Correcting prevalence estimation for biased sampling with testing errors

Area of Science:

• Epidemiology
• Biostatistics
• Infectious Disease Modeling

Background:

• Prevalence estimation is crucial for understanding infection dynamics but is often biased.
• Biases arise from oversampling symptomatic individuals and inaccuracies in diagnostic tests.
• Naïve prevalence estimators can significantly deviate from true infection proportions.

Purpose of the Study:

• To develop a novel method for infection prevalence estimation.
• To mitigate bias introduced by testing errors and symptomatic individual oversampling.
• To account for stratified testing errors in symptomatic and asymptomatic populations.

Main Methods:

• Development of a new statistical procedure for bias reduction in prevalence estimation.
• Incorporation of stratified error rates for diagnostic tests.
• Implementation of easily accessible algorithms with provided code.

Main Results:

• The proposed method significantly reduces bias in prevalence estimation compared to existing approaches.
• Bias is eliminated in certain scenarios by accounting for stratified testing errors.
• Demonstrated effectiveness through formal results, simulations, and real-world COVID-19 data analysis.

Conclusions:

• The new method offers improved accuracy for infection prevalence estimation.
• It provides a robust tool for epidemiological surveillance and public health decision-making.
• The approach is practical and validated on significant public health data.