Session
2023 session 3
Location
Weber State University
Start Date
5-8-2023 9:35 AM
Description
Acute kidney injury (AKI) is a common complication of cardiac surgery that is associated with increased hospital length of stay and mortality risk. The most common diagnostic criteria come from the Kidney Disease Improving Global Outcomes Guidelines. These criteria rely on detecting changes in serum creatinine relative to a baseline measurement and decreased urine output. These markers are meant to capture a decrease in the overall filtration rate. However, the filtration rate may or may not change depending on the baseline health of the individual patient and the type and severity of the injury. In addition, these markers are influenced by factors other than renal function or health, such as the administration of certain drugs or the volume status of the patient. Up to this point, research has primarily focused on the application of serum and urinary biomarkers that increase in response to renal injury or cell death. While these tests have advantages over current methods, the clinical utility of these tools is limited by high cost and a long wait time for results. Thus, there is a need for a noninvasive biomarker that is physiologically linked to the development of AKI and is easy to interpret. Studies have shown that renal hypoxia is a common factor in the development of most forms of AKI. In addition, early studies found there was a strong correlation between urine oxygen partial pressure (PuO2) and renal tissue oxygenation. Others have confirmed that low PuO2 is a risk factor for developing AKI in the setting of cardiac surgery. While we have developed a noninvasive PuO2 monitor, it is necessary to build an algorithm that can interpret this data and help the care team digest the information contained in the signal. Thus, the aim of this work was to test whether we could combine noninvasive PuO2 monitoring and machine learning to predict AKI with similar performance to injury-based biomarkers. The average area under the receiver operator characteristic curve for the machine learning algorithms was less than 0.7, which is the lower end of the predictive performance for injury-based biomarkers. However, we have identified several potential explanations and will work to address them in the future. Developing a noninvasive PuO2 monitor, and related algorithm could potentially improve outcomes in cardiac surgery and more generally in resource limited or austere settings such as during space travel.
Advances in Noninvasive Urine Oxygen Monitoring for Detection of Acute Kidney Injury
Weber State University
Acute kidney injury (AKI) is a common complication of cardiac surgery that is associated with increased hospital length of stay and mortality risk. The most common diagnostic criteria come from the Kidney Disease Improving Global Outcomes Guidelines. These criteria rely on detecting changes in serum creatinine relative to a baseline measurement and decreased urine output. These markers are meant to capture a decrease in the overall filtration rate. However, the filtration rate may or may not change depending on the baseline health of the individual patient and the type and severity of the injury. In addition, these markers are influenced by factors other than renal function or health, such as the administration of certain drugs or the volume status of the patient. Up to this point, research has primarily focused on the application of serum and urinary biomarkers that increase in response to renal injury or cell death. While these tests have advantages over current methods, the clinical utility of these tools is limited by high cost and a long wait time for results. Thus, there is a need for a noninvasive biomarker that is physiologically linked to the development of AKI and is easy to interpret. Studies have shown that renal hypoxia is a common factor in the development of most forms of AKI. In addition, early studies found there was a strong correlation between urine oxygen partial pressure (PuO2) and renal tissue oxygenation. Others have confirmed that low PuO2 is a risk factor for developing AKI in the setting of cardiac surgery. While we have developed a noninvasive PuO2 monitor, it is necessary to build an algorithm that can interpret this data and help the care team digest the information contained in the signal. Thus, the aim of this work was to test whether we could combine noninvasive PuO2 monitoring and machine learning to predict AKI with similar performance to injury-based biomarkers. The average area under the receiver operator characteristic curve for the machine learning algorithms was less than 0.7, which is the lower end of the predictive performance for injury-based biomarkers. However, we have identified several potential explanations and will work to address them in the future. Developing a noninvasive PuO2 monitor, and related algorithm could potentially improve outcomes in cardiac surgery and more generally in resource limited or austere settings such as during space travel.