Acute Kidney Injury (AKI) is a common complication among hospitalized patients and heralds a 3-5 fold increase in mortality, increased costs, and potential lifelong dependence on dialysis. While clinical decision support systems (CDSS) have previously assisted with the medical management of chronic kidney disease, there has been no validated approach for a CDSS to identify AKI and facilitate a medical intervention at an early stage. We propose to develop an AKI detection algorithm using laboratory and bedside measurements that are typically available in electronic medical records. We will validate the algorithm against the judgment of an expert nephrology review panel and use the latest published standards for AKI staging. As a second step, we plan to create models (using both traditional and machine learning approaches) that predict progression of AKI by incorporating comorbidities, admission diagnoses, and exposure to nephrotoxic therapies. Once the electronic criteria for early AKI are defined, we will build and evaluate a hospital-wide CDSS for managing early AKI. The goal is to prevent AKI progression and the associated complications of hypokalemia, acidemia, and medication toxicity.
Hospitalized patients who develop impaired kidney function are at increased risk of dying or developing adverse events. Computerized systems of care are needed to quickly alert hospital care teams to which patients are developing kidney injury in order to tailor medications and other therapies.
Showing the most recent 10 out of 14 publications