Predicting hepatitis B monthly incidence rates using weighted Markov chains and time series methods

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Hepatitis B (HB) poses a significant global mortality risk.
• Accurate disease trend prediction is essential for effective health policy and prevention strategies.
• This study focuses on predicting monthly HB incidence rates.

Purpose of the Study:

• To apply and compare three distinct statistical models for forecasting monthly Hepatitis B incidence.
• To evaluate the predictive accuracy of Weighted Markov Chain (WMC), Holt Exponential Smoothing (HES), and SARIMA models.
• To identify the most effective method for predicting HB trends.

Main Methods:

• A historical cohort study utilizing HB incidence data from Hamadan Province, Iran (2004-2012).
• Application of Weighted Markov Chain (WMC) based on Markov chain theory.
• Implementation of two time series models: Holt Exponential Smoothing (HES) and SARIMA.
• Comparison of model performance based on corrected percentages of predicted incidence rates.

Main Results:

• Monthly incidence rates were categorized into two clusters for Markov chain analysis.
• The Weighted Markov Chain (WMC) achieved 100% prediction accuracy for the first cluster and 0% for the second.
• Holt Exponential Smoothing (HES) showed 84% and 67% accuracy for the two clusters, respectively.
• SARIMA models yielded 79% and 47% accuracy for the clusters.

Conclusions:

• The overall incidence rate of Hepatitis B Virus (HBV) is projected to decline.
• The Holt Exponential Smoothing (HES) model provided the most accurate predictions, considering seasonality and non-stationarity.
• HES is recommended for predicting HB incidence rates due to its superior performance.