Abstract
ABSTRACT
Background and Aims
Myopia is an increasingly important public health concern worldwide, particularly among young adults. While previous research has applied machine learning techniques to identify myopia‐related risk factors, evidence based on traditional statistical models using the same datasets remains limited. This study aimed to estimate the prevalence of self‐reported myopia and examine its associated factors among undergraduate students in the northern region of Bangladesh.
Methods
This secondary analysis utilized cross‐sectional data originally collected between May 17 and June 17, 2024, from 514 undergraduate students enrolled in various academic disciplines. The original study employed a two‐stage sampling technique and collected information using a structured questionnaire covering sociodemographic characteristics, health‐related factors, and lifestyle behaviors. Descriptive statistics were used to estimate prevalence, and binary logistic regression analysis was performed to identify factors associated with myopia. Statistical significance was set at
p
Results
The prevalence of self‐reported myopia among the participants was 43.2%. Factors significantly associated with myopia included a family history of myopia (OR: 5.12; 95% CI: 3.22–8.13), premature birth (OR: 2.52; 95% CI: 1.25–6.03), visual stress (OR: 3.65; 95% CI: 2.37–5.12), uncertain response to visual stress (OR: 4.15; 95% CI: 2.25–6.76) and steroid use (OR: 2.11; 95% CI: 1.13–4.33). The logistic regression model demonstrated good discriminative ability, with an area under the receiver operating characteristic curve (AUC) of 0.836.
Conclusion
This secondary data analysis indicates a high prevalence of myopia among undergraduate students in the northern region of Bangladesh. Several modifiable and non‐modifiable factors, particularly family history, visual stress, and steroid use, were significantly associated with myopia. These findings underscore the importance of targeted public health strategies and support the use of traditional regression models as a transparent and interpretable alternative to machine learning approaches in epidemiological research.