Simulation of pooling optimization methods for differing infection dynamics,sampling practices,and desired outcomes in surveillance testing

Abstract

The reliability of assumptions made about prevalence for pooling optimization varies greatly among target pathogens and surveillance strategies. When prevalence is unknown and difficult to anticipate, surveillance programs risk generating additional costs if pooling is suboptimal. Different methods of approximating optimal pool size (OPS) vary in precision of optimization, required sampling information, and the logistical demands placed on a laboratory workflow. Hence, it can be unclear how to assess compatibility between pooling optimization methods and the priorities of a surveillance program, sampling practices for the target population, and infection dynamics of the target pathogen. Our aim was to determine the relative performance in maximizing testing economy and cost reduction in different surveillance programs by simulating different pooling optimization methods on data from 280 submissions for bovine viral diarrhea virus (BVDV) surveillance (Nebraska Veterinary Diagnostic Center) and 111 submissions for Theileria orientalis surveillance (Virginia Tech Animal Laboratory Services). True prevalence, OPS, and historical prevalence were determined for each submission, and different optimization methods using fixed pool sizes, historical prevalence, and prevalence estimation testing were trialed on the data through Monte Carlo simulations. Contrasting results were observed between the 2 target pathogens, with historical prevalence being the most reliable optimization method for BVDV and the least reliable method for T. orientalis, which required significantly more tests than truly optimized pooling (p <0.05). Our results demonstrate the need to consider the interplay of infection dynamics, sampling practices, and surveillance priorities when selecting a pooling optimization approach.