Table 4.
Predictive performances shown by models in this study compared to those from recent studies.a
| Source | Predictive performanceb | |||
|
|
AUCc | Sensitivity, % | Specificity, % | |
| This study |
|
|
|
|
|
|
CVRd1 (right PI-UtAe) | 0.906 (0.896-0.916) | 91 (85-96) | 97 (90-100) |
|
|
CVR2 (mean PI-UtA) | 0.926 (0.919-0.933) | 95 (91-100) | 100 (100-100) |
|
|
CVR3 (lowest PI-UtA) | 0.970 (0.966-0.974) | 95 (91-100) | 100 (100-100) |
|
|
158-tree random forest | 0.976 (0.967-0.985) | 91 (87-94) | 93 (92-95) |
| Recent studies |
|
|
|
|
|
|
Wright A et al (2019) [26] | N/Af,g | 85 (72-94) | 90 (90-90) |
|
|
Wright D et al (2019) [27] | 0.970 (0.950-0.990) | 93 (76-99) | 90h |
|
|
Tan MY et al (2018) [25] | N/Ag | 90 (80-96) | 90h |
|
|
Sonek J et al (2018) [24] | N/Ag | 85i | 95i |
|
|
Perales A et al (2017) [23] | 0.930i | 81i | 95i |
|
|
Nuriyeva G et al (2017) [22] | 0.888i | 76i | 90i |
|
|
O'Gorman N et al (2017) [21] | 0.987i | 100 (80-100) | 90h |
|
|
Gallo DM et al (2016) [18] | 0.930 (0.892-0.968) | 85 (74-93) | 90h |
|
|
Tsiakkas A et al (2016) [19] | 0.987 (0.980-0.994) | 100 (92-100) | 90h |
|
|
Andrietti S et al (2016) [20] | 0.938 (0.917-0.959) | 82 (70-91) | 90h |
|
|
O'Gorman N et al (2016) [17] | 0.907i | 89 (79-96) | 90h |
|
|
Wright D et al (2015) [16] | 0.811i | 67 (59-74) | 90h |
aModels that showed the best sensitivity and an acceptable specificity in each study.
bPoint and interval estimates.
cAUC: area under the receiver operating characteristic (ROC) curve.
dCVR: classification via regression.
ePI-UtA: pulsatility index of the uterine artery.
fN/A: not applicable because it was not available.
gThis study showed an ROC curve without an AUC statement.
hFixed specificity in order to define sensitivity.
iThis study did not report an interval estimate.