Improving surveillance of hand, foot, and mouth disease in China using ARIMAX model incorporating Baidu search index

Abstract

Introduction: We developed an optimized autoregressive integrated moving average with explanatory variable (ARIMAX) model to predict the incidence trend of hand, foot, and mouth disease (HFMD) using recently updated data available on the internet.

Methodology: We collected HFMD incidence and keyword search data (2011–September 2024) from the National Administration of Disease Prevention and Control and the Baidu Index website. We analyzed the data by selecting relevant keywords through Spearman rank correlation and cross-correlation analysis, constructed a comprehensive search index (CSI), and developed autoregressive integrated moving average (ARIMA) and ARIMAX models; and followed by evaluating their predictive performance.

Results: A total of 11 HFMD keywords constituted the CSI, which showed a strong correlation coefficient of 0.93 with monthly incidence. Both ARIMA(1, 0, 1)(0, 1, 1)(12) and ARIMA(1, 0, 1)(0, 1, 1)(12) + CSI (Lag = 0) passed the least squares test (p < 0.05) and the residual Ljung-Box test (p > 0.05). Their Akaike information criterion (AIC) values were 2333.52 and 2263.41; predicted mean absolute percentage error (MAPE) values were 5.92% and 2.05%; and predicted mean absolute error (MAE) values were 95,858.53 and 38,300.36, respectively.

Conclusions: The ARIMA(1 ,0, 1)(0, 1, 1)(12) + CSI (Lag = 0) model demonstrates outstanding predictive accuracy. When combining the CSI model with the ARIMAX model in China, the fluctuations in HFMD cases caused by public health emergencies (such as the coronavirus disease 2019 (COVID-19) pandemic) will be reflected in the predicted incidence data of this model.