Skip to main content
. 2019 Nov 27;14(11):e0225716. doi: 10.1371/journal.pone.0225716

Table 3. Area under the ROC curve and sensitivity for prediction of preeclampsia with delivery at <34, <37 and ≥37 weeks’ gestation based on pre-specified variables, backward selection and Random forest methods.

Total study population
n = 62 562
Restricted study populationc
n = 58 276
Predictive method Area under the ROCa curve (95% CI) Sensitivity for 10% FPRb (95% CI) Area under the ROCa curve (95% CI) Sensitivity for 10% FPRb (95% CI)
Prediction of preeclampsia <34 weeks
Pre-specified variables 0.68 (0.64–0.72) 30.6 (24.5–37.2) 0.67 (0.63-0-72) 28.8 (22.1–36.3)
Backward selection 0.66 (0.62–0.70) 26.9 (21.1–33.3) 0.64 (0.60–0.69) 28.8 (22.1–36.3)
Random forest 0.58 (0.54–0.62) 18.5 (13.6–24.4) 0.57 (0.53–0.62) 17.6 (12.2–24.2)
Prediction of preeclampsia <37 weeks
Pre-specified variables 0.68 (0.65–0.70) 29.2 (25.2–33.4) 0.68 (0.65-0-71) 29.3 (25.0–34.0)
Backward selection 0.66 (0.63–0.68) 25.8 (22.0–29.8) 0.66 (0.63–0.69) 27.9 (23.6–32.5)
Random forest 0.60 (0.57–0.63) 24.3 (20.6–28.4) 0.60 (0.57–0.62) 21.4 (17.5–25.7)
Prediction of preeclampsia ≥37 weeks
Pre-specified variables 0.67 (0.66–0.69) 28.1 (26.3–30.0) 0.67 (0.66–0.68) 27.6 (25.7–29.5)
Backward selection 0.67 (0.66–0.68) 28.2 (26.4–30.1) 0.67 (0.66–0.68) 27.5 (25.6–29.5)
Random forest 0.61 (0.60–0.62) 22.4 (20.7–24.2) 0.62 (0.60–0.63) 22.7 (20.9–24.5)

a ROC: Receiver operating characteristic

b FPR: False-positive rate

c Restricted study population: Pregnancies without major malformations or treatment with aspirin