Table 3.
Diagnostic accuracy of a multivariable diagnostic prediction model for HIT as determined in the validation dataset (25% of the patients).
| n | TP | FN | TN | FP | Sensitivity (95% CI) | Specificity (95% CI) | PPV | NPV | LR+ | LR− | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| TORADI-HIT algorithm | |||||||||||
| CLIA (SVM) | 324 | 25 | 1 | 282 | 16 | 96 (80, 100) | 95 (91, 97) | 61 (45, 76) | 100 (98, 100) | 17.91 (11.05, 29.02) | 0.04 (0.01, 0.28) |
| PaGIA (GBM) | 319 | 24 | 0 | 280 | 15 | 100 (86, 100) | 95 (92, 97) | 62 (45, 77) | 100 (99, 100) | 19.67 (12.01, 32.20) | 0.00 (0.00, 0.01) |
| ELISA (SVM) | 343 | 25 | 3 | 300 | 15 | 89 (72, 98) | 95 (92, 97) | 62 (46, 77) | 99 (97, 100) | 18.75 (11.26, 31.23) | 0.11 (0.04, 0.33) |
| Current clinical algorithm | |||||||||||
| CLIA | 324 | 21 | 5 | 282 | 16 | 81 (61, 93) | 95 (91, 97) | 57 (39, 73) | 98 (96, 99) | 15.04 (9.01, 25.11) | 0.20 (0.09, 0.45) |
| PaGIA | 319 | 21 | 3 | 243 | 52 | 88 (68, 97) | 82 (78, 87) | 29 (19, 41) | 99 (96, 100) | 4.96 (3.72, 6.63) | 0.15 (0.05, 0.44) |
| ELISA | 343 | 24 | 4 | 292 | 23 | 86 (67, 96) | 93 (89, 95) | 51 (36, 66) | 99 (97, 100) | 11.74 (7.70, 17.89) | 0.15 (0.06, 0.38) |
Abbreviations: TP - true positives, FN - false negatives, TN - true negatives, FP - false positives, PPV - positive predictive value, NPV - negative predictive value, LR+ - positive likelihood ratio, LR− - negative likelihood ratio, SVM - support vector machine, GBM - gradient boosting machine.
The accuracy in the full dataset is given in Table S8 of the supplementary material. Accuracy data of the currently recommended algorithm (4Ts score + immunoassay) are given as comparison.