How Do You Model an Epidemic?

The classic way to model a disease is deceptively simple. You divide a population into a few groups, called compartments, and track how people move between them over time. The best known version is the SIR model: Susceptible people who can still catch the disease, Infectious people who currently have it and can pass it on, and Recovered people who have had it and, for now, cannot catch it again.

Each day, some susceptible people become infectious, and some infectious people recover. Write down the rules for those flows and you have a working model. From a handful of equations you can grow a whole epidemic on a computer and ask what-if questions you could never run on a real city.

The reproduction number, written R0, is the average number of new people one infectious person passes the disease to, in a population where everyone is still susceptible. If R0 is above 1, each case leads to more than one further case and the outbreak grows. If it is below 1, the chain of infections fades out.

That single threshold at 1 is why R0 became famous. It also explains why interventions like distancing, masks, and vaccination all aim at the same target: pushing the effective reproduction number below 1. The goal is not magic; it is arithmetic.

When each case causes more than one further case, numbers do not add, they multiply. That is exponential growth, and human intuition is famously bad at it. A disease that doubles every three days goes from 100 cases to over 100,000 in about a month if nothing changes. Early on, the numbers look small and ignorable, right up until they are not.

This is the deep reason early action matters so much in an outbreak, and why waiting for the situation to 'look bad' can mean waiting too long. The most important decisions often have to be made while the visible numbers are still reassuringly low.

In the episode, Professor Juneja describes leading epidemiological modeling at TIFR during the pandemic, when his team's models were used to track and anticipate how COVID moved through Mumbai, one of the most densely populated cities on earth. Density, crowded housing, and shared services all change how a disease spreads, so an off-the-shelf model built for another city will not simply transfer.

Real models are richer than plain SIR. They add age groups, neighbourhoods, testing delays, and behaviour that shifts as people react to the news. Every addition makes the model more realistic and also more uncertain, because each new piece needs data and assumptions that could be wrong.

A good epidemic model can compare scenarios, estimate when hospitals might come under strain, and show which interventions do the most per unit of disruption. What it cannot do is predict the exact number of cases on a specific future date. The virus, the weather of human behaviour, and plain chance all refuse to be pinned down that precisely.

The honest use of such a model is to narrow the range of plausible futures and reveal which choices hold up across that whole range. Used that way, a model is one of the most powerful tools a decision-maker has. Mistaken for a prophecy, it will eventually embarrass everyone who leaned on it.