A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Age structure is crucial in epidemiology because contact patterns, susceptibility, transmissibility, and outcomes vary substantially by age. Age-structured models partition the population into age strata and allow differential transmission rates and transitions between strata, producing more realistic predictions than homogeneous-mixing models. Age-specific force of infection and next-generation matrices capture how transmission flows between age groups. These models are essential for evaluating childhood disease burden and vaccination strategy optimization.
Build and simulate an age-structured SIR or SEIR model for an infectious disease; compare predictions to a homogeneous-mixing model and real outbreak data.
Age structure matters mainly for childhood diseases (it affects transmission patterns for all infectious diseases). Random mixing between age groups is a reasonable assumption.
From your study of the SIR model, you know that a population can be divided into Susceptible, Infectious, and Recovered compartments, and that transmission depends on contact between susceptible and infectious individuals. The SIR model's implicit assumption of homogeneous mixing — that any person is equally likely to contact any other person — is a useful simplification for getting rough estimates, but it breaks down badly when age-specific differences in contact patterns, immune status, or disease severity are large. Age-structured models replace this single well-mixed population with multiple age strata, each with its own dynamics.
The key input that makes age structuring necessary is empirical: people do not mix randomly by age. Children contact mostly other children (classmates, siblings, playmates); working adults contact other adults in workplaces; elderly individuals have more limited contact networks. This assortative mixing — the tendency to contact others of similar age — was systematically documented in studies like POLYMOD (a large European contact survey), which showed that contact rates form a matrix with strong diagonal dominance. For diseases transmitted by close contact (respiratory infections, childhood illnesses), this means that an outbreak introduced into a school-age population spreads very differently than one introduced into a nursing home, even if R₀ is the same.
The mathematical machinery that captures this is the who-acquires-infection-from-whom (WAIFW) matrix, also called the contact matrix or transmission coefficient matrix β_{ij}. Each entry β_{ij} represents the per-capita rate at which an infectious person in age group j transmits to a susceptible in age group i. The structure of this matrix drives everything: the dominant eigenvalue of the next-generation matrix (derived from the WAIFW matrix combined with age-specific susceptibility and infectious periods) gives R₀ for the age-structured system, and the corresponding eigenvector describes the relative age distribution of cases at the start of an epidemic.
The practical payoff is vaccination strategy design. For measles, vaccinating children is sufficient because children are both the most-exposed and the most-connected group. But the SIR model with homogeneous mixing would give you the same recommendation for every disease with a similar R₀, which is wrong. Age-structured models revealed that for diseases like rubella — where clinical severity peaks in adult women and infection in early pregnancy causes congenital rubella syndrome — the optimal strategy depends on achieving herd immunity in women of childbearing age, which may require targeting adults and not just children. Similarly, models of influenza vaccination show that targeting school-age children (high contact rates, serve as bridges to other groups) can be more effective at reducing overall transmission than targeting the elderly (who face the highest mortality but have low contact rates). These insights are inaccessible to homogeneous-mixing models and represent one of the most direct contributions of mathematical epidemiology to public health policy.