Prevalence Patterns and Onset Prediction of High Myopia for Children and Adolescents in Southern China via Real-World Screening Data: Retrospective School-Based Study

Jieying Guan, Yingting Zhu, Qiuyue Hu, Shuyue Ma, Jingfeng Mu, Zhidong Li, Dong Fang, Xiaohua Zhuo, Haifei Guan, Qianhui Sun, Lin An, Shaochong Zhang, Peiwu Qin, Yehong ZhuoJournal of Medical Internet Research 2023 · DOI

TL;DR

This study leverages a massive real-world dataset of over 1.2 million children and adolescents in Southern China to map the prevalence of high myopia and develop a machine learning model for early prediction. Their random forest model, using age, uncorrected distance visual acuity, and spherical equivalents, achieved impressive accuracy (0.948) and AUC (0.975), offering a promising tool for proactive myopia management.

What Problem Does This Paper Address?

High myopia is a growing global concern, significantly increasing the risk of serious eye complications like retinal detachment and glaucoma later in life. While the prevalence of myopia is well-documented, understanding the specific patterns and, crucially, predicting who is at risk of developing high myopia early on remains a challenge. Current clinical practices often rely on single examinations, making it difficult for parents and clinicians to anticipate future progression. This paper aims to bridge this gap by analyzing large-scale, real-world screening data to identify prevalence trends and build a predictive model for high myopia onset in children and adolescents.

Methods

The researchers conducted a retrospective, school-based study using data from over 1.2 million participants across 13 cities in Guangdong Province, China, spanning from 2019 to 2021.

  • Data Collection: They utilized electronic vision screening data, which included standard metrics like age, sex, uncorrected distance visual acuity (UDVA), and noncycloplegic autorefraction (NCR) for spherical equivalent (SE).
  • Prevalence Analysis: They analyzed the prevalence of high myopia (defined as SE ≤ -6 D) across different age groups, school stages, and city economic development levels (categorized by GDP).
  • Machine Learning for Prediction: To predict high myopia onset, they employed a random forest algorithm. This algorithm was chosen for its robustness and ability to handle complex relationships between features.
  • Feature Selection: Initial analysis identified age, UDVA, and SE as the most predictive factors for high myopia onset. Other factors like mean arterial pressure (MAP) and GDP were found to reduce predictive performance and were excluded from the final model.
  • Model Training and Validation: The data was split into training and testing sets. A 5-fold cross-validation approach was used during training. The model's performance was further validated on an independent hold-out group from a different institution.
  • Evaluation Metrics: Standard metrics like accuracy, precision, recall, and the Area Under the Receiver Operating Characteristic Curve (AUC) were used to assess the model's efficacy.

Datasets

The study utilized a large-scale, real-world dataset comprising 1,285,609 participants aged 6-20 years from 13 cities in Guangdong Province, China, collected between 2019 and 2021. This dataset was curated and filtered from electronic vision screening records. No new benchmark datasets were introduced; instead, the paper leverages existing screening infrastructure.

Key Insights and Ideas

  • High Myopia is Prevalent and Growing: The study confirms a significant prevalence of high myopia in Southern China, with rates varying by age and school stage, highlighting the urgency of the issue.
  • Age is a Critical Predictor: The finding that age is the most significant risk factor for high myopia progression underscores the cumulative nature of environmental and educational pressures contributing to myopia development.
  • Real-World Data is Powerful: The successful application of a machine learning model to predict high myopia onset using readily available screening data (UDVA, SE, age) demonstrates the immense potential of leveraging existing public health infrastructure for proactive eye care.
  • Economic Factors Play a Role (with Nuance): While initial correlations between GDP and high myopia were observed, the relationship became weaker in later years, suggesting that other factors, possibly related to public health interventions or lifestyle changes, might be influencing trends.
  • Machine Learning Offers a Practical Solution: The development of a highly accurate random forest model provides a tangible tool that can be integrated into existing screening programs to flag at-risk individuals for further attention.

