Table 4 Sensitivity and specificity for remote detection of ROP requiring treatment, by three trained readers at three diagnostic cut‐off levels*.
| Reader | Mild or worse ROP | Type 2 prethreshold or worse | ROP requiring treatment | |||
|---|---|---|---|---|---|---|
| Sensitivity | Specificity† | Sensitivity | Specificity‡ | Sensitivity | Specificity | |
| A | 1.000 (−) | 0.513 (0.056) | 1.000 (−) | 0.846 (0.032) | 0.850 (0.080) | 0.960 (0.016) |
| B | 1.000 (−) | 0.553 (0.054) | 0.900 (0.070) | 0.926 (0.022) | 0.850 (0.080) | 0.973 (0.013) |
| C | 1.000 (−) | 0.313 (0.067) | 0.900 (0.070) | 0.933 (0.021) | 0.900 (0.022) | 0.953 (0.017) |
*Results are shown at diagnostic cut‐offs of mild or worse ROP (that is, readers were asked whether image sets showed mild or worse ROP), type 2 prethreshold or worse ROP (that is, readers were asked whether image sets showed type 2 prethreshold or worse ROP), and ROP requiring treatment (that is, readers were asked whether image sets showed ROP requiring treatment). Results are reported as proportion (SE).
†Differences in specificity at diagnostic cutoff of mild or worse ROP are statically significant for reader pairs A v C and B v C (p<0.001).
‡Difference in specificity at diagnostic cut‐off of type 2 prethreshold or worse ROP are statistically significant for reader pairs A v B (p = 0.002) and A v C (p<0.001).