Fig 1.
Flowchart of the study.
Table 1.
Patient characteristics.
Table 2.
Distribution of species of positive culture (≥104), n = 247 (40.2%).
Fig 2.
Scatter plots for urine flow cytometry parameter of the training set, n = 429.
Positive (A) vs. mixed (B) urine culture (≥104) are colored in black. In A: left of the solid line most of the observations showed no growth. Setting a test cut-off for bacteria and leukocytes in urine flow cytometry left of the line will lead to a high negative predictive value (NPV) for urine culture growth; vice versa, cut-off values defined by the dotted line will lead to a high positive predictive value (PPV). For a better graphical representation the number of bacteria and leukocytes, respectively round epithelial cells and epithelial cells (per μL) were ln-transformed.
Fig 3.
A) Sensitivity, B) Specificity, C) negative predictive value (NPV) and D) positive predictive value (PPV) for a positive urine culture for different cut-offs of bacteria and leukocytes (square root-transformed) of the whole data set, n = 613.
Table 3.
Predictive values for different cut-offs (growth probability: Low, medium, high) for urine flow cytometry for the number of bacteria/μL and leukocytes/μL predicting a positive urine culture (≥104), validation set n = 184.
Fig 4.
Validation of the developed CHAID-classification tree with the validation set, n = 184, and a comparison of the classification of the training and validation data set.
The framed nodes predict urine culture growth. The units of bacteria and leukocytes are per μL.
Fig 5.
Nomogram for urine culture growth based on the training sample (n = 429) and predictive values for different predicted possibility cut-offs for a positive test (test pos.) based on the validation sample (n = 184).
N Example: Considering the urine flow cytometry of a patient shows 80 leukocytes/μL and bacteria16/μL. Eighty leukocytes/μL correspond to ≈2.0 points on the score axis, 16 bacteria/μL correspond to ≈2.0 point. Thus, the total score, the sum of the single scores is 4.0 (2.0+2.0). The predicted probability of urine culture growth can be read off the probability axis. Four points on the probability axis correspond to a urine culture growth of about 10%. A test that is defined positive, when the predicted probability of culture growth is greater than 10% (table right corner), has a sensitivity of 98.6%. Thus, urine culture growth is very unlikely and ordering a urine culture not recommended. Remark: The axis of leukocytes and bacteria per μL are (.)0.25- transformed to obtain predictive probabilities between 0.01 and 0.99.