Professor Sanmay Das presented his work on allocating scarce resources in kidney transplant exchanges and homelessness resource allocation. At any given point in time, there are about 100,000 people in the United States waiting for a kidney transplant. However, only 20,000 kidney transplants are performed every year. This number is so low due, in part, to a lack of supply and logistical issues. Kidney transplants are possible from both living and deceased donors, however only 5,500 kidneys are transplanted from living donors a year. Through his work on predicting outcomes of transplants and optimization, professor Das has been working to increase this number.
Most live kidney transplants come from a matched compatible pair, where someone has a tissue match with someone in need and volunteers to donate a kidney. This is a straightforward system, if someone in need of a kidney matches with someone willing to donate, then the transplant is performed. So how could this process be improved if every one of these compatible pairs is already getting a kidney?
There’s another way that a live kidney transplant can occur, in a so-called kidney exchange program. Matching for a kidney is a complicated biological issue, dependent on blood and tissue types. Not everyone who wants to donate a kidney matches with the person they want to donate to (an incompatible pair), but perhaps they will match with someone else who needs a kidney. By matching two or more incompatible pairs who are compatible outside of their pairings, the kidney exchange program is able to facilitate more kidney transplants. At least that’s how it should work in theory. In practice, only 600 of the 5500 live kidney transplants are performed from an exchange program.
Professor Das’ work looks at trying to increase the number of live kidney transplants through a rather unintuitive method: having compatible pairs enter the kidney exchange pool. You might think that compatible pairs would have no reason to enter a kidney exchange program, they are already compatible to perform a transplant after all. The only reason a compatible pair would willingly enter an exchange would be if they could get a better kidney out of the exchange. By predicting the success rates of kidney transplants and forming an optimization problem, Professor Das has been able to increase both the number of transplants and success rates of transplants, almost doubling the number of transplants in simulations. In forming the optimization problem, the question of what exactly we want to optimize remains open. Should we maximize the number of kidney transplants? Should we maximize the quality of kidney transplants? Regardless of the remaining ethical questions, the promise of the technique remains, including compatible pairs in the kidney exchange pool has the ability to increase the number and quality of kidney transplants in large numbers.
In addition, Professor Das also presented his work on optimizing the use of scarce resources for homelessness prevention. His team was able to predict the impact of a household receiving a particular service and optimize the total success rate of the services. In a simulation, they were able to reduce the number of households needing services again by 27%. However, when working with scarce resources, giving assistance to a different household means the household the service was originally going to will no longer get the same support. Professor Das’ algorithm can’t help everyone, but it’s helping some more than it’s hurting others.
Professor Das’ work highlights the impact that prediction and optimization can have when put together. In his work on problems with scarce resources, Professor Das runs into many ethical questions. Optimization tells you how to use resources more efficiently, but it cannot tell you how to use AI for justice or equity. He therefore believes that tackling ethical questions is an increasingly important task for computer scientists.
Sanmay Das is an associate professor in Computer Science and Engineering and the chair of the steering committee of the newly formed Division of Computational and Data Sciences at Washington University in St. Louis. He has been recognized with awards for research and teaching, including an NSF CAREER Award and the Department Chair Award for Outstanding Teaching at Washington University.