According to a research published in the January/February issue of the Annals of Family Medicine, the NoMicro classifier seems to be reliable for screening urine cultures in instances of suspected urinary tract infection in the primary care setting without the requirement for microscopy.
A classifier (NoMicro) that is independent of urine microscopy was redesigned by Gurpreet Dhanda, M.D., and colleagues from the University of Kansas Medical Center in Kansas City. They also retrospectively validated a machine learning prediction model for urine cultures both internally and externally.
The main outcome was pathogenic urine culture growing 100,000 CFU, and the predictors were age, gender, dipstick urinalysis nitrites, leukocytes, clarity, glucose, protein, and blood, dysuria, stomach discomfort, and history of UTI.
The researchers discovered that eliminating microscope features did not significantly impair performance during internal validation (ROC-AUC values for NoMicro/XGBoost and NeedMicro, respectively, were 0.86 and 0.88). Additionally, great performance was attained through external validation.
In their report, the scientists state that “retrospective simulation revealed that NoMicro/random forests can be used to safely delay antibiotics for low-risk patients, hence preventing antibiotic misuse.” “The NoMicro classifier seems suitable for use in routine medical care. It is appropriate to conduct prospective trials to see how employing the NoMicro classifier balances benefits and drawbacks.”
