Predicting Urinary Tract Infections With Interval Likelihood Ratios.

Abstract

Protocols for diagnosing urinary tract infection (UTI) often use arbitrary cutoff values of urinalysis components to guide management. Interval likelihood ratios (ILRs) of urinalysis results may improve the test's precision in predicting UTIs. We calculated the ILR of urinalysis components to estimate the posttest probabilities of UTIs in young children. Review of 2144 visits to the pediatric emergency department of an urban academic hospital from December 2011 to December 2019. Inclusion criteria were age <2 years and having a urinalysis and urine culture sent. ILR boundaries for hemoglobin, protein, and leukocyte esterase were "negative," "trace," "1+," "2+" and "3+." Nitrite was positive or negative. Red blood cells and white blood cells (WBCs) were 0 to 5, 5 to 10, 10 to 20, 20 to 50, 50 to 100, and 100 to 250. Bacteria counts ranged from negative to "loaded." ILRs for each component were calculated and posttest probabilities for UTI were estimated. The UTI prevalence was 9.2%, with the most common pathogen being Escherichia coli (75.2%). The ILR for leukocyte esterase ranged from 0.20 (negative) to 37.68 (3+) and WBCs ranged from 0.24 (0-5 WBCs) to 47.50 (100-250 WBCs). The ILRs for nitrites were 0.76 (negative) and 25.35 (positive). The ILR for negative bacteria on urinalysis was 0.26 and 14.04 for many bacteria. The probability of UTI in young children significantly increases with 3+ leukocyte esterase, positive nitrite results, 20 to 50 or higher WBCs, and/or many or greater bacteria on urinalysis. The probability of UTI only marginally increases with trace or 1+ leukocyte esterase or 5 to 20 WBCs. Our findings can be used to more accurately predict the probability of true UTI in children.