The Burden of Chronic Obstructive Pulmonary Disease in China Predicted by ARIMA and NNAR Models: a Comparative Study

Abstract

Background In China, a country with relatively serious burden of chronic obstructive pulmonary disease (COPD) , COPD is the third leading cause of death, and ranks third among all diseases in terms of disability adjusted life years (DALYs) . Effective prediction of the burden of COPD based on its prevalence, mortality and DALYs rate will provide theoretical support for the formulation of prevention and control measures. Objective To describe and analyze the burden of COPD and its temporal trends from 1990 to 2019, and to forecast the COPD burden between 2020 and 2024 in China, provide a basis for scientific prevention and control of COPD in China. Methods In December 2021, data about indicators measuring the burden of COPD in China from 1990—2019, including COPD prevalence, mortality and DALYs rates, were extracted from the Global Burden of Disease 2019. The average annual percentage change was used to measure the temporal trend of COPD prevalence, mortality and DALYs rates over the period. Autoregressive moving average (ARIMA) and neural network autoregressive (NNAR) models for COPD prevalence, mortality and DALYs rates were constructed based on data from 1990-2016 (training set) , and their predictive performance was tested using data from 2017—2019 (test set) . The relative error, mean absolute percentage error (MAPE) , mean absolute error (MAE) and root mean square error (RMSE) between predicted and actual values of these two models were calculated for comparing their goodness of fit and predictive performance. And the better model was used to estimate the COPD disease burden from 2020 to 2024. Results COPD prevalence, mortality and DALYs rates in China during 1990—2019: (1) The prevalence of COPD in the whole population increased from 2 344.40/100 000 to 3 175.37/100 000, with an average annual growth rate of 1.04%. And the average annual growth rates of COPD prevalence were 0.92% and 1.13% for men and women, respectively. (2) The mortality rate of COPD in the whole population decreased from 105.09/100 000 to 72.94/100 000, with an average annual rate of decrease of 1.29%. And the mortality rates in both men and women showed a decreasing trend, with average annual rates of decrease of 0.83% and 1.83%, respectively. (3) The rate of COPD DALYs in the whole Chinese population decreased from 2 206.55/100 000 to 1 400.71 /100 000, with an average annual rate of decrease of 1.56%. And the rates of DALYs in both men and women showed a decreasing trend, with average annual rates of decrease of 1.37% and 1.86%, respectively. The predicted values of the trends by both ARIMA and NNAR models were basically consistent with the actual values of trends, but the ARIMA model had smaller relative error, MAPE, MAE and RMSE, indicating that it may have better prediction accuracy. And by the ARIMA model, the predicted COPD prevalence in 2020—2024 was 3 229.77/100 000, 3 262.44/100 000, 3 292.38/100 000, 3 322.31/100 000, and 3 352.25/100 000, respectively; the predicted mortality rates were 74.50/100 000, 75.49/100 000, 76.11/100 000, 76.50/100 000, and 76.75/100 000, respectively; the predicted DALYs rates were 1 429.56/100 000, 1 452.07/100 000, 1 469.64/100 000, 1 483.35/100 000, and 1 494.05/100 000, respectively. Conclusion The trend of burden of COPD in China was predicted to be increased from 2020 to 2024 by the ARIMA model. Owing to the high goodness of fit and predictive accuracy demonstrated in the prediction, the ARIMA model may be used as a tool for predicting short-term burden of COPD.

Keywords

• lung diseases, obstructive|burden of illness|cost of illness|arima model|nnar model|prediction