Grade-related gradients and between-school heterogeneity in myopia and astigmatism: a multilevel analysis of 66,977 students in Zhangzhou, China

Myopia and other refractive errors have become major public health concerns among school-aged children and adolescents in China. Large-scale school vision screening data provide an important opportunity to characterize prevalence patterns across educational stages and to assess between-school variation. Such evidence is useful for identifying grades and schools with higher observed burden and for informing school-based surveillance, follow-up, and resource allocation. This study aimed to describe grade-specific prevalence patterns of myopia, high myopia, and astigmatism using large-scale routine school vision screening data, and to quantify between-school heterogeneity through multilevel models, thereby providing evidence for stratified surveillance and resource allocation. This cross-sectional study used de-identified routine school vision screening data from municipal schools in Zhangzhou, China, in 2025. After data cleaning, valid refractive data from 66,977 students across 42 schools were included, covering kindergarten, primary school, middle school, high school, and secondary vocational school. Spherical equivalent was calculated as sphere + 0.5 × cylinder. To harmonize phenotype definitions, a worse-eye approach was adopted: for myopia-related outcomes, the eye with the more negative spherical equivalent was selected, whereas for astigmatism-related outcomes, the eye with the higher absolute cylinder was used. Myopia was defined as worse-eye spherical equivalent ≤ -0.50 D, high myopia as ≤ -6.00 D, and astigmatism as worse-eye |cylinder| ≥ 0.75 D. Astigmatic axis composition was classified as with-the-rule, against-the-rule, or oblique. Prevalence and 95% confidence intervals were estimated using the Wilson method. Two-level logistic regression models with students nested within schools were constructed for primary, middle, and high school stages, and between-school clustering was quantified using the intraclass correlation coefficient and median odds ratio. Refractive outcomes showed marked cross-sectional differences across educational stages. The observed prevalence of worse-eye myopia was 15.6% in kindergarten, 52.3% in primary school, 82.7% in middle school, 92.0% in high school, and 81.8% in vocational school. The observed prevalence of worse-eye high myopia was 0.0%, 1.5%, 9.3%, 19.3%, and 12.7%, respectively. The overall prevalence of worse-eye astigmatism was 73.3%, with the lowest prevalence in primary school (56.3%) and rates exceeding 75% in middle school and above. Cross-sectional grade-sequence analyses showed that myopia was highest in upper primary school and above, exceeding 50% by grade 4, whereas high myopia was most prominent in high school and approached 20% by grade 12. Among participants with astigmatism, the proportion of with-the-rule astigmatism was lower in higher grades, the proportion of against-the-rule astigmatism was highest in high school, and the proportion of oblique astigmatism was relatively similar across grades. Multilevel models showed that between-school heterogeneity varied by outcome and educational stage, with more prominent clustering for astigmatism and high myopia in primary school and clear between-school differences across all outcomes in high school. In this routine school screening dataset, myopia, high myopia, and astigmatism showed clear grade-specific patterns and measurable between-school heterogeneity. Primary school showed the steepest observed grade gradient for myopia, whereas high school had the highest observed burden of high myopia. School type and boarding status, as limited administrative indicators, accounted for only part of the between-school variation. Given the use of non-cycloplegic refraction, absolute prevalence estimates - particularly for myopia defined at -0.50 D - should be interpreted cautiously as screening-based observed prevalence. These findings may support more granular exposure assessment and help prioritize eye-health surveillance and resource allocation for schools with higher observed burden.