People

Collaboration Stories

The purpose of this series is to tell the story of long-lasting, as well as emergent, stories of collaboration among clinicians and methodologists, content experts, and engineers, who have successfully worked on joint projects at the intersection of AI and health. Such collaborations are at the center of e-HAIL’s mission, and we offer these stories as a way of inspiring collaboration in other researchers.

Organ Allocation: From Specific Aims Sprint to Multi-Faceted Collaboration for Direct Clinical Impact

e-HAIL members Mariel Lavieri (College of Engineering), Ji Zhu (Literature, Science, and Arts), and Danielle Haakinson (Transplant Center Director for Michigan Medicine) are leading a project to create a machine learning (ML) model to allocate kidneys for transplants in a manner that improves matching and reduces wait times–ultimately saving lives and improving patients’ quality of life. 

The seeds for this project originated with a January 2023 e-HAIL Specific Aims sprint to brainstorm ideas for creating an AI system to optimize candidate selection of patients on the waitlist to receive a kidney transplant from deceased donors. At the time, the Transplant Center at U-M Hospital had begun actively working on ways to improve matching of donor organs to the most appropriate and needy recipients. Its new team of organ procurement coordinators could operate an AI tool that aids the transplant surgeons in timely identification of candidates who may be optimal for a given donor organ—if such a tool were developed. 

Chris Sonnenday, then-Director of Michigan Medicine’s Transplant Center, presented to 29 e-HAIL members about the main challenges in kidney transplants, including the severe time constraints for decision-making (as little as 30 minutes from time of offer). This places a heavy burden on the provider and introduces significant variation in decision making. A solution, Sonnenday stated, is one that would: improve efficiency and the ability to evaluate offers in a timely manner; make the decision-making process consistent and transparent (e.g. the decision currently varies based on time of day, workload, and surgeon); optimize available information; and overall improve matching of kidneys to patients, potentially increasing organ utilization and improving transplant outcomes.

The current approach to kidney allocation is time consuming and burdensome, explained Sonnenday at the start of the Specific Aims sprint. Medical staff have to review thousands of offers each year and choose the optimal matches. Although each kidney is given an allocation score to rate its matchability, staff have to further evaluate it based on medical comorbidities and other factors not included in the allocation score.

e-HAIL members worked in smaller groups, combining AI/ML experts and health experts including professionals from the Transplant Team, to generate specific aims that would both be interesting from a methodological perspective and also address the specific challenges involved in building a ML-based approach to kidney transplantation. The set of aims proposed by these working groups considered factors beyond simple algorithm generation to include stakeholder involvement, how to evaluate the current process compared to an AI-based process, and definitions of fairness for organ donation.

After the session ended, Lavieri expressed interest in continuing the research. To her, the problem represented unique challenges in using ML for the allocation of limited resources under time constraints, in line with what she has explored in other projects, including on liver transplants. Likewise, Zhu was interested in the project as it connected to his previous work on using statistical methods to help build predictive models, specifically in the health domain. For both Lavieri and Zhu, and their students, the project is a means to improve medical treatment while also strengthening foundational ML.

Over the next few months, e-HAIL facilitated the discussions between Laveri’s team and the Transplant Center, including with incoming Director Dr. Haakinson, and Zhu’s team joined in the course of the process. Through these meetings, they fleshed out the medical need and how artificial intelligence can provide a solution, eventually leading to the hospital securing funding so work could start on the proposed approach. 

The current project focuses on the allocation of marginal kidneys—kidneys that are in less-than-ideal condition, such as those from older donors or individuals with diabetes or renal impairment. Under the current process, such kidneys are often considered unusable and discarded, but recent efforts in the medical field are looking into how to make the best use of these kidneys, particularly in the face of the increasing number of people on waiting lists for kidney transplants. (See, for example, this article in the journal Current Opinion in Organ Transplantation).

e-HAIL has also provided the student funding for Yili Wang, a doctoral student studying under Lavieri, to work on the project. Wang completed her Master’s in biostatistics at U-M before moving to the College of Engineering’s Industrial and Operations Engineering Department for her PhD. She is fascinated by the challenge of working with the medical field, particularly because of the complex, imperfect, real-life systems that do not adhere to theoretical, textbook models. On the KTAlgorithm project, as the project is now called, she is working on modeling the current kidney allocation process using a Markov decision process so that, after they complete a new, ML-based model, they can compare it to the original process.

While the project has made significant progress since its official start in the fall of 2023, the current data set is posing challenges. In particular, the way that kidneys have been matched to potential recipients has changed over time, making it difficult to get consistent data to train a model. At times, there are certain variables missing from the entries. A significant task facing the team is cleaning up the data. After getting IRB approval in April 2024, the team is looking into observing current allocations at MM to better understand the process and get more detailed, cleaner data.

The project, when completed, will present a framework to improve clinicians’ ability to use marginal kidneys. This could potentially have an impact across the transplant field, as kidney transplantation is the most frequent and highest in volume. The waiting list for kidney transplants alone is over 97,000 patients, and U-M’s health system performed a record number of 42,000 transplants in 2022. By bringing together researchers from different fields to solve key medical problems, e-HAIL is bridging the gap between the medical and engineering fields to revolutionize how medicine works.