Time Series Modelling of Syphilis Incidence in China from 2005 to 2012

Abstract

BackgroundThe infection rate of syphilis in China has increased dramatically in recent decades, becoming a serious public health concern. Early prediction of syphilis is therefore of great importance for heath planning and management.MethodsIn this paper, we analyzed surveillance time series data for primary, secondary, tertiary, congenital and latent syphilis in mainland China from 2005 to 2012. Seasonality and long-term trend were explored with decomposition methods. Autoregressive integrated moving average (ARIMA) was used to fit a univariate time series model of syphilis incidence. A separate multi-variable time series for each syphilis type was also tested using an autoregressive integrated moving average model with exogenous variables (ARIMAX).ResultsThe syphilis incidence rates have increased three-fold from 2005 to 2012. All syphilis time series showed strong seasonality and increasing long-term trend. Both ARIMA and ARIMAX models fitted and estimated syphilis incidence well. All univariate time series showed highest goodness-of-fit results with the ARIMA(0,0,1)×(0,1,1) model.ConclusionTime series analysis was an effective tool for modelling the historical and future incidence of syphilis in China. The ARIMAX model showed superior performance than the ARIMA model for the modelling of syphilis incidence. Time series correlations existed between the models for primary, secondary, tertiary, congenital and latent syphilis.

Study of surveillance data for class B notifiable disease in China from 2005 to 2014

“…The decomposition methods have been reported in previous studies. 6,8 Decomposition breaks down the underlying patterns of time series into seasonality and longterm trends. The seasonal trend in the infection series can be expressed using seasonal indices.…”

A comparison of three data mining time series models in prediction of monthly brucellosis surveillance data

“…The estimated values of monthly number of human brucellosis obtained from MARS model did not show a very good concordance with the observed values in some counties including Malayer and Razan. This is important because the bias observed between the estimated and actual values may lead to surveillance artifacts (Zhang et al, ). The three models were consistently biased at the peaks of time series.…”

Time Prediction Models for Echinococcosis Based on Gray System Theory and Epidemic Dynamics

“…The subscripted letter “ s ” shows the seasonal period length. The ARIMA modeling procedure introduced by Box and Jenkins, consists of three iterative steps: identification, estimation, and diagnostic checking [32]. To test the validity of FGM(1,1), an autoregressive integrated moving average is also used to fit the univariate time series model of human Echinococcosis cases in Xinjiang.…”