Amazon Scholar contributes to best student paper award
Paper proposes a method to better and more equitably place COVID vaccine clinics to encourage more vaccinations.
Aravind Srinivasan, an Amazon Scholar and Distinguished University Professor of computer science at the University of Maryland, is an author on a publication that will be honored with a Best Student Paper award at the 2022 International Conference on Autonomous Agents and Multi-Agent Systems in May.
The paper, “Deploying vaccine distribution sites for improved accessibility and equity to support pandemic response”, looks at the optimal distribution of COVID vaccination sites and proposes an approach that distributes the clinics both efficiently and equitably. The paper’s authors include George Z. Li, a UMD student studying under Srinivasan, Ann Li, Madhav Marathe, Leonidas Tsepenekas, who is also one of Srinivasan’s students, and Anil Vullikanti.
Their paper focused on the relatively low rates of vaccination in some parts of the US, as well as the significant demographic disparities in vaccination rates, and examined how to best position vaccination sites to mitigate those trends.
The question of how to optimally distribute a finite set of items — in this case, vaccine clinics — goes to the heart of Srinivasan’s research into combinatorial optimization. However, while he noted the challenge of placing vaccination clinics bears similarities to classical optimization challenges, it also has some novel properties. The first regards the location of the clinics.
“Suppose you want to put 50 fire stations in a small state,” he said. “You look at where the population is located and you make sure everybody is within 10 miles of a fire station, things like that. But the important distinction is that when you model the population, the model is based on the population in their homes.”
That model doesn’t hold for vaccine clinics, which tend to be open during the day, because so many people are mobile during that period. So the team looked at mobility patterns and identified locations people are likely to travel to during the day, for example, a bank or school, and factored that into their model.
A second novel challenge relates to accessibility.
“The accessibility of vaccines was very different between various demographic groups in the US. And this has been well documented for various reasons,” Srinivasan said. “So we focused on developing facility location algorithms that are equitable and accessible for multiple demographics.”
Srinivasan said the team’s key insight was to make sure access was roughly the same across different demographic groups.
The authors note that their model, while specific to COVID, has other applications as well.
“Beyond healthcare, the placement of mobile distribution centers arises in disaster-management settings,” Srinivasan said. “For instance, shelters need to be set up for individuals evacuating during a hurricane or forest fire, who might need food and other basic survival kits. During such large events, mobile sites are also used to place security posts and information kiosks.
“This is work that we hope can have an immediate impact. By using computational tools capable of detecting discernible patterns in demographics, location, and mobility data, we hope to guarantee equal accessibility for anyone wanting to get a vaccination or booster or other vital services.”