Forecasting tuberculosis epidemics using an autoregressive fractionally integrated moving average model: a 17-year...

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Tuberculosis (TB) poses a significant public health challenge in Henan, China.
• Accurate forecasting is crucial for effective TB prevention and control strategies.
• Traditional autoregressive integrated moving average (ARIMA) models may not fully capture long-term data dependencies.

Purpose of the Study:

• To evaluate the autoregressive fractionally integrated moving average (ARFIMA) model for TB forecasting in Henan.
• To compare the predictive accuracy of ARFIMA against ARIMA models.
• To improve TB forecasting by better modeling long-range dependencies and seasonal patterns.

Main Methods:

• Monthly TB incidence data from Henan (January 2007-May 2023) were analyzed.
• Data were split into training (2007-2022) and testing (2022-2023) sets.
• ARIMA and ARFIMA models were developed and their predictive accuracy assessed using multiple error metrics.

Main Results:

• TB incidence in Henan showed an annual reduction of 5.83% with seasonal variations.
• The ARFIMA (2,0,1)(0,0.38,1)₁₂ model demonstrated superior performance over ARIMA (2,0,1)(0,1,1)₁₂.
• ARFIMA yielded lower prediction errors across all evaluated metrics, indicating better capture of long-term dependencies and seasonality.

Conclusions:

• Tuberculosis incidence in Henan exhibits a downward trend with seasonal patterns.
• The ARFIMA model provides more accurate TB forecasts than ARIMA, crucial for public health management.
• Continuous application of ARFIMA is recommended for guiding TB control interventions and timely responses.