Skip to main content
Explore URMC

menu
URMC / Clinical & Translational Science Institute / Stories / March 2019 / UR CTSI-Supported Study Aims to Predict, Prevent Acute Kidney Injury

UR CTSI-Supported Study Aims to Predict, Prevent Acute Kidney Injury

Graphic of kidneysAcute kidney injury, a sudden decline in kidney function, occurs frequently among hospitalized patients with serious, long-lasting effects and even increased risk of death. It’s often preventable, but we currently lack the ability to reliably predict when it will happen and to whom. That is why researchers at the University of Rochester Clinical and Translational Science Institute (UR CTSI) analyzed data from over 34,000 patients to develop a risk score for acute kidney injury that could help doctors intervene and prevent it.

Part of the reason we can’t predict when a patient will develop acute kidney injury is that while some risk factors are known, we often don’t use them in a coordinated way.  For example, machine learning papers often focus on factors that increase risk of acute kidney injury, such as diabetes and medications, but not those that lower that risk. On top of that, most previous studies have looked at single hospitalizations for all patients, many of whom have not been previously hospitalized. By not looking at patients’ past data, those studies missed the opportunity to discover health factors or patterns that reliably precede acute kidney injury.

Samuel Weisenthal, an MD-PhD student at the University of Rochester Medical Center, and Martin Zand, M.D., Ph.D., co-director of UR CTSI, took a different tack, focusing on re-hospitalized patients. The pair and their colleagues analyzed electronic health record data from patients’ prior hospitalizations to identify factors that predict acute kidney injury. From those factors, they used machine learning to developed a risk score that could be calculated for patients at the time of re-hospitalization.

“Developing an accurate risk index for acute kidney injury in re-hospitalized patients could have a major impact on hospital care, particularly if it could allow preventive intervention or better tailored treatments from the time of hospital admission,” said Zand, who is also the senior associate dean for Clinical Research at URMC.

With early risk identification, a variety of preventive strategies can be implemented. For example, acute kidney injury caused by radiocontrast dye or chemotherapy can be prevented by administering fluids or altering a patient’s treatment plan. When these factors are adjusted accordingly, patients fare better and the cost and length of stay can be decreased.

And while such predictive systems require extensive validation before clinical deployment, this work is a step toward creating acute kidney injury predictions, specifically for re-hospitalized patients.

“This study will hopefully help move us in the direction of an automated, locally trained tool that leverages sometimes hidden, longitudinal electronic health record data to estimate acute kidney injury risk without manually ordering tests or collecting and entering data,” said Zand.

Read the full study in PLOS One.

+++

The work and authors detailed above were supported by the University of Rochester CTSA award numbers (UL1 TR002001), as well as UR CTSI’s Career Development Program (KL2 TR001999), Academic Research Track, Translational Biomedical Sciences PhD Program, and MD-PhD program funding (TL1 TR002000) from the National Center for Advancing Translational Sciences of the National Institutes of Health.

Susanne Pritchard Pallo | 3/14/2019

You may also like