Acute kidney injury (AKI) is a life-threatening clinical syndrome prevalent in hospitalized patients (10% to 15% affected), especially among critically ill patients (>50% affected), and has exceeded the annual incidence of myocardial infarction. AKI patients are at much higher risk for developing poor long-term outcomes including incident and progressive chronic kidney disease, cardiovascular disease, and death. Moreover, AKI has complex etiologies, variable pathogenesis, and diverse outcomes. For example, congestive heart failure and dehydration can produce identical changes in serum creatinine (SCr) level and urine output (i.e., biomarkers used to diagnose AKI); however, they differ vastly in their physiological contexts and demand completely opposite treatments.
Recent work has shown that AKI prediction models built on cohort that combines patients with different etiologies can hide subgroups that are more tightly associated with the clinical outcome of interest and conceal unique pathophysiological processes specific to certain subgroups, resulting in poor prediction performance. In this webinar, Liu will first focus on the development and multi-site validation of an AKI prediction model and discuss how its performance was affected by the patient heterogeneity. Liu will then present a new personalized transfer learning framework for more accurate and equitable AKI prediction. Lastly, Liu will discuss in detail the risk factor heterogeneity and interactions discovered at the individual level resulted from the study.
Watch the Recording
Presenter
Dr. Mei Liu is an Associate Professor in the Department of Health Outcomes and Biomedical Informatics (HOBI) at University of Florida. She is also Associate Director of Graduate Education in HOBI, overseeing their Biomedical Informatics MS/PhD programs. Dr. Liu joined University of Florida in October 2022 under the UF AI Initiative from the University of Kansas Medical Center where she served as the co-lead of the University of Kansas CTSI Informatics Core and interim Director of Medical Informatics. Now at UF, she is leading the training component of the UF CTSI Biomedical Informatics Core and serving as dual PI for the PCORnet OneFlorida+ clinical research network. Dr. Liu was trained in computer science and biomedical informatics and has deep expertise in designing and developing AI and machine learning algorithms using electronic health record data for disease risk prediction and predictor discovery.