Johns Hopkins Introduces AI Model to Predict Kidney Disease
Researchers from Johns Hopkins Medicine and Yale University have collaborated to develop a diagnostic model aimed at detecting acute interstitial nephritis (AIN) in patients, potentially enabling earlier diagnosis and treatment.......CONTINUE READING THE ARTICLE FROM THE SOURCE>>>>>
AIN, a common cause of acute kidney injury (AKI), involves inflammation and swelling of specific kidney tissues, often triggered by medications like steroids, proton pump inhibitors, and antibiotics. Early diagnosis and intervention are crucial in reducing the risk of permanent kidney damage in hospitalized patients.
The findings, published in the November 5th issue of the *Journal of the American Society of Nephrology*, build on previous work to create a model using a panel of lab tests documented in electronic medical records. Acute kidney injury affects one in five hospitalized patients, according to the American Kidney Fund.
A major challenge in managing AKI is distinguishing AIN from other causes, as over 90% of AIN patients exhibit no obvious symptoms. Additionally, tests like urine eosinophils, urine microscopy, and imaging have low accuracy for diagnosing AIN.
Misdiagnosis could lead to the discontinuation of life-saving treatments, while delayed diagnosis may result in irreversible kidney damage.
“A patient with AIN should be identified early, as it is one of the treatable causes of AKI,” said Dr. Chirag Parikh, Professor of Medicine and Director of the Division of Nephrology at Johns Hopkins Medicine, and a lead investigator of the study.
Given the difficulty in diagnosing AIN, kidney biopsies are often necessary, though they carry risks. Researchers sought alternative, non-invasive methods to diagnose AIN. In this study, a diagnostic model was created using machine learning techniques, specifically the Least Absolute Shrinkage and Selection Operator (LASSO), which incorporated laboratory tests such as serum creatinine, blood urea nitrogen (BUN), urine protein, and urine specific gravity.
“These laboratory tests make biological sense as they help differentiate AIN from other causes like prerenal azotemia and acute tubular necrosis,” Dr. Parikh explained.
The study involved two patient cohorts who had previously undergone kidney biopsies at either Johns Hopkins Hospital (JHH) or Yale University. The JHH cohort included 1,454 patients from January 2019 to December 2022, while the Yale cohort involved 528 patients scheduled for biopsies between July 2020 and June 2023. The study excluded patients with certain conditions like kidney transplant biopsies, kidney masses, vasculitis, or lupus nephritis.
From the 1,982 patients examined, 22% were diagnosed with AIN. Patients with AIN in both cohorts were more likely to be hospitalized, with higher serum creatinine and BUN-to-creatinine ratios. The diagnostic model improved the accuracy of AIN diagnosis to 77%. After adjusting for prevalence differences between centers, the model’s calibration improved, leading to more precise diagnoses of AIN.