Medical Rationing Using Reserve Systems
- Resources regarding software implementations that we developed to implement reserve systems and the explanation of the policy impact of this research can be reached at www.covid19reservesystem.org.
- After we wrote our paper addressing the rationing of scarce medical resources using reserve systems, we reached out to bioethicists and policymakers.
- Prof. Douglas White and his group at the University of Pittsburgh Medical System adopted with our help a randomized scheme that is based on an underlying simple endogenous reserve system to allocate the scarce antiviral drug Remdesivir to their COVID-19 patients.
- After our proposal is endorsed by the leading bioethicists in the country for medical rationing (see Persad, Peek, and Emanuel (2020), Galiatsatos et al. (2020), Schmidt (2020)), the National Academies of Sciences, Engineering, and Medicine’s (NASEM) COVID-19 vaccine allocation task-force by the testimony of bioethicist Prof. Harald Schmidt who proposed our framework to use equitable vaccine allocation targeting vulnerable communities recommended a 10% over-and-above national reserve for vulnerable communities.
- We wrote the following policy-oriented papers to simulate the dynamic rollout of vaccines to vaccinate individuals in different vulnerability groups under different scenarios including reserve systems to help vulnerable populations. The first two of these papers were presented in our meeting targeted to vaccine-allocation policymakers on Dec 4&9, 2020:
- “Rationing Safe and Effective COVID-19 Vaccines: Allocating to States Proportionate to Population May Undermine Commitments to Mitigating Health Disparities” (with Harald Schmidt, Parag A. Pathak, Michelle Williams, Tayfun Sönmez, Lawrence Gostin) (Nov 2020) |Manuscript from SSRN
- “What Prioritizing Worse-Off Minority Groups for COVID-19 Vaccines Means Quantitatively: Practical, Legal and Ethical Implications” (with Harald Schmidt, Michelle Williams, Parag A. Pathak, Tayfun Sönmez, Lawrence Gostin) (Nov 2020) | Manuscript from SSRN
- “Do Black and Indigenous Communities Receive their Fair Share of Vaccines Under the 2018 CDC Guidelines?” (with Parag A. Pathak, Harald Schmidt, Adam Solomon, Edwin Song, Tayfun Sönmez) (Sep 2020) | Manuscript from NBER | from arXiv
Organization of Living-Donor Paired Kidney Exchanges
The Meta-Level Contribution
- My academic papers and policy interactions along with Alvin E. Roth and Tayfun Sönmez played a key role in the establishment of kidney exchange programs that use economic and optimization-based principles around the world. My contributions in this regard were personally recognized by the scientific background of the 2012 Nobel Memorial Prize in Economics and also by the laureate recognition of the 2014 Frank Edelman Prize in Applied Analytics.
- Here is the link to the presentation speech of Professor Torsten Persson of the Royal Swedish Academy of Sciences at the 2012 Nobel Memorial Prize Ceremony regarding the role of my research on kidney exchange in this award.
- Here are some of my lecture notes explaining the concept of kidney exchange.
- National Science Foundation web-published a story called “Kidney Exchange: A Life-Saving Application of Matching Theory about my research with Alvin E. Roth and Tayfun Sönmez on Kidney Exchange.
- The earlier achievements are surveyed in my survey paper with Tayfun Sonmez. More current developments are in this policy paper
Role in the Establishment of Kidney-Exchange Programs:
- The New England Kidney Exchange Program (NEPKE): Dr. Frank Delmonico, Susan Saidman, Alvin E. Roth, Tayfun Sönmez, and I launched the New England Program for Kidney Exchange (NEPKE) in 2004. This is the first program that uses optimization-based mechanisms to find kidney exchanges. NEPKE became the forerunner of the US National Kidney Exchange Program and dissolved itself in it in 2010. The administrator of NEPKE, Ruthanne Hanto, currently administers the national program.
- Alliance for Paired Donation (APD): Alvin E. Roth, Tayfun Sönmez, and I also helped the launching of the Alliance for Paired Kidney Donation, founded by Dr. Michael Rees through the funding of the University of Toledo and the University of Cincinnati in Ohio. APD is a cross-country kidney exchange registry.
- The US National Kidney Exchange Program: The US National Kidney Exchange Program uses similar principles with the APD exchange system. I served on the advisory board for the development of the program and co-authored the OPTN policy proposal for the principles of the national kidney paired donation program. Here is the US Congress Bill that clarified that paired kidney donations do not violate the National Organ Transplant Act and allowed for the establishment of the National Program. NEPKE dissolved itself in it to become a de-facto national program in the United States: Ruthanne Hanto, the director of NEPKE, became the director of the UNOS National Program,
Adopted Key Contributions
(in Reverse Chronological Order):
- Non-simultaneous non-directed altruistic donor chains: At APKD, we started to implement Never-Ending-Altruistic-Donor Chains (NEAD-Chains), an idea that we developed with Michael Rees and Jon Kopke. Here are the Boston Globe and CNN news stories of the longest NEAD chain until March 2009, which has been documented in our recent NEJM paper. Also, you can watch and read CBS evening news stories featuring Matt Jones, the non-directed altruistic donor of the chain, surgeon Mike Rees, and some of the patients and donors of the chain. Recently a 30-way non-simultaneous non-directed altruistic donor chain has been reported. The National Kidney Registry (not to be confused with the US National Program) is currently the leading organization that facilitates kidney exchanges and uses mostly NEAD chains.
