Decoding disease dynamics

Simulating the spread of a health emergency with epidemic modelling to come up with a solution is easier with this tool, writes Seshadri K S

“When can I get rid of this suffocating mask and move around freely?” is a question that crossed all our minds during the Covid-19 pandemic. If you had asked a doctor or a researcher who studies diseases, they would have replied with a frustrating—“It depends”.

Life is inherently complex. Science is seen as a tool to grapple with complexity and predictions with progressively lower levels of uncertainty. India is a highly stratified society, and coping with unprecedented situations has been insurmountable, even for Science.

BharatSim is a tool that simulates scenarios to understand and manage disease spread. The simulation tool has been designed for India and includes data from various sources. Fundamentally, the tool has three parts: A synthetic population, a simulation engine, and a visualisation platform. The tool is open-source and can be modified for emergent scenarios of interest. 

Agent-based modelling

Imagine going back to the pandemic days in the virtual world. Each one of us can be a “digital agent”. We each have unique characteristics and behaviours and could be individuals in various cities. Depending on the context, we may interact with each other, which could change based on context. Agent-based modelling is a bottom-up approach to simulate scenarios where each agent can be assigned characteristics and interactions can be simulated under specific contexts.

The agent-based approaches are particularly useful in understanding how small-scale individual behaviours result in large-scale patterns and have been used for everything from grazing sheep to predicting election outcomes. Early insights into disease monitoring have strong foundations in theoretical biology.

The SIR model of disease dynamics has three crucial stages: the number of individuals who are Susceptible (S), Infected (I), and have Recovered (R). BharatSim builds on this to add multiple layers of complexity and simulates scenarios such as assigning a value to an agent for having gone to school or being vaccinated. 

Scenario building

BharatSim was developed by researchers at Ashoka University and ThoughtWorks, a leading engineering research company. The tool comprises a synthetic population used in simulating scenarios and validating them with real-world data.

“Among the key sources for this data was the Census of India, the NFHS surveys, the Gridded Population of the World, the India Human Development Survey, and the National Sample Survey,” said Prof. Gautam Menon, the senior author of the work. 

The researchers collated data from these sources and extrapolated information using machine learning techniques. The synthetic population mimics India’s population and socio-economic diversity and thus prides itself on having greater accuracy in predicting outcomes than other models.

Simulations often require high-performance computers to run. BharatSim, however, has been designed to run effectively on various systems, including personal laptops. Users can easily set up and run simulations, analyse results, and visualise data, making it accessible even for those with limited programming skills.

Are models useful?

Theoretical biologists often agree that all models can be wrong, but some are more useful than others. Models, at their best, are simplifications of reality as they cannot always account for long-term changes in behaviour or policy shifts.

They help in thinking about problems from first principles and simulating outcomes. The process is only complete when one goes out, collects real-world data, and tests them against the predictions. The authors use the population of Pune as a case and test the effect of lockdowns. 

They found that introducing a 15-day lockdown, when the infection rates were around 5 per cent of the population, resulted in a flattening curve, with nearly 60 per cent of the population contracting the disease. Next, they determined the effect of vaccination drives by holding infection rates at 1 per cent. Increased vaccination rates resulted in lower infection levels. 

Finally, they examined the spatial heterogeneity of the city and found that densely populated regions had a higher spread of infections. All these patterns were generally true of Pune and other cities. 

Empowering decision-makers

Decision-making in a time of crisis is incredibly challenging. There are simply too many unknowns. Tools such as BharatSim can build on the learnings during Covid-19 and prepare us for challenges should there be another pandemic. Although the tool is new, the researchers who developed it have been toying with it for over a decade, and Covid-19 was coincidental. 

“Having been trained as a physicist, I had experience with complex Monte Carlo simulations, but as I got to learn epidemiology, it seemed that several questions that needed addressing also had much to do with the nature of society, including physical, social, and economic networks, and an understanding of how people might make decisions” explained Prof. Menon, who straddles the fields of Physics and Biology at the Ashoka University.  

A group of 18 researchers were involved in developing the tool. The researchers are keen on taking the science to the policymakers and enabling them to make decisions while articulating the complex nature of the world’s problems. 

“What’s hardest to describe is how people, on their own, change behaviour in response to the inputs they receive, sometimes accurate, sometimes inaccurate. We can describe this in BharatSim at a truly granular level since our agents can make autonomous decisions,” explained Prof Menon.

The researchers will soon study how vector-borne diseases such as Dengue, where the vectors, specific species of mosquitoes, respond to seasonal changes in rainfall, temperature, and humidity.

(The author is an ecologist and faculty at ATREE)