Dynamic linear model and SARIMA: a comparison of their forecasting performance in epidemiology

Area of Science:

• Epidemiology
• Biostatistics
• Public Health Surveillance

Background:

• Public health surveillance systems aim to forecast disease outbreaks accurately.
• Accurate forecasting of epidemiological time series is crucial for effective public health interventions.

Purpose of the Study:

• To evaluate and compare the performance of Seasonal Autoregressive Integrated Moving Average (SARIMA) and Dynamic Linear Models (DLM) for disease surveillance.
• To assess the utility of SARIMA and DLM in estimating case occurrences of notifiable diseases using national surveillance data.

Main Methods:

• Utilized reported cases of malaria and hepatitis A in the United States from January 1980 to June 1995.
• Compared SARIMA and DLM forecasting models based on residual analysis and qualitative aspects relevant to public health.

Main Results:

• Both SARIMA and DLM models demonstrated adequate performance for epidemiological surveillance.
• Forecasting models showed comparable results when substantial historical data (≥52 periods) were available.
• Dynamic Linear Models exhibited advantages in adaptability to various time series and ease of updating with new data.

Conclusions:

• SARIMA and DLM are comparable forecasting techniques for epidemiological time series when sufficient historical data exists.
• Dynamic Linear Models offer greater flexibility and are more easily adaptable for ongoing public health surveillance.
• The study highlights the importance of model selection based on data availability and operational needs in disease surveillance.