Establishment and validation of a risk assessment model for myopia among Chinese primary school students during the COVID-19 pandemic: A lasso regression approach

Abstract

Purpose

To construct a risk assessment model for forecasting the likelihood of myopia in elementary school students.

Design

A cross-sectional study.

Methods

This study utilized convenient sampling and questionnaire survey to collect data from eligible elementary students and their parents during the coronavirus disease 2019 (COVID-19) pandemic period from March to December 2020. The data were divided into training and testing sets in a 7:3 ratio. Lasso regression was employed to screen variables for inclusion in the model to establish a generalized linear model, with a nomogram model as the final result.

Results

The study included 1139 elementary students, comprising 54.5 % male and 45.5 % female participants. A total of 37 variables were obtained, which were analyzed using lasso regression. Cross-validation revealed that the best lambda value was 0.04201788. Five variables affecting myopia were identified: three risk and two protective factors. The three risk factors were student age (OR = 1.32), family location (urban vs. rural, OR = 2.33), and parents' occupation (compared with farmer: worker, OR = 2.03; teacher, OR = 1.62; medical worker, OR = 5.64; self-employed, OR = 1.78; civil servant, OR = 1.65; company employee, OR = 1.45; service industries, OR = 3.38; and others, OR = 3.20). The two protective factors were eye distance score (OR = 0.83) and eye health exercise score (OR = 0.95). The model was verified and showed good accuracy with an AUC of 0.778 and Brier score of 0.122 in addition to satisfactory clinical effects.

Conclusions

The model effectively predicted the risk of myopia in elementary school students during the COVID-19 pandemic. Using this model, high-risk groups can be identified to provide a foundation for early intervention and follow-up, thereby reducing the incidence of myopia in this population.