Table 5.
Diagnostic accuracy measures (with 95% CIs) of five diagnostic prediction models* combined with point of care D-dimer testing, to rule out pulmonary embolism, validated in primary care AMUSE-2 cohort (n=598).
| Measure | Original Wells | Modified Wells | Simplified Wells | Original revised Geneva | Simplified revised Geneva | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ≤5 | ≤3 | ≤3 | ≤1 | ≤2 | =0 | ≤6 | ≤4 | ≤3 | ≤1 | |||||
| Sensitivity | 89% (80% to 95%) | 95% (87% to 98%) | 89% (80% to 95%) | 97% (91% to 99%) | 85% (75% to 92%) | 99% (93% to 100%) | 89% (80% to 95%) | 92%(83% to 97%) | 86% (76% to 93%) | 93% (85% to 98%) | ||||
| Specificity | 59% (54% to 63%) | 50% (46% to 55%) | 59% (54% to 63%) | 32% (28% to 36%) | 59% (55% to 64%) | 21% (18% to 25%) | 54% (44% to 53%) | 38% (34% to 42%) | 59% (55% to 64%) | 36% (32% to 41%) | ||||
| PPV | 23% (18% to 28%) | 21% (17% to 26%) | 23% (18% to 28%) | 17% (13% to 20%) | 23% (18% to 28%) | 15% (12% to 18%) | 21% (17% to 26%) | 13% (10% to 16%) | 23% (18% to 28%) | 17% (13% to 21%) | ||||
| NPV | 97% (95% to 99%) | 99% (96% to 100%) | 97% (95% to 99%) | 99% (96% to 100%) | 97% (94% to 98%) | 99% (95% to 100%) | 97% (95% to 99%) | 97% (94% to 99%) | 97% (94% to 99%) | 97% (94% to 99%) | ||||
| Efficiency | 53% (49% to 57%) | 45% (41% to 49%) | 53% (49% to 57%) | 28% (25% to 32%) | 54% (50% to 58%) | 19% (16% to 22%) | 49% (45% to 53%) | 34% (30% to 38%) | 54% (50% to 58%) | 33% (29% to 37%) | ||||
| Failure rate | 2.5% (1.1% to 4.9%) | 1.5% (0.4% to 3.8%) | 2.5% (1.1% to 4.9%) | 1.2% (0.3% to 4.2%) | 3.4% (1.7% to 6.0%) | 0.9% (0.0% to 4.9%) | 2.8% (1.2% to 5.5%) | 2.9% (1.1% to 6.3%) | 3.1% (1.5% to 5.6%) | 2.6% (0.8% to 5.9%) | ||||
NPV=negative predictive value; PPV=positive predictive value.
*Chosen cut-off values for ruling out pulmonary embolism (“low PE probability”) are 1 point higher or lower than cut-offs recommended for published models.