Abstract

Objective: To study the correlation between expert and non-expert observers in the reporting images for the diagnosis of retinopathy of prematurity (ROP) in a telemedicine setting. Methods: A cross-sectional, multicenter study, consisting of 25 sets of images of patients screened for ROP. They were evaluated by two experts in ROP and 1 non-expert and classified according to telemedicine classification, zone, stage, plus disease and Ells referral criteria. The telemedicine classification was: no ROP, mild ROP, type 2 ROP, or ROP that requires treatment. Ells referral criteria is defined as the presence at least one of the following: ROP in zone I, Stage 3 in zone I or II, or plus + For statistical analysis, SPSS 16.0 was used. For correlation, Kappa value was performed. Results: There was a high correlation between observers for the assessment of ROP stage (0.75; 0.54-0.88) plus disease (0.85; 0.71-0.92), and Ells criteria (0.89; 0.83-1.0). However, interobserver values were low for zone (0.41; 0.27-0.54) and telemedicine classification (0.43; 0.33-0.6). Conclusions: When evaluating telemedicine images by examiners with different levels of expertise in ROP, the Ells criteria gave the best correlation. In addition, stage of disease and plus disease have good correlation among observers. In contrast, the correlation between observers was low for zone and telemedicine classification.