What can zombies teach us about public health?

Why zombies? 

Real-world disease modelling requires evidence to support every assumption and parameter. That scrutiny is important when a model is used to inform decisions, but when the goal is to illustrate how modelling works rather than to predict a real outcome, the same scrutiny becomes an obstacle. 

As a fictional scenario, modelling a zombie outbreak removes the burden of evidence. The modelling team is free to set values in whatever way best illustrates the underlying principles, and the focus remains on the model’s logic rather than where the inputs came from. 

The fictional framing also addresses engagement. Epidemiology can feel remote, technical, or politically charged. A zombie outbreak does not. It invites curiosity rather than defensiveness and provides an accessible entry point into concepts that might otherwise feel out of reach. 

How the model works

The simulation tracks four groups of people: those who are uninfected (susceptible), those who have been bitten but have not yet turned into a zombie (infected), active zombies (zombie), and the dead. At each step, the numbers in each group shift according to rules built into the model, such as how quickly zombies infect people, how effectively people can fight back, and how populations move.

Rather than treating England as a single population in which anyone can infect anyone else, the model divides the country into approximately 500 regions, each running its own simulation. Infection must travel geographically, which means the outbreak spreads outward from its starting point at Heathrow rather than appearing everywhere at once.

Two behaviours make the model particularly interesting: chasing and fleeing. Zombies are drawn towards areas of higher population density, i.e. ‘chasing’. Simultaneously, people tend to move away from areas with high zombie concentrations, i.e. ‘fleeing’. The result of this chasing/fleeing dynamic is that susceptible and infected populations flee to safer areas, which in turn attracts zombies. This feedback loop significantly accelerates the spread, mirroring real patterns of population movement observed during epidemics.

The formal structure is an SIZD model, directly analogous to the SEIR frameworks used in real infectious disease modelling, in which compartment occupancies are governed by a system of nonlinear ordinary differential equations. Running across 500 regions simultaneously, each linked to its neighbours through a geographic adjacency matrix, the full system involves approximately 2,000 coupled equations solved at each time step.

The chasing dynamic was among the more striking findings during development. The zombie attraction mechanic in isolation produced a slow, relatively contained spread. Introducing the human flight response changed not just the rate but the character of the epidemic, movement away from high-concentration areas increased population density at destinations, drawing the infection behind it and creating a self-reinforcing acceleration. This class of emergent behaviour is well documented in real epidemic data but is difficult to capture in models that treat population movement as a fixed background assumption rather than a dynamic variable.

The decision framework 

Participants take on one of four stakeholder roles: Government, Military, Healthcare, or Politician, each with distinct objectives and a fixed starting budget of financial resource and political capital. At five points during the simulation, participants vote on how to respond to the outbreak, with each attempting to achieve their own objectives. The budget is fized and limited, Choices carry costs in money or political capital, and some decisions foreclose options later in the game.

When participants vote in line with their four different roles, the result is typically a compromise that serves none of them particularly well. By the time that becomes apparent, the window for effective intervention has often closed. Successful outcomes almost always reflect commitment to a single coherent strategy, pursued early enough in the simulation to take effect.

The win condition, fewer than two million zombies at day 100, is deliberately arbitrary and presented as such. It is a prompt for a more substantive question: what does success look like when managing a public health emergency? That question has no clean answer in real epidemic management either, and the conversation it generates is often the most valuable part of the exercise.

The decision tree generates approximately 200 theoretical pathways. Financial and political capital constraints reduce accessible routes to between 120 and 150, of which only around 10 to 12 produce a successful outcome. The stakeholder roles are designed to make strategic coherence structurally difficult: Government optimises for expenditure, Military for elimination rate, Healthcare for lives preserved, and Politicians for public approval. These objectives are not merely different in emphasis, they point in genuinely different directions, and aggregate voting rarely converges on the early commitment that success requires.