Key Results

  • Prevalence: The overall prevalence of high myopia was 4.48% in 2019, 4.88% in 2020, and 3.17% in 2021.
  • Age Trends: High myopia prevalence increased significantly from ages 11 to 17.
  • School Stage Trends: Prevalence was highest in high school students, followed by middle school, and then primary school. However, rates decreased across all school stages from 2019 to 2021.
  • Geographic Variation: Coastal and southern cities showed a higher proportion of high myopia.
  • Predictive Model Performance:
    • Accuracy: 0.948 on the test set.
  • AUC: 0.975 on the test set.
  • External Validation: Accuracy of 0.971 and AUC of 0.957.
  • Key Predictors: Age, uncorrected distance visual acuity (UDVA), and spherical equivalents (SE) were identified as the most important predictors.

Limitations and Future Work

  • Limitations:
    • Noncycloplegic Refraction: The study acknowledges that noncycloplegic refraction is not the gold standard for myopia assessment in children. Individuals flagged by the model were referred for cycloplegic examination, but the initial screening relied on noncycloplegic measurements.
  • Limited Risk Factors: The study could not incorporate crucial environmental and behavioral factors (e.g., time spent outdoors, screen time, diet) due to the nature of vision screening data.
  • Cross-Sectional Nature (for some analyses): While the study tracks data over three years, some analyses of prevalence patterns are based on snapshots within those years.
  • Future Work:
    • Smartphone App Development: The authors envision developing a smartphone app that allows users to input age, UDVA, and NCR to predict high myopia risk, making this technology more accessible.
  • Integration into Public Health: They suggest that efficient use of real-world data and predictive models can significantly contribute to the prevention and early diagnosis of high myopia.
  • Further Analysis: Future research could involve more in-depth analysis to incorporate additional factors and refine the methodology.

Summary updated

Abstract

BACKGROUND: Patients with high myopia have an increased lifetime risk of complications. The prevalence patterns of high myopia in children and adolescents in southern China are unclear. Early identification of high-risk individuals is critical for reducing the occurrence and development of high myopia and avoiding the resulting complications. OBJECTIVE: This study aimed to determine the prevalence of high myopia in children and adolescents in southern China via real-world screening data and to predict its onset by studying the risk factors for high myopia based on machine learning. METHODS: This retrospective school-based study was conducted in 13 cities with different gross domestic products in southern China. Through data acquisition and filtering, we analyzed the prevalence of high myopia and its association with age, school stage, gross domestic product, and risk factors. A random forest algorithm was used to predict high myopia among schoolchildren and then assessed in an independent hold-out group. RESULTS: There were 1,285,609 participants (mean age 11.80, SD 3.07, range 6-20 years), of whom 658,516 (51.2%) were male. The overall prevalence of high myopia was 4.48% (2019), 4.88% (2020), and 3.17% (2021), with an increasing trend from the age of 11 to 17 years. The rates of high myopia increased from elementary schools to high schools but decreased at all school stages from 2019 to 2021. The coastal and southern cities had a higher proportion of high myopia, with an overall prevalence between 2.60% and 5.83%. Age, uncorrected distance visual acuity, and spherical equivalents were predictive factors for high myopia onset in schoolchildren. The random forest algorithm achieved a high accuracy of 0.948. The area under the receiver operator characteristic curve (AUC) was 0.975. Both indicated sufficient model efficacy. The performance of the model was validated in an external test with high accuracy (0.971) and a high AUC (0.957). CONCLUSIONS: High myopia had a high incidence in Guangdong Province. Its onset in children and adolescents was well predicted with the random forest algorithm. Efficient use of real-world data can contribute to the prevention and early diagnosis of high myopia.
BibTeX
@article{guan2023prevalence,
  author = {Jieying Guan and Yingting Zhu and Qiuyue Hu and Shuyue Ma and Jingfeng Mu and Zhidong Li and Dong Fang and Xiaohua Zhuo and Haifei Guan and Qianhui Sun and Lin An and Shaochong Zhang and Peiwu Qin and Yehong Zhuo},
  title = {{Prevalence Patterns and Onset Prediction of High Myopia for Children and Adolescents in Southern China via Real-World Screening Data: Retrospective School-Based Study}},
  journal = {Journal of Medical Internet Research},
  year = {2023},
  doi = {10.2196/39507},
}
Figure