Bayesian Modeling of Enteric Virus Density in Wastewater Using Left-Censored Data

Area of Science:

• Environmental microbiology
• Quantitative microbial risk assessment
• Statistical modeling

Background:

• Stochastic models are crucial for microbial risk assessment but struggle with left-censored data (non-detects) common in environmental pathogen analysis.
• Enteric virus densities in water frequently fall below detection limits, posing challenges for accurate quantitative uncertainty analysis.

Purpose of the Study:

• To evaluate a Bayesian model's ability to handle left-censored enteric virus density data in wastewater.
• To determine the minimum number of detects required for accurate estimation of virus density and removal efficiency.

Main Methods:

• Simulated left-censored datasets of enteric virus density were generated with varying sample sizes (12, 24, 48) and numbers of detects.
• A Bayesian model was applied to these censored datasets to estimate virus density parameters.
• Estimation accuracy was assessed using root mean square deviation and Kullback-Leibler divergence for virus density and removal efficiency.

Main Results:

• The accuracy of estimating virus density and removal efficiency was significantly influenced by the number of detected values.
• At least eight detects were found necessary for accurate posterior predictive distribution estimation across tested sample sizes.
• Accurate estimation of virus removal efficiency also required a minimum of eight detects in the treated wastewater dataset.

Conclusions:

• The developed Bayesian model successfully accommodates non-detect data in enteric virus density analysis.
• A minimum of eight detects is recommended for reliable microbial risk assessment involving virus density and removal efficiency.
• This finding provides crucial guidance for designing environmental monitoring studies and interpreting results.