During the height of the COVID-19 pandemic, OSU implemented a longitudinal testing program in which eligible students were required to be tested once per calendar week. Two primary objectives were to control the spread of disease by identifying and isolating students with COVID, and to monitor the prevalence of COVID on campus to inform policy decisions. Daily results were summarized as test-positive rates (# positive tests / # tests taken) which are commonly interpreted as estimates of disease prevalence. However, the test-positive rate is a biased estimator of prevalence under the OSU testing program and many other seemingly innocuous longitudinal testing regimens, especially when they’re effective at controlling disease spread. We illustrate how this bias arises, identify sufficient conditions for prevalence is identifiable from longitudinal testing, and propose a consistent estimator that yields insights into the weekly pattern of COVID infections during the fall 2020 semester.
