Up to 20% of infants and children waiting for a heart transplant die before ever receiving one. This sobering statistic has historically been attributed to a paucity of available donor hearts. But recent work by Michael McCulloch, MD, a pediatric transplant cardiologist at UVA Health Children’s, and his team of doctors and data scientists, challenges this assertion.
For the past 5 years, his team has studied the obstacles surrounding pediatric heart transplantation. Their work suggests that better data visualization and predictive analytics could optimize organ utilization and increase the number of lives saved through transplant.
Big Data & Pediatric Heart Transplant Outcomes
For transplant doctors, each pediatric donor heart offer comes with hundreds of donor, candidate, and offer specific variables. Plus thousands of individual data points to determine whether this offer is the right one for their specific candidate at that particular time. But with only 30 minutes to decide, not every data point can be weighted equally. And at the end of those 30 minutes, the only two answers available are “yes” and “no.”
Out of these data points, which matter most? Which are distractions? Lack of data organization made analysis challenging.
Working in a large academic center allows for cross-collaboration that transcends the medical field. For a true retrospective analysis, combining medical expertise and data science would be needed. And through just a few emails, McCulloch found Michael Porter, PhD, an associate professor in the schools of Data Science and Systems Engineering.
Porter’s work focuses on event prediction, pattern detection, and data linkage. Exactly the data analytics needed to improve pediatric donor heart utilization.
Through this partnership, McCulloch, Porter, and their team analyzed all pediatric donor offers made to pediatric patients awaiting heart transplantation from the last decade, with the intent of increasing donor organ utilization without negatively impacting post-transplant outcomes.
What Determines Whether a Heart Is Accepted?
Between 2010-2020, 5,625 pediatric heart donors resulted in over 30,000 different offers made to 4,905 pediatric patients waiting for a transplant. Surprisingly, nearly 90% of the offers were declined and 40% of these hearts were never used for pediatric waitlisted patients.
Further analyses focused on determining which variables impacted these decisions demonstrated even more unexpected results.
“The number one determiner of whether or not a heart will be accepted by an institution is how many other medical centers have declined it,” McCulloch shares. It’s a single data point pulled from literal mountains of data.
Doctors historically have been able to see how many centers have declined an organ before it’s offered to them. The more centers a heart is declined from, the higher probability it will be declined from the next center where it’s offered.
Objectively, doctors know that a heart being declined from one center doesn’t mean it’s not right for their patient. But this one data point can create doubt and lead to further such decisions.
“You start to wonder, what are these doctors seeing that I’m not,” McCulloch explains. “We know that high rejection leads to high waitlist mortality. There’s a cost to saying no,” he says.
UVA Health Children’s has a very high pediatric heart offer acceptance rate along with excellent short, mid, and long-term post-transplant outcomes. “High acceptance leads to better outcomes,” notes McCulloch.
What Should Determine Whether a Heart Is Used?
Out of all the data reviewed, two data points emerged as centrally important to a heart transplant’s success:
- An ultrasound showing the heart was functioning as needed
- The distance between the donated heart and the patient in need
And when it came to long-term patient outcomes, the amount of time spent waiting for a heart is also a factor.
It can feel easier to say “no” to anything other than an absolute perfect match. “But increased waiting isn’t good for the patient,” McCulloch adds.
The Mountain of Data vs a Dashboard
For most professions where large quantities of data need to be processed quickly, a dashboard is standard practice. Solid organ transplantation remains one of the few where cumbersome spreadsheets with a mix of relevant and irrelevant data points remains the standard practice.
To improve on this process, McCulloch and Porter teamed up with behavioral scientists, pediatric transplant cardiologists from around the country, and members of UNOS (United Network of Organ Sharing — the federally contracted agency responsible for overseeing all solid organ transplantation in the United States). They are simultaneously working to produce a graphic interface capable of displaying this mountain of data in a more individualized, user-friendly fashion, along with mathematical model outputs producing predictive analytics.
McCulloch said, “Our goal is to allow transplant clinicians to evaluate all of the donor, candidate, and offer specific data they want to see in a fashion that is intuitive and efficient for them. This could effectively increase confidence that you have truly evaluated the entirety of an offer. To further support their decision, we are simultaneously producing predictive models informing what the anticipated post-transplant survival would be for their candidate should they accept that particular offer; and if they do not, what the anticipated time to a next offer would be along with the likelihood their candidate would survive until that offer.”
Removing One Piece of Data
For all the greater visibility this plan would give to the relevant data, there’s only one piece of data that McCulloch would actually bury.
“It doesn’t matter how many centers have declined a heart,” he emphasizes. “For the most part, I think ultrasound assessment of the donor heart, the donor size, and distance away are the only three factors that genuinely matter. But everyone has their own way of assessing donor offers” he says, adding that there’s room for this kind of disagreement between experts.
“Being able to see how many other centers have previously declined the donor heart currently being offered to me creates biases that can be difficult to ignore. And there’s no benefit to seeing it.” It is for these reasons McCulloch and Porter have advocated for removing this particular data point.
Filling the Need
Although many have asserted there is a shortage of available pediatric donors, McCulloch believes it is actually a problem of utilization. “Considering the fact that nearly 2/3 of the pediatric donor hearts not utilized for pediatric waitlist patients were successfully transplanted into U.S. adults or Canadian children suggests the more immediately addressable issue is to accept the hearts we have in front of us.”
McCulloch is confident that with his team and a newly acquired grant from the Agency for Healthcare Research and Quality (AHRQ), they will be able to importantly impact the future of pediatric heart transplantation.