Modelling post-lockdown scenarios in cities
The Government of India announced a nation-wide lockdown starting 25 March 2020, an extreme physical distancing measure aimed at reducing the spread of COVID-19 infections. With the number of cases rising, the lockdown period has now been extended to 3 May 2020.
How might the disease evolve once the restrictions are lifted? What would a post-lockdown scenario look like? To answer these questions, researchers at the Indian Institute of Science (IISc) and Tata Institute of Fundamental Research (TIFR) have carried out city-scale simulation experiments and modeled the outcomes of post-lockdown scenarios.
Their predictions, outlined in a working paper, could provide useful insights for public health officials and policymakers on decisions related to easing of restrictions. According to the researchers, unless we continue to aggressively trace and isolate cases, and prevent the influx of new infections, there is likely to be a second wave of infections and the public health threat will continue to persist.
The team simulated the spread of COVID-19 in Bengaluru and Mumbai using an agent-based model, which builds a fine-grained replica of a city, and mimics various interaction spaces such as households, schools, offices and community spaces. “If there are 10 million people in Bengaluru, the city’s model also has that many individuals,” says Rajesh Sundaresan, Professor at the Department of Electrical Communication Engineering, IISc, and the corresponding author of the working paper.
To build these ‘synthetic’ cities, a wide range of parameters were taken into account, such as population, age, household size distribution, number of employees, workplace sizes, commute distances and several others. Individuals were then assigned to the various interaction spaces and their interactions were mapped.
Within these cities, the researchers ‘seeded’ infections and tracked their spread. Some assumptions were built in, based on existing studies and data. The entire population was first assumed to be susceptible to the infection. “Once a person gets exposed, they are assumed as not infective for about 4.5 days, and then asymptomatic and infective on an average for half a day,” says Sandeep Juneja, Professor at the School of Technology and Computer Science, TIFR, who is also a coordinator of the project.
The researchers then tracked how the numbers, including fatalities and hospitalizations, would change under various scenarios where restrictions are phased out. One scenario is a return to normal activity on 4 May 2020, but with ‘case isolation’ continuing. Another considers an ‘odd-even’ strategy between 20 April 2020 and 3 May 2020, where half the workforce returns to their offices, along with other restrictions in place. Yet another scenario looks at a complete lockdown for an indefinite period. In all of these, the researchers assumed that cases would continue to be isolated with 90 percent compliance.
The model forecasts that in a city like Bengaluru, if the lockdown was lifted on 20 April 2020, and normal activity resumes, the number of direct COVID-19 fatalities might increase to levels under a no intervention scenario, but with some delay.
Similarly, the model also forecasts an increase in COVID-19 fatalities in both cities if the lockdown is lifted from 20 April 2020 to 3 May 2020 in the following phases: case isolation, plus home quarantine, plus social distancing of those 65 years and older, plus closure of schools and colleges, and thereafter normal activity resumes but with case isolation.
On the other hand, if the lockdown were to continue indefinitely, the number of direct COVID-19 fatalities will likely be much smaller than in the no intervention scenario, the model predicts.
“The agent-based model gives us enough handle to study a targeted intervention, say what if we just close schools and colleges alone. Or if we have only one half of the workplaces open,” says Sundaresan. “It also gives us a lot more flexibility to test out new interventions before we actually implement them.”
Based on the model, the team has developed an interactive online simulator, a tool that allows one to select different interventions and see how the number of infections, fatalities and hospitalizations might progress over time. The source code has also been made public for anyone to download and use.
These simulations can be extrapolated to any Indian city, says Juneja. “Now we have a clear idea as to what data is needed in each city. Once people in different cities gather this data, they can simply plug it in and see the output,” he says.
The researchers caution, however, that their study only looks at the public health outcomes of interventions and their relaxations, and does not consider economic or ethical issues. It does not also take into account the status of testing strategies or spontaneous changes in people’s behaviour. The case progression in hospitals is also based on available literature which is still evolving.
The researchers add another caveat about the report, which has been prepared to help researchers and public health officials understand the effectiveness of COVID-19 interventions: it should not be used for medical diagnostic, prognostic or treatment purposes or for guidance on personal travel plans.
IISc's activities were supported by the Centre for Networked Intelligence through a Cisco CSR grant.
Ranjini Raghunath is a Communications Officer at the Office of Communications, Indian Institute of Science (IISc), Bengaluru.