Estimated prevalence of syphilis in China based on Bayesian structural time series model

  Objective  To evaluate the performance of Bayesian structural time series (BSTS) model for estimating the prevalence of syphilis in China.

  Methods  The incidence data of syphilis in China from January 2005 to August 2022 were collected. The trend and seasonal components in the syphilis monthly incidence data were analyzed by using software Eviews10. The model was constructed by using software RStudio, and in the model development, the incidence data from January 2005 to December 2021 were used as the training set to fit the BSTS model, whereas the incidence data from January to August 2022 were used as the test set to verify the prediction potential of the model. Subsequently, the prediction accuracy of BSTS model was compared with that of autoregressive integrated moving average (ARIMA) model. The test levels for all statistical analyses were designated as \alpha =0.05.

  Results  Overall, the incidence of syphilis in China showed an upward trend and had a periodicity and seasonality with the trough during January-February and the peak during July-August. On the training and testing sets, the average absolute percentage error of BSTS model (4.95%, 5.73%) was lower than those of ARIMA model (5.44%, 6.52%). Also, it was found that the mean absolute error, root mean square error, and root mean square percentage error of BSTS model were lower than those of ARIMA model. The robustness test indicated the same result. The predicted total case number of syphilis was 719 600 (95%CI: 605 295–826 086) from September 2022 to December 2023 by BSTS model, with a monthly average of 44 975 cases (95%CI: 37 831–51 631).

  Conclusion  In China, syphilis is a seasonal disease with high incidence. The BSTS model can be used for accurately assessing the incidence of syphilis and providing technical support for the accurate prevention and control of syphilis.