Adjusting Incidence Estimates with Laboratory Test Performances: A Pragmatic Maximum Likelihood Estimation-Based...

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Accurate disease incidence data is critical for public policy, especially during pandemics like COVID-19.
• Estimating disease incidence is complicated by imperfect diagnostic tests and variable laboratory performance.
• Existing methods lack pragmatic approaches to adjust for laboratory performance bias in incidence estimation.

Purpose of the Study:

• To develop a statistical method for estimating laboratory performance-adjusted disease incidence.
• To address biases introduced by imperfect diagnostic assays in longitudinal studies.
• To provide a flexible and pragmatic approach applicable to various disease and study settings.

Main Methods:

• Developed a maximum likelihood estimation-based approach using the expectation-maximization algorithm.
• Constructed confidence intervals using bootstrapped-based and large-sample interval estimator methods.
• Evaluated the method's accuracy and convergence through extensive simulations and application to the TrackCOVID study.

Main Results:

• Simulations showed the method converges rapidly and provides accurate estimates across various scenarios.
• Application to the TrackCOVID study indicated that assuming perfect lab performance can lead to inaccurate incidence inferences.
• Both bootstrapped-based and large-sample confidence intervals performed well, with large-sample estimators showing better coverage in extreme scenarios.

Conclusions:

• The developed method effectively adjusts for laboratory performance, yielding more accurate disease incidence estimates.
• This approach is crucial for reliable public health decision-making, particularly when using imperfect diagnostic tests.
• The flexible and pragmatic method can be extended to diverse epidemiological and clinical research settings.