Researchers in Guangzhou, China have used large-scale data analysis from electronic health records to develop an algorithm that can predict high myopia onset among Chinese school-aged children with clinically acceptable accuracy. They trained and validated the algorithm using a large real-world dataset.
This study analyzed 687,063 longitudinal electronic medical records from the largest ophthalmic centers in China, and developed and validated individualized prediction models for myopia prediction based on machine learning techniques. Researchers believe that their algorithm can predict spherical equivalent and onset of high myopia at 18 years of age at a clinically acceptable accuracy as early as 10 years old. However, the accuracy of the prediction is reduced when the targeted prediction time increases. Still, the 95% predicted diopter of refraction was within 0.5D to 0.8D of the true value at eight years.
Large-scale, long-term electronic medical records and machine learning algorithms provide unique opportunities for the development of prediction models for progressive diseases. School-age myopia is the most prevalent eye disease in the Chinese population and the researchers note that their work can help change current approaches used to manage school myopia by pediatric and general ophthalmologists as well as general practitioners and optometrists, who are often the first point of care.
|Lin H, Long E, Ding X, et al. Prediction of myopia development among Chinese school-aged children using refraction data from electronic medical records: A retrospective, multicentre machine learning study. PLOS Medicine. November 6, 2018. [Epub ahead of print].|