Objective To analyze the trends of reported schistosomiasis cases in China from 2004 to 2020, and provide evidence for the improvement of schistosomiasis prevention and control strategies.

Methods Monthly reported schistosomiasis case data in China from 2004 to 2020 were downloaded from National Public Health Sciences Data Center. Joinpoint regression models were constructed to identify turning points in trends of monthly reported schistosomiasis cases. Autoregressive integrated moving average (ARIMA) models of monthly reported schistosomiasis case data in China were developed. Backpropagation (BP) and long short-term memory (LSTM) neural network models were constructed to predict trends of monthly reported schistosomiasis cases in China from 2021 to 2030. Model performance was evaluated by root mean square error (RMSE), mean absolute error (MAE) and the coefficient of determination (R²).

Results From 2004 to 2020, a total of 43 127 schistosomiasis cases were reported cumulatively in China. The Joinpoint model analysis revealed an average monthly decline of 1.48% in the reported schistosomiasis cases, the differences in monthly reported cases were significant average monthly percent change=−1.48%, 95% confidence interval: −2.01, −0.95), P<0.001. Three turning points of the reported case trend were identified in June 2005, March 2014, and June 2017. The optimal time series model was ARIMA (2,1,1)(1,1,1). The LSTM model outperformed the BP model in fitting historical data (RMSE=57.40, MAE=38.31, R²=0.90 vs. RMSE=70.57, MAE=46.81, R²=0.85). For 2021–2030 predictions, mean monthly cases were (16.35±24.30) by the BP model and (17.82±24.41) by the LSTM model, the difference was not significant (t=0.464, P=0.644).

Conclusion Reported schistosomiasis cases in China showed a fluctuating decline from 2004 to 2020. Combined model applications enable more accurate analysis on the incidence trend of the disease.