[The performance of visual photoscreening for Chinese preschool children aged 4 to 5 years]

Objective: To assess the accuracy of photoscreening for detecting refractive amblyopia risk factors (ARFs) in Chinese preschool children aged 4 to 5 years. Methods: A cross-sectional study. Comprehensive ocular examinations were conducted for preschool children in Nanjing, China from September to December, 2016. Photoscreening (Plusoptix A12C) was applied for refractive screening without cycloplegia. Voluntary children and children suspected of eyes abnormalities received cycloplegic retinoscopy (CR). Results of photoscreening and CR were compared using Wilcoxon signed rank test, and Bland-Altman plot were used to assess the agreement between the photoscreener and CR. According to the updated preschool vision screening guidelines from American Association for Pediatric Ophthalmology and Strabismus (AAPOS) in 2013, CR was adopted for identifying children with ARFs, which was considered as a golden standard. Based on the golden standard, the accuracy of 5 sets of referral criteria (including sensitivity standard, Matta/Silbert standard, AAPOS2013 standard, Alaska Blind Child Discovery standard, specificity standard) for photoscreener were tested. Receiver operating characteristics curves were constructed applied to evaluate the quality of the photoscreener in refractive ARFs detection and to find probably the best cut-off points. Results: In total, 1 986 children [mean age, (4.57±0.29) years] received comprehensive examinations, including 1 084 boys and 902 girls. The test ability of photoscreening was 99.04% (1 967/1 986) in the preschool children, and 96.56%(1 827/1 892) of the children got a reliable result within three screening attempts. In 538 children who had data of CR, refractive error of one child exceeded the upper limit of the photoscreener value setting, which was directly categorized as hyperopia, so in the end, 537 children were included to analyze the comparison between the two tests. The measurement values of photoscreening were lower than those of CR in sphere, cylinder and spherical equivalent [(0.75 (0.50, 1.25) D vs. 1.25 (1.00, 1.75) D, Z=-10.36, P<0.01; -0.50 (-0.75, -0.25) D vs. -0.25 (-0.75, 0.00) D, Z=-11.10, P<0.01; 0.63 (0.38, 0.88) D vs. 1.00 (0.75, 1.50) D, Z=-13.33, P<0.01]. The 95% limit of agreement cover rates between the photoscreening and CR in sphere, cylinder and spherical equivalent was 96.28% (517/537), 95.34% (512/537) and 96.65% (519/537), respectively. Based on the golden standard, 47 (8.74%) children had refractive ARFs, and the range of sensitivity, specificity, Youden index, positive predictive values and negative predictive values for detecting refractive ARFs of the 5 common used referral criteria was from 63.83% to 97.87%, from 53.36% to 97.56%, from 0.51 to 0.80, from 16.73% to 74.51% and from 96.57% to 99.62%, respectively. Considering particular refractive ARFs on the basis of the receiver operating characteristic curves, the optimal cut-off point for astigmatism was set at 1.38 D. Conclusion: Photoscreening could be an applicable tool to detect refractive ARFs in preschool children. (Chin J Ophthalmol, 2020, 56: 189-196).