Green Space Morphology and School Myopia in China

Key Points Question What is the association between green space morphology and childhood myopia? Findings In this cohort study including 138 735 students from grades 1 to 4, using high-resolution satellite images, the proportion, aggregation, and connectivity of green space were found to be associated with 2-year changes in school myopia rate. Principal component analysis further supported the finding that overall green space morphology is inversely associated with prevalent and incident myopia. Meaning Careful planning of green space development may be a strategy for myopia prevention.

eMethods.eFigure 1. Heatmap of the Correlation Matrix of Landscape Metrics eFigure 2. Association Between Myopia and Largest Patch Index (LPI) eFigure 3. Association Between Myopia and Area-Weighted Mean of Greenness Area (AREA_AM) eFigure 4. Association Between Myopia and Area-Weighted Mean Shape Index (SHAPE_AM) eFigure 5. Association Between Myopia and Cohesion Index (COHESION) eFigure 6. Association Between Myopia and Patch Density (PD) eFigure 7. Association Between Myopia and Proximity Index (PROX_AM) eTable 1. Characteristics of the Landscape Metrics and Population of the Studied Schools eTable 2. Sensitivity Analyses of the Association Between Myopia-Related Green Space Morphology Index and School-Level Increase in Myopia Rate Using the Linear Regression Model eTable 3. Association Between Myopia-Related Green Space Morphology Index and School-Level Increase in Myopia Rate Using the Linear Regression Model eTable 4. Association Between Myopia-Related Green Space Morphology Index and Myopia Incidence at the Individual-Level Among Students Without Myopia Who Were Followed Using the Mixed-Effects Logistic Regression Model eTable 5. Association Between Myopia-Related Green Space Morphology Index and Myopia Incidence at the Individual-Level Among the Subset Without Myopia Who Completed Questionnaire Using the Mixed-Effects Logistic Regression Model eReferences.
This supplementary material has been provided by the authors to give readers additional information about their work.eMethods.

Campus area detection and buffer zone creation
The boundary of each school campus was manually sketched from the satellite images, and the school area was automatically calculated by ArcGIS (version 10.6, Esri, USA).We included the 110 school campuses (a mean area of 21,800 m 2 ) and surrounding 500-meter buffer zones (an approximately 7-minute walking distance, 1 extracted by FME [Feature Manipulate Engine, Safe Software, Canada)], for the calculation of landscape metrics.The area and distance of green patches in a search radius of 500 meters to reflect the size and proximity of green patches in average, by considering patches' proportional abundance.

Other covariates
Potential confounding factors identified by previous studies, including sociodemographic data, age (of each participant and mean age at each school) and sex (of each participant and percentage of boys at each school), 2 was obtained from school records and included in our regression models.Student density was defined as the number of students divided by the total land area of each school in meters squared.
School socioeconomic ranking, reflecting the proportion of children enrolled from families with higher socioeconomic status, [3][4][5][6] was obtained from the Guangdong Provincial Department of Education. 7 better estimate effects caused by potential confounding factors at the individual level, we randomly sampled 28 among the 110 schools, and then sampled 25% of students per grade in each school, for a total subset of 8,400 students.][10][11] The average outdoor activity time during the school semester days (TSchool) was calculated using the time spent outdoors on weekdays (Twd) and on weekends (Twe): Tschool = (Twd × 5+Twe × 2) ÷ 7.
Average outdoor activity time during the weekends (Twd) was used as a proxy for time spent outdoors during the summer and winter holidays (3 months per school year).
eFigure 4. Association Between Myopia and Area-Weighted Mean Shape Index (SHAPE_AM) A higher value of SHAPE_AM indicates a more irregular shape of green space (value range: SHAPE_AM≥1).SHAPE_AM was not associated with the changes in myopia rate (-0.08, 95% CI: -0.2 to 0.04, p=.06) eFigure 5. Association Between Myopia and Cohesion Index (COHESION) A higher value of COHESION indicates a more connected green space pattern (value range: 0<COHESION<100).The increase in COHESION was associated with a lower change in school myopia rate (-0.5%, 95% CI: -0.8% to -0.2%, p=.004).eFigure 6. Association Between Myopia and Patch Density (PD) A higher value of PD indicates a more fragmented area of green space (value range: PD≥0).PD was positively associated with the change in school myopia rate (0.03%, 95% CI: -0.003% to -0.05%, p=.03).
eFigure 7. Association Between Myopia and Proximity Index (PROX_AM) A higher area-weighted mean of PROX_AM indicates a smaller distance with a less fragmented distribution of green patches within a campus buffer zone (value range: PROX_AM≥0).A 10% increase in PROX_AM was associated with a smaller increase in myopia rate (-0.3%, 95% CI: -0.6% to -1.0%, p=.007).

eFigure 1 .eFigure 2 .
Heatmap of the Correlation Matrix of Landscape Metrics PROX_AM: area-weighted mean of proximity index; AI: aggregation index; PD: patch density; COHESION: cohesion index; PLAND: percentage of landscape; LPI: largest patch index; AREA_AM: area-weighted mean greenness area; SHAPE_AM: area-weighted mean patch shape index.Association Between Myopia and Largest Patch Index (LPI)

eTable 1 .
Characteristics of the Landscape Metrics and Population of the Studied Schools Sensitivity Analyses of the Association Between Myopia-Related Green Space Morphology Index and School-Level Increase in Myopia Rate Using the Linear Regression Model Association Between Myopia-Related Green Space Morphology Index and School-Level Increase in Myopia Rate Using the Linear Regression Model Association Between Myopia-Related Green Space Morphology Index and Myopia Incidence at the Individual-Level Among Students Without Myopia Who Were Followed Using the Mixed-Effects Logistic Regression Model Association Between Myopia-Related Green Space Morphology Index and Myopia Incidence at the Individual-Level Among the Subset Without Myopia Who Completed Questionnaire Using the Mixed-Effects Logistic Regression Model eTable 2. eTable 3. eTable 5.