|
Abstract:
|
We begin with a population of size N and identify a set of observed cases with case count OC. Surveillance systems operate with classification error, including false positives among the observed cases, and excluding false negatives from the observed cases. Hence, observed and true prevalence may differ due to classification error. We determine values for true prevalence that are compatible with fixed values for OC and N by constructing classification tables, each based on choosing values for false positive (FP) and false negative (FN) counts. Under fixed OC and N, each choice of FP and FN describes a combination of true prevalence, predictive value positive (PVP) and sensitivity (Se) that yields the specified values for N and OC. Computing these tables over the full range of values for FN and FP obtains every possible combination of true prevalence and classification yielding our OC and N. By considering tables with specified values for PVP and Se, we can obtain a more narrow range of values for true prevalence reflecting the likely or known performance of the surveillance system.
|
