Beyond Biomarkers: Machine Learning in Diagnosing Acute Kidney Injury

• Published in: Mayo Clinic proceedings Vol. 94; no. 5; pp. 748 - 750
• Main Author: Molitoris, Bruce A.
• Format: Journal Article
• Language: English
• Published: England Elsevier Inc 01.05.2019; Frontline Medical Communications Inc; Elsevier Limited
• Subjects: Acute kidney failure; Analysis; Artificial intelligence; Biological markers; Biomarkers; Care and treatment; Cell cycle; Cell injury; Creatinine; Critical care medicine; Diagnosis; Epithelial cells; Hospital patients; Hospitals; Intelligence gathering; Intensive care; Kidneys; Laboratories; Learning algorithms; Machine learning; Medical diagnosis; Mortality; Nephrology; Patient monitoring equipment; Patients; Physiologic monitoring; Risk factors; Studies; Surveillance; Technology application; Urine
• ISSN: 0025-6196; 1942-5546; 1942-5546
• DOI: 10.1016/j.mayocp.2019.03.017

[...]many individuals donating a kidney for transplant have less than a 0.3 mg/dL increase in their serum creatinine level (the increment in serum creatinine used by the AKIN to diagnose AKI), even though they have lost one-half of their total kidney function. [...]loss of up to 50% of total kidney function in these patients would not have been registered as AKI. [...]one use for the novel biomarkers is to identify "subclinical AKI" as a serum creatinine-negative but biomarker-positive diagnosis, indicating the presence of renal tubular epithelial cell injury. [...]use of these biomarkers is expensive, and this must be considered before considering routine surveillance protocols. [...]attention has turned to other approaches to identify patients with a high probability of AKI and alert the physician to its likely occurrence. [...]the model was computer-calculated, which thus provided near real-time information for surveillance purposes without physician input or time.