Table 4. Diagnostic performance of disseminated intravascular coagulation (DIC) diagnostic classifiers in this study.
| Dataset | DIC criteria | DIC | Non-DIC | AUC (95% CI) |
Diagnostic performance (95% CI) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | Positive LR |
Negative LR |
||||||||
| Set 1 (n = 656) |
ANN* | 0.981 (0.973–0.989) |
89.9 (85.3–93.5) |
96.0 (93.7–97.7) |
92.3 (88.3–95.1) |
94.7 (92.4–96.3) |
22.64 (14.18–36.14) |
0.11 (0.07–0.15) |
|||||
| Positive | 205 | 17 | |||||||||||
| Negative | 23 | 411 | |||||||||||
| ISTH | 0.945 (0.929–0.962) |
82.0 (76.4–86.8) |
93.7 (91.0–95.8) |
87.4 (82.7–90.9) |
90.7 (88.1–92.8) |
13.00 (8.98–18.82) |
0.19 (0.15–0.25) |
||||||
| Score ≥ 5 | 187 | 27 | |||||||||||
| Score < 5 | 41 | 401 | |||||||||||
| JMHW | 0.943 (0.927–0.959) |
91.2 (86.8–94.6) |
84.4 (80.6–87.7) |
75.6 (71.3–79.5) |
94.8 (92.2–96.5) |
5.83 (4.66–7.29) |
0.10 (0.07–0.16) |
||||||
| Score ≥ 7 | 208 | 67 | |||||||||||
| Score < 7 | 20 | 361 | |||||||||||
| JAAM | 0.928 (0.909–0.946) |
94.7 (91.0–97.3) |
79.0 (74.8–82.7) |
70.6 (66.6–74.3) |
96.6 (94.2–98.0) |
4.51 (3.74–5.43) |
0.07 (0.04–0.12) |
||||||
| Score ≥ 4 | 216 | 90 | |||||||||||
| Score < 4 | 12 | 338 | |||||||||||
| Set 2 (n = 217) |
ANN* | 0.968 (0.945–0.986) |
90.0 (81.2–95.6) |
93.4 (87.9–97.0) |
88.9 (80.9–93.8) |
94.1 (89.2–96.9) |
13.7 (7.25–25.87) |
0.11 (0.06–0.22) |
|||||
| Positive | 72 | 9 | |||||||||||
| Negative | 8 | 128 | |||||||||||
| ISTH | 0.946 (0.916–0.976) |
82.5 (72.4–90.1) |
95.6 (90.7–98.4) |
91.7 (83.3–96.0) |
90.3 (85.3–93.8) |
18.84 (8.56–41.46) |
0.18 (0.11–0.29) |
||||||
| Score ≥ 5 | 66 | 6 | |||||||||||
| Score < 5 | 14 | 131 | |||||||||||
ANN, artificial neural network; ISTH, the International Society on Thrombosis and Haemostasis; JMHW, the Japanese Ministry of Health and Welfare; JAAM, the Japanese Association for Acute Medicine; CI, Confidence Interval; AUC, area under curve; PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio.
* Cut-off values for ANN model was determined at 0.501.