Analysis of factors with low positive predictive value in the diagnosis of urinary tract infection by flow cytometry.

PURPOSE

Culture-negative urine specimens can be rapidly screened by urine flow cytometry (UFC), while low positive predictive value (PPV) limits the clinical application. We explored the factors associated with a low PPV.

METHODS

A total of 5095 urine specimens were analyzed with UFC and culture. Diagnostic performance of leukocytes, bacteria, and BACT-info flags was evaluated by sensitivity, specificity, PPV, and negative predictive value (NPV). The association of contaminated culture and squamous epithelial cell count and BACT-info flag was performed by logistic regression analysis.

RESULTS

The NPVs of parallel combination of bacteria and leucocytes were 98.9% in males and 97.9% in females, and PPVs of serial combination were 86.6% and 77.8% in men and women, respectively. The PPV of Gram-negative flag was higher than that of Gram-positive flag. The proportions of contamination in the urine culture results of false positive specimens were 86.9% in males and 98.5% in females at the cutoff points of the serial combination, and these parameters were 53.2% in males and 85.6% in females for the Gram-positive flag. There was a statistically significant association between contaminated cultures and squamous epithelial cells count in females, but not in males. Associations between contaminated cultures and Gram-positive flags or Gram-pos/-neg flags were statistically significant, but there was no association between contaminated cultures and Gram-negative flags.

CONCLUSIONS

A serial combination of leukocytes and bacteria may maximize PPV in the diagnosis of bacterial urinary tract infection by urine flow cytometry, and contamination is the main reason for a low PPV.