A capture-recapture modeling framework emphasizing expert opinion in disease surveillance

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Capture-recapture methods are vital for disease surveillance, estimating total cases from multiple data sources.
• Estimating total disease cases often relies on unverifiable assumptions about surveillance system dependencies.
• Existing methods struggle with unobserved cases and dependency assumptions in surveillance data.

Purpose of the Study:

• To introduce a novel modeling framework for disease surveillance using capture-recapture methods.
• To address the challenge of unverifiable assumptions in estimating disease case counts.
• To provide a flexible and interpretable approach for dependency modeling in surveillance.

Main Methods:

• Advocating a framework focused on a key population-level parameter reflecting surveillance stream dependencies.
• Incorporating expert opinion as prior information to guide estimation.
• Implementing accessible bias corrections and an adapted credible interval approach for inference.
• Applying the framework to human immunodeficiency virus (HIV) surveillance data with three and four streams.

Main Results:

• Successfully estimated the number of human immunodeficiency virus (HIV) positive cases in two real-world datasets.
• Demonstrated the framework's ability to handle realistic, interpretable assumptions under investigator control.
• Enabled principled uncertainty analyses, allowing users to quantify confidence in dependency assumptions.

Conclusions:

• The proposed framework offers a robust and flexible approach to capture-recapture modeling in disease surveillance.
• It enhances the estimation of disease burden by managing dependencies and incorporating expert knowledge.
• The method facilitates more reliable inference and uncertainty quantification for public health applications, particularly for HIV surveillance.