A study on the burden prediction of glaucoma in China based on ARIMA and NNAR models

Objective  To analyze the disease burden of glaucoma in China and its evolution trend from 1991 to 2019 based on an autoregressive integrated moving average (ARIMA) model and a neural network autoregressive (NNAR) model, and predict the future change trend of glaucoma disease burden in China from 2020 to 2050 using the better model.

Methods  Data on the age-standardized prevalence of glaucoma, years lost due to disability (YLD) rate, disability adjusted life year (DALY) rate in China from 1991 to 2019 were extracted from the 2019 Global Burden of Disease Database. Based on those data, an ARIMA model and an NNAR model were constructed. The performance of the two models in terms of fitting and prediction was evaluated and compared using metrics including the mean absolute percentage error (MAPE), mean absolute error (MAE), and root mean square error (RMSE). The superior model was selected to forecast the disease burden of glaucoma in China from 2020 to 2050.

Results  From 1991 to 2019, the age-standardized prevalence of glaucoma in China across all age groups increased from 45.91 per 100 000 population to 94.08 per 100 000 population, with an average annual growth rate of 3.62%. Additionally, the annual average growth rates of glaucoma age-standardized prevalence for males and females were 3.24% and 4.03%, respectively. The years lost due to YLD rate and DALY rate of glaucoma in China across all age groups increased from 5.20 per 100 000 population to 7.91 per 100 000 population, with an average annual change rate of 1.80%. The age-standardized YLD rate and DALY rate of glaucoma in males and females showed an increasing trend by 1.50% and 2.16% per year on average, respectively. The trends predicted by both the ARIMA and NNAR models were essentially in line with the actual observed trends. Moreover, the MAPE, MAE, and RMSE values of the ARIMA model were all lower than those of the NNAR model, suggesting that the ARIMA model provided better prediction performance and more accurate results. According to the ARIMA model prediction, the age-standardized prevalence rate, YLD rate, and DALY rate of glaucoma in China in 2020 were 128.70 per 100 000 people, 10.63 per 100 000 people, and 10.63 per 100 000 people, respectively, which were projected to increase to 942.17 per 100 000 people, 10.87 per 100 000 people, and 10.87 per 100 000 people, respectively, by 2050.

Conclusions  Trend of the burden of glaucoma in China is predicted to increase from 2020 to 2050. The ARIMA model shows higher accuracy in forecasting compared to the NNAR model and can provide an effective reference for predicting the burden of glaucoma.