Table 1 General characteristics of the 6,321 participants

Table 2 Comparison of accuracy of diff erent screening tests for children of all ages

Table 3 The best cutoff points of combination measure for children of all ages

More and more attention was paid to myopia of children due to the high incidence and its increasing tendency^{(1,6-9,11,12)}. Some potential associations between myopia and major age-related eye diseases were reported, such as nuclear cataract, posterior subcapsular cataract and primary open-angle glaucoma⁽¹³⁾. So it is critical to detect children with true myopia or with high risk of myopia in early age accurately, which would lead to an early treatment and help to relieve the long-term damage. However, the result of UCVA was subject to be affected by environmental factors, such as light intensity and reminding from people around in reality, and often could not detect those with myopia accurately, which means quite a few of missed diagnosis. Cycloplegic refraction with retinoscopy and subjective  refraction were usually used in most of professional hospitals, but cycloplegic refraction with retinoscopy needs high professional skills and rich experience, while subjective refraction without cycloplegia is not accurate enough for children, so both of these two methods are not suitable to be used during large-scale screening. By contrast, NCAR test is more objective, therefore, it is necessary to combine NCAR test in order to increase the accuracy of myopia screening^(14-16). For controlling and even lowering the incidence of myopia, a program on establishing individual archives of ocular refractive development of children, which was included in the agenda of local government to enhance the public health system, was launched in Jiading District in 2011 and completed at the end of 2013. According to our study on this program, the SP of UCVA combined with NCAR was 85.0%, which was much higher than that of the two tests alone. Meanwhile, the SE of the combination kept similar to that of UCVA and NCAR alone. In other words, the combination measure achieved an increased accuracy, which was in accordance with some researches before^(10,17,18). Furthermore, our population-based study had a larger sample size, which promised a reliable result. 
In addition, the accuracy for children aged between 9 and 12 was relatively higher than that of other ages, indicating that the combination was more practical for children in this age group. The accuracy for those aged 6 and 7 was much lower than that for other ages, due to the stronger refraction adjustment ability and poor cooperation of younger children⁽¹⁹⁾. Furthermore, we explored that the best cutoff points were age dependent for the measure of UCVA combined with NCAR. For children aged from 6 to 9 and 15, the best cutoff points were the same. Similarly, children aged 12 and 13 also shared the same best cutoff point. For other ages, the best cutoff points varied. With the progress of dynamic monitoring of children refractive development in the future, we would further examine the criteria and the associated accuracy for myopia screening. To find the risk factors of myopia and deliver targeted intervention after vision screening is important, some researches have reported the potential risk factors, such as lack of outdoor activities, heavy homework, wrong posture of reading and writing, long-time playing electronic products, and so on^(8,9). In the following study, we would explore the myopia-related risk factors and feasible intervention based on the dynamic screening.
In conclusion, with an increased accuracy, the combination of UCVA and NCAR was more feasible to apply in large-scale myopia screening. The referral criteria of myopia for primary and junior middle school children should be age-determined.