Resolving parameter uncertainty in SIR models through population-level serological surveillance: A synthetic study.

Area of Science:

• Epidemiology
• Mathematical Modeling
• Public Health Surveillance

Background:

• Epidemic models are crucial for understanding disease spread, but rely on surveillance data that captures only a fraction of actual infections.
• Ignoring or improperly accounting for underdetection leads to significant parameter errors (>1000%) and model non-identifiability.
• Transmission rates and detection ratios become mathematically confounded when underdetection is treated as an unknown parameter.

Purpose of the Study:

• To address the fundamental challenges of underdetection in epidemic modeling.
• To demonstrate how integrating seroprevalence data can resolve parameter uncertainty and improve model accuracy.
• To establish a framework for enhanced pandemic preparedness through integrated surveillance.

Main Methods:

• Utilized Bayesian inference on synthetic SIR (Susceptible-Infectious-Recovered) model data.
• Investigated the impact of ignoring underdetection versus explicitly modeling it with unknown detection ratios.
• Assessed the effect of incorporating population-level seroprevalence measurements.

Main Results:

• Models ignoring underdetection exhibit parameter errors exceeding 1000%.
• Explicitly modeling underdetection leads to confounded transmission and detection parameters, resulting in non-identifiable models.
• Integrating seroprevalence data significantly reduces parameter uncertainty by orders of magnitude.
• Accurate inference of transmission dynamics, peak timing, and outbreak size is achievable with integrated serological data.

Conclusions:

• Serological surveillance integration is a mathematical necessity for accurate epidemic modeling.
• This approach resolves fundamental issues of parameter error and non-identifiability caused by underdetection.
• Integrating seroprevalence data is a strategic investment for robust pandemic preparedness.