METHODS: We used pre-COVID-19 pulmonary tuberculosis (PTB) data (2007-2018) to fit SARIMA, Prophet, and LSTM models, assessing their ability to predict PTB incidence trends. These models were then applied to compare the predicted PTB incidence patterns with actual reported cases during the COVID-19 pandemic (2020-2023), using deviations between predicted and actual values to reflect the impact of COVID-19 countermeasures on PTB incidence.
RESULTS: Prior to the COVID-19 outbreak, PTB incidence in China exhibited a steady decline with strong seasonal fluctuations, characterized by two annual peaks-one in March and another in December. These seasonal trends persisted until 2019. During the COVID-19 pandemic, there was a significant reduction in PTB cases, with actual reported cases falling below the predicted values. The disruption in PTB incidence appears to be temporary, as 2023 data indicate a gradual return to pre-pandemic trends, though the incidence rate remains slightly lower than pre-COVID levels. Additionally, we compared the fitting and forecasting performance of the SARIMA, Prophet, and LSTM models using RMSE (root mean squared error), MAE (mean absolute error), and MAPE (mean absolute percentage error) indexes prior to the COVID-19 outbreak. We found that the Prophet model had the lowest values for all three indexes, demonstrating the best fitting and prediction performance.
CONCLUSIONS: The COVID-19 pandemic has had a temporary but significant impact on PTB incidence in China, leading to a reduction in reported cases during the pandemic. However, as pandemic control measures relax and the healthcare system stabilizes, PTB incidence patterns are expected to return to pre-COVID-19 levels. The Prophet model demonstrated the best predictive performance and proves to be a valuable tool for analyzing PTB trends and guiding public health planning in the post-pandemic era.