There are two viable routes to a successful outcome. The first involves early investment in vaccine development, sustained through a period of continued deterioration, with the intervention arriving in time to reverse the trajectory. The second involves aggressive early military action and acceptance of the associated civilian casualties. Both require an unambiguous strategic orientation from the first decision point, before the full costs of any given path are visible, a condition that multi-stakeholder voting is unlikely to produce, and one that reflects documented challenges in real-world health policy decision-making.

What you see 

As the simulation runs, the outbreak is visible as a heat map spreading across England, showing the geographic progression of infection in real time. Alongside it, a graph tracks the competing trajectories of the susceptible population, the infected, active zombies, and the dead. The effect of each decision point is visible almost immediately.

The simulation runs across 100 days. Success is assessed against a single criterion: fewer than two million active zombies. A result above that threshold is a loss, regardless of the trajectory at that point or how close the outcome came to the target.

The heat map and trajectory graph serve a specific function: they render the model’s equations in a form that requires no mathematical background to interpret. The geographic spread of infection, the acceleration of zombie population growth, and the effect of interventions on the trajectory all become legible through the visualisation rather than through the underlying system of differential equations.

Although the simulation presents as a live computation, every pathway through the decision tree is pre-computed before the session begins. When participants select an option, the model retrieves the corresponding pre-run result rather than recalculating in real time. The participant experience is identical, but the architecture is fundamentally different, and it is an architecture with direct application in client-facing model delivery. Lightweight browser and tablet environments cannot execute 2,000 coupled differential equations in the background, but they can serve pre-computed results instantly. Pre-computing the full solution space and filtering by user input allows analytically rigorous models to feel responsive and intuitive without requiring dedicated server-side infrastructure on the client side.

From fiction to practice

This zombie simulation is not a novelty. It is a demonstration of what modelling can do.

Every technical choice in its construction, the compartmental model structure, the geographical stratification, the stakeholder-weighted decision framework, the pre-computed delivery architecture, reflects an approach applied to real problems in health economics, market access, and public health strategy. The fictional setting removes barriers that can limit engagement. Once those barriers are removed, what remains is the logic of the model itself and that logic transfers directly to real-world decision-making.

For pharmaceutical and biotech organisations 

The same modelling principles that drive the zombie simulation underpin the epidemiological and health economic models used to support HTA submissions, market access strategy, and value communication. By running hundreds of interconnected population models rather than treating a country as a single homogeneous group, geographical stratification allows modelling teams to capture the heterogeneity that national-level analyses routinely miss.

The stakeholder decision framework illustrates something equally relevant in commercial settings: the same underlying evidence can lead to very different conclusions depending on the perspective from which it is evaluated. Government bodies, payers, clinicians, patients, and commercial teams are not arriving at different conclusions by accident. Each is optimising for different objectives. Designing communications and models that account for that, and speak to each stakeholder’s priorities, is a core component of an effective market access strategy.

The experience also demonstrates the value of interactive formats in environments where conventional presentations struggle to achieve meaningful engagement. In a crowded conference setting, or in a stakeholder workshop where participants arrive with established views, an activity that places people in the decision-maker’s seat can encourage more active participation than slides. The format is also more memorable, which matters when the goal is to shift understanding rather than simply deliver information.

For public health and health system organisations

The model architecture reflects approaches used in real-world pandemic planning and disease surveillance modelling. The spatial heterogeneity that makes the simulation visually compelling is the same property that allows real infectious disease models to capture the differential timing of outbreaks across regions, the role of urban density in amplifying transmission, and the effect of population movement on how quickly an infection escapes its initial geography.

The simulation also illustrates, in a format accessible to non-specialist audiences, why models matter and what they can and cannot tell us. Every output is conditional on its inputs: a model does not describe reality itself, but rather the consequences that would follow if reality behaved according to the model’s assumptions. That distinction is crucial and is frequently lost in public communication of modelling results.

For organisations working on public health communication, health literacy, or stakeholder engagement — with patients, policymakers, commissioners, or the general public — interactive, narrative-driven modelling of this kind offers a practical route to making complex evidence legible without requiring a background in epidemiology.

To find out more about how interactive modelling could support your work, contact us: enquiries@heor.co.uk.