Capturing the pool dilution effect in group testing regression: A Bayesian approach

Abstract

Group (pooled) testing is becoming a popular strategy for screening large populations for infectious diseases. This popularity is owed to the cost savings that can be realized through implementing group testing methods. These methods involve physically combining biomaterial (eg, saliva, blood, urine) collected on individuals into pooled specimens which are tested for an infection of interest. Through testing these pooled specimens, group testing methods reduce the cost of diagnosing all individuals under study by reducing the number of tests performed. Even though group testing offers substantial cost reductions, some practitioners are hesitant to adopt group testing methods due to the so-called dilution effect. The dilution effect describes the phenomenon in which biomaterial from negative individuals dilute the contributions from positive individuals to such a degree that a pool is incorrectly classified. Ignoring the dilution effect can reduce classification accuracy and lead to bias in parameter estimates and inaccurate inference. To circumvent these issues, we propose a Bayesian regression methodology which directly acknowledges the dilution effect while accommodating data that arises from any group testing protocol. As a part of our estimation strategy, we are able to identify pool specific optimal classification thresholds which are aimed at maximizing the classification accuracy of the group testing protocol being implemented. These two features working in concert effectively alleviate the primary concerns raised by practitioners regarding group testing. The performance of our methodology is illustrated via an extensive simulation study and by being applied to Hepatitis B data collected on Irish prisoners.

Reconstructing prevalence dynamics of wildlife pathogens from pooled and individual samples

“…It builds on existing methods for the analysis of pooled samples and advances them by providing a solution for dealing with the dilution effect where biomarker concentration in the pooled sample decreases or turns negative due to mixing with negative or lower-concentration samples [18]. The approach also provides a way to deal with non-standard family distributions of the biomarker, thereby making it more flexible than existing methods [13]. Encouragingly, we found that the model is quite robust against misspecifications of the underlying biomarker distribution.…”

“…This quantitative information offers opportunities, however, that can address both limitations of the binary approach. Although few studies have developed methods to use the full quantitative test results for estimating prevalence from pooled samples [12, 19], the work by [13] in particular has shown how promising this approach can be. They used a Bayesian mixture model approach to estimate prevalence, taking into account the dilution effect based on the distribution of biomarker values (e.g.…”

“…The goal of this model is to estimate the proportion of positive bats using the Ct value of a pooled sample. The analysis of pooled samples can be challenging, leading to a large body of studies on pooled testing (also called group testing or composite testing, depending on the field) addressing the different problems related to pooled samples [11][12][13]. Most studies have focused on pooled testing in the context of laboratory assay cost reduction, where the main challenge is to find the optimal number of samples to pool given an expected proportion of positives [3].…”

“…Most studies on the analysis of pooled samples focus on testing protocols for cost reduction, with the goal of eventually identifying the positive individuals [3,14,15]. Perhaps for this reason, few methods have been developed for explicitly using pooled samples to estimate prevalence in the population [10,12,[16][17][18][19], and even fewer have attempted to use the actual concentration of the infectious agent in the pooled sample to estimate how many of the contributing individuals are positive [12,13]. Furthermore, to our knowledge no methods exist that are able to estimate the proportion of positive individuals using concentration when the underlying distribution of test values does not follow a standard-family (e.g.…”

“…Gaussian) distribution, even though this is the most common situation, especially for wildlife populations [20, 21]. Few methods exist that can incorporate such distributions, and to our knowledge none provide a method for numerically calculating the full probability distribution of test values, instead using approximation methods [13, 19]. Leveraging the information present in the concentration of the infectious agent in pooled samples instead of only using binary negative/positive information can lead to significant improvements in the estimation of prevalence, particularly in the case of disease surveillance in wildlife populations.…”