Abstract
Purpose :
To analyze differences in ROP severity in patients diagnosed with treatment-requiring ROP (TR-ROP) by examiner.
Methods :
Fundus photographs of all babies undergoing ROP screening were obtained as part of the Imaging and Informatics in ROP (i-ROP) cohort study. An ROP severity score (1-9) was generated for each image using methods previously published. Images were analyzed for this study for babies who were diagnosed with TR-ROP by an examining i-ROP physician based on the ophthalmoscopic exam. We analyzed differences in ROP severity by treating physician, adjusting for birthweight and gestational age. Examiners who diagnosed fewer than 12 babies in the study were excluded.
Results :
191 eyes diagnosed with TR-ROP from 5 participating i-ROP centers met inclusion criteria for this study. The mean i-ROP score for treatment requiring ROP patients was 7.30 (standard deviation of 1.85). A multi-level linear regression analysis was conducted to determine whether there were differences between examiners in the level of severity diagnosed as TR-ROP. One examiner (E) had a higher mean ROP severity score than the others (P<0.01). The mean for each examiner was 6.64 (Examiner A, standard deviation of 1.91), 7.27 (Examiner B, standard deviation of 1.86), 7.21 (Examiner C, standard deviation of 1.65), 7.48 (Examiner D, standard deviation of 1.88), 8.24 (Examiner E, standard deviation of 1.70), and 6.60 (Examiner F, standard deviation of 2.34).
Conclusions :
There are several key findings to this study: 1) We found that experts treated at different levels of vascular severity in the diagnosis of TR-ROP in the i-ROP cohort study. 2) There was a wide variance around the severity score at time of diagnosis of TR-ROP for each examiner, suggesting that factors other than the level of vascular severity may play a role in the diagnosis of TR-ROP. Future work may illuminate these factors that are relevant for the diagnosis of TR-ROP and how the level of vascular severity should factor into treatment decisions based on avoidance of unfavorable anatomic outcomes, and optimization of visual outcomes.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.