Table 3.
Hypothetical example of the impact on estimated sensitivity and specificity of using different accuracy assessment interval lengths
| Unobserved truth | ||||||
|---|---|---|---|---|---|---|
| Cancer | No cancer | Total | ||||
| Screen + | 8 | 20 | 28 | |||
| Screen − | 2 | 970 | 972 | |||
| Total | 10 | 990 | 1000 | |||
| Sensitivity | 80.0% | |||||
| Specificity | 98.0% | |||||
| Observed with 6-month accuracy assessment interval | ||||||
| Cancer | No cancer | Total | TP misclassified as FP | 2.400 | ||
| Screen + | 5.604 | 22.396 | 28 | TN misclassified as FN | 0.194 | |
| Screen − | 2.194 | 969.806 | 972 | FP misclassified as TP | 0.004 | |
| Total | 7.798 | 992.202 | 1000 | FN misclassified as TN | 0 | |
| Sensitivity | 71.9% | |||||
| Specificity | 97.7% | |||||
| Observed with 12-month accuracy assessment interval | ||||||
| Total | Cancer | No cancer | Total | TP misclassified as FP | 0.800 | |
| Screen + | 7.210 | 20.790 | 28 | TN misclassified as FN | 0.485 | |
| Screen − | 2.485 | 969.515 | 972 | FP misclassified as TP | 0.010 | |
| Total | 9.695 | 990.305 | 1000 | FN misclassified as TN | 0 | |
| Sensitivity | 74.4% | |||||
| Specificity | 97.9% | |||||
FP=false positive; FN=false negative; TP=true positive; TN=true negative
1
We assume that during a 6-month accuracy assessment interval, 0.02% of the no-cancer group develops and is diagnosed with cancer and that 70% of the screen-positive group receives a cancer-confirming follow-up test.
2
We assume that during a 12-month accuracy assessment interval 0.05% of the no-cancer group develops and is diagnosed with cancer and that 90% of the screen-positive group receives a cancer-confirming follow-up test.