(Back row, from left) Michael Porter, PhD, Michael McCulloch, MD, and Jerome Dixon; (front row) Sara Riggs, PhD, and Firezer Haregu, MD
Michael McCulloch, MD
Michael McCulloch, MD, has been part of the Division of Pediatric Cardiology at UVA Health Children’s Hospital since 2017. He serves as a professor of pediatrics and the medical director of the Pediatric Pulmonary Hypertension and Mechanical Circulatory Support Programs. Additionally, he is a member of the Heart Failure and Heart Transplantation Team. Dr. McCulloch provides care for pediatric patients with a range of acquired and congenital heart conditions, as well as for children and young adults with Duchenne muscular dystrophy, in both inpatient and outpatient settings.
As a physician-scientist, Dr. McCulloch is actively involved in research regarding pulmonary hypertension, hypertrophic cardiomyopathy, Duchenne muscular dystrophy, and optimizing the pediatric heart transplant allocation process.
We recently sat down with Dr. McCulloch to learn more about his study titled, “Improving Pediatric Donor Heart Utilization with Predictive Analytics.” His goal for the study is to optimize pediatric heart donor utilization, donor and candidate matching, and improve waitlist/post-transplant survival.
Q. What do you hope to learn from this study? Will the results potentially change the standard of care for future patients?
McCulloch: Heart transplantation is the only option for children with end stage heart failure or inoperable congenital heart defects, but between 10-20% of those patients on the waitlist die before ever receiving one. Despite this fact, nearly 90% of all offers of pediatric donor hearts are refused and ultimately 40% of these hearts go unused for pediatric candidates.
Our project started over three years ago and has produced the country’s largest and most granular database, including all available data from every pediatric donor heart offer made to every pediatric patient waitlisted for a heart transplant since January 1, 2010. Our early work demonstrated that despite the mountains of data available to pediatric transplant clinicians determining whether a pediatric heart donor offer is appropriate at that time for that waitlisted patient (e.g. vital signs, laboratory data, cause of death), the one variable which seems to effect offer acceptance more than any other is the number of times that organ has already been refused. This novel finding is despite multiple publications demonstrating that prior organ refusal has no bearing on post-transplant outcomes.
Our goal is to optimize pediatric donor utilization, donor/candidate matching and improve waitlist/post-transplant survival by producing not only an improved user interface to update the woefully outdated spreadsheet format currently employed for all organ offers, but to also produce a predictive model utilizing all donor (e.g. size, cause of death, heart function parameters)/candidate (e.g. size, current clinical status, time on the waitlist)/offer (e.g. distance between donor and candidate) specific data available up to that time to help a clinician determine the expected post-transplant survival if they accept the offer and the expected time to next offer and likelihood of candidate waitlist survival if they do not accept that offer. By improving a clinician’s ability to confidently assess the aforementioned mountain of data in the 30 minutes available to make these decisions, we anticipate fundamentally changing how pediatric donor heart offers are utilized.
Q. What is novel about your study design/treatment/intervention?
McCulloch: To start, our team represents everything that is great about working within a university system. My co-PI Michael Porter, PhD, is a member of the University of Virginia’s schools of both Data Science and Systems Engineering, who I was able to find with just a few emails focused on finding a partner with experience handling such large amounts of data. Together, we have been able to capitalize on the incredibly bright and motivated student population available at UVA to get the project to where it is today.
We started the project by finding 24 students through UVA’s Internship Placement Program who helped dig through millions of documents to find and extract very specific data regarding donor heart function measurements in a countless number of different formats that precluded the use of document scanning programs. We have also been fortunate to secure the work of three different Data Science graduate students and two different teams of undergraduates working on capstone projects.
One of these projects was run by Sarah Riggs, PhD, from UVA’s School of Engineering, and helped us generate an individualizable end-user interface that was presented to and universally applauded by a large contingency of the UNOS administration; she will continue working with us throughout the remainder of the project. In addition, we have also been fortunate to recruit the work of Jerome Dixon who has his Masters of Science in both Information Technology Management and Decision Analytics and Firezer Haregu, MD, one of my pediatric heart transplant colleagues, to advance the project even further.
This has all been possible due to foundation and pilot grants through programs such as iTHRIV, Engineering in Medicine, and the Jefferson Trust, not to mention a lot of hard work from everyone involved. Ongoing success has allowed us to get the attention of and then partner with the research team from the United Network for Organ Sharing (UNOS), the agency contracted by the federal government’s Organ Procurement and Transplantation Network (OPTN), which oversees all solid organ transplantation across the United States. We are now partnered with the team from UNOS; a behavioral scientist from Carnegie Mellon University with a unique skill set in donor offer behavioral economics; and 37 pediatric transplant cardiologists from all 11 UNOS regions around the country, representing programs that perform over 80% of annual pediatric heart transplants. This team is tasked with achieving the four specific aims outlined in our recently-funded Agency for Healthcare Research and Quality (AHRQ) grant, which were to:
- Construct and validate predictive models to support transplant offer decision making;
- Develop a custom simulated offer environment within the UNOS platform;
- Develop a dashboard to better display the millions of datapoints from the hundreds of donor, candidate, and offer specific variables involved with each offer; and
- Evaluate the impact of both the predictive models and dashboard on actual decision- making practices by sending mock donor offers to the aforementioned transplant cardiologists from around the country.
We are incredibly honored to represent the first time any of these processes have been undertaken and to do so with this large of a scale.
Q. How will you ensure inclusion of a diverse subject population?
McCulloch: We are in the unique position to have data from the entire population of pediatric donors and pediatric heart transplant candidates/recipients in the United States since January 1, 2010, regardless of age, gender, race, socioeconomic status or region of the country.
Filed Under: Research