Table 2 Performance of selection and merged models on test data subset.

From: Identification of clinical and urine biomarkers for uncomplicated urinary tract infection using machine learning algorithms

Data set

Algorithm

AUC

PPV

NPV

LR+

LR-

F1 score1

Selected predictors

POETIC UTI classification, UTI prevalence 42.6%

Clinical markers with cloudiness

RF + RFE

0.72 (0.60–0.85)2

0.65 (0.44–0.82)

0.79 (0.59–0.91)

2.55 (1.4–4.6)

0.37 (0.18–0.75)

0.69 (0.57–0.81)

Cloudiness

Clinical markers with turbidity

RF + RFE

0.73 (0.60–0.85)

0.76 (0.50–0.92)

0.73 (0.56–0.86)

4.38 (1.6–11.7)

0.49 (0.31–0.80)

0.65 (0.52–0.78)

Turbidity

Immunological markers

RF + RFE

0.82 (0.69–0.94)

0.68 (0.45–0.85)

0.75 (0.56–0.88)

2.88 (1.41–5.92)

0.45 (0.25–0.80)

0.67 (0.54–0.80)

IL-1β and MMP9

Selected clinical with cloudiness + selected immunological markers

RF

0.82 (0.70–0.95)

0.75 (0.51–0.90)

0.76 (0.58–0.89)

3.82 (1.72–9.52)

0.42 (0.23–0.73)

0.70 (0.58–0.82)

Cloudiness, IL-1β and MMP9

Selected clinical with turbidity + selected immunological markers

RF

0.76 (0.63–0.90)

0.67 (0.43–0.85)

0.73 (0.54–0.86)

2.61 (1.30–5.59)

0.52 (0.30–0.86)

0.64 (0.51–0.77)

Turbidity, IL-1β and MMP9

Clinical markers with cloudiness

SVM + RFE

0.73 (0.61–0.85)

0.65 (0.44–0.82)

0.79 (0.59–0.91)

2.55 (1.4–4.6)

0.37 (0.18–0.75)

0.69 (0.57–0.81)

Cloudiness

Clinical markers with turbidity

SVM + RFE

0.86 (0.76–0.96)

0.76 (0.50–0.92)

0.73 (0.56–0.86)

4.38 (1.6–11.7)

0.49 (0.31–0.80)

0.65 (0.52–0.78)

Turbidity and age category

Immunological markers

SVM + RFE

0.81 (0.68–0.94)

0.82 (0.55–0.95)

0.76 (0.58–0.88)

6.29 (2.04–19.36)

0.43 (0.26–0.73)

0.70 (0.58–0.82)

MMP9, NGAL, IL-8/CXCL8 and IL-1β

Selected clinical with cloudiness + selected immunological markers

SVM

0.82 (0.70–0.94)

0.79 (0.54–0.93)

0.77 (0.59–0.89)

5.00 (1.93–13.23)

0.40 (0.23–0.71)

0.71 (0.58–0.83)

Cloudiness, MMP9, NGAL, IL-8/CXCL8 and IL-1β

Selected clinical with turbidity + selected immunological markers

SVM

0.79 (0.66–0.92)

0.70 (0.47–0.86)

0.77 (0.58–0.90)

3.04 (1.52–6.24)

0.39 (0.31–0.74)

0.70 (0.58–0.82)

Turbidity, age category, MMP9, NGAL, IL-8/CXCL8 and IL-1β

  1. 1F1 score: harmonic mean of precision and recall.
  2. 295% confidence interval of the performance metric.
  3. AUC: Area under the curve.
  4. PPV: Positive predictive value.
  5. NPP: Negative predictive value.
  6. LR+ and LR−: positive and negative likelihood ratio.
  7. SVM: support vector machine.
  8. RF: random forest.
  9. RFE: recursive feature elimination.
  10. IL: interleukin.
  11. MMP: matrix metalloproteinase.
  12. NGAL: neutrophil gelatinase-associated lipocalin.
  13. CXCL: the chemokine (C-X-C motif) ligand.
Back to article page