An epidemic model says: on average, each infected person contacts 12 others per day. But averages hide structure. A schoolteacher contacts 30 children in a classroom. A remote worker contacts 2 household members. A nurse contacts 50 patients, each briefly. The same “12 contacts” produces radically different transmission chains depending on who contacts whom, where, and for how long.
A large-scale actor-based model of the Dutch population (arXiv:2603.06235) simulates individuals with age, residence, and movement patterns drawn from demographic and residential registry data. Contacts occur across three venues — households, workplaces, and schools — with stochastic interactions reflecting actual mobility patterns rather than average mixing rates.
The key finding: geographic and demographic profiles of initial cases determine transmission trajectories. A novel respiratory pathogen seeded in densely populated western municipalities spreads as an epidemic hub — fast, connected, hard to contain. The same pathogen seeded in a rural area may burn out before reaching the hub. Where the first case happens isn't just an initial condition. It's a structural determinant of the epidemic's character.
The insight for intervention: isolation and travel restrictions work not by reducing a uniform transmission rate but by severing specific edges in the contact graph. The model can distinguish between closing schools (removes classroom contacts but not household contacts) and restricting travel (removes inter-municipality links but not local ones). The same reduction in average contacts produces different epidemic outcomes depending on which contacts are removed. The average is not the action.