the model. They proved that solutions remain positive and bounded, meaning the equations never produce negative numbers of people or unbounded populations, and they demonstrated global existence of solutions, ensuring the model is well-posed over time. These are not merely technical formalities; they guarantee that the model’s outputs are biologically meaningful and that numerical simulations rest on solid theoretical ground. The researchers then derived the basic reproduction number, R0, using the next-generation matrix method, a standard but powerful technique that computes the expected number of secondary infections produced by a single infected individual introduced into a wholly susceptible population.

The stability analysis forms the theoretical heart of the paper. The team identified two possible long-term outcomes of the epidemic: a disease-free equilibrium, where infection vanishes from the population, and an endemic equilibrium, where the disease persists at a steady level. Using a symmetry condition on the transmission rates, they established the global stability of both equilibria. In practical terms, this means that when R0 falls below one, the disease dies out regardless of initial conditions, and when R0 exceeds one, the infection persists in the population no matter how the outbreak begins. Such global results are considerably stronger than local stability findings, which only guarantee behavior near the equilibrium points, and they give public health modelers confidence that the qualitative predictions hold across a wide range of epidemic scenarios.

To understand which parameters matter most, the researchers performed a global sensitivity analysis using the Latin Hypercube Sampling and Partial Rank Correlation Coefficient method, commonly abbreviated as LHS-PRCC. Unlike one-at-a-time sensitivity tests, this approach varies all parameters simultaneously across their plausible ranges, revealing how uncertainty in each parameter propagates to uncertainty in R0. The analysis produced a clear and somewhat counterintuitive result: the transmission coefficient within the low-risk group, denoted beta-11, emerged as the strongest driver increasing R0, while the quarantine rate in the low-risk group, q1, had the greatest impact in reducing it. In other words, in this model, transmission among people who behave cautiously, perhaps because they mix widely while maintaining partial precautions, exerts an outsized influence on overall epidemic potential, and improving quarantine compliance in that same group offers the biggest lever for suppression.

The second major contribution of the study is its optimal control framework. The researchers formulated the problem of deploying two time-dependent interventions, vaccination and awareness campaigns, across the two behavioral groups, and they sought the intervention schedule that minimizes infections while keeping costs manageable. Solving such problems analytically requires Pontryagin’s Maximum Principle, a cornerstone of optimal control theory that converts the minimization problem into a system of adjoint equations coupled to the original epidemic dynamics. The resulting optimality system characterizes the intervention intensities that best balance the epidemiological benefit of reducing infections against the economic cost of implementing vaccination drives and awareness programs at each moment of the outbreak.

Numerical simulations brought the theoretical framework to life and yielded strategically important insights. The simulations showed that awareness-based interventions are highly effective in the early stages of an epidemic, when informing the public about risks and protective behaviors can rapidly bend the transmission curve before large numbers of people have been infected. As the epidemic progresses, however, the combined strategy of vaccination and awareness achieved the best overall performance, reducing both the cumulative number of infections and the total cost of the intervention program. This finding aligns with the intuitive logic of outbreak response: behavioral messaging works quickly but requires reinforcement, while vaccination builds durable protection that compounds over time, and the two tools complement each other when deployed in coordination.

Perhaps the most striking result concerns the allocation of intervention effort between the two groups. Under the specific transmission matrix, epidemiological parameters, and objective-function weights considered in the study, the optimal control solution directed greater intervention effort toward the low-risk group. This outcome may seem paradoxical, since high-risk individuals are typically prioritized in real-world vaccination campaigns. The authors are careful to address this directly: the greater allocation to the low-risk group is a model-dependent finding, arising from the particular parameter values and cost structure they examined, and it should not be interpreted as general public-health guidance to prioritize low-risk populations over clinically high-risk populations. The result instead illustrates a deeper principle, that heterogeneity in transmission can substantially influence the optimal allocation of interventions between population groups, and that control strategies optimized for one behavioral structure may be suboptimal for another.

The broader significance of this work lies in its demonstration that behavioral heterogeneity is not a nuisance to be averaged away but a structural feature that reshapes epidemic dynamics and control policy. When a population splits into groups with different contact patterns, quarantine compliance, and responsiveness to public health messaging, the single-number R0 of a homogeneous model gives way to a transmission matrix, and the levers that most effectively suppress the outbreak may sit in unexpected places. Sensitivity analysis becomes an essential tool for locating those levers, and optimal control theory provides a principled way to distribute finite resources, vaccines, funding, and public attention, across the groups where they will do the most good.

For policymakers and modelers preparing for future epidemic threats, the study offers both a methodological template and a caution. The template is the combination of rigorous dynamical analysis, global sensitivity testing, and optimal control formulation, which together transform a descriptive model into a decision-support tool. The caution is that optimal allocations are contingent on the assumed parameters and objective weights, meaning that real-world deployment requires careful calibration to local epidemiological data and ethical constraints. As the authors emphasize, their finding on low-risk allocation is a property of their specific model configuration, not a prescription. What endures is the framework itself, which shows that acknowledging how differently people behave is not a complication of epidemic modeling but the very key to designing interventions that work.

Subject of Research: A two-group SEIQR epidemiological model of COVID-19 with heterogeneous behavior, analyzing stability, sensitivity, and optimal vaccination and awareness control strategies.

Article Title: Two-group SEIQR epidemiological model with heterogeneous behavior

Article References: Al-Arydah, M., Berhe, H. W., & Tridane, A. (2026). Two-group SEIQR epidemiological model with heterogeneous behavior. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14399-9

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14399-9

Keywords: COVID-19, SEIQR model, epidemiological modeling, behavioral heterogeneity, basic reproduction number, optimal control, vaccination, awareness campaigns, sensitivity analysis, Pontryagin's Maximum Principle, global stability, BMC Infectious Diseases