Electronic Diagnostic Model Predicts Acute Interstitial Nephritis in Patients

It is typically associated with the use of medications such as steroids, proton pump inhibitors, and antibiotics. Studies show that AKI, which involves a sudden decline in kidney function, affects about 20% of hospitalized patients. One of the key challenges in managing AKI is distinguishing AIN from other causes of kidney injury. This is complicated by the fact that over 90% of AIN patients show no obvious symptoms, and common diagnostic methods, including urine eosinophil counts, urine microscopy, and imaging tests, have poor accuracy. Misdiagnosing AIN can result in the premature discontinuation of essential treatments like immune checkpoint inhibitors or antibiotics, potentially leading to permanent kidney damage if the condition is not promptly identified. Given the difficulty of diagnosing AIN, a kidney biopsy is often required, though it is an invasive procedure with its own risks. To address this challenge, researchers have developed a diagnostic model using lab tests from electronic medical records, which could significantly improve early detection of AIN in patients.

In the study, researchers from Johns Hopkins Medicine (Baltimore, MD, USA) and Yale University (New Haven, CT, USA) developed a diagnostic model to predict AIN in patients using a machine learning technique called least absolute shrinkage and selection operator (LASSO). The laboratory tests used in the model included serum creatinine, blood urea nitrogen (BUN), urine protein levels, and urine specific gravity (the density of urine compared to water). The study involved two patient cohorts, both of which had previously undergone kidney biopsies at Johns Hopkins Hospital (JHH) or Yale University. The JHH cohort consisted of 1,454 patients who had a native kidney biopsy between January 2019 and December 2022, while the Yale cohort included 528 patients scheduled for clinical kidney biopsy between July 2020 and June 2023. Patients who did not have a serum creatinine value within a year before their biopsy, were undergoing kidney allograft biopsies, or had known vasculitis or lupus nephritis were excluded from the study.

A total of 1,982 patients were analyzed, with 22% diagnosed with AIN. The study found that patients with AIN were more likely to be hospitalized and had higher serum creatinine levels and a higher blood urea nitrogen-to-creatinine ratio. The diagnostic model improved the accuracy of AIN diagnosis to 77%. However, there were differences in the prevalence of AIN between the two cohorts. After adjusting for prevalence at the individual centers, the model's calibration improved significantly, leading to more accurate diagnoses. The findings, published in the Journal of the American Society of Nephrology, suggest that this diagnostic model could assist clinicians in determining whether a kidney biopsy is necessary in patients with AKI and help guide treatment decisions for AIN. The formula for predicting AIN is available on MDCalc.