Table 3.
Prediction of pre-eclampsia in Avon Longitudinal Study of Parents and Children (ALSPAC) and Southampton Women’s Survey (SWS) with model 2* at 28 weeks’ gestation for fixed sensitivities using imputed datasets
| Sensitivity | Risk score threshold | Specificity (95% CI) | Positive predictive value (95% CI) | Negative predictive value (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | No (%) testing positive | No (%) true positive | No (%) false positive | No (%) true negative | No (%) false negative |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ALSPAC | |||||||||||
| 0.95 | 0.008 | 0.37 (0.31 to 0.42) | 0.04 (0.03 to 0.04) | 1.00 (0.99 to 1.00) | 1.50 (1.37 to 1.64) | 0.14 (0.08 to 0.23) | 8316 (64.1) | 297 (2.3) | 8019 (61.9) | 4633 (35.7) | 16 (0.1) |
| 0.99 | 0.003 | 0.15 (0.13 to 0.17) | 0.03 (0.02 to 0.03) | 1.00 (1.00 to 1.00) | 1.16 (1.13 to 1.20) | 0.07 (0.02 to 0.21) | 11068 (85.4) | 310 (2.4) | 10758 (83.0) | 1894 (14.6) | 3 (0.0) |
| SWS (without recalibration, using ALSPAC risk score thresholds) | |||||||||||
| 1.00 (0.98 to 1.00) | 0.008 | 0.31 (0.29 to 0.33) | 0.04 (0.03 to 0.05) | 1.00 (1.00 to 1.00) | 1.44† (1.40 to 1.50) | 0.02† (0.00 to 0.32) | 2086 (69.6) | 87 (2.9) | 2000 (66.7) | 913 (30.4) | 0 (0.0) |
| 1.00 (1.00 to 1.00) | 0.003 | 0.10 (0.09 to 0.11) | 0.03 (0.03 to 0.04) | 1.00 (1.00 to 1.00) | 1.11† (1.09 to 1.13) | 0.06† (0.00 to 0.90) | 2699 (90.0) | 87 (2.9) | 2612 (87.1) | 300 (10.0) | 0 (0.0) |
| SWS (after recalibration) | |||||||||||
| 0.95 | 0.011 | 0.47 (0.40 to 0.55) | 0.05 (0.04 to 0.06) | 1.00 (0.99 to 1.00) | 1.82 (1.56 to 2.11) | 0.10 (0.04 to 0.26) | 1617 (53.9) | 83 (2.8) | 1534 (51.2) | 1379 (46.0) | 4 (0.1) |
| 0.99 | 0.006 | 0.33 (0.22 to 0.44) | 0.04 (0.03 to 0.05) | 1.00 (1.00 to 1.00) | 1.49 (1.28 to 1.73) | 0.04 (0.00 to 0.27) | 2037 (67.9) | 86 (2.9) | 1951 (65.1) | 961 (32.0) | 1 (0.0) |
*Includes maternal covariates, mean arterial pressure at first visit, and observed value of mean arterial pressure at 28 weeks.
†Continuity correction applied in calculation of likelihood ratios in multiply imputed datasets where 2×2 contingency tables of test v disease status had zero cells. Constant of 0.5 has been added to all cells in these cases.