Sample size calculations for disease freedom and prevalence estimation surveys

Area of Science:

• Epidemiology
• Biostatistics
• Veterinary Public Health

Background:

• Estimating disease prevalence in populations, especially with clustered data, presents statistical challenges.
• Existing methods may not adequately account for zero prevalence, variability among clusters, or uncertainty in diagnostic tests.
• Accurate sample size calculations are crucial for effective public health and animal health surveillance.

Purpose of the Study:

• To develop a Bayesian approach for sample size calculations in disease prevalence estimation and disease freedom surveys.
• To provide a flexible model that handles cluster-level prevalence, including zero prevalence and variability.
• To incorporate uncertainty in diagnostic test accuracy and within-cluster prevalences.

Main Methods:

• Developed a Bayesian hierarchical model for cluster-level prevalence estimation.
• Utilized a predictive approach for sample size determination in surveys.
• Applied the method to two survey types: disease freedom and prevalence estimation surveys.
• Implementation suggested using the emBedBUGS library in Splus/R with WinBUGS.

Main Results:

• The Bayesian approach provides a robust framework for sample size calculations in complex survey designs.
• The model successfully accommodates clusters with zero prevalence and varying prevalences among infected clusters.
• Sample size calculations can be determined to achieve high predictive probability for detecting excessive cluster-level prevalence in disease freedom surveys.

Conclusions:

• The proposed Bayesian method offers a statistically sound and practical approach to sample size determination for disease prevalence studies.
• This method enhances the reliability of human and animal health surveys by accounting for key sources of uncertainty.
• The approach is adaptable for both confirming disease absence and estimating disease burden in clustered populations.