Urinary Squamous Epithelial Cells Do Not Accurately Predict Urine Culture Contamination, but May Predict Urinalysis Performance in Predicting Bacteriuria

Abstract

Abstract Objectives The presence of squamous epithelial cells ( SEC s) has been advocated to identify urinary contamination despite a paucity of evidence supporting this practice. We sought to determine the value of using quantitative SEC s as a predictor of urinalysis contamination. Methods Retrospective cross‐sectional study of adults (≥18 years old) presenting to a tertiary academic medical center who had urinalysis with microscopy and urine culture performed. Patients with missing or implausible demographic data were excluded (2.5% of total sample). The primary analysis aimed to determine an SEC threshold that predicted urine culture contamination using receiver operating characteristics ( ROC ) curve analysis. The a priori secondary analysis explored how demographic variables (age, sex, body mass index) may modify the SEC test performance and whether SEC s impacted traditional urinalysis indicators of bacteriuria. Results A total of 19,328 records were included. ROC curve analysis demonstrated that SEC count was a poor predictor of urine culture contamination (area under the ROC curve = 0.680, 95% confidence interval [ CI ] = 0.671 to 0.689). In secondary analysis, the positive likelihood ratio ( LR +) of predicting bacteriuria via urinalysis among noncontaminated specimens was 4.98 (95% CI = 4.59 to 5.40) in the absence of SEC s, but the LR + fell to 2.35 (95% CI = 2.17 to 2.54) for samples with more than 8 SEC s/low‐powered field (lpf). In an independent validation cohort, urinalysis samples with fewer than 8 SEC s/lpf predicted bacteriuria better (sensitivity = 75%, specificity = 84%) than samples with more than 8 SEC s/lpf (sensitivity = 86%, specificity = 70%; diagnostic odds ratio = 17.5 [14.9 to 20.7] vs. 8.7 [7.3 to 10.5]). Conclusions Squamous epithelial cells are a poor predictor of urine culture contamination, but may predict poor predictive performance of traditional urinalysis measures.