- The NEAD-chain idea is based on the fact that chain transplants initiated by non-directed altruistic donors need not be done simultaneously. This idea was proposed in our AJT paper.
- Gains from larger exchanges: We showed in our AER paper “Efficient Kidney Exchange: Coincidence of Wants in Markets with Compatibility-Based Preferences” that using at-most 4-way exchanges, almost all gains from kidney exchange can be exploited. Based on this, we started to implement priority mechanisms using at most 4-way kidney exchanges in NEPKE and APKD ( see a related news story). The national program also uses 3-way exchanges.
- Optimization and software: We have also authored the optimization software used in NEPKE and APKD.
- Earlier optimization and two-way exchanges: In our JET paper “Pairwise Kidney Exchange“, besides our mechanism design approach, we propose using combinatorial optimization and graph theoretic techniques developed by Edmonds (1965) on organizing kidney exchanges. After we published `Pairwise Kidney Exchange’ as an NBER working paper in the summer of 2004, Johns Hopkins team published a paper in 2005 in the Journal of American Medical Association with simulations using the generalized version of Edmonds’ (1965) algorithm that we proposed in `Pairwise Kidney Exchange’. Consequently, in 2005, The Johns Hopkins University Transplant Center adopted a pairwise kidney exchange scheme based on Edmonds’ algorithm.
- Simultaneous non-directed altruistic and deceased donor chains: In our QJE paper “Kidney Exchange“, we propose the idea of a “w-chain exchange.” Non-directed altruistic donor chain exchanges are based on the same idea, and this second idea was developed by Johns Hopkins. Recently, Johns Hopkins University conducted the first 5-way non-directed donor chain exchange, in which a non-directed altruistic donor donates a kidney to the patient of the first pair, the donor of the first pair donates a kidney to the patient of the second pair, the donor of the second pair donates a kidney to the patient of the third pair, the donor of the third pair donates a kidney to the patient of the fourth pair, and finally the donor of the fourth pair donates a kidney to a waiting list patient without a donor.
Adoption of Children in State-wide Adoption Network of Pennsylvania
Onur Kesten at Carnegie Mellon (now at the University of Sydney) has initiated work on “adoption of kids” with his then-student Vince Slaugh (now at Cornell, Ph.D. in OM) along with his OM colleague Mustafa Akan and me in connection with State-wide Adoption Network (SWAN) of Pennsylvania to improve recommendation systems for suggesting families for adoptive children. Previous recommendation tools have been unsuccessful to be used by social workers distributed to different parts of the state, who are looking for best fits for the children up for adoption in the state network. They often by-passed recommendations made by the tools and used their own limited network for finding families. We have made some improvements to their system that aims to attain higher usage rates by social workers.
Here is the paper that explains those improvements (published in the Interfaces/INFORMS Journal on Applied Analytics):
- The Pennsylvania Adoption Exchange Improves Its Matching Process” (with Vincent W. Slaugh, Mustafa Akan, and Onur Kesten)
School Choice in Turkey
The Ministry of Education of Turkey planned to implement a centralized school allocation scheme for public high schools, starting from 2018, based on student-school addresses and other idiosyncratic priority determinants and student GPA, etc. for each local school and exam scores for national exam schools.
- I, together with Tayfun Sonmez and Umut Dur, have presented to the Minister in January 2018 how a school-choice scheme can be implemented in Turkey. Bahcesehir University, with the help of our colleague Can Erbil, organized and sponsored this meeting.
- The slides of this presentation are here (in Turkish).
- After our interactions in 2018, the Ministry’s Allocation Unit used the student-proposing deferred-acceptance (DA) algorithm.
- For 2018, priorities of students at schools were largely based on exam scores for national exam schools and street addresses, GPA, etc for local schools.
- Despite our objections, the ranking of the school in the student’s list was used as a tie-breaker ahead of some other factors. For example, a student who ranked a school at 2nd place in her list would get higher priority than a student who ranked this school at 3rd place in his list if higher-order priority criteria of these two students were the same.
- If the priorities were purely lexicographic based on how students ranked the schools and then other factors, DA turned into the old “Boston” algorithm, aka immediate acceptance (IA) algorithm, which is known to be inferior to the textbook DA based on its manipulability and unfairness. Although the problems with the 2018 Ministry algorithm were not as severe, still it was not strategy-proof.
- After our further interactions in 2019 along with ITU economist Sinan Ertemel and Umut Dur’s presentation to the new Ministry administration, according to their new rulebook, the new Ministry Administration started implementing a system where priorities are not a function of how the students ranked the schools.
- See the 2019 rulebook (in Turkish).