Kansas City-based University of Kansas Medical Center researchers created an artificial intelligence model that can predict urinary tract infection without a microscopy, WFMZ reported Jan. 23.
The study, published in Annals of Family Medicine, created a classifier called NoMicro that does not depend on urine microscopy to predict urine cultures. The model used predictor variables including age, gender, dipstick urinalysis nitrites, leukocytes, clarity, glucose, protein, blood, dysuria, abdominal pain and a history of urinary tract infections. The retrospectively validated model used 80,387 emergency department data sets and 472 primary care data sets to prove accuracy.
Researchers say NoMicro could be used to safely withhold antibiotics for low-risk patients and prevent antibiotic overuse.