Estimating the Temporal Epidemiological Trends of Tuberculosis Incidence by Using an Advanced Theta Method.

We aimed to assess the temporal epidemiological trends in tuberculosis (TB) by use of an advanced Theta method. The TB incidence data from Tianjin, Heilongjiang, Hubei, and Guangxi provinces in China, spanning January 2005 to December 2019, were extracted. We then constructed and compared various modeling approaches, including the seasonal autoregressive integrated moving average (SARIMA) model, the Theta model, the standard Theta model (STM), the dynamic optimized Theta model (DOTM), the dynamic standard Theta model (DSTM), and the optimized Theta model (OTM). During 2005-2019, these four provinces recorded a total of 2,068,399 TB cases. Analyses indicated that TB exhibited seasonality, with prominent peaks in spring and winter, and a slight downward trend was seen in incidence. In the Tianjin forecast, the OTM consistently demonstrated superior performance with the lowest values across metrics, including mean absolute deviation (0.159), mean absolute percentage error (7.032), root mean square error (0.21), mean error rate (0.068), and root mean square percentage error (0.093), compared with those of SARIMA (0.397, 16.654, 0.436, 0.169, and 0.179, respectively), Theta (0.166, 7.248, 0.231, 0.071, and 0.102, respectively), DOTM (0.169, 7.341, 0.234, 0.072, and 0.102, respectively), DSTM (0.169, 7.532, 0.203, 0.072, and 0.092, respectively), and STM (0.165, 7.218, 0.231, 0.070, and 0.101, respectively). Similar results were also observed in the other provinces, emphasizing the effectiveness of the OTM in estimating TB trends. Thus, the OTM may serve as a beneficial and effective tool for estimating the temporal epidemiological trends of TB.