基于ARIMA与NNAR模型的中国食管癌疾病负担预测

  目的  探索中国食管癌疾病负担时间序列特征，并进行预测。   方法  收集1990—2019年中国食管癌发病率、死亡率、伤残调整寿命年(disability adjusted life year, DALY)等疾病负担数据，基于1990—2016年数据建立自回归移动平均(autoregressive integrated moving average, ARIMA)模型、神经网络自回归(neural network autoregression, NNAR)模型，通过平均误差率(modulation error ratio, MER)、平均绝对百分比误差(mean absolute percentage error, MAPE)、平均绝对误差(mean absolute error, MAE)和均方根误差(root mean squared error, RMSE)比较2017—2019年预测数据与实际数据以验证模型预测性能，并预测2020—2024年食管癌疾病负担。   结果  1990—2019年中国食管癌疾病负担整体呈波动上升趋势，发病率上升33.26%，死亡率上升21.26%，DALY率上升6.66%。ARIMA模型与NNAR模型的预测值和实际值动态趋势基本一致，选择其中更优模型预测得到2020—2024年中国食管癌发病率分别为20.375/10万、21.057/10万、21.380/10万、21.341/10万和21.080/10万；死亡率分别为18.834/10万、19.647/10万、20.407/10万、20.889/10万和20.988/10万。DALY率分别为418.192/10万、431.123/10万、442.780/10万、452.376/10万和459.358/10万。   结论  中国食管癌疾病负担在2020—2024年将上升。NNAR模型在拟合中国食管癌疾病负担应用中具有良好预测性能与精度，可为疾病负担短期预测提供借鉴方法。

Prediction of disease burden of esophageal cancer in China based on ARIMA and NNAR models

  Objective  To explore the time series characteristics of the disease burden of esophageal cancer in China and predict the disease burden of esophageal cancer.   Methods  The incidence, mortality, and disability adjusted life year (DALY) of esophageal cancer in China from 1990 to 2019 were collected. autoregressive integrated moving average (ARIMA) model and neural network autoregression (NNAR) model were established based on the data from 1990 to 2016, and model prediction performance was verified by comparing 2017-2019 forecast data with actual data through mean absolute percentage error (MAPE), modulation error ratio (MER), mean absolute error (MAE) and root mean squared error (RMSE). The better model was applied to predict the disease burden of esophageal cancer from 2020 to 2024.   Results  From 1990 to 2019, the overall disease burden of esophageal cancer in China showed a fluctuating upward trend, with the incidence rate rising by 33.26%, the mortality rate rising by 21.26%, and the DALY rate rising by 6.66%. The predicted values of disease burden by ARIMA model and NNAR model were basically consistent with the actual dynamic trend. The incidence rate of esophageal cancer in China from 2020 to 2024 would be 20.375/100 000, 21.057/100 000, 21.380/100 000, 21.341/100 000, 21.080/100 000, and mortality rate would be 18.834/100 000, 19.647/100 000, 20.407/100 000, 20.889/100 000, 20.988/100 000, and the DALY rate would be 418.192/100 000, 431.123/100 000, 442.780/100 000, 452.376/100 000, and 459.358/100 000.   Conclusions  The disease burden of esophageal cancer in China will increase slightly from 2020 to 2024. The NNAR model demonstrates good prediction performance and accuracy in simulating the disease burden of esophageal cancer in China, and provides a reference method for short-term prediction of the disease burden.