Objective To understand the epidemiological and etiological characteristics of hand foot and mouth disease (HFMD) in Yunnan province during 2010−2024 and provide evidence for the prevention and control of HFMD.

Methods Descriptive analyses were performed to analyze the temporal, spatial and population distributions and etiological characteristics of HFMD. Spatial autocorrelation analysis was conducted to evaluate the spatiotemporal clustering of HFMD incidence. Seasonal autoregressive integrated moving average (SARIMA), long short-term memory (LSTM), and SARIMA-LSTM hybrid models were constructed using monthly HFMD incidence data in Yunnan from 2010 to 2023. Model performance was internally evaluated by using monthly incidence data in 2024.

Results The annual average incidence of HFMD in Yunnan during 2010−2024 was 149.55/100 000. The disease mainly affected children aged 1～5 years. Significant spatial autocorrelation was observed in most years. In the training set, the SARIMA model achieved root mean square error（RMSE）of 3.79, mean absolute error（MAE）of 2.62, and mean absolute percentage error（MAPE）of 44.80%, while in the test set, the mean error was −4.27, RMSE was 6.14, MAE was 4.57, and MAPE was 81.59%. The LSTM model with a 9-month lookback had the best performance (RMSE=5.38, MAE=3.90, MAPE=64.00%, R²=0.37). The SARIMA-LSTM hybrid model using stacked ensemble demonstrated superior performance (MAE=2.72, RMSE=3.55, MAPE=44.86%).

Conclusion The incidence of HFMD was high in Yunnan during 2010-2024, with obvious area specific clustering. The SARIMA-LSTM hybrid model effectively fitted overall trends and seasonal fluctuations of monthly HFMD incidence.