Confidence intervals for prevalence estimates from complex surveys with imperfect assays

Area of Science:

• Biostatistics
• Epidemiology
• Survey Methodology

Background:

• Estimating disease prevalence with confidence intervals is crucial for public health.
• Existing methods are limited to either complex surveys with perfect assays or simple surveys with imperfect assays.
• A gap exists in methodology for complex surveys involving imperfect diagnostic assays.

Purpose of the Study:

• To develop and evaluate novel statistical methods for estimating disease prevalence in complex surveys with imperfect diagnostic assays.
• To compare the performance of new methods against established techniques under various simulation scenarios.
• To apply the developed methods to real-world epidemiological data.

Main Methods:

• Utilized the melding method to integrate gamma intervals for directly standardized rates.
• Incorporated adjustments for imperfect assays by estimating sensitivity and specificity.
• Conducted simulations to assess method performance, including coverage probabilities.

Main Results:

• One proposed method demonstrated at least nominal coverage across all simulated scenarios.
• The new methods showed superior performance compared to established methods, particularly in low prevalence settings.
• In certain simulations, the developed methods guaranteed coverage, unlike competing approaches with significantly lower coverage.

Conclusions:

• The developed statistical methods effectively address the challenge of estimating disease prevalence in complex surveys with imperfect assays.
• These methods provide more reliable confidence intervals, especially in scenarios with low disease prevalence.
• The approach was successfully applied to a SARS-CoV-2 seroprevalence survey in the US adult population.