Skip to main content
menu
URMC / Clinical & Translational Science Institute / Stories / February 2024 / Building Better Tools to Predict Kidney Injury in Kids

Building Better Tools to Predict Kidney Injury in Kids

Children who come into the ICU for any number of reasons may wind up with acute kidney injury, a dangerous condition in which kidneys can no longer filter waste from the blood. If doctors catch the warning signs early, there are several things they can do to prevent acute kidney injury in kids, but it can be difficult to predict which patients are at risk.

Adam Dziorny, MD, PhD
“I think that the KL2 has been a great opportunity especially to protect time for a junior investigator to develop a research program and to move it forward. I think the UR CTSI resources are really important for that,” Dziorny said.

In July 2022, Dziorny was awarded a KL2 Career Development Award from the UR CTSI. He ended the two-year award early to begin work on the new K23 Mentored Patient-Oriented Research Career Development Award in January 2024.

“Adam is a rising star in clinical informatics. This project has all the elements of high-impact clinical research that can be translated into practice and affect how we practice medicine,” said UR CTSI Co-Director Martin Zand, MD, PhD, who serves as Dziorny’s primary mentor on both the KL2 and new K23 awards. “Acute kidney injury is a serious problem in the ICU, increasing patient risk of other complications and decreasing survival. Predicting and preventing kidney injury in the ICU proactively would be a game changer.”

Expanding on work supported by the UR CTSI KL2 Career Development Program, Adam Dziorny, MD, PhD, assistant professor of Pediatric Critical Care and Biomedical Engineering, is building a tool that can identify children who are at risk for developing acute kidney injury and alert health care providers.

“We have patients who are admitted into the ICU who have a risk of acute kidney injury depending on things like their initial presentation, their vital signs, and their underlying diagnosis,” Dziorny said. “It's very difficult for a human to tease all that out, but that's a task that a computer could be quite good at. I can take all that data, integrate it, and then come up with a prediction score.”

With a new career development award from the National Institute of Diabetes and Digestive and Kidney Diseases, Dziorny will validate and refine an existing algorithm that was developed and tested at the Children’s Hospital of Los Angeles. To ensure the algorithm will accurately predict acute kidney injury in children beyond Los Angeles, he will test it on two large data sets that include data from children across the country, one of which he helped to develop during his fellowship training.

“One of the big challenges with machine learning and artificial intelligence methods is having enough data and enough variability in that data such that it comes from multiple centers and sites so you can actually apply it to different patients around the country and not just the population that you're focused on,” said Dziorny.

Once the algorithm is refined and trained on these large datasets, Dziorny will test the algorithm on real-time data extracted from the medical record 12 hours after a child is admitted to the UR Medicine Golisano Children’s Hospital. Two and a half days later, the algorithm’s predictions will be compared to what really happened and will be tweaked to improve accuracy.

For the final step in the process, Dziorny will design a method to alert health care providers to a patient’s predicted risk. This step sounds simple, but it has been the downfall of similar systems in the past.

“Oftentimes people will buy or build these algorithms and dump them into the medical record without thinking about what will work best for the users—the health care providers,” Dziorny said. “They often don't get used or are used improperly, so people get burned out and wind up just ignoring them.”

To avoid that pitfall, Dziorny will collaborate with Marc Lande, MD, MPH, chief of Pediatric Nephrology at URMC, to get input from pediatric nephrologists on what will work best for them: where these alerts will be most useful and how they will fit in their workflow. Naveen Muthu, MD, a cognitive informaticist at Children’s Healthcare of Atlanta and mentor on the grant, will also guide the global user-centered design process.

In the end, Dziorny hopes not only to create a tool that predicts and prevents acute kidney injury in pediatric ICU patients but also to create a process that can be used to build similar tools for other patient populations and health issues, like sepsis or weaning patients off certain medications or ventilators.

In addition to Muthu and Zand, who is also a dean’s professor of Medicine and senior associate dean of Clinical Research at URMC, the grant will also include co-mentors James McMahon, PhD, endowed chair for Innovation in Health Care at the UR School of Nursing and co-director of Dissemination and Implementation Science at UR CTSI, and Nelson Sanchez-Pinto, MD, MBI, associate professor of Pediatrics and Preventive Medicine at Northwestern University Feinberg School of Medicine.

###

Research reported in this article is supported by the National Institute of Diabetes and Digestive and Kidney Diseases under award number K23DK138299 and was previously supported by the University of Rochester CTSA award KL2 TR001999 from the National Center for Advancing Translational Sciences of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Susanne Pritchard Pallo | 2/28/2024

You may also like